designing silvopastoral systems for the amazon: a framework ...

1

DESIGNING SILVOPASTORAL SYSTEMS FOR THE AMAZON: A

FRAMEWORK FOR THE EVALUATION OF NATIVE SPECIES

Thesis submitted in partial fulfilment of the degree Doctor of Philosophy at UCL by

Lucy Jayne Dablin

2

SIGNED DECLARATION

I, Lucy Jayne Dablin, confirm that the work presented in this thesis is my own. Where

information has been derived from other sources, I confirm that this has been

indicated in the thesis.

3

ACKNOWLEDGEMENTS

This work was possible due to the kind help and support of my supervisors, Dr Mark

Lee, Professor Simon Lewis, Dr William Milliken, Dr Alex Monro and my parents. I

would like to thank the many people who volunteered their efforts for this research.

This work was supported by the Natural Environment Research Council

[NE/L002485/1]; the Emily Holmes Memorial Scholarship; and the Bentham Moxon

Trust.

4

ABSTRACT

The integration of trees into pastures, termed silvopasture systems, could make

livestock farming more economically and environmentally sustainable. A framework

to assess the forage potential of native tree species would assist in developing

regionally appropriate silvopasture for pastures in the Amazon. Five tree species of

with reported silvopastoral potential, Senegalia loretensis, Ceiba pentandra,

Erythrina berteroana, Inga edulis and Leucaena leucocephala, were selected to

investigate their potential use in a silvopastoral system. A 26-month field experiment

planted 900 trees of the five species to assess tree mortality, biomass production

and palatability for cattle. Five monospecific replicates of each species and a control

with no trees planted were installed on a Brachiaria brizantha pasture in Madre de

Dios, Peru. After 18 months, S. loretensis and C. pentandra showed double the rate

of mortality compared to the other species and were deemed inappropriate for

continued consideration. Destructive harvests of the remaining three species showed

that total biomass production and hence carbon storage was greater for I. edulis (7.2

Mg dry mass ha-1) than E. berteroana (4.4 Mg dry mass ha-1) or L. leucocephala (4.3

Mg dry mass ha-1) at 24 months. These biomass gains were not at the expense of

grass: edible biomass within the reach of cattle was 1.5 Mg ha-1 in the no tree

control, whereas E. berteroana produced 2.3 Mg ha-1, I. edulis 2.3 Mg ha-1 and L.

leucocephala 2.2 Mg ha-1. Cattle were introduced into each treatment to assess

palatability. Cattle consumed 99% of E. berteroana, 75% of I. edulis and 80% of L.

leucocephala. E. berteroana had a mortality six times greater than other species, but

5

all species have silvopastoral potential. This evaluation of tree species supports

regionally appropriate silvopastoral systems and shows that silvopasture can

increase productivity on existing cleared land and can contribute to mitigating climate

change.

6

IMPACT STATEMENT

Silvopastoral farming systems, where trees are planted in cattle pasture, could make

livestock farming more economically and environmentally sustainable when

compared with traditional models of cattle farming. In some areas, a barrier to the

adoption of silvopasture is a lack of knowledge of which environmentally adapted

trees have forage potential and whether these trees can meet the needs of animal

production. This thesis describes a framework that can be used to evaluate the

silvopastoral suitability of native and locally adapted tree species. The framework

presents a set of design considerations and recommends that silvopastoral trials be

carried out in a well-replicated scientifically robust design over a minimum of a two-

year timeframe. This approach has been designed for on-farm use in a way that can

be adapted by practitioners in different parts of the world.

The framework was applied in the context of a typical Amazon pasture in Madre de

Dios, Peru. Five locally adapted trees were chosen following a review of scientific

literature that suggested locally adapted species that could survive in the pasture

environment and were palatable for cattle. When compared with conventional

models of farming in the region, the silvopastoral system provided gains in

productivity, accruing as increased total biomass, an increase in available forage for

cattle and the sequestration of carbon. A statistical method to provide non-

destructive assessments of the biomass production and carbon sequestration

potential of the tree components of three species is presented.

7

An accomplishment of this study was the establishment of a 3-hectare experimental

silvopastoral system. This serves as a demonstration of the procedures advocated in

this thesis and can continue inspire those who seek to implement silvopastoral

systems. Site visits by stakeholders may increase the adoption of more sustainable

cattle farming techniques. It is hoped that the dissemination of this research via

scientific journals will contribute to a growing body of evidence that supports the

adoption of more environmentally friendly practices in livestock production, of

interest to extension workers, policy makers, governments and businesses.

This work is an attempt to find sustainable ways to intensify the current extensive

livestock production in the Amazon using natural technology and simultaneously

increase the provision of ecosystem services from these landscapes. It is envisaged

that widespread adoption of silvopastoral systems on existing cleared land could

reduce the need for further deforestation, reverse the trend of pasture degradation

and soil erosion, increase biodiversity in agricultural landscapes and contribute to the

mitigation of climatic change. This work is a preliminary assessment but it could be

of use to future researchers and extension workers who may be able to address the

shortcomings of this study and further improve the recommendations for the design

of silvopastoral systems. Finally, the ultimate impact that this study could hope to

achieve is that the communities that rely on cattle farming in the Amazon adopt

these more profitable and sustainable approaches, and that this results in a decline

in the rate of pasture degradation and a decline in the rate of deforestation of the

Amazon.

8

TABLE OF CONTENTS

DESIGNING SILVOPASTORAL SYSTEMS FOR THE AMAZON: A FRAMEWORK

FOR THE EVALUATION OF NATIVE SPECIES ....................................................... 1

SIGNED DECLARATION ....................................................................................... 2

ACKNOWLEDGEMENTS ....................................................................................... 3

ABSTRACT ............................................................................................................ 4

IMPACT STATEMENT ........................................................................................... 6

TABLE OF CONTENTS .......................................................................................... 8

TABLE OF FIGURES ........................................................................................... 13

TABLE OF TABLES .............................................................................................. 20

TABLE OF APPENDICES .................................................................................... 25

CHAPTER I .......................................................................................................... 35

1. Livestock Production .............................................................................. 35

2. Introduction to Silvopastoral Systems .................................................... 37

3. Selection and Management of Species for Silvopastoral Systems ........ 41

4. Traditional Livestock Production Models And The Potential For

Silvopastoral Systems: Influence On Soil Health ....................................... 51

5. Evaluating the Forage Potential of Tree Species ................................... 55

6. Summary and Rationale......................................................................... 60

References ................................................................................................. 62

CHAPTER II ......................................................................................................... 87

1. Objective ................................................................................................ 87

9

2. The Purpose .......................................................................................... 89

3. The Site.................................................................................................. 89

3.1 Site Description ................................................................................ 90

3.2 Floristic Description of Site ............................................................... 92

3.3 Climatic Description of Site ............................................................... 93

3.4 Edaphic Description of Site .............................................................. 95

3.5 Social and Economic History ............................................................ 96

3.6 Future of the Amazon ....................................................................... 97

3.7 Site Preparation ................................................................................ 98

4. Species ................................................................................................ 102

4.1 Evaluation of Desirable Plant Traits and Candidate Trees in

Silvopastoral Systems .......................................................................... 102

4.2. Species Information ....................................................................... 108

4.2.1 Inga edulis ............................................................................... 108

4.2.2 Ceiba pentandra ...................................................................... 112

4.2.3 Erythrina berteroana ................................................................ 114

4.2.4 Senegalia loretensis ................................................................. 116

4.2.5 Leucaena leucocephala ........................................................... 117

4.2.6 Brachiaria brizantha ................................................................. 118

4.3 Design and Management of Tree Nursery ...................................... 120

4.4 Treatment and Sowing of Seeds .................................................... 124

5. Planting Design .................................................................................... 128

6. Experimental Design ............................................................................ 130

6.1 Planting Sequence ......................................................................... 132

References ............................................................................................... 136

10

CHAPTER III ...................................................................................................... 152

Abstract .................................................................................................... 152

1. Introduction .......................................................................................... 153

2. Methods ............................................................................................... 155

2.1. Censuses of Tree Growth .............................................................. 156

2.2. Soil Chemical Analysis .................................................................. 158

2.3 Pollard Treatment ........................................................................... 160

2.4. Effects of Trees on Rates of Pasture Regrowth ............................ 160

3. Results ................................................................................................. 161

4. Discussion ............................................................................................ 177

References ............................................................................................... 185

CHAPTER IV ...................................................................................................... 193

Abstract .................................................................................................... 193

1. Introduction .......................................................................................... 194

2. Methods ............................................................................................... 199

2.1 Measurements of Tree Biomass ..................................................... 199

2.2 Data Exploration ............................................................................. 202

2.3 Model Comparison and Selection ................................................... 203

2.4 Model Validation ............................................................................. 203

2.5 Allometric Models .......................................................................... 204

2.6 Calculation of Carbon Stocks ......................................................... 205

3. Results ................................................................................................. 206

3.1 Measurements of Tree Biomass ..................................................... 206

3.2 Data Exploration ............................................................................. 210

3.3 Model Selection .............................................................................. 213

11

4. Discussion ............................................................................................ 218

References ............................................................................................... 229

Appendices .............................................................................................. 243

CHAPTER V ....................................................................................................... 265

Abstract .................................................................................................... 265

1. Introduction .......................................................................................... 266

2. Methods ............................................................................................... 270

2.1 Experimental Design and Cattle Herd ............................................ 270

2.2. Browsing Detection Using Destructive Measurements .................. 272

2.3. Browsing Detection Using Non-Destructive Measurements .......... 274

2.4 Browsing Detection Using Leaf Counts .......................................... 276

2.5 Measurement Of Browsing Impact ................................................. 276

2.6 Estimation Of Grass Forage ........................................................... 277

2.7 Effect Of Trees and Cattle on Soil Bulk Density ............................. 278

3. Results ................................................................................................. 280

4. Discussion ............................................................................................ 289

References ............................................................................................... 299

CHAPTER VI ...................................................................................................... 307

1. Summary of Research Findings ........................................................... 307

2. Comparison with Previous Research ................................................... 311

3. Theoretical Implications of Findings ..................................................... 315

4. Limitations of the Study ........................................................................ 317

5. Future Research .................................................................................. 321

5.1 Direct Extensions of the Study ....................................................... 321

5.2 Broader Issues for Future Work ..................................................... 321

12

7. Conclusion ........................................................................................... 323

References ............................................................................................... 324

13

TABLE OF FIGURES

Figure 1. Theoretical framework for the evaluation of candidate species for

agroforestry systems ......................................................................................... 88

Figure 2. Location of the study site in Madre de Dios, Peru ..................................... 91

Figure 3. Experimental site in 2002 with the future location of the plots projected .. 92

Figure 4. Experimental site in 2015 with the future location of the plots projected. .. 92

Figure 5. Average (solid line), maximum (dotted line), and minimum (dashed line)

monthly temperature (°C) from the period 1970-2000 for the study site derived

from the WorldClim dataset (Hijmans et al., 2005). ........................................... 94

Figure 6. Average monthly precipitation (mm) from the period 1970-2000 for the

study site derived from the WorldClim dataset (Hijmans et al., 2005). .............. 94

Figure 7. Average monthly level of solar irradiation (kJ m-2 day-1) from the period

1970-2000 for the study site derived from the WorldClim dataset (Hijmans et al.,

2005). ................................................................................................................ 95

Figure 8. Experimental site in October 2016 with the future plots projected. ........ 100

Figure 9. A monumental task. All site clearance was carried out by hand to minimise

disturbance to the soils. .................................................................................. 100

Figure 10. An international team of volunteers involved helped with the pile and

burn. ................................................................................................................ 101

Figure 11. When the site had been cleared of branches, the weeds at the site had

grown to be unmanageable. A glyphosphate herbicide was applied across the

site and the vegetation had to be cleared again. Clearance of this vegetation

14

was done using a hoe and rake. All residues were taken off site in

wheelbarrows. ................................................................................................. 101

Figure 12. Mind map of desirable plant traits in the context of developing

silvopastoral systems at a site in the Amazon that could be advantageous in

achieving the goal of increasing the productivity per hectare through the use of

silvopastoral systems compared with traditional methods of cattle production in

the region. ....................................................................................................... 104

Figure 13. Resources for identifying potential species for inclusion in agroforestry

systems. .......................................................................................................... 105

Figure 14. Inga edulis growth (Height) over time (Months), from: (1) Sobanski and

Marques (2014), (2) Alegre (2005), (3) Kanmegne et al., (2000), (4) Rezende &

Vieira (2019), (5) Butterfield (1996), (6) Butterfield (1995), (7) Gonzalez and

Fisher (1994) ................................................................................................... 110

Figure 15. Biomass after 11 years (1993 – 2004) in five replications of two

treatments, T1 (control) and T6 (interplanted with Inga edulis), in Mg ha-1.

Biomass of weeds and Inga was measured in 1 x 3 m subplots within treatment

plots and converted to a per hectare basis. Biomass of Terminalia amazonia is

the sum of crown foliage and bole wood biomasses of a sample tree for each

treatment plot. Crown and bole biomasses were estimated from volumes

calculated from diameter and height measurements multiplied by foliage density

and wood density, respectively. Once the tree biomass was obtained, biomass

was converted to a per hectare basis depending on the number of trees in the

plot. Total biomass is the sum of the weed and Inga biomass plus the T.

amazonia biomass on a per hectare basis (from Nichols and Carpenter, 2006).

........................................................................................................................ 111

15

Figure 16. Little is known about Senegalia loretensis, but despite it’s sharp spines

the leaves are highly palatable to cattle. ......................................................... 117

Figure 17. (Clockwise from left to right) The first attempt at digging the well, the base

of the steel poles, which contained a 2.5 cm iron bar, alignment of the steel

poles for the nurseries. .................................................................................... 121

Figure 18. The nursery mesh was elevated upon steel poles above the plants, with

enough room to allow air to circulate and to enable easy access for watering the

plants. ............................................................................................................. 122

Figure 19. Some prepared soil for filling the bags for the plants. .......................... 123

Figure 20. View over the nurseries with 10,000 bags of soil. Borders made from

trees define the beds. The bags of soil are arranged in rows to give easy access

for weeding and watering of the plants. In the background is the water tower that

enabled water to the site. ................................................................................ 124

Figure 21. Seed pods and leaves of C. pentandra. ............................................... 125

Figure 22. I edulis .................................................................................................. 126

Figure 23. Seeds of L. leucocephala ..................................................................... 126

Figure 24. Seedlings of L. leucocephala ............................................................... 127

Figure 25. Beds of S. loretensis (foreground) and C. pentandra (background). .... 127

Figure 26. E. berteroana was planted from stakes, pictured here 2 months post

planting in association with the immature pasture grass. ................................ 128

Figure 27. The randomised replicate plot layout, where EB is Erythrina berteroana,

IE is Inga edulis, LL is Leucaena leucocephala, SL is Senegalia loretensis, CP is

Ceiba pentandra and C is a control of grass. .................................................. 131

16

Figure 28. Map of experimental site. 0.09 ha plots as red rectangles, tree rows as

black dashed lines. Tree rows were continuous between plots. There were no

tree lines on the red lines. ............................................................................... 132

Figure 29. Satellite image of experimental site, 0.09 ha parcels as red rectangles.

Image taken in April 2018. Image by Maxar Technologies. ............................. 132

Figure 30. The plastic tubes used to mark the 30 m intervals of the plot corners and

in the background weeds responding to the glyphosphate applied across the

site. ................................................................................................................. 133

Figure 31.Team planting L. leucocephala in January of 2017. .............................. 135

Figure 32. Planting lines of I. edulis........................................................................ 135

Figure 33. Satellite image from April 2017 with the parcels projected, showing faint

tree lines and the establishing grass. .............................................................. 136

Figure 34. Relative growth rate for changes in height of five tree species after

planting at six (C2), twelve (C3) and eighteen (C4) month intervals post-planting

for tree species Leucaena leucocephala, Senegalia loretensis, Inga edulis,

Ceiba pentandra and Erythrina berteroana. .................................................... 164

Figure 35. Relative growth rate for changes in root collar diameter (RCD) and of five

tree species at between planting and six (C2), twelve (C3) and eighteen (C4)

month intervals post-planting for tree species Leucaena leucocephala,

Senegalia loretensis, Inga edulis, Ceiba pentandra and Erythrina berteroana 165

Figure 36. Height (cm) and standard deviation of five tree species growing in a

Brachiaria brizantha cv. Marandú pasture as of census 1. March 2017, 2. July

2017, 3. March 2018 and 4. July 2018. ........................................................... 166

17

Figure 37. RCD (mm) and standard deviation of five tree species growing in a

Brachiaria brizantha cv. Marandú pasture as of census 1. March 2017, 2. July

2017, 3. March 2018 and 4. July 2018. ........................................................... 167

Figure 38. DBH (mm) of five tree species growing in a Brachiaria brizantha cv.

Marandú pasture as of census 2. July 2017, 3. March 2018 and 4. July 2018.

........................................................................................................................ 168

Figure 39. Rows of I. edulis .................................................................................... 169

Figure 40. Row of Ceiba pentandra ....................................................................... 170

Figure 41. Replicate of Senegalia loretensis .......................................................... 171

Figure 42. Leucaena leucocephala in March 2017 (above) and October 2017 (below)

........................................................................................................................ 172

Figure 43. Proportion of alive, dead and trees classified as regrowth of five tree

species in a Brachiaria brizantha cv. Marandú pasture in the SE Amazon 18

months after planting at 0.5 x 7.5 m ................................................................ 173

Figure 44. Mean dry weight biomass and standard deviation (S.D.) of pollarded

biomass (over 2 m) of 18-month-old trees 0.5 x 7.5 m planting design (2667

trees ha-1) for individual trees, sample size represented in parenthesis. Means

sharing the same superscript are not significantly different from each other

(Tukey's HSD, P<0.05) ................................................................................... 175

Figure 45. Mean total dry matter biomass accumulation (Mg ha-1 year-1) and

standard deviation (S.D.) of 30-day regrowth of Brachiaria brizantha cv.

Marandú collected from different treatments, sample size represented in

parenthesis. Means sharing the same superscript are not significantly different

from each other (Tukey's HSD, P<0.05) ......................................................... 176

18

Figure 46. Mean total dry matter biomass accumulation (Mg ha-1 year-1) and standard

deviation (S.D.) of six months regrowth of Brachiaria brizantha cv. Marandú

following the pollarding of associated tree species to 2 m, sample size

represented in parenthesis. Means sharing the same superscript are not

significantly different from each other (Tukey's HSD, P<0.05) ........................ 177

Figure 47. Diagram representing the location of measurements on tree, where CD is

crown diameter, CDep is crown depth, CH is crown height, H is height, RCD is

root collar diameter, SD is stump diameter and DBH is diameter at breast

height. ............................................................................................................. 200

Figure 48. The log transformation of x (root collar diameter (mm) + height (cm))

derived using a linear model for the log transformation of 1) y as total biomass

(g, R2 = 0.77, P <0.01) and 2) y as trunk biomass (g, R2 = 0.76, P<0.01) for

Erythrina berteroana following the equations presented in Table 13............... 216

Figure 49. The log transformation of crown diameter (cm) derived using a linear

model for the log transformation of branch biomass (g, R2 = 0.73, P<0.01) and

leaf biomass (g, R2 = 0.58, P<0.01) for Erythrina berteroana following the

equations presented in Table 13. .................................................................... 216

Figure 50. The log transformation of root collar diameter (mm) derived using a linear

model for the log transformation of total biomass (g, R2 = 0.70, P<0.01), trunk

biomass (g, R2 = 0.71, P<0.01), branch biomass (g, R2 = 0.48, P<0.01) and leaf

biomass (g, R2 = 0.48, P<0.01) for Inga edulis following the equations presented

in Table 13. ..................................................................................................... 217

Figure 51. The log transformation of x where 1) x is root collar diameter (mm) +

height (cm) derived using a linear model for the log transformation of total

biomass (g, R2 = 0.86, P<0.01), trunk biomass (g, R2 = 0.89, P<0.01) and leaf

19

biomass (g, R2 = 0.61, P<0.01) and 2) x is crown diameter (cm) derived using a

linear model for the log transformation of branch biomass (g, R2 = 0.51, P<0.01)

for Leucaena leucocephala following the equations presented in Table 13. ... 217

Figure 52. Map of the cattle trial from east to west Phase I (red), Phase II (blue) and

Phase III (yellow). ............................................................................................ 272

Figure 53. Cattle browsing Inga edulis .................................................................. 281

Figure 54. Browsing line where cattle consumed all the foliage they could reach. . 282

Figure 55. Post-browsing, an Erythrina berteroana row is missing several trees ... 282

Figure 56. Leucaena leucocephala defoliated on low branches ............................. 283

Figure 57. Percentage change of Erythrina berteroana biomass for four assessment

methodologies used to detect biomass removal following cattle browsing. ..... 285

Figure 58. Percentage change of Inga edulis biomass components between

methodologies used to detect biomass removal following a cattle-browsing

experiment. ..................................................................................................... 286

Figure 59. Percentage change of Leucaena leucocephala biomass components

between methodologies used to detect biomass removal following a cattle-

browsing experiment. ...................................................................................... 286

20

TABLE OF TABLES

Table 1. Plantation tree species killed and damaged by exposure to approximately

80-100 cattle in a 3-hour period. Data provided by Camino Verde (2018). . ... 107

Table 2. Criteria used in the selection of the final five species for the present

experiment. Where ++ is best suited to -- worse suited, + or – are weak

evidence to support either way, and ? indicate there was no information found

linking the species to the criteria. .................................................................... 107

Table 3. Plant height, soil cover and dry matter yield of forage grasses Brachiaria

brizantha cv. Marandú, B. humidicola, Paspalum atratum BRA-9610, P.

guenoarum BRA-3824, P. regnelli BRA-0159, P. plicatulum BRA-9661 and

Hemarthria altissima established under a 12 year old rubber plantation

reproduced from Costa et al. (2001). .............................................................. 119

Table 4. Data for the first day of each month from a site in Madre de Dios, Peru (lat -

12.53945°, long -69.17288°) in 2016, including the altitude and azimuth of the

sun at 12:00, the shadow cast by an object of 2 m height and the daylight

duration. .......................................................................................................... 130

Table 5. Plant growth census dates. ...................................................................... 157

Table 6. Height (cm), root collar diameter (RCD mm), and diameter at breast height

(DBH mm) of five tree species in a Brachiaria brizantha cv. Marandú pasture,

with census number (C), time after planting in months (t), mean average (Mean)

standard deviation (S.D.), annual relative growth rate (RGR), n as number of

individuals sampled and λ as annual mortality (Shiel et al., 1995). Means

21

sharing the same superscript are not significantly different from each other

within a census interval (Tukey's HSD, P<0.05).............................................. 163

Table 7. Descriptive statistics of biomass parameters of individual harvested

Erythrina berteroana used in constructing allometric equations, including sample

size (n), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m) and

diameter at breast height (DBH, 1.3 m). Mean, standard deviation (SD),

minimum (min), maximum (max), lower 25th percentile (25th perc), upper 75th

percentile (75th perc), as well as coefficients of skewness and kurtosis are

presented. ....................................................................................................... 207

Table 8. Descriptive statistics of biomass parameters of individual harvested Inga

edulis used in constructing allometric equations, including sample size (n), root

collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m) and diameter at

breast height (DBH, 1.3 m). Mean, standard deviation (SD), minimum (min),

maximum (max), lower 25th percentile (25th perc), upper 75th percentile (75th

perc), as well as coefficients of skewness and kurtosis are presented. .......... 208

Table 9. Descriptive statistics of biomass parameters of individual harvested

Leucaena leucocephala used in constructing allometric equations, including

sample size (n), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m)

and diameter at breast height (DBH, 1.3 m). Mean, standard deviation (SD),

minimum (min), maximum (max), lower 25th percentile (25th perc), upper 75th

percentile (75th perc), as well as coefficients of skewness and kurtosis are

presented. ....................................................................................................... 209

Table 10. Spearman’s correlation of E. berteroana biomass components. Total

biomass (Total), trunk biomass (Trunk), branch biomass (Branch), total foliar

biomass (Leaf), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m),

22

diameter at breast height (DBH, 1.3 m), height, crown volume (CVol) and crown

diameter (CD).................................................................................................. 211

Table 11. Spearman’s correlation of I. edulis biomass components. Total biomass

(Total), trunk biomass (Trunk), branch biomass (Branch), total foliar biomass

(Leaf), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m),

diameter at breast height (DBH, 1.3 m), height, crown volume (CVol) and crown

diameter (CD).................................................................................................. 211

Table 12. Spearman’s correlation of L. leucocephala biomass components. Total

biomass (Total), trunk biomass (Trunk), branch biomass (Branch), total foliar

biomass (Leaf), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m),

diameter at breast height (DBH, 1.3 m) height, crown volume (CVol) and crown

diameter (CD).................................................................................................. 212

Table 13. Log-linear allometric equations demonstrated the lowest Furnival’s Index

(F.I.) against linear or quadratic models for biomass components of 0.5 x 7.5 m

alley rows of Erythrina berteroana, Inga edulis and Leucaena leucocephala 6

months post pollarding to a height of 2 m. Where ln(x) is the log transformation

of the following measurements used as predictors of biomass: Root collar

diameter (RCD mm, 0.05 m), height (cm) and crown diameter (CD, cm). ln(y) is

the log transformation of total biomass (g), trunk biomass (g), branch biomass

(g) or leaf biomass (g). Where b0 is the intercept coefficient, b1 and b2 are the

slope coefficients, all of which are reported with standard error (S.E.) and

significance (P). Lambda (λ) and standard deviation (S.D.) as back

transformation according to Marklund (1987), root mean square error of model

(Model rMSE) and results of K-fold model cross validation (k=10) reported as

root mean squared error (CV rMSE). .............................................................. 215

23

Table 14. Total above ground biomass fraction and carbon (C) stock (Mg ha-1) after

24 months, including previously removed biomass > 2 m at 18 months (Chapter

III) of Erythrina berteroana, Inga edulis and Leucaena leucocephala when 2667

trees ha-1 (0.5 x 7.5 m), using the IPCC default C fraction for tropical forests of

0.47. Where n is sample size and MAI is Mean Annual Increment (Mg ha-1 year-

1) (Eggleston et al., 2006)................................................................................ 218

Table 15. Pairing combinations and random rotation of cattle (8 cattle numbered 1-8)

between four treatments in three, eight day phases of a cattle trial. ............... 271

Table 16. Dry foliar biomass for individual trees (dry matter g tree-1), standard

deviation (S.D.) and number of individuals sampled (n) prior to and post cattle

browsing either mean total tree foliar biomass or foliar biomass harvested <1.6

m on potential browse species Erythrina berteroana, Inga edulis and Leucaena

leucocephala. Significant differences (P<0.05) from prior value to post value

following cattle browsing denoted by an asterisk (*). ....................................... 280

Table 17. Mean average (Av) and standard deviation (S.D.) of change per individual

tree in crown volume (M3) and leaf emergence height (LEH, cm) prior to and

post a cattle browsing experiment on individuals (n) of potential browse species,

Erythrina berteroana, Inga edulis and Leucaena leucocephala. Significant

differences between sampling period prior to and post (P<0.05) denoted by an

asterix (*). ........................................................................................................ 284

Table 18. Mean leaf count per tree (number of leaves tree-1) and standard deviation

(S.D.) prior to and post a cattle-browsing trial on individuals (n) of potential

browse species Erythrina berteroana, Inga edulis and Leucaena leucocephala.

Significant differences (P<0.05) between leaves counted prior to and post

browsing are denoted by an asterix (*). ........................................................... 285

24

Table 19. Impacts measured following a cattle trial on three potential browse species

Erythrina berteroana, Inga edulis and Leucaena leucocephala. Impact types

were classified as browsing-type damage (Browsed) or breakage-type damage

(Broken) of two categories, branches (all species) and leaves (I. edulis).

Average diameter (mm), average height (cm) and maximum height (cm) and

respective standard deviations (S.D.) reported on a number (n) of individual

trees. Significant differences between species (P<0.05) denoted by asterisk (*).

........................................................................................................................ 287

Table 20. Edible and non-edible fractions of mean average dry weight grass biomass

(Mg ha-1) produced in four treatments, reported for sampling carried out prior to

and post the introduction of cattle. .................................................................. 289

25

TABLE OF APPENDICES

Appendix 1. Table. The wet to dry relationship for biomass components of

subsamples.taken from Inga edulis, Erythrina berteroana and Leucaena

leucocephala, where R2 is the coefficient of determination, RSE is the residual

square error and n is the sample size. ............................................................ 243

Appendix 2. Figure. Total biomass (g) and root collar diameter (mm) of Inga edulis

plotted using the linear regression line (y = aX + b) showing clockwise from left

(a) scatter plot and linear regression line, (b) departure from normal Q-Q plot

and (c) residuals vs. leverage ......................................................................... 244

Appendix 3. Figure. Relationship between total biomass (g) and root collar diameter

(mm) for Inga edulis plotted using the quadratic regression model (equation y =

ax2 + bx + c) showing clockwise from left (a) scatter plot and polynomial

regression line, (b) departure from normal Q-Q plot and (c) residuals vs.

leverage .......................................................................................................... 245

Appendix 4. Figure. Relationship between total biomass (g) and root collar diameter

(mm) for I. edulis plotted using the log-linear transformation ln(y) = a + b * lnx

showing clockwise from left (a) scatter plot and regression line, (b) departure

from normal Q-Q plot and (c) residuals vs. leverage ....................................... 246

Appendix 5. Table. Allometric equations following the linear model for biomass

components of 0.5 x 7.5 m alley rows of Erythrina berteroana 6 months post

pollarding to a height of 2 m. Where (x) is the measurement used as a predictor

of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height

26

(DBH (mm), 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3)

and crown diameter (CD cm), (y) is the log transformation of total biomass(g),

trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept

coefficient, b1 as the slope coefficient, both reported with standard error (S.E.)

and significance (P). Coefficient of determination (R2), root mean square error of

model (Model RMSE, which is the same as Furnival’s Index for untransformed

data) and results of K-fold model cross validation (k=10) reported as root mean

squared error (CV RMSE). .............................................................................. 247



pollarding to a height of 2 m. Where (x) is the root collar diameter (RCD mm,

0.05 m), or RCD and Height (cm) as predictors of biomass. (y) is the log

transformation of total biomass (g), trunk biomass (g), branch biomass (g) or

leaf biomass (g), b0 as the intercept coefficient, b1 and b2 as the slope coefficient,

both reported with standard error (S.E.) and significance (P). Coefficient of

determination (R2), root mean square error of model (Model RMSE, which is the

same as Furnival’s Index F.I. for untransformed data) and results of K-fold

model cross validation (k=10) reported as root mean squared error (CV RMSE).

........................................................................................................................ 248


components of 0.5 x 7.5 m alley rows of Inga edulis 6 months post pollarding to

a height of 2 m. Where (x) is the measurement used a predictor of biomass:

Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm,

1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and crown

diameter (CD cm), (y) is the log transformation of total biomass (g), trunk

27

biomass (g), branch biomass (g) or leaf biomass (g), b0 as the intercept








a height of 2 m. Where (x) is the root collar diameter (RCD mm, 0.05 m), or

RCD and Height (cm) as predictors of biomass, (y) is the log transformation of

total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0

as the intercept coefficient, b1 and b2 as the slope coefficient, both reported with

standard error (S.E.) and significance (P). Coefficient of determination (R2), root

mean square error of model (Model RMSE, which is the same as Furnival’s

Index F.I. for untransformed data) and results of K-fold model cross validation

(k=10) reported as root mean squared error (CV RMSE)................................ 250


components of 0.5 x 7.5 m alley rows of Leucaena leucocephala 6 months post

pollarding to a height of 2 m. Where (x) is the measurement used as predictor of

biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height

(DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and

crown diameter (CD cm), (y) is the log transformation of total biomass (g), trunk




28







0.05 m), or RCD and Height (cm) as predictors of biomass, (y) is the log


leaf biomass (g). b0 as the intercept coefficient, b1 and b2 as the slope coefficient,





........................................................................................................................ 252

Appendix 11. Table. Allometric equations following a quadratic model for biomass




(DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV, cm3)

and crown diameter (CD cm), (y) is the log transformation of total biomass (g),

trunk biomass (g), branch biomass (g) or leaf biomass (g), b0 as the intercept




29



Appendix 12. Table. Allometric equations following the quadratic model for biomass










........................................................................................................................ 254



a height of 2 m. Where (x) is the measurement used as a predictor of biomass:

Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm,

1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV, cm3) and crown

diameter (CD cm), (y) is the log transformation of total biomass (g), trunk







30



a height of 2 m. Where (x) is the root collar diameter (RCD mm, 0.05 m), or

RCD and Height (cm) as predictors of biomass, (y) is the log transformation of

total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0

as the intercept coefficient, b1 and b2 as the slope coefficient, both reported with

standard error (S.E.) and significance (P). Coefficient of determination (R2), root

mean square error of model (Model RMSE, which is the same as Furnival’s

Index F.I. for untransformed data) and results of K-fold model cross validation

(k=10) reported as root mean squared error (CV RMSE)................................ 256





(DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and

crown diameter (CD cm), (y) is the log transformation of total biomass (g), trunk










31








........................................................................................................................ 258

Appendix 17. Table. Allometric equations following the log-linear model for biomass


pollarding to a height of 2 m. Where ln(x) is the log transformation of the

following measurements used as predictors of biomass: Root collar diameter

(RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter

(SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm) ln(y) is

the log transformation of total biomass (g), trunk biomass (g), branch biomass

(g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient,


determination (R2), lambda (λ) and standard deviation (S.D.) as back

transformation according to Marklund (1987), Furnival’s Index (F.I.) according to

Kershaw et al. (2016), root mean square error of model (Model RMSE) and

results of K-fold model cross validation (k=10) reported as root mean squared

error (CV RMSE). ............................................................................................ 259



pollarding to a height of 2 m. Where ln(x) is the log transformation of root collar

32

diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass.

ln(y) is the log transformation of total biomass (g), trunk biomass (g), branch

biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope

coefficient, both reported with standard error (S.E.) and significance (P).

Coefficient of determination (R2), lambda (λ) and standard deviation (S.D.) as

back transformation according to Marklund (1987), Furnival’s Index (F.I.)

according to Kershaw et al. (2016), root mean square error of model (Model

RMSE) and results of K-fold model cross validation (k=10) reported as

root mean squared error (CV RMSE). ............................................................. 260



a height of 2 m. Where ln(x) is the log transformation of the following

measurements used as predictors of biomass: Root collar diameter (RCD mm,

0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm,

0.3 m), crown volume (CV cm3) and crown diameter (CD cm) ln(y) is the log

transformation of total biomass(g), trunk biomass (g), branch biomass (g) or leaf

biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient, both

reported with standard error (S.E.) and significance (P). Coefficient of





error (CV RMSE). ............................................................................................ 261



33

a height of 2 m. Where ln(x) is the log transformation of root collar diameter

(RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass. ln(y) is

the log transformation of total biomass(g), trunk biomass (g), branch biomass (g)

or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient,






error (CV RMSE). ............................................................................................ 262



pollarding to a height of 2 m. Where ln(x) is the log transformation of the

following measurements used as predictors of biomass: Root collar diameter

(RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter

(SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm) ln(y) is

the log transformation of total biomass(g), trunk biomass (g), branch biomass (g)

or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient,






error (CV RMSE). ............................................................................................ 263

34



pollarding to a height of 2 m. Where ln(x) is the log transformation of root collar

diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass.

ln(y) is the log transformation of total biomass(g), trunk biomass (g), branch

biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope

coefficient, both reported with standard error (S.E.) and significance (P).

Coefficient of determination (R2), lambda (λ) and standard deviation (S.D.) as

back transformation according to Marklund (1987), Furnival’s Index (F.I.)

according to Kershaw et al. (2016), root mean square error of model (Model

RMSE) and results of K-fold model cross validation (k=10) reported as

root mean squared error (CV RMSE). ............................................................. 264

35

CHAPTER I

AMAZON PASTURES: THE

POTENTIAL FOR SILVOPASTORAL SYSTEMS

1. Livestock Production

The world’s population is predicted to reach 9.6 billion by the year 2050

(Alexandratos and Bruinsma, 2012). Meat and milk currently provide a third of global

protein intake and demand for livestock products is expected to increase by 70 to

80% by the year 2050 (Rosegrant et al., 2009; Herrero, 2015). Pasturelands cover

35 million km2, or 68% of the total global land area dedicated to agricultural

production (White et al., 2000). The rising demand for livestock products is likely to

be met with a concurrent expansion of livestock production into wilderness areas

(Thornton and Herrero, 2010).

To date, deforestation has resulted in the loss of more than a third of all global forest

(Hansen et al., 2013). An example of this is the 413,506 km2 that have been

deforested since 1988 in the Brazilian Amazon to be replaced by pastures for the

production of cattle (Downing, 2019; PRODES, 2019). Central and Latin America

have one of the highest rates of land conversion for livestock production (Wassenaar

et al., 2007). It is estimated that the livestock sector drives climatic change

contributes 14.5% of the total anthropogenic greenhouse gases (GHG) emissions

(Gerber et al., 2013). It is a sector that may struggle to adjust to climatic change, as

36

livestock producers must contend with predictions of increased pressure on

productive land, water shortages and declining forage production under rising

temperatures (Lee et al., 2017; Fellman et al., 2018).

In order to meet the growing demand for livestock products, efforts should focus on

intensifying land that is already dedicated to livestock production. This would have

the dual benefit of avoiding further emissions of GHGs and ensure the preservation

of biodiversity and wilderness areas (Godde, 2018). Increasing the productivity per

hectare of land in a sustainable way is known as sustainable intensification

(Carvalho et al., 2018). Interest has been growing in the potential of sustainable

intensification to integrate multiple land uses that can result in increases in economic

and environmental benefits from the same unit of land (Schultze-Kraft et al., 2018).

Integrating cattle with crops, rotational stocking systems and the use of improved

forages have led to advances in the intensification of cattle farming (Bogaerts et al.,

2017). Farms that integrate trees and shrubs in a practice known as silvopasture

have been documented to have substantial increases in on-farm productivity and

observed benefits to biodiversity (Murgueitio et al., 2011). A critical challenge facing

humanity is to reconcile the production of food for a growing population with the

protection of natural ecosystems while confronted with uncertain predictions of the

impacts of climatic change (Rojas-Downing et al., 2017).

37

2. Introduction to Silvopastoral Systems

Agroforestry systems, as defined by the Food and Agriculture Organisation, are

“traditional and modern land-use systems where trees are managed together with

crops and/or animal production systems in agricultural settings” (FAO, 2016). Many

terms are used to describe arrangements where trees are incorporated into livestock

production systems. Hecht (1982) makes a distinction between “silvopastoral”

systems, “a situation where animals graze a ground cover crop grown under

plantation trees,” and “agropastoral” systems, which “encompass livestock as well as

trees grown for food (human or animal)”. Both may be viewed as “integrated farming

systems” (Hecht, 1982). For the purposes of this discussion, silvopastoral systems

are defined according to Kaur et al. (2002) as an agroforestry practice that integrates

livestock, forage production, and forestry on the same land-management unit.

The incentive to increase production on existing lands is growing as property rights

in the Amazon are strengthening and landowners reach the boundaries of their

properties. Brazil’s commitment to the 15th United Nations Conference on Climate

Change in 2009 resulted in the Low Carbon Agriculture plan, a national credit

incentive that has supported the conversion to silvopastoral systems of

approximately 2 million hectares (Embrapa, 2016; Santos et al., 2018). Many

systems aim to coproduce timber, and often favor exotics such as Tectona grandis

and Eucalyptus.

Farmers can manage trees to provide a range of products and services.

Management of trees commonly include planting arrangements such as living fence

posts, fodder banks or alley farming, a term defined by Kang et al. (1990) as “the

38

cultivation of food crops between hedgerows of multipurpose trees” (Russo and

Botero, 1994; Sanchez, 2002). Appropriate management of agroforestry systems

can help farmers increase production of crops, fuelwood, timber and fodder (Russo

and Botero, 1994). Additionally, a range of ecosystem services are associated with

the presence of trees in agricultural systems, some of which directly benefit the

productivity of agri-ecosystems, wheras others accrue as less-tangible global good

by, for example, increasing the amount of carbon dioxide sequestered.

The presence of trees helps to maintain and improve soil fertility (Hohnwald et al.,

2005; Murgueitio et al., 2011). Trees can act as ‘nutrient pumps’, as they have the

advantage of accessing nutrients from deep soil horizons (Scholes and Archer 1997;

Sanchez, 2002), that are recycled into the system as leaf or root litter and can

contribute to crop available nutrients (Menezes et al., 2002). Nitrogen fixing trees are

commonly employed in agricultural systems as they can increase the nitrogen inputs

of a system (Dommergues, 1987). Some tree legumes have an advantageous

relationship with rhizobia; bacteria that infect roots and form nodules which fix

nitrogen (Downie, 2014). The extent to which tree legumes can access or contribute

nutrients to a given system is dependent on, amongst other factors, mycorrhizal

associations, soil, climate, species and management (Szott et al., 1991). Trees also

promote complex soil food webs (Montagnini and Ashton, 1999) and increase the

habitat available to ecosystem engineers such as decomposers, predators and

parasitoids that control harmful insects (Giraldo et al., 2011; Murgueitio et al., 2011).

Silvopastoral systems can offer improved production and sustainability over

traditional pasture systems. In temperate climates, some silvopastoral systems have

39

been productive for over 4,500 years (Stevenson and Harrison, 2014). The presence

of shade trees can reduce heat stress in cattle and may increase their weight, milk

production or reproductive success. The inclusion of trees can provide a useful

adaptation among smallholder livestock farmers in response to climatic changes. For

example, in times of drought, fodder production by trees may continue when pasture

productivity decreases due to water stress and therefore the (Altieri et al., 2015).

Silvopastoral systems also generate additional products that can increase a farmer’s

income, such as timber, fuelwood, fruits and fodder for cattle. In the Amazon there

are low stocking densities, typically less than one cattle per hectare (Hohnwald et al.,

2006). A pasture which incorporated tree legumes (Leucaena leucocephala) in

Western Australia reported record liveweight gains for cattle and a stocking rate of 3-

4 animals per hectare (Jones, 1986 in Russo and Botero 1994). In the South Eastern

United States the native and endangered longleaf pine Pinus palustris was used

within a silvopastoral system which subsequently produced more profits than either

traditional ranching or traditional forestry (Stainback and Alavalapati, 2004). In this

approach, a threatened species was incorporated into a silvopastoral system which

resulted in both a profitable farming systems and the continued promotion of the

existence of Pinus palustris.

Silvopastoral systems also have considerable potential as a landscape level tool for

conservation. The increased complexity of silvopastoral systems results in a greater

provision of ecosystem services and increased habitat for wildlife (Mcadam et al.

2007). In areas of fragmented forests, silvopastoral systems can increase

connectivity and act as wildlife corridors (Ibrahim et al., 2006). Silvopastoral systems

may sequester more carbon than traditional pasture (Haile et al. 2010). The inclusion

40

of certain forages containing tannins may reduce methane emissions from livestock

production (Hess et al., 2005) and control livestock parasites (Nguyen et al., 2005).

Silvopastoral systems represent an opportunity for sustainable intensification and

livelihood diversification, while simultaneously providing benefits for the provision of

ecosystem services and biodiversity conservation. Such systems can challenge the

traditional view of livestock production as the antithesis of biodiversity conservation.

Livestock production in the Amazon is constrained by the complex agro-ecological

problem of early pasture degradation. Within ten years pasture may reach an

advanced stage of degradation (Martínez and Zinck, 2004). Some direct causes at

the farm level of this phenomenon include decreasing soil fertility, water stress,

insect pests, inadequate grass species, weed invasion and overstocking, whereas

underlying causes of pasture degradation include low return yields, lack of technical

knowledge, perverse policy environments, and insecure property rights (Hohnwald,

2016). Estimates of the extent of pasture degradation are not consistent, but authors

indicate that up to half of the pastures sown in Brazil’s rain forest may be degraded

to some degree (Serrao et al., 1993). The presence of forage trees can enhance the

productivity of pasture systems compared to grass monocultures, with effects

varying dependent on the arrangement of trees, selection of species and soil

conditions of the site (Szott et al., 1991).

Relatively few studies have attempted to increase the productivity of Amazonian

livestock production systems using silvopastoral techniques. A successful

rehabilitation from bulldozed land to a productive silvopastoral system is reported for

a study in Yurimaguas in 1988. The experiment established Centrosema

41

macrocarpum as a forage ground cover crop between peach palm Bactris gasipaes.

Eighteen months after establishment, the Centrosema macrocarpum was grazed.

Average increase in live-weight gains of the cattle was 445 g-1animal-1 per day, an

improvement over traditional grazing systems in the region. Soil bulk density,

mechanical resistance, acidity and aluminium saturation decreased considerably

(Arevalo et al., 1998). Silvopasture can therefore present a unique opportunity to

rehabilitate degraded pastureland and increase the productivity of marginal land.

Such studies reinforce the potential for silvopastoral systems to outcompete

traditional grazing management in the tropics while increasing the sustainability of

livestock production. A framework to assist in the evaluation of candidate species for

inclusion in silvopastoral systems could help interested stakeholders participate in

trialling species they identify as having potential, while generating data that

documents the success of the species.

3. Selection and Management of Species for Silvopastoral Systems

Certain characteristics are shared by species that perform well in silvopastoral

systems; they fix nitrogen, are fast growing, demonstrate high foliage productivity

and nitrogen content, are palatable and readily eaten by animals when fresh,

nutritious for livestock, exhibit tannin content lower than 5%, can withstand trampling,

coppicing, heavy grazing pressure or even complete defoliation, show rapid rates of

regrowth from numerous growing points on the remaining stems, develop a vigorous

tap root, retain their leaves in the dry season, have small leaflets and open crowns

which allow light to reach the ground (Devendra, 1990; Arevalo et al., 1998;

Sanchez, 2002; Vitti et al., 2005; ; Hohnwald, 2016).

42

Appropriate species for silvopastoral systems may be specific to the local

environmental and climatic variables that can affect production. What is proven to

work in one context can fail in another. For example, Leucaena leucocephala is a

nitrogen fixing tree leguminous native to the Yucutan peninsular of Mexico and used

worldwide as a forage tree. When trialed in the acidic Utisols of Yurimaguas in the

Peruvian Amazon failed to be productive due to high levels of aluminum saturation

(Sanchez, 1987 in Loker, 1994). On acidic, low base status soils species such as

Fleminga macrophylla and Erythrina poeppigiana have been shown to be more

productive than L. leucocephala (Kass et al., 1992).

Physical and chemical soil characteristics effect the plants and species they support.

It is important to understand the effects of soil characteristics on a species success.

Degraded, acidic soils are prevalent across the Americas, yet a paucity of studies

have evaluated the adaptability native tree species to these conditions for

consideration in incentive-supported reforestation schemes (Tilki and Fisher, 1998).

The majority of research has focused on Costa Rica, with results demonstrating the

importance of using locally adapted species for rehabilitation of degraded lands

(Montagnini and Sancho, 1990; Perez et al., 1993; González and Fisher, 1994).

Despite the paucity of studies that consider the Amazon, much of the knowledge that

has been acquired regarding the impact of soils and associated biota on trees is

transferrable. For example, data acquired by Carpenter et al. (2004) suggested

survival and growth of trees in their study were negatively correlated with soil

erosion. This observation may be due to the negative relationship between

arbuscular mycorrhizal fungi and levels of soil erosion (Carpenter et al., 2001). Trees

43

associated with arbuscular mycorrhizal fungi have shown improved utilisation of

phosphorous, a nutrient that is frequently limiting to plant growth in the tropics.

Inoculation with appropriate strains of arbuscular mycorrhizal fungi is important as it

can significantly increase plant survival and productivity.

Characteristics that may increase success of a species in the context of pasture

lands in the Amazon include tolerance to high levels of irradiance and aluminium,

fast growth and biomass production that enables trees to outcompete weeds and

certain species may be able provide an advantageous association with neighbouring

trees or crops through their ability to fix atmospheric nitrogen. Nichols et al. (2001)

studied techniques to enhance the growth of Terminalia amazonia on degraded

Utisols in Costa Rica. Four years after planting, T. amazonia inter-planted with the

leguminous nitrogen fixing tree Inga edulis (with no fertiliser applied) was 15% taller

than all other treatments, including T. amazonia treated with annual applications of

fertiliser. Transfer of nitrogen from nitrogen fixing leguminous trees to pasture

grasses has also been recorded (Sierra and Nygren, 2006; Sierra et al., 2007).

Experimental evaluation of local species led one Colombian farmer to discover the

silvopastoral benefits of an endemic and rare legume, Mimosa trianae, which

outperformed the other species in the trial and shows promise for the development of

silvopastoral systems of the Andean foothills (Calle et al., 2012). Native species

consistently show promising results in trials, but are frequently overlooked.

Butterfield (1995) screened 84 species of hardwoods for forestry development

potential in Costa Rica and found that of the species whose performance was in the

top 25%, 67% were native. In one of the only studies of its kind, research that

44

considered capoeira (secondary vegetation) species as supplementary forages in

Northeastern Brazil highlighted many species that could be of potential use to

silvopastoral arrangements in the Amazon. Hohnwald (2016) used a randomized

block design to assess relative palatability, height, biomass production, and chemical

composition of foliage. Their results highlighted the potential of unexplored species

present in diversity of the capoeira for formal integration into livestock production.

The promising results of this research suggest similarly useful but unexplored plant

capital may exist across Amazonia.

The International Livestock Centre for Africa defines four management strategies for

silvopastoral systems: fodder tree banks, intensive feed gardens, alley grazing via a

rotational system or alley grazing via a permanent system (Reynolds and Cobbina,

1992). ‘Fodder tree banks’ or ‘tree protein banks’ are closely spaced planting

arrangements at high densities. Livestock access to the trees can be controlled by

fencing. Fodder banks and intensive feed gardens incur additional management

inputs such as labour but offer opportunities for greatly increased yields due to

higher planting densities (Reynolds and Cobbina, 1992). More than 10,000 fodder

shrubs or trees per ha-1 may be considered an ‘intensive’ silvopastoral system

(Murgueitio et al., 2011). Hedge (1983) recommended a density of 10 to 15 trees per

m2 for Leucaena leucocephala (Devendra, 1990). Pathak et al. (1980) report higher

yields from trees at a density of 40,000 trees ha-1 than 15,000 trees ha-1 (Sanchez,

2002). Savory and Breen (1979) found a planting density of 60,000 trees ha-1

produced greater yields than 10,000 or 30,000 trees ha-1 (Sanchez, 2002).

Intensive feed gardens are cut and carry systems similar to fodder tree banks.

45

However, in the intensive feed garden system, livestock are kept separate from the

forage plants. Livestock can be housed and fed in stalls, or pasture grazed and

offered supplementary forage. An example of management of this system may

include closely spaced hedgerows of forage plants (2 - 4 m spacing) with grass sown

between rows (Reynolds and Cobbina, 1992).

A rotational system of alley grazing uses a wider spacing between hedgerows (3 - 4

m) that livestock graze directly. Livestock are rotated between plots of forage plants,

allowing plants time to rest. For forage tree Gliricidia sp., a suitable rotation may be 2

weeks of direct grazing followed by 8 to 10 weeks of rest period (Reynolds and

Cobbina, 1992). The alley grazing-rotational system has been used successfully in

El Hatico nature reserve in Colombia, where conventional ranching management of

pasture was replaced with an intensive silvopastoral system with a resultant increase

in milk production of 130% (Murgueitio et al., 2011).

A permanent system of alley grazing may use the widest hedgerow spacing (7-10

m). Free ranging livestock graze the system directly. Defoliation intensity of fodder

trees is offset by the higher palatability of grasses. Knowledge of the effect of trees

on animal nutrition and palatability is critical as care must be taken to manage the

exposure of livestock to excessive intake of secondary compounds (e.g. Mimosine in

the highly palatable Leucaena leucocephala) (Reynolds and Cobbina, 1992).

Of particular relevance to the alley grazing systems are the interactions between

trees and pasture grasses. In tree-pasture systems there may be competition

between the trees and the pasture for light, moisture and nutrients (Sousa et al.,

46

2015). Selecting deep-rooted tree species may reduce this competition. In addition,

shading caused by trees can have negative impacts on the productivity of pasture,

which may reduce potential stocking capacity for livestock production (Parsons et al.,

1983; Jackson and Ash, 1998). For this reason, in alley systems, hedgerows of

forage trees can be planted on east-west lines to minimise pasture shading. Pasture

productivity may also be greater in silvopastoral systems (Sousa et al., 2015). As an

additional benefit, grass species adapted to shade generally have improved nutrient

composition than species tolerant to high levels of irradiance (Russo and Botero,

1994). At a planting density of 6 x 6 m in Costa Rica, the legume tree Erythrina

poeppigana (subfamily Fabeaceae) was shown to positively affect yields of some

adjacent pasture grasses species such as Panicum maximum, Brachiaria brizantha,

B. humidicoia and Cynodon nlemfuensis, but decrease yields of other species such

as B. dictyoneura and Pennisetum purpureum (Russo and Botero, 1994). In practice,

any positive effect of trees on pasture productivity may be negated at high stocking

densities combined with low canopy cover as livestock may seek shade and

therefore inhibit tree growth and degrade shaded grasses and soil.

A range of destructive and non-destructive techniques have been developed to

determine pasture productivity. Non-destructive techniques can also give good

estimates of pasture yield; Mannetje (2000) categorised these indirect methods as

visual estimation (including the pasture plate meter and pasture capacitance meter

techniques) and estimation based on grass height and density measurements.

Destructive techniques are generally accepted to be a more reliable indicator of

pasture productivity, but it is important to include an adequate number of samples

(Wilm et al., 1944; Frame, 1981). Grazed swards display inherent variation, which

47

means a large number of samples are required (Shaw et al., 1976; Vermeire et al.,

2002). Sampling is usually done by quadrats; it has been found that, particularly in

row sown pastures, rectangular quadrats exhibit decreased variation than square

ones (Mannetje, 1978). The cutting height of samples also affects estimates of

pasture yield. Thomas and Laidlaw (1981) recommend that for forage estimates,

cutting height be selected at minimum grazing height used by the type of animals the

experiment considers and the experiment’s objectives. This is in contrast to Hodgson

(1979) who suggests cutting at ground level. Ultimately, the technique used to

determine pasture productivity in a study will be determined by time and cost

constraints.

Cattle browse on trees and shrubs, and trees and shrubs can be managed as

resources for cattle production. There is need for more research into which species

cattle browse, and the amount and type of browse consumed by cattle of each

species. To determine how to best document the interaction of cattle with browse at

the species level it is necessary to review the existing literature on how plants

respond and adapt to the herbivory. Research on herbivory is highly relevant to the

present study, yet such information is scarce, particularly at the species level.

Therefore, the following discussion seeks to understand the underlying processes by

which plants respond to defoliation. Trees and shrubs show varying levels of

adaptive responses to defoliation, which may be a result of herbivory by arboreal or

terrestrial animals, or by humans harvesting foliage or wood for a variety of purposes

(e.g. coppicing) (Hardesty and Box, 1988). Defoliation may affect the productivity of

trees or cause mortality in species that are not adequately adapted to the type of

defoliation they are subjected to. The response to defoliation is species specific, and

48

can also be influenced by the type of defoliation, intensity and/or frequency of

defoliation and environmental factors such as water availability in the period of

defoliation. Management of species for their foliage or wood therefore relies on an

understanding of the implications of different management strategies on productivity.

Erdmann et al. (1993) highlight the difficulty of comparing the biomass productivity of

trees between studies due to inter-study inconsistencies and incomplete information,

as important parameters such as the age of plants, cutting heights, plant densities

and plant component biomass are not always specified (Erdmann et al., 1993).

The majority of work on defoliation has focused on effects of defoliation frequency

and defoliation intensity. However, the ‘cutting’ height has also been shown to be of

importance in the management of some species. For example, Sesbania grandiflora

experiences mortality following repeated cutting of the main stem (Ella et al., 1989),

but has been reported to survive well if only side branches are removed (Stür et al.,

1994). Dry matter productivity for Leucaena leucocephala was shown to vary with

cutting height by Osman (1981). Osman (1981) reported greater yields from a cutting

height of 90 cm compared to heights of 15 cm or 150 cm. Similarly, variations in yield

as a response to cutting height have been shown to exist in other forage trees such

as Calliandra calothyrsus (Catchpoole and Blair, 1990). Other species, such as

Gliricidia sepium, exhibit less response to cutting height. Erdmann et al. (1993)

determined a lower yield at a cutting height of 0 cm but no significant difference in

dry matter yield between a cutting height of 25 cm or 100 cm.

Defoliation intensity can be conceptualized as a measure of the amount of

vegetation removed relative to the total plant biomass in a defoliation event.

49

Defoliation frequency is the time between defoliation events. Intensity and frequency

of defoliation are often combined and compared in studies. Abbott et al. (1993)

developed a ‘defoliation index’, where visual estimates of manual defoliation

intensities were made and combined with defoliation frequency. For example, a

defoliation index of 100 would be derived from four seasonal annual defoliations of

25%, or one single defoliation event of 100%. They conclude, in the case of

Eucalyptus marginata, that repetitive, low-intensity defoliation has a greater impact

on growth (measured by stem diameter) than defoliation events at low frequency but

with high-intensity. Teague (1989) found Acacia karroo trees response to more

frequent defoliations (8 weeks versus 12 weeks) resulted in a higher yield, and that

they adjusted to cope with frequent defoliation. In contrast, Erdmann et al. (1993)

found Gliricidia sepium produced low biomass when cut at 3-week intervals, which is

consistent with the findings of Ella et al. (1989) and Duguma et al. (1988) for

Gliricidia sepium and also those of Guevarra et al. (1978) for Leucaena

leucocephala.

Studies that neglect to compare different defoliation frequencies and intensities may

overlook important inter-species variation important for the appropriate management

of individual species. By comparing growth rates with change in mean growth rates

management strategies can be tailored to optimise the cutting interval and result in

higher productivity and less risk of mortality (Stür et al., 1994). Ella et al. (1991)

demonstrated the benefit to productivity of a long period before the first cut or

harvest of forage trees by showing that age at first harvest was positively related to

yield at subsequent harvests. They found shorter cutting intervals were positively

correlated with an increase in leaf biomass, whereas longer cutting intervals

50

generally stimulated production of woody biomass (Ella et al., 1991). There are

significant differences in the effects of browsing or coppicing on species (Hardesty

and Box, 1988). Browsing is often less severe than coppicing as the ratio of wood to

leaf area removed is generally lower. In turn, this may mean there are more buds

and leaf regrowth area left from which trees may regenerate foliage. As Hardesty

and Box (1988) assert in their study of the effects of coppicing and browsing on

species in North-eastern Brazil, there is weak evidence to support the assumption

that defoliation may stimulate growth which exceeds that of plants which do not

experience defoliation. This response is highly dependent on the species and the

context in which they are found.

Only by measuring these characteristics and adaptations can species and

appropriate management strategies be identified which are likely to result in

increased productivity of browse components for animal consumption, of interest to

the development of silvopastoral systems. Furthermore, context specific factors such

as soil health, climate or presence of pests may result in considerable intra-species

variations and should not be ignored. Trials that evaluate these characteristics of

species must be carried out in a context appropriate to the application of the results.

Methodologies that experimentally evaluate the impact of browsing on foliage are

needed in order to measure the context specific potential production of browse.

51

4. Traditional Livestock Production Models And The Potential For Silvopastoral

Systems: Influence On Soil Health

Silvopastoral systems have been proposed as a method for rehabilitating degraded

soils as they incorporate trees, which have been shown to maintain and improve soil

characteristics (Hohnwald et al., 2005; Murgueitio et al., 2011). Trees can benefit the

physical properties of soils and can reduce soil erosion events caused by runoff

(Murgueitio et al., 2011) and leaching (Bishop, 1983), enhance water infiltration

(Ilstedt et al., 2007), counteract compaction (Murgueitio et al., 2011) and increase

soil aeration (Bishop, 1983). Silvopastoral systems may therefore present a positive

contribution to soil health compared to current models of livestock production in the

Amazon, which typically result in pasture degradation (Martínez and Zinck, 2004).

The magnitude of the effect of livestock on soils has been related to stocking rate,

grazing system, soil texture and soil moisture content (McCalla et al., 1984; Van

Haveren, 1983; Martínez and Zinck, 2004). The success of post pasture

abandonment recovery of soils has been shown to vary with the prior intensity of

pasture use (Uhl et al., 1988). An important factor to consider when selecting

species for rehabilitation potential is the influence of a species on soil properties

(Tilki and Fisher, 1998).

Attempts to quantify levels of pasture degradation have used soil physical

parameters such as bulk density and infiltration rate as indicators of degradation.

Bulk density is related to soil compaction and measurements are simple and cost-

efficient when compared to other methods developed to measure compaction such

as computerised tomography (Pedrotti et al., 2005) or sensor based technology

(Quraishi and Mouazen, 2013). Bulk density is the weight of dry soil divided by a

52

known given volume; it is typically measured by extracting soil cores in a cylinder of

known volume, Guo et al. (2016) used a cylinder with a volume of 100 cm3 and

Pedrotti et al. (2005) a volume of 90 cm3. It is generally accepted that larger volume

cylinders reduce edge effects of this sampling technique.

Soil with a bulk density greater than 1.6 g cm3 begins to restrict root growth in some

species. An increase of 30% in bulk density compared to forest has been used as a

threshold value corresponding to severe soil degradation of pastures (Hecht, 1982).

Increased bulk density is positively related to increases in the intensity of grazing

regimes. High bulk density has a negative effect on pasture production and has been

shown to reduce the protein content of pasture grasses and is associated with

increased presence of weeds (McCarty and Mazurak 1976; Martínez and Zinck

2004). Reátegui et al. (1990) found a 20% increase in bulk density when

continuously grazed pastures were compared to pastures protected from grazing in

the Peruvian Amazon. In another study into Amazon pasture degradation, Moraes et

al. (1996) found after nine years of grazing a 26% increase in bulk density compared

to forest. Stephenson and Veigel (1987) found it took 16 months to recover initial

bulk density once grazing had ceased, which is consistent with Alegre and Cassel

(1996) who found the effect of grazing on bulk density was reversible once livestock

were removed. McCalla et al. (1984) found bulk density can predict infiltration rate

across a variety of soils and ground cover characteristics. Viana et al. (2014) found

soil bulk density was variable across a gradient of land rehabilitation in the Amazon.

Bulk density was highest in degraded pasture, decreased in two rehabilitation trials

using woody species and was lowest in natural forest. Total soil nitrogen

concentration increased along the same gradient. They proposed degradation and

53

rehabilitation measurements of Amazon pastures can be derived from measures of

soil total nitrogen and bulk density (Viana et al., 2014).

Water infiltration is a measure of the time taken for water to infiltrate a soil. It is an

important factor governing the runoff of water over land. If infiltration rate is low then

the probability of flooding and soil erosion are increased. Measurement of infiltration

rates is commonly carried out using either single or double ring methodologies.

However, results may be affected by soil disturbance from ring installation and an

unrealistic water pressure (Benegas et al., 2014). The presence of livestock in

pastures has been shown to decrease infiltration rate, whereas tree presence has

been shown to increase infiltration rate (Benegas et al., 2014). Basic infiltration has

been shown to decrease 98% when grazed pastures are compared to intact forest in

studies on fine textured soils (Martínez and Zinck, 2004). Martínez and Zinck (2004)

observed that the majority of changes in infiltration rate occur within the first 3 years

of pasture establishment. Alegre and Cassel (1996) found that, in the Amazon,

infiltration rates were decreased after five years of pasture establishment. Benegas

et al. (2014) found infiltration rates in pastures were positively correlated with

proximity to trees.

Soil organic matter (SOM) refers to the fraction of soil that is derived from living or

once living plant or animal organisms. SOM has an influence on the physical,

chemical and biological properties of soils (Stolt and Lindbo, 2010). Soil SOM

contributes to the quantity and diversity of microbes in soils. A relatively high soil

organic matter is considered beneficial for agriculture as the presence of SOM can

increase the physical properties of the soil, such as the aeration and water infiltration

54

rate. SOM can alter the soils chemical properties, a soil with high SOM is more likely

to have a higher cation exchange capacity than a soil with low SOM, and be more

likely to resist changes of pH. Buschbacher et al. (1988) found that biomass was not

correlated with nutrient content in regenerating pastures, but did not consider soil

organic matter in their study. This contrasts with the findings of Montagnini and

Sancho (1990), who studied the impacts of native trees on the abandoned pasture

soils of Costa Rica. They used the Walkley-Black acid digestion method to determine

soil organic matter. After 2.5 years they reported an increase in soil organic matter

from 4.83% in a grass control to 5.31-6.60% in the upper 15 cm of soil beneath

native trees. They found a positive relationship between the soil organic matter

content and sum of bases, which highlights the importance of soil organic matter to

cation retention capacity. They observed that land use practices that negatively

affect the level of organic matter, such as the use of fire, were less sustainable than

practices that have been shown to recover soil organic matter such as planting trees

which can increase soil organic matter through the inputs of leaf litter and root

turnover.

Sustainable systems are often managed with the goal of balancing the nutrient cycle.

It has been proposed that trees can act as ‘nutrient pumps’, accessing reservoirs of

nutrients that may be present in lower soil horizons and distributing them to the

surface as litter. Kumar et al. (1998) present indicative results that when soil is

associated with trees it has a higher organic carbon, nitrogen, phosphorous and

potassium content when compared with soils without trees. Nitrogen fixing species

increased soil nitrogen, whereas levels of soil organic carbon were highest under

Leucaena leucocephala and Acacia auriculiformis. However, as discussed by Wang

55

et al. (1991), there are noticeable differences in the nutrient cycling of systems

where the tree component is produced with the intention of improving soil

characteristics and systems where biomass is produced with the intention of removal

from the site. Nutrient concentrations are not consistently distributed between

different plant components and in addition this distribution may vary considerably

between species (Wang et al., 1991). This assertion is consistent with the findings of

Achat et al. (2015), who found removal of biomass reduced the amount of organic

matter in soils, reduced total and available nutrients in soils and resulted in increased

acidification of the soils. Boddey et al. (2004) studied nitrogen cycling in grazed

Brachiaria pastures in Brazil and found cattle unevenly deposit nitrogen acquisitions,

the highest concentrations were found in rest areas and proximate to drinking

troughs. This means nutrient accounting related to the removal of biomass is a

complex, but important requisite of designing any truly ‘sustainable’ system. In the

context of silvopastoral systems, understanding the nutrient cycle is additionally

complicated by the animal component that may result in nutrient removal or their

unequal deposition (Boddey et al., 2004).

5. Evaluating the Forage Potential of Tree Species

Tree forages are an important source of feed for ruminants around the world (Dynes

et al., 2002). In order to intensify production of livestock systems it is necessary to

understand how the quantity and quality of inputs govern outputs. Therefore, it is

crucial to evaluate forages to understand their value to livestock. Forage value,

sometimes referred to as nutritive value, refers to the relationship between a unit of

forage intake and a unit of animal productivity (Dynes et al., 2002). ‘Feeding value’ is

56

used to describe the amount of voluntary intake of forage by an animal and the

forage nutritive value. Ulyatt (1973) calculated that variation in intake could account

for up to 70% of variation in feeding value. Weston (1982; 1985; 1996) proposed the

“forage consumption constraint” in reference to the phenomena whereby animals’

intake is less than required for optimal output. It was proposed the constraints

include environmental stress, forage components which complicate removal from the

rumen, low palatability and high harvesting expenditure. Understanding constraints

to intake of forages is important to the goal of increasing animal productivity, as

forage of high nutritive value is of limited use if it is not palatable to animals.

Palatability can be conceptualised as an animal’s choice of feed, and the feed intake

displayed by an animal. The amount of forage eaten by animals has not been related

to any specific or fixed pattern of its nutrient composition (Crampton, 1957).

Therefore, palatability can only be evaluated through feeding trials, but this gives no

indication as to the digestibility of the forage. Digestibility is “a measure of the

availability of nutrients” (Khan et al., 2003); yet, digestibility is not always an accurate

predictor of nutritive value, or vice versa (Dynes et al., 2002). In vivo, in vitro and in

sacco techniques are methods that have been used to evaluate the digestibility of

forages. The ‘total collection’ technique is commonly considered the most reliable

measure of digestibility (Khan et al., 2003). This approach uses feeding trials and

records feed intake, refusals and faecal output. Minson et al. (1964) described

digestibility as a function of the percentage of faecal output relative to the weight of

organic matter consumed. Feeding trials are costly, time consuming, and while a 7 to

21 day period is allowed for cows to adjust to a feed, their intake may still be affected

by environmental stress (Khan et al., 2003). This method of estimating digestibility

57

also suffers from an inherent animal-to-animal variation of 1 to 4% (Khan et al.,

2003). Other in vivo approaches include the use of markers by identifying either a

naturally indigestible constituent of the feed or applying external substances. The

presence of markers can then be identified from faecal samples.

In vitro digestibility and in sacco degradability can be used to indicate the nutritional

value of forage (Rittner and Reed, 1992). Tilley and Terry (1963) used a method for

in vitro digestibility by incubating the forage in rumen fluid followed by acid pepsin

and measuring the amount of forage degraded over time. A modification to this

technique is measuring the gas production of this process which has the additional

benefit of accounting for microbial growth and soluble and insoluble fractions in feed,

including tannins (Theodorou et al., 1994). The nylon bag technique is an in sacco

technique which additionally provides some measure of kinetic digestion processes

in ruminants. Both in vitro and in sacco experiments require the use of fistulated

animals and experiments demonstrate poor reproducibility, in part due to variations

in the rumen fluid used. Furthermore, results yielded by this methodological

approach offer no indices of palatability or animal production.

While there are many factors that influence the efficiency of digestion, one of the

most significant is the chemical composition of the feed. This is of particular

relevance when considering forages derived from tree foliage. Many tree species

develop high levels of secondary structures in their foliage to deter herbivores.

Chemical compounds can alter taste and odor which make them unpalatable or limit

their nutritional value (Hess et al., 2005). These compounds make multiple demands

on the animal. For example, they can affect methanogenesis in cattle (Sun et al.,

58

2012). Carulla et al. (2005) found condensed tannins in Acacia species decreased

hydrogen and resulted in a 13-24% drop in methanogenesis. Voluntary intake of a

forage may also change over time, presumably as a response to post-ingestion

feedbacks (Maasdorp et al., 1999).

Hydrolysable tannins and condensed tannins are the two major groups of plant

tannins generally found in tree foliage (Reed, 1995). Hydrolysable tannins are

potentially toxic to ruminants whereas condensed tannins may reduce the

digestibility of a forage (Kumar, 1992). In some cases, the presence of condensed

tannins can have beneficial effect on protein digestion and metabolism. Barry and

Blaney (1987) report a beneficial effect of condensed tannins when foliar

concentration is 2-3% of dry matter. The rumen escape theory suggests this

beneficial effect is due to tannins forming complexes with proteins which escape

ruminal degradation and become available in the lower gastrointestinal tract (Reed,

1995). However, Vitti et al. (2005) assert that tannin concentrations above 5% are

deleterious to metabolism.

One of the most widely planted forage tree legumes, Leucaena leucocephala, offers

an example of the complexity of plant chemistry in relation to animal production. L.

leucocephala contains the toxic secondary compound mimosine in its leaves and

pods, at dry matter concentrations reported as 12% in growing tips and 3-5% in

young leaves (Monoj and Bandyopadhyay, 2007). For this reason, it is generally

recommended that L. leucocephala comprise no greater than 20-30% of ruminant

diet. Ruminant animals can be inoculated with microbes shown to be able to break

down mimosine to non-toxic compounds. This suggests the existence of other rumen

59

organisms capable of degrading other phenolic compounds (Devendra, 1990). Due

to the multiple interactions between plants, chemistry and livestock it is difficult to

ascertain the value of a new species for forage value through chemical analysis

alone (El Hassan et al. 2000). Feeding trials may be considered the most

appropriate way of ascertaining the effect of forage on animal nutrition (Kumar,

1992).

Maasdorp et al. (1999) employed 20 cows in a feeding trial where amount of forage

consumed and milk production offered insights into the forage value of different trees

to ruminants. They found supplementary feeding of Leucaena leucocephala and

Acacia boliviana increased milk yields, but that Calliandra calothyrsus did not have a

significant effect. They reported a strong association between milk yield and the

organic matter degradability of tree forages. Forage content of ytterbium-precipitated

polyphenols and neutral detergent fiber were both negatively correlated with milk

yield. When measured via fodder intake trees exhibited different levels of palatability.

The palatability of Calliandra calothyrsus appeared to improve with time. The study

showed the effect of different tree legumes on cattle milk production. It also

highlighted the divergence between measurable chemical constituents and observed

livestock production.

An ongoing challenge of forage evaluation is to identify plant characteristics that can

be used to predict the value of a forage species for animal feed. Many farmers are

aware of potential forages for cattle, reflected in the common practice of allowing

cattle to graze on fallow lands and roadsides. Farmers may also exhibit preferences

towards some trees and remove others (Fujisaka et al., 2000). An improved

60

understanding of native species management requirements and forage quality with

regard to livestock productivity may provide evidence necessary for practitioners to

demonstrate the efficacy of such systems over conventional management.

6. Summary and Rationale

The Amazon rain forest represents a vast natural capital that has been under-

evaluated and under-utilised in the context of livestock production. Many studies that

aim to develop agroforestry or silvopastoral systems have highlighted the importance

of evaluating native species adapted to the local environmental conditions to

increase chances of success. The current paradigm of livestock production in the

Amazon is essentially a consumptive land use that drives forest loss and reduces the

provision of ecosystem services. Silvopastoral systems represent an opportunity for

sustainable intensification while simultaneously rehabilitating ecosystem functions.

This thesis aims to develop a framework that could be used to assist the decision

making of how to integrate trees into agriculture to meet the needs of a given

landowner. It is envisaged that this framework would be transferrable between

locations and could outline some key questions that arise when attempting to

identify, at a given site, appropriate agroforestry or silvopastoral arrangements. The

framework could describe methods for the appropriate selection of species, planting

design and management of resources. This framework will be introduced in Chapter

II, in which the decision-making process for the design of the present study will be

outlined with a detailed description of the study site.

61

This thesis aims to address the contextual limitations of previous research into

silvopastoral systems by developing a novel silvopastoral system that could be

adopted in the pasturelands of the Amazon. The objective of this silvopastoral

system is to increase the per hectare productivity of edible tree forage for cattle and

carbon sequestration potential per hectare when compared with conventional

pasture models in the region.

In order to achieve this, this thesis will need to address the lack of knowledge of

locally adapted forage plants in the Amazon. It is beyond the scope of this thesis to

provide an overview of all potential forage plants. Therefore, this thesis selects five

tree species that are native or locally adapted and have some positive reference to

palatability in the literature or in local knowledge. These five species will be planted

in an experimental trial in a pasture that is typical of the region. Chapter III will

discuss the growth, biomass production and effects of these species on soil

chemistry and a commonly used pasture grass of the Amazon.

The most comprehensive assessment of carbon stocks is derived from the

destructive sampling and combustion of individual biomass components. A less

labour intensive method is to choose an easily measurable part of a tree that has

some relationship to total biomass (usually the diameter at breast height, measured

at 1.3 m above the ground) and develop allometric equations that relate this

measurement to total biomass, which in turn can be used to estimate carbon

sequestration. This approach can has also been used to estimate the forage

production of trees and shrubs. As both the sequestration of carbon and forage

production are objectives of this thesis, allometric equations are presented for each

62

of the species that showed good adaptations to the pasture environment. The

allometric equations are calculated for each biomass component, and the fit of linear,

quadratic and log linear models for each of these components are compared in

Chapter IV.

Complex methodologies and resources that are beyond the reach of a farmer

support the current research into the forage potential of plants. Less work has

concentrated on how measurements can be taken before a plant is exposed to cattle

and repeated afterwards to measure the removal of biomass by cattle. Such an

approach could be a valuable on-farm tool for a preliminary assessment of the

forage potential of trees. Chapter V describes and compares four such

methodologies for estimating the removal of foliage by cattle and the impact of the

cattle on the trees. The total available biomass of tree and grass foliage per hectare

is estimated for each of the treatments.

Chapter VI concludes the present study, and summarises some of the major findings

and limitations of this study. The present work is considered in the context of the

existing literature. Finally, future research is described as both direct extensions of

the present work and the broader issues that need to be addressed to support the

transition of silvopastoral systems from theory to practice in the Amazon rainforest.

References

Abbott, I., Van Heurck, P., Burbidge, T., 1993. Impact of frequency

and intensity of defoliation on growth of Jarrah (Eucalyptus

63

marginata): an experimental study with saplings. Forest Ecology and

Management 56, 175–183. https://doi.org/10.1016/0378-1127(93)90111-Y

Achat, D.L., Deleuze, C., Landmann, G., Pousse, N., Ranger, J.,

Augusto, L., 2015. Quantifying consequences of removing harvesting

residues on forest soils and tree growth – A meta-analysis. Forest

Ecology and Management 348, 124–141.

https://doi.org/10.1016/j.foreco.2015.03.042

Alegre, J.C., Cassel, D.K., 1996. Dynamics of soil physical properties

under alternative systems to slash-and-burn. Agriculture, Ecosystems &

Environment 58, 39–48. https://doi.org/10.1016/0167-8809(95)00654-0

Alexandratos, N. and Bruinsma, J., 2012. World agriculture towards 2030/2050: the

2012 revision. ESA Working paper No. 12-03. Rome, FAO.

Altieri, M.A., Nicholls, C.I., Henao, A., Lana, M.A., 2015.

Agroecology and the design of climate change-resilient farming

systems. Agronomy for Sustainable Development 869–890.

https://doi.org/10.1007/s13593-015-0285-2

Arevalo, L.A., Alegre, J.C., Bandy, D.E., Szott, L.T., 1998. The

effect of cattle grazing on soil physical and chemical properties in a

silvopastoral system in the Peruvian Amazon. Agroforestry Systems 40,

109–124. https://doi.org/10.1023/a:1006075114659

https://doi.org/10.1016/0378-1127(93)90111-Y


https://doi.org/10.1016/0167-8809(95)00654-0

https://doi.org/10.1007/s13593-015-0285-2

https://doi.org/10.1023/a:1006075114659

64

Barry, T.N., Blaney, B.J., 1987. Secondary compounds of forages. In Hacker, J.B.

and Ternouth, J.H. (eds) The Nutrition Of Herbivores, Sydney, Academic Press 91-

119

Benegas, L., Ilstedt, U., Roupsard, O., Jones, J., Malmer, A., 2014.

Effects of trees on infiltrability and preferential flow in two

contrasting agroecosystems in Central America. Agriculture, Ecosystems

& Environment 183, 185–196. https://doi.org/10.1016/j.agee.2013.10.027

Bishop, J.P., 1983. Tropical forest sheep on legume forage/fuelwood

fallows. Agroforestry Systems 1, 79–84.

https://doi.org/10.1007/BF00596350

Boddey, R.M., MacEdo, R., Tarré, R.M., Ferreira, E., De Oliveira,

O.C., De, C., Cantarutti, R.B., Pereira, J.M., Alves, B.J.R.,

Urquiaga, S., 2004. Nitrogen cycling in Brachiaria pastures: The key

to understanding the process of pasture decline. Agriculture,

Ecosystems and Environment 103, 389–403.

https://doi.org/10.1016/j.agee.2003.12.010

Bogaerts, M., Cirhigiri, L., Robinson, I., Rodkin, M., Hajjar, R.,

Costa Junior, C., Newton, P., 2017. Climate change mitigation through

intensified pasture management: Estimating greenhouse gas emissions on

cattle farms in the Brazilian Amazon. Journal of Cleaner Production


https://doi.org/10.1007/BF00596350


65

162, 1539–1550. https://doi.org/10.1016/J.JCLEPRO.2017.06.130

Buschbacher, R., Uhl, C., Serrao, E.A.S., 1988. Abandoned Pastures in

Eastern Amazonia. II. Nutrient Stocks in the Soil and Vegetation.

Journal of Ecology 76, 682–699.

Butterfield, R.P., 1995. Promoting biodiversity: advances in

evaluating native species for reforestation. Forest Ecology and

Management 75, 111–121.

Calle, Z., Murgueitio, E., Chará, J., 2012. Integrating forestry,

sustainable cattle-ranching and landscape restoration. Unasylva 63,

31–40.

Carpenter, F.L., Mayorga, S.P., Quintero, E.G., Schroeder, M., 2001.

Land-use and erosion of a Costa Rican ultisol affect soil chemistry,

mycorrhizal fungi and early regeneration. Forest Ecology and


Carpenter, L.F., Nichols, D.J., Eduer, S., 2004. Early growth of

native and exotic trees planted on degraded tropical pasture. Forest



Carulla, J.E., Kreuzer, M., Machmüller, A., Hess, H.D., 2005.

https://doi.org/10.1016/J.JCLEPRO.2017.06.130


66

Supplementation of Acacia mearnsii tannins decreases methanogenesis

and urinary nitrogen in forage-fed sheep. Australian Journal of

Agricultural Research 56, 961. https://doi.org/10.1071/AR05022

Catchpoole, D., Blair, G., 1990. Forage tree legumes. I. Productivity

and N economy of Leucaena, Gliricidia, Calliandra and Sesbania and

tree/green panic mixtures. Australian Journal of Agricultural Research

41, 521. https://doi.org/10.1071/AR9900521

Crampton, E.W., 1957. Interrelations between digestible nutrient and

energy content, coluntary dry matter intake and the overall feeding

value of forages. Journal of Animal Science 16, 546–552.

https://doi.org//1957.163546x

Devendra, C., 1990. Shrubs and tree fodders for farm animals :

proceedings of a workshop in Denpasar, Indonesia, 24-29 July 1989,

International Development Research Centre, Ottawa.

Dommergues, Y.R., 1987. The role of biological nitrogen fixation in

agroforestry, in: Steppler, H.A., Nair, P.K.R. (Eds.), Agroforestry: A

Decade of Development. ICRAF, Nairobi, pp. 245–272.

Downie, J.A., 2014. Legume nodulation. Current Biology 24, R184–R190.

https://doi.org/10.1016/j.cub.2014.01.028

https://doi.org/10.1071/AR05022

https://doi.org/10.1071/AR9900521

https://doi.org/1957.163546x

https://doi.org/10.1016/j.cub.2014.01.028

67

Downing, T.E., 2019. Development or destruction: the conversion of tropical forest to

pasture in Latin America. Routledge.

Duguma, B., Kang, B.T., Okali, D.U.U., 1988. Effect of pruning

intensities of three woody leguminous species grown in alley cropping

with maize and cowpea on an alfisol. Agroforestry Systems 6, 19–35.

https://doi.org/10.1007/BF02220106

Dynes, R.A., Henry, D.A., Masters, G., 2002. Characterising Forages

for Ruminant Feeding. Proceedings of the 2002 International Symposium on “Recent

Advances in Animal Nutrition,” New Delhi, India.

El Hassan, S.M., Lahlou Kassi, A., Newbold, C.J., Wallace, R.J., 2000.

Chemical composition and degradation characteristics of foliage of

some African multipurpose trees. Animal Feed Science and Technology

86, 27–37. https://doi.org/10.1016/S0377-8401(00)00158-9

Ella, A., Blair, G.J., Stür, W.W., 1991. Effect of age of forage tree

legumes at the first cutting on subsequent production. Tropical

Grasslands - Forrajes Tropicales 25, 275–280.

Ella, A., Jacobsen, C., Stur, W.W., Blair, G., 1989. Effect of plant

density and cutting frequency on the productivity of four tree

legumes. Tropical Grasslands - Forrajes Tropicales 23, 28–34.

https://doi.org/10.1007/BF02220106

https://doi.org/10.1016/S0377-8401(00)00158-9

68

Erdmann, T.K., Nair, P.K.R., Kang, B.T., 1993. Effects of cutting

frequency and cutting height on reserve carbohydrates in Gliricidia

sepium (Jacq.) Walp. Forest Ecology and Management 57, 45–60.

https://doi.org/10.1016/0378-1127(93)90161-F

Fellmann, T., Witzke, P., Weiss, F., Van Doorslaer, B., Drabik, D., Huck, I., Salputra,

G., Jansson, T. and Leip, A., 2018. Major challenges of integrating agriculture into

climate change mitigation policy frameworks. Mitigation and adaptation strategies for

global change, 23(3), pp.451-468.

Food and Agriculture Organisation of the United Nations, F., 2016.

Agroforestry [WWW Document]. URL www.fao.org/forestry/agroforestry/en/

(accessed 2.13.16).

Frame, J., 1981. Herbage mass, in: Hodgson, J., Baker, R.D., Davis,

A., Laidlaw, A.S., Leaver, J.. (Eds.), Sward Measurement Handbook. Br.

Grassl. Soc., Berkshire, U.K., pp. 39–67.

Fujisaka, S., Escobar, G., Veneklaas, E.J., 2000. Weedy fields and

forests: interactions between land use and the composition of plant

communities in the Peruvian Amazon. Agriculture, Ecosystems &

Environment 78, 175–186. https://doi.org/10.1016/S0167-8809(99)00122-X

Gerber, P.J., Steinfeld, H., Henderson, B., Mottet, A., Opio, C., Dijkman, J., Falcucci,

A. and Tempio, G., 2013. Tackling climate change through livestock: a global

https://doi.org/10.1016/0378-1127(93)90161-F

http://www.fao.org/forestry/agroforestry/en/

https://doi.org/10.1016/S0167-8809(99)00122-X

69

assessment of emissions and mitigation opportunities. Food and Agriculture

Organization of the United Nations (FAO).

Giraldo, C., Escobar, F., Chara, J.D., Calle, Z., 2011. The adoption

of silvopastoral systems promotes the recovery of ecological processes

regulated by dung beetles in the Colombian Andes. Insect Conservation

and Diversity 4, 115–122.

https://doi.org/10.1111/j.1752-4598.2010.00112.x

González, J.E., Fisher, R.F., 1994. Growth of native forest species

planted on abandoned pasture land in Costa Rica. Forest Ecology and

Management 70, 159–167. https://doi.org/10.1016/0378-1127(94)90083-3

Guevarra, A.B., Whitney, A.S., Thompson, J.R., 1978. Influence of

Intra-Row Spacing and Cutting Regimes on the Growth and Yield of

Leucaena1. Agronomy Journal 70, 1033.

https://doi.org/10.2134/agronj1978.00021962007000060034x

Guo, L., Wu, G., Li, Y., Li, C., Liu, W., Meng, J., Liu, H., Yu, X.,

Jiang, G., 2016. Effects of cattle manure compost combined with

chemical fertilizer on topsoil organic matter, bulk density and

earthworm activity in a wheat–maize rotation system in Eastern China.

Soil and Tillage Research 156, 140–147.

https://doi.org/10.1016/j.still.2015.10.010

https://doi.org/10.1111/j.1752-4598.2010.00112.x

https://doi.org/10.1016/0378-1127(94)90083-3



70

Haile, S.G., Nair, V.D., Nair, P.K.R., 2010. Contribution of trees to

carbon storage in soils of silvopastoral systems in Florida, USA.

Global Change Biology 16, 427–438.

https://doi.org/10.1111/j.1365-2486.2009.01981.x

Hansen, M.C., Potapov, P. V, Moore, R., Hancher, M., Turubanova, S.A.,

Tyukavina, A., Thau, D., Stehman, S. V, Goetz, S.J., Loveland, T.R.,

Kommareddy, A., Egorov, A., Chini, L., Justice, C.O., Townshend,

J.R.G., 2013. High-resolution global maps of 21st-century forest cover

change. Science (New York, N.Y.) 342, 850–3.

https://doi.org/10.1126/science.1244693

Hardesty, L.H., Box, T.W., 1988. Defoliation Impacts on Coppicing

Browse Species in Northeast Brazil. Journal of Range Management 41,

66–70.

Hecht, S., 1982. Amazonia, Agriculture and Land Use Research:

Proceedings of the International Conference. CIAT.

Hedge, N., 1983. Leucaena forage management in India. In Leucaena

research in the Asian--Pacific region, in: Proceedings of a Workshop

Held in Singapore,. International Development Research Centre, Ottawa,

Ont., Canada. IDRC-21 Le. pp. 73–78.

https://doi.org/10.1111/j.1365-2486.2009.01981.x

https://doi.org/10.1126/science.1244693

71

Herrero, M., Wirsenius, S., Henderson, B., Rigolot, C., Thornton, P., Havlík, P., de

Boer, I. and Gerber, P.J., 2015. Livestock and the environment: what have we

learned in the past decade?. Annual Review of Environment and Resources, 40,

pp.177-202.

Hess, H., Stuerm, C., Tiemann, T., Lascano, C., Kreuzer, M., 2005. Effects of

mixtures of tropical legumes with contrasting tannin contents as supplements to low-

quality grass diets on ruminal fermentation in vitro, In: Tielkes, E., Hülsebusch, C.,

Häuser, I., Deininger, A., Becker, K., (eds.). The Global food & product chain:

Dynamics, innovations, conflicts, strategies: Tropentag 2005: International research

on food security, natural resource management and rural development: Book of

abstracts, University of Hohenheim, Stuttgart October 11-13, 2005. Universität

Hohenheim, Tropenzentrum, Centre for Agriculture in the Tropics and Subtropics,

Stuttgart, DE. p. 1.

Hodgson, J., 1979. Nomenclature and definitions in grazing studies.

Grass and Forage Science 34, 11–17.

https://doi.org/10.1111/j.1365-2494.1979.tb01442.x

Hohnwald, S., Rischkowsky, B., Schultze-Kraft, R., Rodrigues-Filho, J.

A., Camarão, A.P., 2005. Experiences with establishing legumes as part

of a ley pasture in a low-input farming system of the eastern Amazon,

Brazil. Pasturas Tropicales 27, 2–12.


72

Hohnwald, S., Rischkowsky, B., Camarão, A.P., Schultze-Kraft, R.,

Rodrigues Filho, J.A., King, J.M., 2006. Integrating cattle into the

slash-and-burn cycle on smallholdings in the Eastern Amazon, using

grass-capoeira or grass-legume pastures. Agriculture, Ecosystems &

Environment 117, 266–276. https://doi.org/10.1016/J.AGEE.2006.04.014

Hohnwald, S., 2016. Relative palatability and growth performance of

capoeira species as supplementary forages in the NE-Amazonia.

Agriculture, Ecosystems & Environment 218, 107–115.


Ibrahim, M., Villanueva, C., Casasola, F., Rojas, J., 2006. Sistemas

silvopastoriles como una herramienta para el mejoramiento de la

productividad y restauración de la integridad ecológica de paisajes

ganaderos. Pastos y Forrajes 29, 383–419.

Ilstedt, U., Malmer, A., Verbeeten, E., Murdiyarso, D., 2007. The

effect of afforestation on water infiltration in the tropics: A

systematic review and meta-analysis. Forest Ecology and Management

251, 45–51. https://doi.org/10.1016/j.foreco.2007.06.014

Jackson, J., Ash, A.J., 1998. Tree-grass relationships in open

eucalypt woodlands of Northeastern Australia: influence of trees on

pasture productivity, forage quality and species distribution.

Agroforestry Systems 40, 159–176.

https://doi.org/10.1016/J.AGEE.2006.04.014



73

https://doi.org/10.1023/A:1006067110870

Kang, B.T., Reynolds, L., Atta-Krah, A.N., 1990. Alley Farming. Advances in

Agronomy Volume 43. Elsevier.

https://doi.org/10.1016/S0065-2113(08)60481-2

Kass, M., Benavides, J., Romero, F., Pezo, D., Speedy, A., Pugliese,

P.L., 1992. Lessons from main feeding experiments conducted at CATIE

using fodder trees as part of the N-ration., in: Legume Trees and

Other Fodder Trees as Protein Sources for Livestock. FAO Animal

Production and Health Paper. Food and Agriculture Organization of the

United Nations, pp. 161–175.

Kaur, B., Gupta, S.R., Singh, G., 2002. Carbon storage and nitrogen

cycling in silvopastoral systems on a sodic soil in northwestern

India. Agroforestry Systems 54, 21–29.

Khan, M.A., Mahr-Un-Nisa, S.M. and Sarwar, M., 2003. Techniques measuring

digestibility for the nutritional evaluation of feeds. International Journal of Agriculture

and Biology, 5(1), pp.91-94.

Kumar, M.B., George, S.J., Jamaludheen, V., Suresh, T.K., 1998.

Comparison of biomass production, tree allometry and nutrient use

efficiency of multipurpose trees grown in woodlot and silvopastoral

experiments in Kerala, India. Forest Ecology and Management 112,

https://doi.org/10.1023/A:1006067110870

https://doi.org/10.1016/S0065-2113(08)60481-2

74

145–163. https://doi.org/10.1016/S0378-1127(98)00325-9

Kumar, R., 1992. Anti-nutritional factors, the potential risks of

toxicity and methods to alleviate them, in: Expert Consultation on

Legume Trees and Other Fodder Trees as Protein Sources for Livestock,

Kuala Lumpur (Malaysia), 14-18 Oct 1991. FAO Animal Production and

Health Paper (FAO), 1992.

Lal, R., 2004. Agricultural activities and the global carbon cycle.

Nutrient Cycling in Agroecosystems 70, 103–116.

https://doi.org/10.1023/B:FRES.0000048480.24274.0f

Lee, M.A., Davis, A.P., Mizeck G.G.C., Manning, P., 2017. Forage

quality declines with rising temperatures, with implications for

livestock production and methane emissions. Biogeosciences 14,

1403–1417. https://doi.org/doi:10.5194/bg-14-1403-2017

Loker, W.M., 1994. Where’s the beef?: Incorporating cattle into

sustainable agroforestry systems in the Amazon Basin. Agroforestry

Systems 25, 227–241. https://doi.org/10.1007/BF00707462

Maasdorp, B.V., Muchenje, V., Titterton, M., 1999. Palatability and

effect on dairy cow milk yield of dried fodder from the forage trees

Acacia boliviana, Calliandra calothyrsus and Leucaena leucocephala.

Animal Feed Science and Technology 77, 49–59.

https://doi.org/10.1016/S0378-1127(98)00325-9

https://doi.org/10.1023/B:FRES.0000048480.24274.0f

https://doi.org/doi:10.5194/bg-14-1403-2017

https://doi.org/10.1007/BF00707462

75

https://doi.org/10.1016/S0377-8401(98)00232-6

Mannetje, L. ’t., 2000. Measuring biomass of grassland vegetation, in:

Mannetje, L.’t., Jones, R.M. (Eds.), Field and Laboratory Methods for

Grassland and Animal Production Research. C.A.B.I. Publishing. Univ.

Press, Cambridge, UK, pp. 151–177.

Mannetje, L. ’t., 1978. Measuring quantity of grassland vegetation,

in: L.’t., M. (Ed.), Measurement of Grassland Vegetation and Animal

Production. Commonwealth Agricultural Bureaux, Bucks, England, pp.

63–95.

Martínez, L.J., Zinck, J.A., 2004. Temporal variation of soil

compaction and deterioration of soil quality in pasture areas of

Colombian Amazonia. Soil and Tillage Research 75, 3–17.


Mcadam, J.H., Sibbald, A.R., Teklehaimanot, Z., Eason, W.R., 2007.

Developing silvopastoral systems and their effects on diversity of

fauna. Agroforestry Systems 70, 81–89.

https://doi.org/10.1007/s10457-007-9047-8

McCalla, G.R., Blackburn, W.H., Merrill, L.B., 1984. Effects of

Livestock Grazing on Infiltration Rates, Edwards Plateau of Texas.

Journal of Range Management 37, 265–269.

https://doi.org/10.1016/S0377-8401(98)00232-6


https://doi.org/10.1007/s10457-007-9047-8

76

McCarty, M.K., Mazurak, A.P., 1976. Soil Compaction in Eastern

Nebraska after 25 Years of Cattle Grazing Management and Weed Control.

Journal of Range Management 29, 384 – 386.

Menezes, R.S.C., Salcedo, I.H., Elliott, E.T., 2002. Microclimate and

nutrient dynamics in a silvopastoral system of semiarid northeastern

Brazil. Agroforestry Systems 56, 27–38.

https://doi.org/10.1023/A:1021172530939

Minson, D.J., Harris, C.E., Raymond, W.F., Milford, R., 1964. The

digestibility and voluntary intake of s22 and h.1 ryegrass, s170 tall

fescue, s48 timothy, s215 meadow fescue and germinal cocksfoot. Grass

and Forage Science 19, 298–305.


Monoj, K.G., Bandyopadhyay, S., 2007. Mimosine Toxicity-A Problem of

Leucaena Feeding in Ruminants. Asian Journal of Animal and Veterinary

Advances 2, 63–73.

Montagnini, F., Sancho, F., 1990. Impacts of Native Trees on Tropical

Soils: A Study in the Atlantic Lowlands of Costa Rica. Ambio 19,

386–390.

https://doi.org/10.1023/A:1021172530939


77

Montagnini, F., Ashton, M., 1999. The Silvicultural Basis For

Agroforestry Systems. CRC Press.

Moraes, J.F.L. de, Volkoff, B., Cerri, C.C., Bernoux, M., 1996. Soil

properties under Amazon forest and changes due to pasture installation

in Rondônia, Brazil. Geoderma 70, 63–81.

Murgueitio, E., Calle, Z., Uribe, F., Calle, A., Solorio, B., 2011.

Native trees and shrubs for the productive rehabilitation of tropical

cattle ranching lands. Forest Ecology and Management 261, 1654–1663.


Nguyen, T.M., Binh, D. Van, Ørskov, E.R., 2005. Effect of foliages

containing condensed tannins and on gastrointestinal parasites. Animal

Feed Science and Technology 121, 77–87.

https://doi.org/10.1016/j.anifeedsci.2005.02.013

Nichols, J.D., Rosemeyer, M.E., Carpenter, F.L., Kettler, J., 2001.

Intercropping legume trees with native timber trees rapidly restores

cover to eroded tropical pasture without fertilization. Forest Ecology

and Management 152, 195–209.

https://doi.org/10.1016/S0378-1127(00)00603-4

Osman, A.M., 1981. Effects of cutting interval on the relative dry

matter production of four cultivars of leucaena. Leucaena Res. Rep. 2,



https://doi.org/10.1016/S0378-1127(00)00603-4

78

33–38.

Parsons, A.J., Leafe, E.L., Collett B., Penning, P.D., Lewis, J.,

1983. The Physiology of Grass Production Under Grazing. II.

Photosynthesis, Crop Growth and Animal Intake of Continuously-Grazed

Swards. Journal of Applied Ecology 20, 127–139.

Pathak, P.S., Rai, P., Roy, R.D., 1980. Forage production from koo-bab

ool (Leucaena leucocephala (Lam.) de Wit). 1. Effect of plant density,

cutting intensity and interval. Forage Research 6, 83–90.

Pedrotti, A., Pauletto, E.A., Crestana, S., Holanda, F.S.R., Cruvinel,

P.E., Vaz, C.M.P., 2005. Evaluation of bulk density of Albaqualf soil

under different tillage systems using the volumetric ring and

computerized tomography methods. Soil and Tillage Research 80,

115–123. https://doi.org/10.1016/j.still.2004.03.003

Perez, J.B., Boraemisza, E., Sollins, P., 1993. Identificacion de

especies forestales acumuladoras de aluminio en una plantación

forestal experimental ubicada en Sarapiquı, Costa Rica. Agronomıa

Costarricense 17, 99–103.

PRODES 2019 Programa de Cálculo do Desflorestamento da Amazônia, Amazon

Deforestation Calculation Program [WWW]

URL: http://www.obt.inpe.br/prodes/index.php [accessed September 25 2019]


http://www.obt.inpe.br/prodes/index.php

79

Quraishi, M.Z., Mouazen, A.M., 2013. A prototype sensor for the

assessment of soil bulk density. Soil and Tillage Research 134,

97–110. https://doi.org/10.1016/j.still.2013.07.011

Reátegui, K., Ruiz, R., Cantera, G., Lascano, C., 1990. Persistencia

de pasturas asociadas con diferentes manejos del pastoreo en un

Ultisol arcilloso de Puerto Bermúdez, Perú. Tropical Grasslands -

Forrajes Tropicales 12, 16–24.

Reed, J.D., 1995. Nutritional toxicology of tannins and related

polyphenols in forage legumes. Journal Of Animal Science 73, 1516–28.

Reynolds, L., Cobbina, J., 1992. Unit 4: The Integration of Livestock

Production in Alley Farming, in: Tripathi, B.R., Psychas, P.J. (Eds.),

The AFNETA Alley Farming Training Manual - Volume 1: Core Course in

Alley Farming. International Institute of Tropical Agriculture,

Ibadan.

Rittner, U., Reed, J.D., 1992. Phenolics and in-vitro degradability of

protein and fibre in West African Browse. Journal of the Science of

Food and Agriculture 58, 21–28. https://doi.org/10.1002/jsfa.2740580105

Rojas-Downing, M.M., Nejadhashemi, A.P., Harrigan, T., Woznicki, S.A.,

2017. Climate change and livestock: Impacts, adaptation, and


https://doi.org/10.1002/jsfa.2740580105

80

mitigation. Climate Risk Management 16, 145–163.

https://doi.org/10.1016/J.CRM.2017.02.001

Rosegrant, M. W., Fernandez, M., Sinha, A., Alder, J., Ahammad, H., de Fraiture,

Charlotte, Eickhour, B., Fonseca, J., Huang, J., Koyama, O., Omezzine, A. M.,

Pingali, P., Ramirez, R., Ringler, C., Robinson, S., Thornton, P., van Vuuren, D.,

Yana-Shapiro, H. 2009. Looking into the future for agriculture and AKST. In

McIntyre, B. D., Herren, H. R., Wakhungu, J., Watson, R. T. (Eds.). International

Assessment of Agricultural Knowledge, Science and Technology for Development

(IAASTD): Agriculture at a Crossroads, global report. Washington, DC, USA: Island

Press. pp.307-376.

Russo, R., Botero, R.B., 1994. Nitrogen fixing trees for animal

production on acid soils. Powell, Mark H.(ed.). Workshop on Nitrogen

Fixing Trees for Acid Soils (1994, Turrialba, Costa Rica). Nitrogen

fixing trees for acid soils: A field manual. 1994.

Sanchez, F.J.S., 2002. Integrating fodder trees into animal production

systems in the tropics. Tropical and Subtropical Agroecosystems 1,

1–11.

Sanchez, P., 1987. Soil productivity and sustainability in

agroforestry systems, in: Agroforestry: A Decade of Development.

ICRAF, Nairobi, pp. 205–223.

https://doi.org/10.1016/J.CRM.2017.02.001

81

Savory, R., Breen, J., 1979. The production of Leucaena leucocephala

in Malawi. (iii) Plant population studies. United Nations Development

Programme/Food and Agriculture Organization Project MLW /757020. (No.

17).

Scholes, S.J., Archer, S.R., 1997. Tree-grass Interactions in

Savannas. Annual Review of Ecology and Systematics 28, 517–544.

Serrao, E.D., Uhl, C., Nepstad, D., 1993. Deforestation for pasture in

the humid tropics: Is it economically and environmentally sound in the

longterm?, in: Proceedings of the XVII International Grassland

Congress, Rockhampton, Australia and Palmerston North, New Zealand. p.

2215−2221.

Shaw, N.H., Mannetje, L., Jones, R.M., 1976. Pasture Measurements, in:

Shaw, N.H., Bryan, W.. (Eds.), Tropical Pasture Research. Principles

and Methods. Commonwealth Agricultural Bureaux, Bucks, England, pp.

235–262.

Sierra, J., Daudin, D., Domenach, A.M., Nygren, P., Desfontaines, L.,

2007. Nitrogen transfer from a legume tree to the associated grass

estimated by the isotopic signature of tree root exudates: A

comparison of the 15N leaf feeding and natural 15N abundance methods.

European Journal of Agronomy 27, 178–186.

https://doi.org/10.1016/j.eja.2007.03.003

https://doi.org/10.1016/j.eja.2007.03.003

82

Sousa, L.F., Mauricio, R.M., Paciullo, D.S.C., Silveria, S.R.,

Riberio, R.S., Calsavara, L.H., Moreira, G.R., 2015. Forage intake,

feeding behavior and bio-climatological indices of pasture grass,

under the influence of trees, in a silvopastoral system. Tropical

Grasslands - Forrajes Tropicales 3, 129.

https://doi.org/10.17138/TGFT(3)129-141

Stainback, A.G., Alavalapati, J.R.R., 2004. Restoring longleaf pine

through silvopasture practices: an economic analysis. Forest Policy

and Economics 6, 371–378. https://doi.org/10.1016/j.forpol.2004.03.012

Stephenson, G.R., Veigel, A., 1987. Recovery of Compacted Soil on

Pastures Used for Winter Cattle Feeding. Journal of Range Management

1, 46–48.

Stevenson, A.C., Harrison, R.J., 2014. Ancient Forests in Spain: A

Model for Land-use and Dry Forest Management in South-west Spain from

4000 BC to 1900 AD. Proceedings of the Prehistoric Society 58,

227–247. https://doi.org/10.1017/S0079497X00004175

Stolt, M.H., Lindbo, D.L., 2010. Interpretation of Micromorphological

Features of Soils and Regoliths, Interpretation of Micromorphological

Features of Soils and Regoliths. Elsevier.

https://doi.org/10.1016/B978-0-444-53156-8.00017-9

https://doi.org/10.17138/TGFT(3)129-141

https://doi.org/10.1016/j.forpol.2004.03.012

https://doi.org/10.1017/S0079497X00004175

https://doi.org/10.1016/B978-0-444-53156-8.00017-9

83

Stür, W.W., Shelton, H.M., Gutteridge, R.C., 1994. Defoliation

management of forage tree legumes in Forage tree legumes in tropical

agriculture. Cab International.

Sun, X.Z., Waghorn, G.C., Hoskin, S.O., Harrison, S.J., Muetzel, S.,

Pacheco, D., 2012. Methane emissions from sheep fed fresh brassicas

(Brassica spp.) compared to perennial ryegrass (Lolium perenne).

Animal Feed Science and Technology 176, 107–116.


Szott, L.T., Fernandes, E.C.M., Sanchez, P.A., 1991. Soil-plant

interactions in agroforestry systems. Forest Ecology and Management

45, 127–152. https://doi.org/10.1016/0378-1127(91)90212-E

Teague, W.R., 1989. Effect of intensity and frequency of defoliation

on aerial growth and carbohydrate reserve levels in Acacia karroo

plants. African Journal of Range and Forage Science.

Theodorou, M.K., Williams, B.A., Dhanoa, M.S., McAllan, A.B., France,

J., 1994. A simple gas production method using a pressure transducer

to determine the fermentation kinetics of ruminant feeds. Animal Feed

Science and Technology 48, 185–197.

https://doi.org/10.1016/0377-8401(94)90171-6


https://doi.org/10.1016/0378-1127(91)90212-E

https://doi.org/10.1016/0377-8401(94)90171-6

84

Thomas, H., Laidlaw, A.S., 1981. Planning, design, and establishment

of experiments, in: Hodgson, J., Baker, R.D., Davis, A., Laidlaw,

A.S., Leaver, J.D. (Eds.), Sward Measurement Handbook. Br. Grassl.

Soc, Berkshire, U.K., pp. 15–38.

Thornton, P.K. and Herrero, M., 2010. The inter-linkages between rapid growth in

livestock production, climate change, and the impacts on water resources, land use,

and deforestation. The World Bank.

Tilki, F., Fisher, R.F., 1998. Tropical leguminous species for acid

soils: Studies on plant form and growth in Costa Rica. Forest Ecology

and Management 108, 175–192.

https://doi.org/10.1016/S0378-1127(98)00225-4

Tilley, J.M.A., Terry, R.A., 1963. A two-stage technique for the in

vitro digestion of forage crops. Grass and Forage Science 18, 104–111.


Uhl, C., Buschbacher, R., Serrao, E.A.., 1988. Abandoned Pastures in

Eastern Amazonia. I. Patterns of Plant Succession. Journal of Ecology

76, 663–681.

Ulyatt, M.J., 1973. The feeding value of herbage, in: Butler, G.W.,

Bailey, R.W. (Eds.), Chemistry and Biochemistry of Herbage. Academic

Press, London, pp. 131–178.

https://doi.org/10.1016/S0378-1127(98)00225-4


85

Van Haveren, B.P., 1983. Soil Bulk Density as Influenced by Grazing

Intensity and Soil Type on a Shortgrass Prairie Site. Journal of Range

Management 5, 586–588. https://doi.org/http://doi.org/10.2307/3898346

Vermeire, L.T., Ganguli, A.C., Gillen, R.L., 2002. A Robust Model for

Estimating Standing Crop across Vegetation Types. Journal of Range


Viana, R.M., Ferraz, J.B.S., Neves, A.F., Vieira, G., Pereira, B.F.F.,

2014. Soil quality indicators for different restoration stages on

Amazon rainforest. Soil and Tillage Research 140, 1–7.


Vitti, D.M.S.S., Abdalla, A.L., Bueno, I.C.S., Filho, J.C.S., Costa,

C., Bueno, M.S., Nozella, E.F., Longo, C., Vieira, E.Q., Filho,

S.L.S.C., Godoy, P.B., Mueller-Harvey, I., 2005. Do all tannins have

similar nutritional effects? A comparison of three Brazilian fodder

legumes. Animal Feed Science and Technology 119, 345–361.


Wassenaar, T., Gerber, P., Verburg, P.H., Rosales, M., Ibrahim, M. and Steinfeld,

H., 2007. Projecting land use changes in the Neotropics: The geography of pasture

expansion into forest. Global Environmental Change, 17(1), pp.86-104.

https://doi.org/http:/doi.org/10.2307/3898346



86

Wang, D., Bormann, F.H., Lugo, A.E., Bowden, R.D., 1991. Comparison of

nutrient-use efficiency and biomass production in five tropical tree

taxa. Forest Ecology and Management 46, 1–21.

https://doi.org/10.1016/0378-1127(91)90241-M

Weston, R.H., 1996. Some aspects of constraint to forage consumption

by ruminants. Australian Journal of Agricultural Research. 47, 175–198.

Weston, R.H., 1985. The regulation of feed intake in herbage-fed

ruminants. Proceedings of the Nutrition Society of Australia. 10, 55–62.

Weston, R.H., 1982. Animal factors affecting voluntary feed intake,

in: Hacker, J.B. (Ed.), Nutritional Limits to Animal Production from

Pasture. Commonwealth Agricultural Bureaux, Slough, pp. 183–198.

White, R.P., Murray, S., Rohweder, M., Prince, S.D. and Thompson, K.M.,

2000. Grassland ecosystems (p. 81). Washington, DC: World Resources Institute.

Wilm, H.G., Costello, D.F., Klipple, G.E., 1944. Estimating forage

yield by the double-sampling method. Journal of the American Society

of Agronomy.

https://doi.org/10.1016/0378-1127(91)90241-M

87

CHAPTER II

FRAMEWORK AND METHODOLOGY

1. Objective

The objective of this chapter is to outline a theoretical framework that can be used to

evaluate the potential of tree species for a purpose driven agroforestry system and

demonstrate how this framework can be employed in a practical setting. The chapter

will describe the establishment of a silvopastoral system based on the principles

presented in the framework and outline a practical example of how the decision

making process has been used in the design of a field experiment. The framework

consists of four interacting components: the purpose, the species, the site and the

planting design (Figure 1).

88

Figure 1. Theoretical framework for the evaluation of candidate species for agroforestry systems

The purpose can be described as the reason that necessitates the implementation of

an agroforestry system. The subsequent components of species, site and planting

design are independent and complex issues that must be overcome in order to

achieve the purpose. Sanchez (1995) assert that the biophysical determinant of

successful agroforestry systems is the appropriate management between trees and

crops for light, water and nutrient competition. Even if the purpose is the plantation of

a single tree species, consideration can be given to additional species that can

increase the benefits provided per hectare, such as ground cover leguminous plants

that may increase soil fertility or mychorrizae that may benefit the species of interest.

The next section aims to put these components into the context of the purpose of

designing a silvopastoral system for the increase in productivity per hectare of edible

biomass production for cattle in the southwestern Amazon.

89

2. The Purpose

The purpose of this study was to identify native potential forage trees in the Amazon

that could increase the production per hectare of edible biomass for cattle. Systems

that manage tree or shrub and animal components are known as silvopastoral

systems. There is no minimum tree or shrub component of a silvopastoral systems,

which may be almost all pasture with a few trees for shade or timber and little

management, it may be a eucalyptus plantation where the manager sees cattle as a

useful byproduct that controls grass, or an intensive silvopastoral system that aims to

provide balanced nutrients to cattle using with strips of forage legumes and highly

productive African grasses, but without a major canopy element. In general, while

silvopastoral systems include both animal and plant elements, they may be managed

predominantly for the production of timber or livestock. The purpose of the current

study is an empirical evaluation of the potential of tree species to increase per

hectare productivity of edible biomass when compared to the status quo (no trees),

under the assumption that an increase in edible biomass can result in an increase in

stocking density and therefore an increase in livestock production per hectare.

3. The Site

In Figure 1, the ‘site’ refers to the biotic and abiotic resources available, and can also

be described as the balance of inputs and available to the system and the

constraints placed upon the system. The conditions of the site will be either

advantageous or disadvantageous to the set of species of under consideration. In

order to design an optimal system it is important to understand how the site, the

species and the management of the species interact. This understanding can be

90

gained via observation of landscapes, consultation with local ecological knowledge,

a review of the literature and the use of pilot studies. For the best chances of

success, agroforestry systems could include diverse species and arrangements and

be subject to adaptive management.

3.1 Site Description

The present study is situated in Madre de Dios, Peru (Figure 2). Madre de Dios is a

part of the Tropical Andes Biodiversity Hotspot and in the Southwestern Amazonian

Moist Forests Global 200 Ecoregion (Olson and Dinerstein, 2002; Myers, 2001). The

site is located 8 km from the region’s capital of Puerto Maldonado in an area known

as the Cachuela.

91

Figure 2. Location of the study site in Madre de Dios, Peru

The site was a cattle pasture from 1970-2000 and in 2000 the site was abandoned.

The Cachuela is an agricultural region prone to seasonal flooding, with a mosaic of

secondary forests, farms and cattle pastures. A total of 4 hectares of land was

available of which 2.7 was to establish an experimental silvopastoral system and 1.3

was to develop the tree seedling nurseries, cattle corrals and associated

infrastructure. In its entirety the plot was 100 x 400 m (Figure 3, Figure 4).

92

Figure 3. Experimental site in 2002 with the future location of the plots projected

Figure 4. Experimental site in 2015 with the future location of the plots projected.

3.2 Floristic Description of Site

The site was historically a cattle pasture that had been given to 15 years of natural

regeneration and species that had been propagated by the cattle dominated, notably

Psidium guajava and Genipa americana. Additional species at the site included

Ormosia coccinea, Cecropia spp., Attalea butyracea, Ficus spp., Moena alcanforada,

Apeiba membranacea, Ceiba spp. and Ochroma pyramidale.

The most eastern point of the site began approximately 150 m west of the river

Madre de Dios and extended 400 m away from the river. The site was on a mid to

high floodplain that may have formed during the Younger Dryas which continued

until approximately 3.7 ka BP (Nikitina et al., 2011). The site did not flood from the

93

river but was prone to inundation from rainfall events. Floods from the adjacent river

were experienced less than 100 m from the site. On the northern edge of the site

there was a seasonal lake, fed by rainfall and seasonally connected to a network of

swamps and clear water streams.

3.3 Climatic Description of Site

From 1970-2000 the annual mean temperate at the site was 25.3 °C with a mean

diurnal range of 10.3 °C (Figure 5). The maximum temperature recorded in the

hottest month during this time period was 31.6 °C and the minimum temperature

recorded in the coldest month was 18.2 °C. On average, annual precipitation was

2122 mm, the highest level of precipitation in the wettest month was 319 mm and

that of the driest month was 55 mm (Figure 6). Solar irradiation on average was

14400 kJ m-2 day-1, with an annual maximum of 16081 kJ m-2 day-1 and a

minimum of 12946 kJ m-2 day-1 (Figure 7). All climate data was derived for the

years 1970-2000 for a 1 km2 resolution from the WorldClim dataset (Hijmans et al.,

2005).

94

Figure 5. Average (solid line), maximum (dotted line), and minimum (dashed line) monthly temperature (°C) from the period 1970-2000 for the study site derived from the WorldClim dataset (Hijmans et al., 2005).

Figure 6. Average monthly precipitation (mm) from the period 1970-2000 for the study site derived from the WorldClim dataset (Hijmans et al., 2005).

15

17

19

21

23

25

27

29

31

33D

egre

es

Cel

siu

s

0

50

100

150

200

250

300

350

(mm

)

95

Figure 7. Average monthly level of solar irradiation (kJ m-2 day-1) from the period 1970-2000 for the study site derived from the WorldClim dataset (Hijmans et al., 2005).

3.4 Edaphic Description of Site

The sites physiography is a recent alluvium (Osher and Buol, 1998). The depth of

the water table varied from 0.5 - 1 m in the rainy season, to 7 - 10 m in the dry

season. The soil at the site could be described as a Gleysol or Fluvisol (FAO, 2015)

or Alfizol or Entisol (Eswaran et al., 2002). There were no clear distinctions of the soil

horizon, which was grey or red brown and dominated by clay particulates. This

resulted in low soil porosity and a slow infiltration rate. Frequent and high intensity

precipitation events combined with a history of cultivation could have contributed to

the highly weathered soils (Moreira et al., 2006). The pH of the site was acidic with a

site average of pH 5.5.

12000

12500

13000

13500

14000

14500

15000

15500

16000

16500(k

J m

-2d

ay-1

)

96

3.5 Social and Economic History

Archeological evidence points to human habitation in the Madre de Dios region for

over 1,000 years (Mann, 2008). Petroglyphs found along the Palotoa, Shinkebenia

and Urubamba and in the Pongo de Mainique suggest these areas were important

for early trade. The Inca’s were well aware of the populations in the forested regions

of Amarumayo, as it was then known. Missionaries, botanists and those in search of

Paititi were the first ‘explorers’ of the region of Madre de Dios, which was inhabited

at the time. The first successful colonization occurred in 1902 as a result of the

rubber boom, when the first government of Madre de Dios was formed in the newly

founded town of Puerto Maldonado. The memory of the rubber trade continues in

modern indigenous memory, and the tribes who overthrew the slave masters and

fled to the forests continue to live in voluntary isolation in the headwaters of the Las

Piedras, Manu and Alto Purus rivers in Madre de Dios.

The first road connecting Cusco to Puerto Maldonado, the capital of Madre de Dios,

was a horse path in use since the 1930’s, and the first road that was constructed to

the region was in the 1960’s. In the period of 1980-1990 the Peruvian government,

led by Alan Garcia, introduced policy incentives to develop agricultural expansion in

the Amazon basin (Chávez Michaelsen et al., 2013). In the same period, the forests

of Madre de Dios became major sources of tropical timber. The practice of selective

logging for high value timber species continues to be a major driver of economic

activity. Other major non-timber economies in the region are the production of

approximately 10 percent of the world’s Brazil nuts and as a popular destination for

tourists who visit the rainforest (Duchelle et al., 2012; Kirkby et al., 2011). The road

that connects Puerto Maldonado to Cusco was continued on to the Brazilian and

97

Bolivian frontiers and was paved in the early 2000’s and converted into the major

Initiative for the Integration of the Regional Infrastructure of South America (IIRSA)

project, also known as the Interoceanic highway, that was completed in 2011. Since

the early 2000’s, Puerto Maldonado has become a popular destination for the

profitable and often illegal activity of the extraction of alluvial gold deposits (Scullion

et al., 2014). Property rights in the region continue to undergo reform (Perz et al.,

2016). This process is slow due to changing responsibility between different

government departments. An arson attack on the Ministry of Agriculture in 2002 and

on the Regional Government in 2008 destroyed many records of land ownership and

was a setback to the formalisation of property rights in the region.

3.6 Future of the Amazon

The climate in the Amazon is changing. 2016 was the hottest year since 1950 (0.9

°C + 0.3 °C) and this has been related to the burning and clearing of the forest and

pastures (Marengo et al., 2018; van der Werf et al., 2010). Soares et al. (2006) used

deforestation rates prior to 2004 to produce a business‐as‐usual scenario that

estimated that 47% of the Brazilian Amazon would be deforested by 2050 (Soares-

Filho et al., 2006). Deforestation is known to reduce evapotranspiration and

precipitation. Consequently the frequency and duration of drought is increasing

(Panisset et al., 2017). Spracklen and Garcia-Carreras (2015) estimated that using

the pre-2004 rates of deforestation, as a business-as-usual scenario would lead to

an 8.1 ± 1.4% reduction in annual mean rainfall in the Amazon basin by 2050. Lima

et al. (2014) found that regional deforestation increases the length of the dry season

and the seasonal amplitude of water flow in the southwestern Amazon. Lawrence

98

and Vandecar (2015) found that tropical deforestation results in a warmer and drier

climate at the local scale and predict that negative impacts could extend beyond the

tropics. Scientists have considered the possibility of a ‘tipping point’, which is

described as a critical threshold at which an irreversible drought occurs as a result

either of global warming or a 25% or 40% loss of the total forest area of the Amazon

(Lenton et al., 2008; Nobre et al., 2016; Lovejoy and Nobre, 2018). Authors who

have modelled climate scenarios for the Amazon using the CPTEC model suggest it

becomes drier as a result of reduced precipitation and that in place of tropical

rainforest there will be an increase in savannah ( Sampaio et al., 2007; Sampaio et

al., 2019). While the future of the Amazon’s response to global warming and

deforestation is unknown, it is known that this is a high-risk scenario. Authors have

attempted to calculate the value of the ecosystem services the Amazon contains and

the financial ramifications of losing the provision of these services. Rammig et al.

(2018) predict that the loss of the Amazon in Brasil will result in short term gains of

USD 576 to 2,880 billion and a long-term loss of USD 1,367 to 6,928 billion. It is not

possible to predict how future governments of Amazon nations will respond to what

is now viewed as a global environmental crisis. However, the demands of a

globalised economy increasingly transcend governments and can place unforeseen

demands on natural resources, such as a US-China trade war resulting in increased

demand for Brazilian soy that may result in a surge in deforestation (Fuchs et al.,

2019).

3.7 Site Preparation

The decision was taken to clear the site by hand to simulate a traditional ‘slash and

burn’. The alternative would have been the use of heavy machinery such as tractors.

99

However, this would have disturbed the soils. Site clearing began in May 2016. All

undergrowth and small trees were removed by machete, and then the trees left

standing were felled with a chainsaw. All fallen organic material was dried in the sun

for three months. Prior to burning a 10 m fire break was cut around the perimeter of

the site and cylinders of water were positioned every 100 m or so to control any

unintended spread of fire. The site was burned in October 2016 (Figure 8). The fire

did not spread beyond the site. However, two weeks later an uncontrolled fire from a

farm 400 m away reached the edge of the site. Remaining fallen trunks were cut with

a chainsaw, piled and burned (Figure 9, Figure 10). A glyphosate herbicide was

applied across the site in December 2016 (Figure 11). Parcels were cleared of all

vegetation by raking and hoeing. Trunk stumps were cut to ground level with a

chainsaw prior to planting.

100

Figure 8. Experimental site in October 2016 with the future plots projected.

Figure 9. A monumental task. All site clearance was carried out by hand to minimise disturbance to the soils.

101

Figure 10. An international team of volunteers involved helped with the pile and burn.

Figure 11. When the site had been cleared of branches, the weeds at the site had grown to be unmanageable. A glyphosphate herbicide was applied across the site and the vegetation had to be cleared again. Clearance of this vegetation was done using a hoe and rake. All residues were taken off site in wheelbarrows.

102

4. Species

4.1 Evaluation of Desirable Plant Traits and Candidate Trees in Silvopastoral

Systems

Kattge et al. (2011) define Plant traits as “the morphological, anatomical,

physiological, biochemical and phenological characteristics of plants and their

organs”. Westoby and Wright (2006) explains their importance as “control ecosystem

processes and define habitat and resources for other taxa; thus, they are a high

priority for understanding the ecosystem at a site”. The identification and selection of

plants for a specific purpose is a critical evaluation of plant traits that determine their

suitability for the intended purpose.

In this study we use the example of developing a silvopastoral system that, in the

context of our study site in the Amazon, can increase the productivity of cattle

production systems over monoculture approaches. In this study, desirable plant traits

are those that can be identified as helping to achieve this goal (Figure 12). In the

Bragantina region of the Brazilian Amazon, Hohnwald et al. (2016) considered the

following criteria of decreasing importance to select species of potential forage value

for an experiment in north eastern Pará, Brazil. “good adaptations to environmental

conditions (poor, acid, and compacted soils), tropical climate (with up to 3 months

long dry seasons), periodical slashings and fires, high leafy biomass production,

highly preferred by cattle (Hohnwald 2002; Hohnwald et al., 2015), fast recuperation

abilities after defoliation, free accessibility on smallholdings, and high nutrient

contents in leaves.” Promising species could be identified by crude protein and fiber

values similar to that of forage grasses, and species to be avoided included those

103

with high amounts of condensed tannins. Pennington (1998) selected species from

the genus Inga with growth chatacteristics suitable for forestry or agroforestry use,

such as land reclamation, alley-cropping systems and fuelwood. The characteristics

they selected for were fast growth rates, high production of leafy and woody

biomass, ability to establish and flourish on marginal sites, and ability to compete

with fast growing secondary vegetation. Within the genus they distinguish between

species that are non-pioneer light demanding gap species and shade tolerant plants

from the forest understory, the latter being less desirable due to their intolerance to

the sun and slow establishment. Román-Dañobeytia et al., (2015) used the following

criteria to select species for the rehabilitation of sites degraded by alluvial mining:

seed availability, ease of propagation, rapid growth in open areas, and wide

geographical distribution across tropical America (Román et al., 2012). There has

been a call for an exploration of the agroforestry potential of native species as they

can provide unique ecosystem services and increase diversity (Butterfield, 1995;

Fisher, 1995; Murgueitio et al., 2011).

104

Figure 12. Mind map of desirable plant traits in the context of developing silvopastoral systems at a site in the Amazon that could be advantageous in achieving the goal of increasing the productivity per hectare through the use of silvopastoral systems compared with traditional methods of cattle production in the region.

The list of desirable traits in Figure 12 is presented as an example of how species

selection was informed in this study. Information was gathered from three categories

of resources: Published knowledge, unpublished knowledge and knowledge gained

by observation (Figure 13). Benavides (1995) outlined three steps for the

identification and characterization of potential forage species (1) surveys of cattle

workers to find out if animals prefer certain woody species (2) direct observation of

animals browsing preference (3) secondary information from existing literature

(Benavides, 1995).

105

Figure 13. Resources for identifying potential species for inclusion in agroforestry systems.

Local ecological and environmental knowledge (LEK) is a valuable resource for the

selection of species. LEK is a resource recognised within the academic realm and

has been used to manage wild plants in Albania (Tomasini and Theilade, 2019),

develop knowledge of plant phenology in the Brazilian Cerrado (Campos et al.,

2018), and surveys with farmers have highlighted important local forage species in

West Africa (Naah and Guuroh, 2017). Linstädter et al. (2013) held a group interview

with three herdsmen of sheep and goats to estimate palatability and nutritive value

for 62 species on a pasture. Informants classified species into “refused, little or

occasionally eaten, regularly eaten, highly favoured” and were asked to rank their

perception of the species’ nutritive value as “bad, medium, good”. Bahru et al. (2014)

carried out an ethnobotanic study of forage plants in Ethiopia with 96 informants.

One hundred twenty-six forage species of 90 genera and 43 families were collected

in the study area. Naah and Guuroh (2017) used a technique known as free listing to

elicit local environmental knowledge of fodder plants for goats and sheep in Ghana

and Burkina Faso and found local reports of fodder potential for 95 genera and 52

plant families. Dechnik-Vázquez et al. (2019) search for potential new useful species

106

for silvopastoral purposes in Chiapas, Mexico. They used field observation by

searching for plants with browsing marks, interviews with stakeholders,

bromatological analysis and vegetative propagation tests. They describe cattle as

generalist browsers.

In some landscapes local people have decades or even centuries of experience

managing ecosystems for the production of resources that are of interest to humans.

The management of cattle in Madre de Dios is a recent occurrence, itself being only

recently colonized, and is predominantly practiced by migrant farmers from the

Andes. The present study was time limited, and doubtless would have benefitted

from a structured and formal review of LEK in the region. Nine informal and

unstructured discussions with cattle farmers were held in preparation for the species

selection in this study. A systematic review of LEK of forage plants in the Amazon is

needed to understand the knowledge farmers may have regarding potential browse

species. Such a review is currently lacking in the published literature.

Expert input for this study came from the author’s supervisory team who first

witnessed cattle browsing on Inga edulis. On a separate occasion the author

observed a plantation of three species of Inga that had been destroyed by cattle.

Additionally, information was requested from a local agroforestry practitioner who

reported the unintended access of 80-100 cattle to their plantation site for

approximately a three-hour period. The data they collected for their records is

presented below (Table 1). However, it should be noted that this was not a replicated

browsing trial by cattle, and some areas were more accessible than others.

Furthermore, no distinction was made on the type of damage sustained by the

107

plants, which may have included trampling or browsing. However, it was reported

that the I. edulis was palatable as cattle were observed to browse on this species.

Direct and indirect observation of browsing preference can be elicited using means

such as species identification and frequency of use as an indication of palatability.

Indirect observations of browsing habits can include monitoring of browsing damage,

and types of damage, on trees and shrubs. The final criteria used in the evaluation

and selection of species for the present study is synthesised in Table 1.

Table 1. Plantation tree species killed and damaged by exposure to approximately 80-100 cattle in a 3-hour period. Data provided by Camino Verde (2018). .

Species # of plants % plants

killed % plants damage

Inga edulis 12719 14 16

Musa sp. 1748 13 2

Theobroma grandiflora 268 6 4

Cocos nucifera 171 9 3

Euterpe oleracea 100 13 6

Hymenaea courbaril 50 14 4

Dipteryx odorata 100 6 6

Morus nigra 728 1 2

Calliandra angustifolia 875 1 1

Myrciaria dubia 100 4 1

Cedrela odorata 97 3 5

Carapa guianensis 152 1 3

Astrocaryum murumuru 48 2 0

Myroxylon balsamum 52 0 0

Genipa americana 115 0 0

Table 2. Criteria used in the selection of the final five species for the present experiment. Where ++ is best suited to -- worse suited, + or – are weak evidence to

108

support either way, and ? indicate there was no information found linking the species to the criteria.

Criteria Leucaena

leucocephala

Inga edulis

Senegalia loretensis

Ceiba pentandra

Erythrina berteroana

Palatable to Cattle ++ ++ ++ ++ ++

Nutritious to Cattle ++ ? ? ? ?

Non-toxic to Cattle - ? ? ? ?

Resilient to browsing ++ + ? ++ ?

Fast Growing ++ ++ ? ++ ?

High proportion of edible biomass - ++ ? -- ?

Can withstand high planting density ++ ++ ? - ++

Can colonise degraded soils ? ++ ? + ?

Can withstand full sun ++ ++ + ++ ?

Resilient to fire ? ? ? ? ? Adapted to site conditions ? ++ ? ++ ++

Fix atmospheric Nitrogen ++ ++ + -- ++ Low susceptibility to pests ++ ++ ? ? ?

Accessible stock ++ ++ ++ ++ ++

Easy to Grow ++ ++ ++ ++ ++

4.2. Species Information

4.2.1 Inga edulis

Species of the genus Inga are nitrogen fixing leguminous trees (subfamily

Mimosoideae) with a range from Mexico to Uruguay (Pennington, 1998). Authors

report they are fast growing (Figure 14), demonstrate excellent biomass production,

coppicing ability, produce copious amounts of leafy biomass which helps control

weed species and are well adapted to acid soil conditions (Szott et al., 1991; Leblanc

et al., 2006). Inga species are depicted in the Mochica pottery of Peru, dated from

100-800AD, and some species have been domesticated and are now widely

distributed (Leon, 1998). One domesticated species of Inga is Inga edulis, cultivated

109

for their edible fruit. Farmers identified I. edulis as “useful” in previous studies in the

Peruvian Amazon (Fujisaka et al., 2000). I. edulis has been used widely in the

recovery of degraded and abandoned lands in the Chanchamayo regions of Peru,

where it has been used on floodplain land, shade for coffee, recovery of nutrient poor

soils following demanding crops such as pineapple or ginger, areas with compacted

soils and scarce organic matter, recuperation of areas dominated by weed growth

(including various aggressive grasses) and as a pioneer species on land abandoned

by landslides (Leon et al., 2016)

Dechnik-Vázquez et al. (2019) found I. edulis was ‘eagerly eaten’ by cattle as

reported by cattle herders in La Sepultura Biosphere Reserve in Chiapas, Mexico.

They found 78% neutral detergent fiber, 64.6% acid detergent fiber, 57.6% non-

structural carbohydrates and 0.31 g/100g of condensable tannins in the foliage of I.

edulis. Staver (1989) considered I. edulis in association with Desmodium ovalifolium

as a “planted fallow suitable for grazing”, and suggested I. edulis ‘grew more

vigorously’ than Leucaena sp., Gliricidia sepium, and Calliandra calothyrsus when

planted in observation plots. Hohnwald (2010) reported that I. edulis was palatable

on a scale of very palatable, palatable or medium palatability. They report that in a

browsing trial by cattle I. edulis plants may have been too small to be detected by the

cattle among the tall adjacent grass. Hohnwald (2016) reported I. edulis as having

‘poor palatability’ for cattle.

110

Figure 14. Inga edulis growth (Height) over time (Months), from: (1) Sobanski and Marques (2014), (2) Alegre (2005), (3) Kanmegne et al., (2000), (4) Rezende & Vieira (2019), (5) Butterfield (1996), (6) Butterfield (1995), (7) Gonzalez and Fisher (1994) .

Authors have described plantations of I. edulis in planting arrangements of 3 x 3 m

(Gonzalez and Fisher, 1994; Butterfield, 1995; Haggar et al., 1997; Carpenter, 2006;

Sobanski and Marques, 2014), 3 x 1 m (Kanmegne, 2000), and 1.5 x 1.5 m (Alegre

et al., 2005). I. edulis has also been described as a contour hedgerow planted at a

spacing of 0.5 m (Alegre and Rao, 1996) . It is widely reported to be fast growing, in

the Ecuadorian amazon the trunk diameter of I. edulis grew a mean of 5 cm year-1

(Bishop, 1983). Kanmegne (2000) reports that in Southern Cameroon on a Ferric

Acrisol soil type, I. edulis grew to 3.7 m in 12 months and by 20 months had reached

a height of 7.7 m. The same study coppiced trees to 5 cm from ground level. After 9

months of regrowth, I. edulis produced 26.8 Mg ha-1 of leafy biomass. Alegre (2005)

found that I. edulis, grown in association with Centrosema macrocarpum, had

0

1

2

3

4

5

6

7

8

9

12(1

)

12,(

1)

12,(

2)

12,(

2)

12,(

3)

18,(

4)

20,(

3)

24,(

2)

24,(

2)

30,(

4)

36,(

4)

36,(

5)

36,(

6)

36,(

5)

36,(

2)

36,(

2)

36,(

7)

36,(

5)

42,(

4)

60,(

4)

Heig

ht

(m)

Months After Planting, (Reference)

111

increased height compared with monocultures of I. edulis. Sobanski and Marques

(2014) in Southern Brazil’s Atlantic Forest associated I. edulis with Urochloa cf.

humidicola or Urochloa arrecta on either Cambisol or Greysol soil, and plots were

either tilled or untilled. They found soil conditions were the principal predictor of

seedling performance. Fisher (1995) found I. edulis increased soil Nitrogen from 3.0

g kg-1 in 1987 to 3.4 g kg-1 by 1992 and extractable Phosphorous from 5.5 mg kg-1 to

7.2 mg kg-1 in the same 7-year period.

Figure 15. Biomass after 11 years (1993 – 2004) in five replications of two treatments, T1 (control) and T6 (interplanted with Inga edulis), in Mg ha-1. Biomass of weeds and Inga was measured in 1 x 3 m subplots within treatment plots and converted to a per hectare basis. Biomass of Terminalia amazonia is the sum of crown foliage and bole wood biomasses of a sample tree for each treatment plot. Crown and bole biomasses were estimated from volumes calculated from diameter and height measurements multiplied by foliage density and wood density, respectively. Once the tree biomass was obtained, biomass was converted to a per hectare basis depending on the number of trees in the plot. Total biomass is the sum of the weed and Inga biomass plus the T. amazonia biomass on a per hectare basis (from Nichols and Carpenter, 2006).

112

Nichols and Carpenter (2006) found that Terminalia amazonia showed higher rates

of biomass accumulation when interplanted with I. edulis than in a monoculture

control in 4 of 5 replicates (Figure 15). ‘Pioneer’ or ‘colonising’ tree species may

open up a niche in Brachiaria brizantha pastures which may result in soil moisture,

shade and lower levels of soil compaction (Khurana and Singh, 2001; Garen et al.,

2011). Otsamo (1998) found that Anisoptera marginata grew fastest when

interplanted with the nurse tree Paraserianthes falcataria on an Imperata cylindrica

grassland site in Indonesia. Similarly, the dense shade and ability to suppress weeds

may give I. edulis a unique advantage in the pasture environment and

simultaneously open a niche for additional biodiversity.

4.2.2 Ceiba pentandra

Ceiba pentandra (L.) Gaertn, also known as the kapok tree, is an anemochorous

Bombacaceae (Wikander, 1984). It is a deciduous broadleaved tree with alternate

palmate leaves composed of 5 to 9 leaflets each up to 15 cm long. It is a species

known to associate with arbuscular mychorrizae (Iversen et al., 2007). Román-

Dañobeytia et al. (2015) used C. pentandra in a study of rehabilitation of sites

degraded by alluvial mining in Madre de Dios. They considered C. pentandra to be a

mid successional species as they find they are capable of colonizing open areas but

are generally long‐ lived and grow taller than short-lived pioneer species (Román-

Dañobeytia et al., 2012). It is a species that has been reported to grow rapidly

(Richards, 1952 in Condit et al., 1993). Fertilizer application at an experimental site

in Kerala, India, was shown to “increase” seedling root collar diameter but have “little

effect” on height (Dileep et al., 1994). C. pentandra is a traditionally used forage in

Cambodian goat production. In one feeding trial in Cambodia the inclusion of C.

113

pentandra foliage increased live weight gains of goats by 26% over a 32-week period

(Kouch, Preston and Hieak, 2006). In a study by Kongmanila and Ledin (2009) in

Laos, C. pentandra foliage had a condensed tannin (CT) content in the leaves and

petioles (as dry matter) of 300 g/kg, and 231 g/kg of dry matter in the stem. When

offered to goats, the percentage intake of C. pentandra biomass was 3.8% of the

animal’s body weight, which was less (P<0.05) than Artocarpus heterophyllus (4.4%)

or Erythrina variegata (4.0%) but more (P<0.05) than Ficus racemosa (3.4%),

Mangifera indica (2.7%) or Ziziphus jujuba Mill (2.5%). Thanu, Kadirvel and

Ayyaluswami (1983) fed C. pentandra seeds to day-old New Hampshire chicks. A

control group where no seeds were fed gained 305 g, whereas birds on the raw seed

diet gained 243 g and those on seeds that had been oven dried gained 127 g. In

India, C. pentandra seeds were ground and mixed into linseed oilcakes fed to

Kumauni bullocks with no observed adverse effects (Sahai and Kehar, 1968). In

Ghana, Ansah and Nagbila (2011) report farmers use C. pentandra as fodder, and

also use the fresh pounded leaves medicinally to expel retained placentas in cattle.

There is a precedent for this species in plantations. Majid et al. (1998) grew C.

pentandra in Kampung Pasir, Malaysia, and found that after 23 months trees grown

in association with Calopogonium muconoides were significantly taller (380 cm) and

had a larger diameter (72 mm) than those grown in association with Centrosema

pubescens (height 280 cm / RCD 68 mm) or Puereria phaseoloides (height 220 cm /

RCD 55 cm). At Indira Gandhi Agricultural University, C. pentandra was harvested 9

years after planting at 4 x 4 m, 4 x 6 m and 4 x 8 m. The 4 x 4 m spacing had

reached a DBH of 16.56 cm and a height of 8.29 m, the 4 x 6 m had reached a DBH

of 16.28 cm and a height of 7.49 m and the 4 x 8 m had reached a DBH of 17.04 and

114

a height of 7.55 m (Gawali et al., 2015). This species was documented by Toledo-

Aceves and Swaine (2007) in the Bobiri Forest Reserve in Ghana to be susceptible

to fungal infections. They planted C. pentandra at 0.7 x 0.7 m and after 12 months

found that 39% of the C. pentandra seedlings had died. Another mortality rate,

reported by d’Oliveira (2000) for saplings planted in primary forest, was 10% after 12

months, which rose to an accumulated mortality of 25% after 24 months, which

d’Oliveira (2000) speculated was due to canopy closure.

4.2.3 Erythrina berteroana

Erythrina is a pantropical genus of flowering plants in the family Leguminosae

(subfamily Faboideae). Some species are used as living fences and they have

gained popularity as a species for alley cropping. They are easily propagated.

Camacho et al. (1993) reports that in Costa Rica, production can be 12-16 Mg ha

year-1 with pH of 4.0 and 50% aluminium saturation. Haque et al. (1986) report that

Erythrina are adapted to acid utisols. This is confirmed by a study by Salazar and

Palm (1987) that demonstrated that a species of Erythrina tolerated 75% aluminium

saturation.

Muschler et al. (1993) compared E. berteroana Urban, E. fusca Loureiro and

Gliricidia sepium (Jacq.) Steud in Talamanca, southern Costa Rica. E. berteroana

was found to have the greatest crown diameter (CDiam 2.18

m), crown depth (CDep 2.43 m) than E. fusca (CDiam 1.52 m, CDep 1.68 m) or G.

sepium (CDiam 130 m, CDep 2.65 m). E. berteroana was also reported to produce

the most foliar biomass (1100 g) when compared with E. fusca (800 g) or G. sepium

115

(500 g). Szott et al. (1991) reported that Erythrina sp. have “excellent biomass

production”, coppicing ability and are “well adapted” to acid soils. They report from

Yurimagas, in the northern Peruvian Amazon, that the leaves of Erythrina

mineralized 40% of their initial phosphorus and calcium contents and 75% of their

magnesium and potassium contents within 4 weeks. After 20 weeks there was 25%

or less of initial mineral levels. In the same experiment, after 4 weeks, Inga edulis

had 80% remaining of fallen leaf mass, by 20 weeks this had decreased to 70%.

Salazar et al. (1993) found that grain and stover (above ground biomass) of rice

production in the first cropping phase was greater under Erythrina sp.(4861 kg ha-1)

and Leucaena leucocephala (5289 kg ha-1) than that from Inga edulis (4171 kg ha-1).

By the third cropping phase there were (400 kg ha-1) of weed production in the I.

edulis plots, compared to 1200 kg ha-1 in the Erythrina sp. or 1300 kg ha-1 in the L.

leucocephala plots. Salazar et al. (1993) speculate that the greater biomass

increases in the first crop were due to rapid decomposition of Erythrina sp. and L.

leucocephala leaves and rapid transfer of the nutrients to the associated crops,

whereas I. edulis decomposed more slowly as evidenced by it’s greater weed

suppression, and may therefore have resulted in a delayed nutrient transfer to the

associated crops

Standley et al. (1946) describe how in some species of Erythrina, young leaves are

cooked and eaten like a vegetable. In Guatemala and El Salvador the young shoots

of E. berteroana are popular and their consumption is said to induce a relaxing sleep

(Morton, 1994). Standley et al. (1946) describe how other species of Erythrina have

toxic alkaloids in concentrations that enable the use of crushed branches as a fish

116

poison. In seeds the effects of the toxins are reported to be paralysis or death by

respiratory failure (Cowan et al., 1982). Leaves have been reported to be eaten by

cattle, goats and rabbits without ill effect (Cordero et al., 2003). Payne (1991) found

that b –erythroidine from Erythrina leaves were hydrogenated in the rumen of goats

and detectable in the milk. The same study found partial genetic control of the

alkaloid content.

In Costa Rica, Ibrahim and Mannetje (1998) found that E. berteroana could produce

an annual ‘edible’ yield of more than 11 Mg ha–1 dry matter when pruned every four

months. Experiments carried out at CATIE field sites in Costa Rica found that a

pruning interval of four months maintained the proportion of E. berteroana edible to

non edible biomass, and that when planted ‘similarly to sugarcane’ dry matter yields

of 16.4 Mg ha-1 were obtained after the first year (Pezo et al., 1990; Kass et al.,

1992). Pezo et al. (1990) suggest that E. berteroana had ‘higher’ nutritional quality

than E. fusca or E. costarricenses and that there is the potential for generic

improvement of the forage quality of E. berteroana.

4.2.4 Senegalia loretensis

Senegalia loretensis (J.F.Macbr) is synonymous with Acacia loretensis (J.F. Macbr.)

and Acacia riparia var. angustifoliola Kuntze. It is a tree in the Fabaceae family

(subfamily Mimosoideae). It is an evergreen tree with oblong unequal and small

bipinnate leaflets arranged on a secondary axis. It has scattered prickles. It is native

to Bolivia, Brazil North, Brazil Northeast, Ecuador and Peru. Little literature exists

117

with regard to this species. However, the author observed that it is a common tree

proximate to the study site and is highly palatable to cattle (Figure 16).

Figure 16. Little is known about Senegalia loretensis, but despite it’s sharp spines the leaves are highly palatable to cattle.

4.2.5 Leucaena leucocephala

Leucaena leucocephala (Lam.) de Wit (leucaena) is a shrub or tree in the

Leguminosae (subfamily Mimosoideae) family. It is native to Central America and is

used across the tropics as a multipurpose tree legume. It is long-lived and able to

withstand pruning, pollarding, coppicing and direct browsing. L. leucocephala is used

as a shade tree for plantation crops, and is also well known for its forage value,

despite containing the toxic compound mimosine (Dalzell et al., 2012). L.

leucocephala is promoted as a tree of high potential for agroforestry and

silvopastoral systems (Mauricio et al., 2019). In 1990, L. leucocephala was estimated

to cover 2-5 million hectares worldwide (Brewbaker and Sorensson, 1990). Szott et

al. (1991) conducted 6 years of experimental trials in Yurimaguas, Peru, and found

118

that L. leucocephala and its more acid-tolerant relative, L. diversifolia, failed to

develop adequately and speculate this may have been due to due to aluminum

toxicity.

4.2.6 Brachiaria brizantha

Brachiaria brizantha (Hochst. ex A. Rich.) Stapf. is in the family Poaceae. It is native

to Botswana, Cameroon, Côte D'Ivoire, Ethiopia, Ghana, Guinea, Kenya, Malawi,

Mozambique, Namibia, Nigeria, Sierra Leone, South Africa, Tanzania, Uganda,

Zaire, Zambia and Zimbabwe. Brachiaria species were first introduced to South

America during the 18th century when B. mutica was used as bedding on ships.

Subsequently the Food and Agriculture Organization introduced species of

Brachiaria in 1952 (Serrão et al., 1971). B. brizantha became a popular species due

to its antibiotic resistance to the spittlebug, which feeds on the xylem of Brachiaria

species causing chlorosis and necrosis (CIAT, 1992).

‘Marandú’ (CIAT 6294, BRA000591, CPI 81408, ILCA 16550) is a hairy Kenyan

accession that has good spittlebug resistance. An experiment to see which grass

would be the most appropriate for a silvopastoral system under a 12 year old rubber

plantation in Porto Velho, Rondônia, Brazil recommended B. brizantha cv. Marandú

(Costa et al., 2001). A random block design was employed with three replicates.

Grasses were sown in four rows of 4 m length with a spacing of 0.5 m. Harvest was

performed at 12 week intervals in the wet season and 16 week intervals in the dry

season. In total six evaluations were realized, four during the rainy season and two

in the dry season see Table 3 (Costa et al., 2001)

119

.

Table 3. Plant height, soil cover and dry matter yield of forage grasses Brachiaria brizantha cv. Marandú, B. humidicola, Paspalum atratum BRA-9610, P. guenoarum BRA-3824, P. regnelli BRA-0159, P. plicatulum BRA-9661 and Hemarthria altissima established under a 12 year old rubber plantation reproduced from Costa et al. (2001).

At low N inputs, dry matter yields have been higher in shade than in full sunlight

(Dias-Filho and De Carvalho, 2000). Marandú is intolerant to waterlogging and

flooding (Beloni et al., 2017; Dias-Filho and De Carvalho, 2000). Many species of

Brachiaria are reported to be allelopathic and the Marandú cultivar has been

associated with negative effects on associated Desmodium adscendens, Sida

rhombifolia and Vernonia polyanthes (Favaretto et al., 2018; Souza Filho et al.,

1997). Dupas et al. (2010) found that production of B. brizantha cv. Marandú

improves with N input and fertilization that resulted in an average production of 24.9

Mg ha-1 in the wet season and 8.1 Mg ha-1 in the dry season. Perondi and Oliveira

(2007) found that the annual production was of 23.7 Mg ha-1 of which 8.5 Mg ha-1

was produced in the dry season and 15.1 Mg ha-1 was produced in the wet season

120

4.3 Design and Management of Tree Nursery

A nursery where plants can be grown and monitored is essential prior to the

collection of seeds, as often seeds must be planted immediately. For this study, the

key elements of the nursery included access to water, control of sun exposure, and

an appropriate medium in which to grow the plants. The present site had no existing

well and therefore it was necessary to dig a well (Figure 17). It was also necessary to

build a structure upon which to house a water tank as this facilitated water pressure

with which to water the plants. Plants were watered in the evening due to the high

levels of solar irradiation at the site. The plants in this project were to be housed in

two separate structures (nurseries) of approximately 10 x 20 m each. Due to the

complete site clearance, there was no natural shade. Therefore, a shade nursery

netting of 80% was used to shade the plants. In order to erect this hollow galvanized

steel poles were cut to size and holes was drilled 2.5 cm from one end of each pole.

To stabilize the pole the hollow base of the pole was then fitted in a coaxial fashion

over an iron bar that had been set proud of a concrete base (Figure 17). Further

cement was added to fill the hole to ground level. Through the holes in the poles a

thick galvanized wire was run around the entire perimeter of each nursery, and

tightened (Figure 18). A polystyrene ball was placed on the top of each pole to

protect the mesh from tearing on the pole during extreme weather events.

121

Figure 17. (Clockwise from left to right) The first attempt at digging the well, the base of the steel poles, which contained a 2.5 cm iron bar, alignment of the steel poles for the nurseries.

122

Figure 18. The nursery mesh was elevated upon steel poles above the plants, with enough room to allow air to circulate and to enable easy access for watering the plants. Finally, the nursery mesh was sewn around the galvanized wire and pulled taught at

either end with guide ropes. The beds in which the plants would be accommodated

were on the ground, and defined as 1.20 in width and the length of the nursery. The

beds were lined with perforated plastic. This was to reduce the likelihood of the

plants setting root into the ground, and also to reduce the incidence of weeds

growing up into the bags. A more preferable technique is that of growing plants in

root trainers off the ground. However, such resources were not available for the

current project.

The next task involved preparing the potting medium (Figure 19). Topsoil was

harvested from areas of the site outside of the experimental zone that were identified

as having preferable topsoil. There was a notable lack of good quality soil at the site,

which may be been due to the site history, which was of a cattle pasture and

123

regeneration of secondary forest. The topsoil was very clay heavy and was mixed

with sand to decrease the compaction of the bags that occurs with repeated

watering. The eventual mix did not include any resources that were not found at the

site. This was because this stage was also intended to stimulate how the plants may

establish at the site. The use of fertilisers or enriched soils might have had a

confounding effect on the experiment. A total of 10,000 bags of soil were filled, with

the intention of producing 4,500 plants of five species (Figure 20). As 900 plants of

each species were required of the same age and treatment, an excess of 1,100 trees

were planted per species for a total of 2,000 per species.

Figure 19. Some prepared soil for filling the bags for the plants.

124

Figure 20. View over the nurseries with 10,000 bags of soil. Borders made from trees define the beds. The bags of soil are arranged in rows to give easy access for weeding and watering of the plants. In the background is the water tower that enabled water to the site.

4.4 Treatment and Sowing of Seeds

In September 2016, seed pods of C. pentandra were collected (Figure 21). The

seeds were separated from the fiber and soaked in water for 24 hours. Seeds were

then placed individually into the bags of soil. Seeds of I. edulis were collected,

soaked in a mash of roots from the parent tree for 24 hours, drained and individually

placed into bags of soil (Figure 22). Seeds of L. leucocephala were scarified, soaked

for 24 hours and individually placed into bags of soil (Figure 23, Figure 24). In

October, seeds of S. loretensis were collected, soaked in water for 24 hours and

sown individually into bags of soil (Figure 25). In February 2017, cuttings of E.

berteroana were harvested using a machete and cut into 1 m lengths; these were

planted directly into the planting lines at a depth of 30-40 cm (Figure 26). With the

125

exception of E. berteroana, all trees were grown from seed, kept in 80% shade and

watered daily. The moisture in the bags of soil determined the watering regime.

Plants were removed from the shade one month prior to planting and the watering

regime reduced.

Figure 21. Seed pods and leaves of C. pentandra.

126

Figure 22. I edulis

Figure 23. Seeds of L. leucocephala

127

Figure 24. Seedlings of L. leucocephala

Figure 25. Beds of S. loretensis (foreground) and C. pentandra (background).

128

Figure 26. E. berteroana was planted from stakes, pictured here 2 months post planting in association with the immature pasture grass.

5. Planting Design

The present study envisaged a silvopastoral system in which a tree and shrub

element would be directly browsed by cattle. Tree and shrub elements can be

thought of as ‘vertical’ forage, whereas grasses and ground cover can be described

as a as ‘horizontal’ forage. There is competition between proximate plants for water,

nutrients and sun. Therefore, it can be conceptualised that the production of foliar

biomass and the leaf area of a vertical forage element will reduce the sun exposure

of adjacent horizontal forage. To minimise the shading effect of the vertical forage

the design of agroforestry may incorporate an east to west planting line, or ‘alley

rows’. Species can respond to different planting densities and arrangements by

adapting their ratio of wood to leaf production. To optimise production, trees in rows

129

should be planted as densely as possible without compromising the production of

foliar biomass. The aim is to compensate for the effect of the reduction of horizontal

foliage by producing more vertical foliage. Furthermore, if dense initial planting

proves problematic this can be amended by management techniques such as

thinning, pollarding or the application of fertilisers. The altitude of the sun relative to

the site can be considered when designing appropriate planting arrangements. In

this study where the aim was to maximise available forage to cattle, allowing the

trees to grow taller than the cattle could reach would reduce the light available for the

associated grass. If the maximum height of browsing cattle is 2 meters height, then

this value of 2 m can be used in conjunction with the altitude angle of the sun at the

site to determine the optimum width of the alley row (Leroy et al., 2009; Reid and

Ferguson, 1992). This can be calculated by:

(Equation 1)

Length of shadow = Height of vertical forage Tan(altitude angle of sun)

The above equation can be expanded to include, for example, the light interception

of the trees, and the response to shade of the associated horizontal forage element

(Dupraz et al., 2019; Zhang et al., 2019).

At the site in the present study, trees of all species were planted at a spacing of 0.5 x

7.5 m in east to west lines. The planting of 0.5 m in rows was chosen based on the

experience of previous authors with I. edulis (Alegre and Rao, 1996; Villachica,

1996; Leblanc et al., 2006). A distance of 7.5 m was left between the three rows of

each parcel. A planting arrangement of 0.5 x 7.5 m is 2,667 trees per hectare.

Daylight duration was at a maximum in January and minimum in July (Table 4). The

130

length of a shadow of an object of 2 meters height was at its shortest in November

and its longest in July.

Table 4. Data for the first day of each month from a site in Madre de Dios, Peru (lat -12.53945°, long -69.17288°) in 2016, including the altitude and azimuth of the sun at 12:00, the shadow cast by an object of 2 m height and the daylight duration.

First day of Altitude (°) Azimuth (°)

Shadow of object at 2 m

height (m) Daylight Duration

January 78.58 203.47 0.40 12h51m29s

February 84.99 207.63 0.18 12h38m52s

March 84.21 331.84 0.20 12h20m34s

April 72.09 344.03 0.65 11h58m49s

May 61.51 346.65 1.09 11h39m49s

June 54.81 349.75 1.41 11h26m33s

July 54.09 352.38 1.45 11h24m36s

August 59.29 352.08 1.19 11h34m43s

September 68.55 344.02 0.79 11h52m41s

October 77.6 317.07 0.44 12h13m8s

November 80.14 257.01 0.35 12h34m3s

December 77.61 220.04 0.44 12h49m5s

6. Experimental Design

The experiment used a randomised block design with five replicates. There were six

treatments in the experiment, with each treatment associated with a pasture of

Brachiaria brizantha cv. Marandú. The treatments were (1) Leucaena leucocephala,

(2) Inga edulis, (3) Ceiba pentandra, (4) Senegalia loretensis, (5) Erythrina

berteroana, (6) Control of Brachiaria brizantha cv. Marandú with no tree association.

Ideally the experiment would have been arranged as a 6 x 5 plot rectangle to

minimise edge effects. However, the site area dedicated to the system was 300 x

100 m so the parcel layout was 3 x 10 plots. A moisture gradient was suspected so

the site was divided into 5 blocks of 6. Thirty plots of 30 x 30 m (0.09 ha) were

installed in January 2017 and randomly allocated between the 6 treatments species

131

using the statistics program R (version 3.2.2 for OS X)(Figure 27). Each plot

therefore encompassed 180 trees, distributed in three alley rows of 0.5 x 7.5 m, and

each row of trees was 30 m long and contained 60 trees. The rows between each

replicate were separated by 15 m on the north and south side. The rows were

continuous between plots and species (Figure 28). In other words, there was a 0.5 m

break between one species and the next in a row (Figure 29). To account for the

potential edge effects of this design, the last 5 trees in each row were excluded from

subsequent censuses of plant growth.

Figure 27. The randomised replicate plot layout, where EB is Erythrina berteroana, IE is Inga edulis, LL is Leucaena leucocephala, SL is Senegalia loretensis, CP is Ceiba pentandra and C is a control of grass.

132

Figure 28. Map of experimental site. 0.09 ha plots as red rectangles, tree rows as black dashed lines. Tree rows were continuous between plots. There were no tree lines on the red lines.

Figure 29. Satellite image of experimental site, 0.09 ha parcels as red rectangles. Image taken in April 2018. Image by Maxar Technologies.

6.1 Planting Sequence

A map of the planting area was made in ArcGIS (ArcGIS Developer V.1.2). Due to

the shape of the site it was necessary to stagger the parcels in order to

accommodate them. The coordinates of the corners of each of the parcels were then

transferred onto a Trimble Geo 7X handheld GPS and the 60 points were located to

1-2 m accuracy. A 100 m tape measure was then used to corroborate the spacing of

133

the points. All points were marked plastic tubes that were 4 meters long and buried 1

m into the ground at each point (Figure 30). Using a 30 m tape measure three

additional points were located in each parcel at 7.5 m. These 7.5 m points were the

planting lines within each plot and these points were marked with wooden stakes. A

plastic twine was then held taught between the two stakes and any remaining

vegetation was cleared from the planting lines. Small, flat, wooden splints were

placed in the ground at 0.5 m increments using a 30 m tape measure on the ground.

Manual post diggers were used to dig holes, aiming to keep the wooden splint in the

middle of the hole to be dug. The top 5 cm of soil was returned to the base of each

hole at the time of planting.

Figure 30. The plastic tubes used to mark the 30 m intervals of the plot corners and in the background weeds responding to the glyphosphate applied across the site.

Planting of L. leucocephala began on the 20th of January 2017 and was completed

on the 26th January 2017 (Figure 31). Planting of I. edulis began on the 1st February

134

2017 and was completed on February 6th 2017 (Figure 32). Planting of C. pentandra

began on the 6th February 2017 and finished on the 11th February 2017. Planting of

S. loretensis began on the 11th February 2017 and was completed on the 18th

February 2017. E. berteroana were harvested and planted between the 20th

February 2017 and the 1st March 2017. From April – May 2017 the grass Brachiaria

brizantha cv. Marandú was sown by hand from west to east across the entire

experiment. Working in pairs, one person would drive a 2 m wooden stake with a

pointed end approximately 2.5 cm into the ground at a spacing of 25 x 25 cm. A

person would follow behind dropping approximately 0.05 g of seed in each hole. This

procedure was followed across all 30 parcels (Figure 33).

135

Figure 31.Team planting L. leucocephala in January of 2017.

Figure 32. Planting lines of I. edulis.

136

Figure 33. Satellite image from April 2017 with the parcels projected, showing faint tree lines and the establishing grass.

References

Alegre, J.C., Rao, M.R., 1996. Soil and water conservation by contour hedging in the

humid tropics of Peru. Agriculture, Ecosystems & Environment 57, 17–

25. https://doi.org/10.1016/0167-8809(95)01012-2

Beloni, T., Pezzopane, C.G., Rovadoscki, G.A., Fávero, A.P., Dias-Filho, M.B.,

Santos, P.M., 2017. Morphological and physiological responses and the recovery

ability of Paspalum accessions to water deficit and waterlogging. Grass and Forage

Science 72, 840–850. https://doi.org/10.1111/gfs.12281

Benavides, J.E., 1995. Tropical Feeds and Feeding Systems Research on Forage

Trees, in: FAO (Ed.), Electronic Conference on Tropical Feeds and Feeding

Systems.

https://doi.org/10.1016/0167-8809(95)01012-2

https://doi.org/10.1111/gfs.12281

137

Bishop, J.P., 1983. Tropical forest sheep on legume forage/fuelwood fallows.

Agroforestry Systems 1, 79–84. https://doi.org/10.1007/BF00596350

Brewbaker, J.L., Sorensson, C.T., 1990. New tree crops from interspecific leucaena

hybrids, In: Janick, J., Simon, J.E. (Eds) Timber Press, Portland, OR.

Butterfield, R.P., 1995. Promoting biodiversity: advances in evaluating native species

for reforestation. Forest Ecology and Management 75, 111–

121. https://doi.org/10.1016/0378-1127(95)03535-I

Campos, L.Z., Nascimento, A.L.B., Albuquerque, U.P., Araújo, E.L., 2018. Use of

local ecological knowledge as phenology indicator in native food species in the

semiarid region of Northeast Brazil. Ecological Indicators 95, 75–

84. https://doi.org/10.1016/J.ECOLIND.2018.07.016

Chávez Michaelsen, A., Huamani Briceño, L., Fernandez Menis, R., Bejar Chura, N.,

Valera Tito, F., Perz, S., Brown, I., Domínguez Del Aguila, S., Pinedo Mora, R.,

Alarcón Aguirre, G., 2013. Regional Deforestation Trends within Local Realities:

Land-Cover Change in Southeastern Peru 1996–2011. Land 2, 131–

157. https://doi.org/10.3390/land2020131

CIAT, 1992. Pastures for the Tropical Lowlands, CIAT’s Contribution. Centro

Internacional de Agricultura Tropical, Cali, Colombia.

Condit, R., Hubbell, S.P., Foster, R.B., 1993. Identifying fast-growing native trees

https://doi.org/10.1007/BF00596350

https://doi.org/10.1016/0378-1127(95)03535-I

https://doi.org/10.1016/J.ECOLIND.2018.07.016

https://doi.org/10.3390/land2020131

138

from the neotropics using data from a large, permanent census plot. Forest Ecology

and Management 62, 123–143. https://doi.org/10.1016/0378-1127(93)90046-P

Cordero, J., Boshier, D., Barrance, A., Centro Agronomico Tropical de Investigacion

y Ensenanza., Oxford Forestry Institute., Great Britain. Forestry Research

Programme., 2003. Arboles de Centroamerica : un manual para extensionistas.

Centro Agronomico Tropical de Investigacion y Ensenanza.

Costa, N.L., Paulino, V.T., Igreja, A.C.M., Townsend, C.R., Magalhães, J.A., Pereira,

R.G., Paulino, T.S., 2001. Agronomic evaluation of forage grasses under mature

rubber plantation. EMBRAPA, Roraima.

Cowan, R.S., Allen, O.N., Allen, E.K., 1982. The Leguminosae: A Source Book of

Characteristics, Uses, and Nodulation. Taxon 31,

133. https://doi.org/10.2307/1220602

Dalzell, S.A., Burnett, D.J., Dowsett, J.E., Forbes, V.E., Shelton, H.M., 2012.

Prevalence of mimosine and DHP toxicity in cattle grazing Leucaena leucocephala

pastures in Queensland, Australia. Animal Production Science 52,

365. https://doi.org/10.1071/AN11236

Dechnik-Vázquez, Y.A., García-Barrios, L., Ramírez-Marcial, N., van Noordwijk, M.,

Alayón-Gamboa, A., 2019. Assessment of browsed plants in a sub-tropical forest

frontier by means of fuzzy inference. Journal of Environmental Management 236,

163–181. https://doi.org/10.1016/J.JENVMAN.2019.01.071

https://doi.org/10.1016/0378-1127(93)90046-P

https://doi.org/10.2307/1220602

https://doi.org/10.1071/AN11236

https://doi.org/10.1016/J.JENVMAN.2019.01.071

139

Dias-Filho, M.B., De Carvalho, C.J.R., 2000. Physiological and morphological

responses of Brachiaria spp. to flooding. Pesquisa Agropecuaria Brasileira 35,

1959–1966. https://doi.org/10.1590/S0100-204X2000001000006

Dileep, M., Sudhakara, K., Santhoshkumar, A. V., Nazeema, K.K., Ashokan, P.K.,

1994. Effect of seed size, rooting medium and fertilizers on the growth of seedlings

of Ceiba pentandra (Linn.) Gaertn., Indian Journal of Forestry. Bishen Singh

Mahendra Pal Singh.

D’Oliveira, M.V.N., 2000. Artificial regeneration in gaps and skidding trails after

mechanised forest exploitation in Acre, Brazil. Forest Ecology and Management 127,

67–76. https://doi.org/10.1016/S0378-1127(99)00117-6

Duchelle, A.E., Guariguata, M.R., Less, G., Albornoz, M.A., Chavez, A., Melo, T.,

2012. Evaluating the opportunities and limitations to multiple use of Brazil nuts and

timber in Western Amazonia. Forest Ecology and Management 268, 39–

48. https://doi.org/10.1016/J.FORECO.2011.05.023

Dupraz, C., Wolz, K.J., Lecomte, I., Talbot, G., Vincent, G., Mulia, R., Bussière, F.,

Ozier-Lafontaine, H., Andrianarisoa, S., Jackson, N., Lawson, G., Dones, N.,

Sinoquet, H., Lusiana, B., Harja, D., Domenicano, S., Reyes, F., Gosme, M., Van

Noordwijk, M., 2019. Hi-sAFe: A 3D Agroforestry Model for Integrating Dynamic

Tree–Crop Interactions. Sustainability 11, 2293. https://doi.org/10.3390/su11082293

https://doi.org/10.1590/S0100-204X2000001000006

https://doi.org/10.1016/S0378-1127(99)00117-6

https://doi.org/10.1016/J.FORECO.2011.05.023

https://doi.org/10.3390/su11082293

140

Eswaran, H., Rice, T., Ahrens, R., Stewart, B.A. (Eds.), 2002. Soil Classification.

CRC Press. https://doi.org/10.1201/9781420040364

FAO, 2015. World reference base for soil resources, International soil classification

system for naming soils and creating legends for soil maps.

Favaretto, A., Scheffer-Basso, S.M., Perez, N.B., 2018. Allelopathy in Poaceae

species present in Brazil. A review. Agronomy for Sustainable Development 38,

22. https://doi.org/10.1007/s13593-018-0495-5

Fisher, R.F., 1995. Amelioration of degraded rain forest soils by plantations of native

trees. Soil Science Society of America Journal 59, 544–

549. https://doi.org/10.2136/sssaj1995.03615995005900020039x

Fuchs, R., Alexander, P., Brown, C., Cossar, F., Henry, R.C., Rounsevell, M., 2019.

Why the US–China trade war spells disaster for the Amazon. Nature 567, 451–

454. https://doi.org/10.1038/d41586-019-00896-2

Fujisaka, S., Escobar, G., Veneklaas, E.J., 2000. Weedy fields and forests:

interactions between land use and the composition of plant communities in the

Peruvian Amazon. Agriculture, Ecosystems & Environment 78, 175–

186. https://doi.org/10.1016/S0167-8809(99)00122-X

Garen, E.J., Saltonstall, K., Ashton, M.S., Slusser, J.L., Mathias, S., Hall, J.S., 2011.

The tree planting and protecting culture of cattle ranchers and small-scale

https://doi.org/10.1201/9781420040364

https://doi.org/10.1007/s13593-018-0495-5

https://doi.org/10.2136/sssaj1995.03615995005900020039x

https://doi.org/10.1038/d41586-019-00896-2

https://doi.org/10.1016/S0167-8809(99)00122-X

141

agriculturalists in rural Panama: Opportunities for reforestation and land restoration.

Forest Ecology and Management 261, 1684–

1695. https://doi.org/10.1016/j.foreco.2010.10.011

Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high

resolution interpolated climate surfaces for global land areas. International Journal of

Climatology 25, 1965–1978. https://doi.org/10.1002/joc.1276

Ibrahim, M., Mannetje, L.T., 1998. Compatibility, persistence and productivity of

grass-legume mixtures in the humid tropics of Costa Rica. 1. Dry matter yield,

nitrogen yield and botanical composition. Tropical Grasslands. 2, 96–104.

Iversen C.M., McCormack M.L., Powell A.S., Blackwood C.B., Freschet G.T., Kattge

J., Roumet C., Stover D.B., Soudzilovskaia N.A., Valverde-Barrantes O.J., van

Bodegom P.M., Violle C. (2017) A global Fine-Root Ecology Database to address

belowground challenges in plant ecology. New Phytologist. doi:10.1111/nph.14486.

Kass, M., Benavides, J., Romero, F., Pezo, D., Speedy, A., Pugliese, P.L., 1992.

Lessons from main feeding experiments conducted at CATIE using fodder trees as

part of the N-ration. In: Legume Trees and Other Fodder Trees as Protein Sources

for Livestock. FAO Animal Production and Health Paper. Food and Agriculture

Organization of the United Nations, pp. 161–175.

Kattge, J., Díaz, S., Lavorel, S., Prentice, I.C., Leadley, P., Bönisch, G., Garnier, E.,

Westoby, M., Reich, P.B., Wright, I.J., Cornelissen, J.H.C., Violle, C., Harrison, S.P.,


https://doi.org/10.1002/joc.1276

142

Van Bodegom, P.M., Reichstein, M., Enquist, B.J., Soudzilovskaia, N.A., Ackerly,

D.D., Anand, M., Atkin, O., Bahn, M., Baker, T.R., Baldocchi, D., Bekker, R., Blanco,

C.C., Blonder, B., Bond, W.J., Bradstock, R., Bunker, D.E., Casanoves, F.,

Cavender-Bares, J., Chambers, J.Q., Chapin, F.S., Chave, J., Coomes, D.,

Cornwell, W.K., Craine, J.M., Dobrin, B.H., Duarte, L., Durka, W., Elser, J., Esser,

G., Estiarte, M., Fagan, W.F., Fang, J., Fernández-Méndez, F., Fidelis, A., Finegan,

B., Flores, O., Ford, H., Frank, D., Freschet, G.T., Fyllas, N.M., Gallagher, R. V.,

Green, W.A., Gutierrez, A.G., Hickler, T., Higgins, S.I., Hodgson, J.G., Jalili, A.,

Jansen, S., Joly, C.A., Kerkhoff, A.J., Kirkup, D., Kitajima, K., Kleyer, M., Klotz, S.,

Knops, J.M.H., Kramer, K., Kühn, I., Kurokawa, H., Laughlin, D., Lee, T.D.,

Leishman, M., Lens, F., Lenz, T., Lewis, S.L., Lloyd, J., Llusià, J., Louault, F., Ma, S.,

Mahecha, M.D., Manning, P., Massad, T., Medlyn, B.E., Messier, J., Moles, A.T.,

Müller, S.C., Nadrowski, K., Naeem, S., Niinemets, Ü., Nöllert, S., Nüske, A., Ogaya,

R., Oleksyn, J., Onipchenko, V.G., Onoda, Y., Ordoñez, J., Overbeck, G., Ozinga,

W.A., Patiño, S., Paula, S., Pausas, J.G., Peñuelas, J., Phillips, O.L., Pillar, V.,

Poorter, H., Poorter, L., Poschlod, P., Prinzing, A., Proulx, R., Rammig, A., Reinsch,

S., Reu, B., Sack, L., Salgado-Negret, B., Sardans, J., Shiodera, S., Shipley, B.,

Siefert, A., Sosinski, E., Soussana, J.F., Swaine, E., Swenson, N., Thompson, K.,

Thornton, P., Waldram, M., Weiher, E., White, M., White, S., Wright, S.J., Yguel, B.,

Zaehle, S., Zanne, A.E., Wirth, C., 2011. TRY - a global database of plant traits.

Global Change Biology 17, 2905–2935. https://doi.org/10.1111/j.1365-

2486.2011.02451.x

Khurana, E., Singh, J.S., 2001. Ecology of seed and seedling growth for

conservation and restoration of tropical dry forest : a review. Environmental

https://doi.org/10.1111/j.1365-2486.2011.02451.x

https://doi.org/10.1111/j.1365-2486.2011.02451.x

143

Conservation 28, 39–52. https://doi.org/10.1017/S0376892901000042

Kirkby, C.A., Giudice, R., Day, B., Turner, K., Soares-Filho, B.S., Oliveira-Rodrigues,

H., Yu, D.W., 2011. Closing the ecotourism-conservation loop in the Peruvian

Amazon. Environmental Conservation 38, 6–

17. https://doi.org/10.1017/S0376892911000099

Kongmanila, D., Ledin, I., 2009. Chemical Composition of Some Tropical Foliage

Species and Their Intake and Digestibility by Goats. Asian-Australasian Journal of

Animal Sciences 22, 803–811. https://doi.org/10.5713/ajas.2009.80589

Lawrence, D., Vandecar, K., 2015. Effects of tropical deforestation on climate and

agriculture. Nature Climate Change 5, 27–36. https://doi.org/10.1038/nclimate2430

Leblanc, H.A., Nygren, P., McGraw, R.L., 2006. Green mulch decomposition and

nitrogen release from leaves of two Inga spp. in an organic alley-cropping practice in

the humid tropics. Soil Biology and Biochemistry 38, 349–

358. https://doi.org/10.1016/J.SOILBIO.2005.05.012

Leon, J.G., Reynel, C.R., Pennington, T.D., 1996. Agroforesteria y recuperacion de

areas degradadas con arboles de Inga (Pacae, Guaba) en el valle de

Chanchamayo, Departamento de Junin (Peru), Universidad Nacional Agraria La

Molina, Lima, Peru

https://doi.org/10.1017/S0376892901000042

https://doi.org/10.1017/S0376892911000099

https://doi.org/10.5713/ajas.2009.80589

https://doi.org/10.1038/nclimate2430

https://doi.org/10.1016/J.SOILBIO.2005.05.012

144

Leon, J., 1998. History of the utilization of INGA as fruit trees in Mesoamerica and

Peru. Chapter 1 in The Genus Inga, Utilization, T.D. Pennington and E.C.M.

Fernandez (editors). The Royal Botanic Gardens, Kew, Continental Printing,

Belgium.

Leroy, C., Sabatier, S., Wahyuni, N.S., Barczi, J.-F., Dauzat, J., Laurans, M., Auclair,

D., 2009. Virtual trees and light capture: a method for optimizing agroforestry stand

design. Agroforestry Systems 77, 37–47. https://doi.org/10.1007/s10457-009-9232-z

Lima, L.S., Coe, M.T., Soares Filho, B.S., Cuadra, S. V., Dias, L.C.P., Costa, M.H.,

Lima, L.S., Rodrigues, H.O., 2014. Feedbacks between deforestation, climate, and

hydrology in the Southwestern Amazon: implications for the provision of ecosystem

services. Landscape Ecology 29, 261–274. https://doi.org/10.1007/s10980-013-

9962-1

Mann, C.C., 2008. Ancient Earthmovers of the Amazon. Science 321, 1148–1152.

Marengo, J.., Souza, C.., Thonicke, K., Burton, C., Halladay, K., Betts, R.A., Alves,

L.., Soares, W.B., 2018. Changes in Climate and Land Use Over the Amazon

Region: Current and Future Variability and Trends. Frontiers in Earth Science 6, 1–

21. https://doi.org/10.3389/feart.2018.00228

Mauricio, R.M., Ribeiro, R.S., Paciullo, D.S.C., Cangussú, M.A., Murgueitio, E.,

Chará, J., Estrada, M.X.F., 2019. Silvopastoral Systems in Latin America for

Biodiversity, Environmental, and Socioeconomic Improvements, in: Agroecosystem

https://doi.org/10.1007/s10457-009-9232-z

https://doi.org/10.1007/s10980-013-9962-1

https://doi.org/10.1007/s10980-013-9962-1

https://doi.org/10.3389/feart.2018.00228

145

Diversity. Elsevier, pp. 287–297. https://doi.org/10.1016/B978-0-12-811050-8.00018-

2

Moreira, F.M.S., Siqueira, J.O., Brussard, L., 2006. Soil biodiversity in Amazonian

and other Brazilian ecosystems, Soil Biodiversity in Amazonian and other Brazilian

Ecosystems. CABI Pub.

Morton, J.F., 1994. Pito (Erythrina berteroana) and chipilin (Crotalaria longirostrata),

(fabaceae) two soporific vegetables of Central America. Economic Botany 48, 130–

138. https://doi.org/10.1007/BF02908199

Murgueitio, E., Calle, Z., Uribe, F., Calle, A., Solorio, B., 2011. Native trees and

shrubs for the productive rehabilitation of tropical cattle ranching lands. Forest

Ecology and Management 261, 1654–


Muschler, R.G., Nair, P.K.R., Meléndez, L., 1993. Crown development and biomass

production of pollarded Erythrina berteroana, E. fusca and Gliricidia sepium in the

humid tropical lowlands of Costa Rica. Agroforestry Systems 24, 123–

143. https://doi.org/10.1007/BF00706887

Naah, J.B.S.N., Guuroh, R.T., 2017. Factors influencing local ecological knowledge

of forage resources: Ethnobotanical evidence from West Africa’s savannas. Journal

of Environmental Management 188, 297–

307. https://doi.org/10.1016/J.JENVMAN.2016.11.064

https://doi.org/10.1016/B978-0-12-811050-8.00018-2

https://doi.org/10.1016/B978-0-12-811050-8.00018-2

https://doi.org/10.1007/BF02908199


https://doi.org/10.1007/BF00706887


146

Nichols, J.D., Carpenter, F.L., 2006. Interplanting Inga edulis yields nitrogen benefits

to Terminalia amazonia. Forest Ecology and Management 233, 344–


Nikitina, D., Remizova, L., Losco, R.L., 2011. A Preliminary Investigation of the Soils

and Geomorphology of a Portion of the Madre de Dios Region, Peru. Soil Horizons

52, 40. https://doi.org/10.2136/sh2011.2.0040

Panisset, J.S., Gouveia, C.M., Libonati, R., Peres, L., Machado-Silva, F., França,

D.A., França, J.R.A., 2017. Evaluation of last extreme drought events in Amazon

basin using remotely sensing data. 19th EGU General Assembly, EGU2017,

proceedings from the conference held 23-28 April, 2017 in Vienna, Austria., p.16144

19, 16144.

Payne, L., 1991. The Alkaloids of Erythrina: Clonal Evaluation and Metabolic

Fate. Ph.D. Dissertation. Department of Chemistry. Louisiana State Univérsity. 160

p.

Pennington, T.D., 1998. Chapter 2. Growth and Biomass Production of INGA

Species in The Genus Inga, Utilization, T.D. Pennington and E.C.M. Fernandez

(editors). The Royal Botanic Gardens, Kew, Continental Printing, Belgium.

Perz, S.G., Espin, J., Castillo, J., Chavez, A., Rojas, R.O., Barnes, G., 2016. Ideal

type theories and concrete cases in land science: A multi-step appraisal of the


https://doi.org/10.2136/sh2011.2.0040

147

evolutionary theory of land rights in Madre de Dios, Peru. Land Use Policy 58, 9–

20. https://doi.org/10.1016/J.LANDUSEPOL.2016.07.008

Pezo, D., Kass, M., Benavides, J., Romero, F. and Chaves, C., 1990. Potential of

legume tree fodders as animal feed in Central America. In Shrubs and tree fodders

for farm animals: proceedings of a workshop in Denpasar, Indonesia, 24-29 July

1989. IDRC, Ottawa, ON, CA.

Rammig, A., Lapola, D.M., Pinho, P., Quesada, C.A.N., Brown, I.F., Kruijt, B.,

Premebida, A., Ometto, J.P.H.B., Marengo, J.A., Vergara, W., Nobre, C.A., 2018.

Estimating the likelihood of an Amazon forest dieback and potential socio-economic

impacts. 20th EGU General Assembly, EGU2018, Proceedings from the conference

held 4-13 April, 2018 in Vienna, Austria

Reid, R., Ferguson, I.S., 1992. Development and validation of a simple approach to

modelling tree shading in agroforestry systems. Agroforestry Systems 20, 243–

252. https://doi.org/10.1007/BF00053142

Román-Dañobeytia, F., Huayllani, M., Michi, A., Ibarra, F., Loayza-Muro, R.,

Vázquez, T., Rodríguez, L., García, M., 2015. Reforestation with four native tree

species after abandoned gold mining in the Peruvian Amazon. Ecological

Engineering 85, 39–46. https://doi.org/10.1016/j.ecoleng.2015.09.075

Román-Dañobeytia, F.J., Levy-Tacher, S.I., Aronson, J., Rodrigues, R.R.,

Castellanos-Albores, J., 2012. Testing the Performance of Fourteen Native Tropical

https://doi.org/10.1016/J.LANDUSEPOL.2016.07.008

https://doi.org/10.1007/BF00053142

https://doi.org/10.1016/j.ecoleng.2015.09.075

148

Tree Species in Two Abandoned Pastures of the Lacandon Rainforest Region of

Chiapas, Mexico. Restoration Ecology 20, 378–386. https://doi.org/10.1111/j.1526-

100X.2011.00779.x

Sahai, B., Kehar, N.D., 1968. Investigations on subsidiary feeds. Kapok (Ceiba

pentandra) seed as a feed for livestock. Indian Journal of Veterinary Science and

Animal Husbandry 38, 670–673.

Salazar, A., Szott, L.T., Palm, C.A., 1993. Crop-tree interactions in alley cropping

systems on alluvial soils of the Upper Amazon Basin. Agroforestry Systems 22, 67–

82. https://doi.org/10.1007/BF00707472

Sampaio, G., Nobre, C., Heil Costa, M., Satyamurty, P., Silveira Soares-Filho, B.,

Cardoso, M., 2007. Regional climate change over eastern Amazonia caused by

pasture and soybean cropland expansion. Geophysical Research Letters 34,

17709. https://doi.org/10.1029/2007GL030612

Sampaio, G., Borma, L.S., Cardoso, M., Alves, L.M., von Randow, C., Rodriguez,

D.A., Nobre, C.A., Alexandre, F.F., 2019. Assessing the Possible Impacts of a 4 °C

or Higher Warming in Amazonia, in: Climate Change Risks in Brazil. Springer

International Publishing, Cham, pp. 201–218. https://doi.org/10.1007/978-3-319-

92881-4_8

Sanchez, P.A., 1995. Science in agroforestry. Agroforestry Systems 30, 5–

55. https://doi.org/10.1007/BF00708912

https://doi.org/10.1111/j.1526-100X.2011.00779.x

https://doi.org/10.1111/j.1526-100X.2011.00779.x

https://doi.org/10.1007/BF00707472

https://doi.org/10.1029/2007GL030612

https://doi.org/10.1007/978-3-319-92881-4_8

https://doi.org/10.1007/978-3-319-92881-4_8

https://doi.org/10.1007/BF00708912

149

Scullion, J.J., Vogt, K.A., Sienkiewicz, A., Gmur, S.J., Trujillo, C., 2014. Assessing

the influence of land-cover change and conflicting land-use authorizations on

ecosystem conversion on the forest frontier of Madre de Dios, Peru. Biological

Conservation 171, 247–258. https://doi.org/10.1016/J.BIOCON.2014.01.036

Serrão, E.A.D., Simão Neto, M., 1, 1971. Informações sobre duas espécies de

gramíneas forrageiras do gênero Brachiaria na Amazônia : B.decumbens Stapf e

B.ruziziensis Germain et Evrard, in: Estudos Sobre Forrageiras Na Amazônia. p. 3.

Soares-Filho, B.S., Nepstad, D.C., Curran, L.M., Cerqueira, G.C., Garcia, R.A.,

Ramos, C.A., Voll, E., McDonald, A., Lefebvre, P., Schlesinger, P., 2006. Modelling

conservation in the Amazon basin. Nature 440, 520–

523. https://doi.org/10.1038/nature04389

Souza Filho, A.P.S., Rodrigues, L.R., Rodrigues, T.D., 1997. Inibição da germinação

e alongamento da radícula de invasoras de pastagens pelos extratos aquosos de

gramineas forrageiras tropicais. Pasturas Tropicales 19, 45–50.

Spracklen, D. V., Garcia-Carreras, L., 2015. The impact of Amazonian deforestation

on Amazon basin rainfall. Geophysical Research Letters 42, 9546–

9552. https://doi.org/10.1002/2015GL066063

Staver, C., 1989. Shortened bush fallow rotations with relay-cropped Inga edulis and

Desmodium ovalifolium in wet central Amazonian Peru. Agroforestry Systems 8,

https://doi.org/10.1016/J.BIOCON.2014.01.036

https://doi.org/10.1038/nature04389

https://doi.org/10.1002/2015GL066063

150

173–196. https://doi.org/10.1007/BF00123120

Szott, L.T., Palm, C.A., Sanchez, P.A., 1991. Agroforestry In Acid Soils Of The

Humid Tropics. Advances in Agronomy 45, 275–301. https://doi.org/10.1016/S0065-

2113(08)60043-7

Toledo-Aceves, T., Swaine, M.D., 2007. Effect of three species of climber on the

performance of Ceiba pentandra seedlings in gaps in a tropical forest in Ghana.

Journal of Tropical Ecology 23, 45–52. https://doi.org/10.1017/S0266467406003725

Tomasini, S., Theilade, I., 2019. Local ecological knowledge indicators for wild plant

management: Autonomous local monitoring in Prespa, Albania. Ecological Indicators

101, 1064–1076. https://doi.org/10.1016/J.ECOLIND.2019.01.076

Torres-Sovero, C., González, J.A., Martín-López, B., Kirkby, C.A., 2012. Social–

ecological factors influencing tourist satisfaction in three ecotourism lodges in the

southeastern Peruvian Amazon. Tourism Management 33, 545–

552. https://doi.org/10.1016/J.TOURMAN.2011.06.008

Villachica, H., de Carvalho, J.E.U., Mercado, R.S., Izquierdo, J. and Egg, A.B.,

1996. Frutales y hortalizas promisorios de la Amazonia (Vol. 44). Lima: Tratado de

Cooperacion Amazonica, Secretaria Pro-Tempore.

Westoby, M., Wright, I.J., 2006. Land-plant ecology on the basis of functional traits.

Trends in Ecology & Evolution 21, 261 268.

https://doi.org/10.1007/BF00123120

https://doi.org/10.1016/S0065-2113(08)60043-7

https://doi.org/10.1016/S0065-2113(08)60043-7

https://doi.org/10.1017/S0266467406003725


https://doi.org/10.1016/J.TOURMAN.2011.06.008

151

https://doi.org/10.1016/J.TREE.2006.02.004

Wikander, T., 1984. Mecanismos de Dispersion de Diasporas de una Selva Decidua

en Venezuela. Biotropica 16, 276. https://doi.org/10.2307/2387936

Zhang, W., Wang, B.J., Gan, Y.W., Duan, Z.P., Hao, X.D., Xu, W.L., Li, L.H., 2019.

Different tree age affects light competition and yield in wheat grown as a companion

crop in jujube-wheat agroforestry. Agroforestry Systems 93, 653–

664. https://doi.org/10.1007/s10457-017-0160-z

https://doi.org/10.1016/J.TREE.2006.02.004

https://doi.org/10.2307/2387936

https://doi.org/10.1007/s10457-017-0160-z

-152-

CHAPTER III

TREE GROWTH IMPACTS ON GRASS AND SOILS IN FIVE MONOSPECIFIC

SILVOPASTORAL SYSTEMS IN AMAZONIA

Abstract

The integration of trees in pastures, known as silvopastoral systems, relies on

selection of species that are appropriate for a given site. This study seeks to support

the development of silvopastoral systems by critically evaluating the growth of trees

and pasture in a Brachiaria brizantha cv. Marandú pasture in the southwestern

Amazon. Leucaena leucocephala, Inga edulis, Erythrina berteroana, Ceiba

pentandra, Senegalia loretensis and a no-tree control treatment were planted in

monospecific alleys of 0.5 x 7.5 m in five replicates of 0.9 ha random block plots.

Three censuses were carried out at 6-month intervals to evaluate the relative growth

rates of tree species to 18 months. I. edulis and L. leucocephala were the tallest

species after 18 months (443 cm, P<0.05) and I. edulis had a increased root collar

diameter (6.9 cm, P<0.05) and diameter at breast height (3.6 cm, P<0.05) when

compared to other species. 1m2 quadrats were used to evaluate the rate of grass

regrowth between treatments. Prior to pollarding the grass regrowth in I. edulis plots

was 50% less than that of any other treatment. Subsequently, all tree species were

pollarded to 2 m at 18 months. Post-pollard, all tree species treatments except C.

pentandra had a negative impact on the accumulation of grass biomass when

-153-

compared to the control treatment. I. edulis produced more pollard biomass than

other tree species (0.75 Mg ha-1year-1, P<0.05). Soil nitrogen, phosphorous and

organic matter measured at 18 months showed no effect of treatment. I. edulis, L.

leucocephala and E. berteroana are recommended for further investigation at the

present site.

1. Introduction

Livestock production currently contributes around 14.5% to total global greenhouse

gas emissions and is a leading driver of global land cover change (Hardin, 1968;

Steinfeld et al., 2006; McMichael et al., 2007; Lambin and Meyfroidt, 2011; Gerber,

2013; Grossi et al., 2019). Global meat production is projected increase 15% by

2027 to meet growing demand (OECD‐ FAO, 2018). Rising temperatures are

predicted to decrease the forage quality of grasses and this may lead to increased

methane outputs, thereby generating a positive feedback effect that could

exacerbate climatic change (Lee et al., 2017). The use of livestock production

systems that incorporate trees or shrubs, also known as silvopastoral systems, are

associated with increased environmental benefits than conventional livestock

production (Nepstad et al., 2006; Schiermeier, 2019). The presence of trees in

pastures helps maintain soil functioning, hosts biodiversity and is increasingly being

recognized as providing an important function in resilience to climate change in rural

farming communities (Reed, 2013; Altieri et al., 2015). Silvopastoral systems have

been used to implement landscape scale conservation interventions, to increase

livestock or timber productivity per hectare of land and have been found to sequester

-154-

more carbon than the traditional grass monoculture approach to livestock production

(Pagiola et al., 2004; Haile et al., 2010; Tonucci et al., 2011; Broom et al., 2013).

In some regions of the Amazon, the current paradigm of livestock production is of a

cattle stocking density of less than 1 ha-1 ( Valentim and Andrade, 2009; Bowman et

al., 2012; Reis, 2017). Cattle contribute to unequal nutrient deposits and pastures

can reach an advanced stage of degradation in ten years (Boddey et al., 2004;

Martínez and Zinck, 2004). Cattle pastures are viewed as the antithesis of the

ecosystem services provided by intact forests, yet these degraded pastures

represent an excellent opportunity for rehabilitation using silvopastoral systems.

There are a paucity of studies that attempt to explore potential native forages that

would be suitable for silvopastoral systems in the Amazon biome (Hohnwald et al.,

2016). To enhance the success of a plantation requires careful selection of the tree

species to match the environmental conditions of a potential site (Butterfield, 1995).

In many cases, the forage potential of locally adapted species is unknown and there

is a need to identify native trees and shrubs for forage production (Murgueitio et al.,

2011). The pasture environment in the Amazon presents a unique challenge as the

environmental conditions include poor, compacted soils with low pH, high levels of

exchangeable aluminum, high levels of irradiation and predominantly monocultures

of exotic African grasses such as Brachiaria sp. that may have deleterious effects on

adjacent vegetation ( Moraes et al., 1996; Loker et al., 1997; Nesper et al., 2015).

Authors have called for the exploration of the agroforestry potential of native species

as they may be well adapted to their environment and may be able to increase the

-155-

provision of ecosystem services (Butterfield, 1995; Fisher, 1995; Murgueitio et al.,

2011). The present study seeks to test the potential of candidate species for

silvopastoral systems in the SW Amazon via a 21-month field experiment in a

Brachiaria brizantha cv. Marandú pasture. This experiment compares the

establishment success and growth of five tree species and their impact on the

production of pasture grass and soil chemical parameters nitrogen (N), phosphorous

(P) and organic matter (OM). It is hypothesized that:

(a) I. edulis will demonstrate the fastest relative growth rate

(b) L. leucocephala will demonstrate the highest mortality rate

(c) The presence of trees will positively effect soil N and OM and negatively effect

soil P.

(d) I. edulis will produce the greatest quantity of biomass when harvested by a

pollard > 2 m.

(e) The rate of grass growth will be negatively influenced by the presence of all

tree species.

2. Methods

A detailed description of experimental design and planting arrangement for this

experiment can be found in Chapter II and briefly described here. The experiment

used a randomised block design with five replicates. There were six treatments in

the experiment, with each treatment involving a pasture dominated by Brachiaria

brizantha cv. Marandú. The treatments were (1) Leucaena leucocephala, (2) Inga

edulis, (3) Ceiba pentandra, (4) Senegalia loretensis, (5) Erythrina berteroana, (6)

Control of Brachiaria brizantha cv. Marandú with no tree association. These

-156-

treatments were randomly allocated between thirty plots of 30 x 30 m (0.09 ha). Tree

treatments were planted in three rows of 0.5 x 7.5 m. There were a total of 180 trees

per plot.

2.1. Censuses of Tree Growth

All trees were individually labelled. The first census of plant growth was carried out

immediately after the trees had been planted in March 2017, and represents the

initial conditions (C1). The census was subsequently repeated in August 2017 (C2),

March 2018 (C3) and August 2018 (C4), for a total of three, six-month intervals of

plant growth between the censuses (Table 5). A 2.5 m strip around each plot was

excluded from all censuses to reduce edge effects. All trees were monitored during

the first census in March 2017 and the last census in August 2018. During the

census of August 2017 and March 2018, only the 63 trees in the centre of each plot

were monitored (21 from each of the three rows). Fewer trees were monitored during

the August 2017 and March 2018 census due to a lack of resources.

-157-

Table 5. Plant growth census dates.

Census Year Month

Months post-

planting

Interval between census

C1 2017 March 0 - C2 2017 August 6 6 C3 2018 March 12 6 C4 2018 August 18 6

Measurements were taken with a digital Vernier calliper and a 5 m measuring tape.

The following measurements were taken: height from ground to apical meristem of

tallest branch (height, measured in cm), root collar diameter (RCD, measured in mm)

and diameter at 130 cm from the ground (DBH, measured in mm). The tree was

reported as ‘alive’ (i.e. healthy), ‘regrowth’ or ‘dying’. The regrowth category was

employed when the majority (50-100% of the branch and stem material) of the tree

had died and then had begun to regrow. Trees in the regrowth or dying category

were excluded from subsequent analysis. The cause of tree mortality was reported if

known. A replicate of I. edulis was excluded from the analysis due to heavy and

persistent flooding of the plot at the time of experimental establishment. A replicate

of C. pentandra was also excluded from the analysis due to frequent and complete

defoliation by leafcutter ants.

Statistical Analysis

An annual mean relative growth rate was calculated for each census interval to

facilitate comparison between species according to South (1995):

(Equation 2)

-158-

MRGR = (ln(X)2 – ln(X)1)/(t2 – t1)

where X1 and X2 is the size at the beginning t1 and end t2 of the census interval.

Annual mortality (λ) was calculated for each census according to equation 7 in Shiel

et al. (1995):

(Equation 3)

λ = (logeN0 – logeN1)/t

Where N0 is the population count at the beginning of a census and N1 is the

population count at the end of the census and t is the measurement interval.

Statistical analysis of growth rates of trees excluded trees that had been placed into

the regrowth or dying categories. All analyses were carried out using the program R

(3.2.2). Shapiro-Wilk tests were used to test the normality of the distribution of all

continuous measurement variables (height, RCD, and DBH). For each census, a

one-way ANOVA was used to test for differences between each of the

measurements and differences between species (categorical). A Tukey’s Honest

Significant Differences test was used post-hoc to compare the differences between

mean measurements for all species in each census.

2.2. Soil Chemical Analysis

The three tree species that had exhibited the lowest mortality to the time of January

2019 were I. edulis, E. berteroana and L. leucocephala. These species were

therefore chosen as the most likely to have had an effect on the soil chemistry. Soil

chemical analysis was carried out on a total of 16 plots composed of four plots of the

-159-

control treatment of no trees and four plots of each tree treatment. Total nitrogen (N),

phosphorous (P) and organic matter (OM) samples were taken from 0-10 cm of the

soil. Total N, P and soil organic matter were available at the laboratory and due to

limited resources these were chosen as priorities, as the nitrogen fixing species may

increase N, P was expected to decrease under plantation conditions and OM to

increase as a result of input via litterfall. However, due to unforeseen conditions the

samples were taken sooner than originally envisaged (samples were taken in

January 2019, at 24 months post establishment of the experiment, but had originally

been planned for 31 months). For the 20 samples that were taken from within the

tree line the sample was taken from between adjacent two trees (0.25 m from the

base of the two neighboring trees). The samples were mixed and three subsamples

were taken. The same procedure was used between the rows of trees where

samples were taken at a distance of 3.75 m from the tree line. Samples were sent for

analysis in the Laboratorio de Analysis de Suelos de la facultad de Agronomia of the

La Molina University in Lima.


The response variables of N, P and OM were tested for normality using Shapiro-Wilk

test. Differences in N, P, and OM (continuous) between treatments (categorical)

were tested by an ANOVA. To test for an effect of distance from the tree line on N, P

and OM, ANOVAs were run for each tree species with the distance from the tree line

(1 m or 3.75 m) as a categorical predictor variable.

-160-

2.3 Pollard Treatment

Trees of all species were pollarded at 18 months post planting, immediately following

the fourth census of plant growth. This was due to the observation that the grass

production near the base of some species of trees was being negatively effected,

presumably by the competition for light. In order to reduce competition all trees were

pollarded to 2 m. The pollarded biomass of approximately 20 trees per replicate was

collected and weighed. After each sampling event, three subsamples were taken

from a composite of plant material from each species. Subsamples were air dried for

48 hours then oven dried at 65 °C until a constant dry weight was reached. All

harvested plant material was removed from the experimental site. Differences in the

amount of pollarded biomass harvested (continuous variable) were tested between

treatments (categorical variable) using an ANOVA.

2.4. Effects of Trees on Rates of Pasture Regrowth

Quadrats (1 m2) were situated at random either at 1-2 m or 3.25-3.75 m from the

middle row (of three rows of trees within each parcel) within each plot and no less

than 7.5 m from the plot edge to reduce edge effects. The distances of 1-2 m or

3.25-3.75 m were chosen to test for the effect of the proximity of trees on the grass.

The quadrats were located at random in control parcels with no trees. Prior to the

pollarding of all trees to 2 m the rate of pasture regrowth was monitored for a 30-day

period in the first week of July 2018. In all plots, the grass in the eight quadrats was

cut to 10 cm before and after a 30-day period. The accumulated biomass was

weighed fresh, air-dried for 48 hours then oven dried at 65 °C until a constant dry

weight was reached.

-161-

To test an effect of pollarding on grass production, the clipped parcels were left to

regrow for six months until the first week of January 2019, when the accumulated

biomass was clipped to 10 cm and weighed. All biomass was weighed fresh, air-

dried for 48 hours then oven dried at 65 °C until a constant dry weight was reached.


The two monitoring periods of grass biomass were treated separately. The dry

weight of the grass was inspected for outliers and normality using density and scatter

plots. The Shapiro-Wilk test suggested that the dry weights of the grass biomass

were normally distributed. One-way ANOVAs were used to test for the effect of the

categorical predictor variable treatment on the continuous response of grass dry

weight. Where significant effects were found, these were explored with the post-hoc

analysis Tukey’s Honest Significant Differences test. The effect on grass dry weight

of distance from the tree line was tested within treatments by following the same

procedure.

3. Results

After 18 months, I. edulis and L. leucocephala were the tallest species and both

reaching similar mean heights, for I. edulis this was 442.8 ± 103.5 (S.D.) and for L

leucocephala this was 443.1 ± 111 (S.D.) (Table 6). When compared to that of L.

leucocephala, I. edulis had a diameter at breast height (DBH) that was 44%

increased in size and a root collar diameter (RCD) that was 20% increased in size

(Table 6). I. edulis, L. leucocephala, and E. berteroana were the three of the five

-162-

species that consistently showed growth in all parameters over the 18 months of the

experiment (Table 6). All species demonstrated the highest growth rates in 6 months

following planting followed by a decline in the growth rate with the exception of E.

berteroana. E. berteroana showed the slowest relative growth rate of height (0.8)

and RCD (0.2) between the first and second census, and subsequently was the only

species to demonstrate increased growth in the interval between the second and

third census (Figure 34, Figure 35, Figure 36). Between the first and second census,

C. pentandra showed the greatest increase in RCD (324%) followed by A.

loretenesis (266%) (Figure 37). Between the third and fourth census these species

had the lowest increase of any species (1 and 2% respectively). In the same inter-

census interval C. pentandra and A. loretenesis displayed a decline in average

height as a result of dieback, but not a decline in average RCD or DBH (Figure 36,

Figure 37, Figure 38). Pictures of the plantation give an idea of the scale and layout

of the experiment (Figures 39-42).

-163-

Table 6. Height (cm), root collar diameter (RCD mm), and diameter at breast height (DBH mm) of five tree species in a Brachiaria brizantha cv. Marandú pasture, with census number (C), time after planting in months (t), mean average (Mean) standard deviation (S.D.), annual relative growth rate (RGR), n as number of individuals sampled and λ as annual mortality (Shiel et al., 1995). Means sharing the same superscript are not significantly different from each other within a census interval (Tukey's HSD, P<0.05)

Species C t Height (cm) RCD (mm) DBH (mm)

λ Mean S.D. RGR n Mean S.D. RGR n Mean S.D. RGR n

S. loretensis 1 1 73.3c 23.5 715 6.9c 2.2 715 0 2 6 197.9b 59 2 261 25.0a 7.6 2.6 262 9.9d 4 222 0.008 3 12 239.4c 86.4 0.4 189 31.5a 9.8 0.5 189 13.4d 5.2 0.6 170 0.089 4 18 113.6c 114 -1.5 371 32.0d 10.4 0 372 14.8d 7 0.2 137 0.601

C. pentandra 1 1 77.5b 16.4 709 10.9a 2.8 709 0 2 6 202.9b 30.4 1.9 252 46.1c 8 2.9 252 19.2c 6.7 247 0 3 12 227.7c 41.7 0.2 252 53.2c 13.3 0.3 252 22.5a 9.9 0.3 250 0 4 18 209.1b 67.1 -0.2 619 53.0c 13.8 0 619 23.7a 10 0.1 540 0.549

E. berteroana 1 1 69.0d 5.9 891 36.4d 14.8 872 0

2 6 101.1c 25 0.8 258 40.9d 16 0.2 258 10.0bd 1.9 36 0.274

3 12 202.4d 73 1.4 263 50.1c 16.7 0.4 262 19.7b 7.9 1.4 209 0.244

4 18 224.7d 87.6 0.2 509 59.2e 20.2 0.3 510 23.1b 9.5 0.3 413 0.48

I. edulis 1 1 79.3b 17.2 699 10.1b 2.7 700 0.025

2 6 194.3b 41.7 1.8 254 30.2b 8.3 2.2 254 11.2b 3.3 240 0.05

3 12 373.3b 68.2 1.3 230 57.7b 17.7 1.3 231 30.2c 9 2 230 0.148

4 18 442.8a 103.5 0.3 553 68.9b 22.2 0.4 553 36.2c 13.1 0.4 551 0.205

L. leucocephala 1 1 112.0a 25 887 8.4a 2.2 887 0

2 6 281.4a 64.3 1.8 429 24.5a 6.1 2.1 429 13.3a 5 424 0.005

3 12 409.7a 103 0.8 225 34.4a 11.1 0.7 226 20.7ab 6.8 0.9 222 0.079

4 18 443.1a 111 0.2 483 44.3a 18.4 0.5 507 23.1ab 8.7 0.2 503 0.281

-164-

Figure 34. Relative growth rate for changes in height of five tree species after planting at six (C2), twelve (C3) and eighteen (C4) month intervals post-planting for tree species Leucaena leucocephala, Senegalia loretensis, Inga edulis, Ceiba pentandra and Erythrina berteroana.

-2

-1

0

1

2

3

C2 C3 C4

Rela

tive

Gro

wth

Rate

of

Heig

ht

Census

S. loretensis C. pentandra

E. berteroana I. edulis

-165-

Figure 35. Relative growth rate for changes in root collar diameter (RCD) and of five tree species at between planting and six (C2), twelve (C3) and eighteen (C4) month intervals post-planting for tree species Leucaena leucocephala, Senegalia loretensis, Inga edulis, Ceiba pentandra and Erythrina berteroana

0

1

2

3

C2 C3 C4

Rela

tive

Gro

wth

Rate

of

RC

D

Census

L. leucocephala S. loretensis

C. pentandra E. berteroana

I. edulis

-166-

Figure 36. Height (cm) and standard deviation of five tree species growing in a Brachiaria brizantha cv. Marandú pasture as of census 1. March 2017, 2. July 2017, 3. March 2018 and 4. July 2018.

-167-

Figure 37. RCD (mm) and standard deviation of five tree species growing in a Brachiaria brizantha cv. Marandú pasture as of census 1. March 2017, 2. July 2017, 3. March 2018 and 4. July 2018.

-168-

Figure 38. DBH (mm) of five tree species growing in a Brachiaria brizantha cv. Marandú pasture as of census 2. July 2017, 3. March 2018 and 4. July 2018.

-169-

Figure 39. Rows of I. edulis

-170-

Figure 40. Row of Ceiba pentandra

-171-

Figure 41. Replicate of Senegalia loretensis

172

Figure 42. Leucaena leucocephala in March 2017 (above) and October 2017 (below)

173

Figure 43. Proportion of alive, dead and trees classified as regrowth of five tree species in a Brachiaria brizantha cv. Marandú pasture in the SE Amazon 18 months after planting at 0.5 x 7.5 m

After 18 months, I. edulis was the species with the least number of individuals placed

in the regrowth category (Figure 43). C. pentandra had the lowest rate of mortality.

There was considerable damage to all species by rodents and in many cases this

contributed to the mortality of the trees. The rodents stripped away the outer

cambium in a ring around the bottom of the plant, restricting the plant’s access to

water. Where this occurred, the species would attempt to regrow from the base.

C. pentandra reached a maximum average height of 228 cm after 12 months and

subsequently declined in height between 12 to 18 months as a result of dieback. By

0%

100%

E. berteroana I. edulis S. loretensis L.leucocephala

C. pentandra

% o

f c

ate

tgo

ry

Alive Regrowth Dead

174

the fourth census, it was observed that the majority of C. pentandra had dropped

their leaves and dieback had occurred from the apical meristem towards the ground.

When C. pentandra is healthy it exhibits a green trunk, yet by the fourth census, the

trees exhibited a brown or black dieback, followed by a vivid red then yellow

discolouration. This phenomenon was seen in all replicates of C. pentandra. There

were no visible above ground pests and no common leaf discolouration. There was

no local plant pathologist to consult. One replicate of C. pentandra had been

excluded at the start of the experiment due to repeated total defoliation by ants of the

genus Atta that resulted in the mortality of all trees in the plot. S. loretensis survived

well until the 12 month census, at which point the trees began to die. It was observed

that S. loretensis demonstrated increased height and survival when adjacent to I.

edulis.

When the levels of harvested tree biomass were compared, I. edulis produced 1.1

Mg ha-1 of biomass (Figure 44). This was significantly more than L. leucocephala

(0.7 Mg ha-1, P<0.05), and more than double the biomass production of E.

berteroana, C. pentandra or S. loretensis. It was observed at I. edulis produced

substantial foliar and woody biomass, whereas the majority of L. leucocephala

biomass was woody.

175

Figure 44. Mean dry weight biomass and standard deviation (S.D.) of pollarded biomass (over 2 m) of 18-month-old trees 0.5 x 7.5 m planting design (2667 trees ha-1) for individual trees, sample size represented in parenthesis. Means sharing the same superscript are not significantly different from each other (Tukey's HSD, P<0.05)

Prior to the pollarding event the rate of grass regrowth was measured for 30 days. I.

edulis was found to be the only treatment that negatively impacted the rate of grass

regrowth (Figure 45, P<0.01). Grass regrowth in the I. edulis treatment showed less

than half the growth of any other treatment. There was no effect of the distance from

the tree line on the rate of grass regrowth in any species. Six months following the

following the pollard the grass regrowth was reassessed. The control treatment (no

trees) and C. pentandra had the greatest accumulation of grass biomass of all the

tree treatments (Figure 46). No effect of distance from the tree line was found for any

of the treatments.

176

Figure 45. Mean total dry matter biomass accumulation (Mg ha-1 year-1) and standard deviation (S.D.) of 30-day regrowth of Brachiaria brizantha cv. Marandú collected from different treatments, sample size represented in parenthesis. Means sharing the same superscript are not significantly different from each other (Tukey's HSD, P<0.05)

177

Figure 46. Mean total dry matter biomass accumulation (Mg ha-1 year-1) and standard deviation (S.D.) of six months regrowth of Brachiaria brizantha cv. Marandú following the pollarding of associated tree species to 2 m, sample size represented in parenthesis. Means sharing the same superscript are not significantly different from each other (Tukey's HSD, P<0.05)

There was no effect of treatment on the soil chemical parameters nitrogen (N),

phosphorous (P) or organic matter (OM). There was no effect of the distance from

the tree line on N, P or OM for E. berteroana, I. edulis or L. leucocephala.

4. Discussion

Trees and grass compete for nutrients, light, and water. In this experiment, all

planted trees were competing with a pasture grass, Brachiaria brizantha cv. Marandu

for resources. Inga edulis was the species of tree that was expected to perform well

in the pasture environment and therefore it was hypothesised that I. edulis would

178

have the fastest relative growth rate. I. edulis produces a high proportion of foliar to

total biomass. The foliage of I. edulis has the highest leaf area and light interception

of the tree species in this experiment. In a B. brizantha pasture environment this

adaptation favours I. edulis, as the Brachiaria genus is averse to shade and

therefore the ability of a tree to generate enough shade to protect its roots from

competition with B. brizantha could be strategically important to the practice of

pasture reclamation. B. brizantha is highly invasive and can spread and rapidly

dominate landscapes. Multiple studies report that pasture grasses have a negative

effect on growth and survival of trees in pastures (Ferguson et al., 2003; Hooper et

al., 2002; Hooper et al., 2005). African grasses of the Brachiaria genus contain

phytotoxic substances that produce an allelopathic effect that inhibits the growth of

adjacent plants (Kobayashi and Kato-Noguchi, 2015; Sesto Cabral et al., 2011;

Pivello et al., 1999; Ganade and Brown, 2002).

It is possible that I. edulis was able to avoid the effect of the allelopathic chemicals in

the B. brizantha grass by shading out the proximate grasses and opening up a niche

within the grass. ‘Pioneer’ or ‘colonising’ tree species open a niche which in a

silvopastoral system may include relief from the grass, increased shade, increased

soil moisture and decreased soil compaction (Khurana and Singh, 2001; Garen et

al., 2011). Otsamo (1998) found that Anisoptera marginata grew fastest using the

nurse tree Paraserianthes falcataria on an Imperata cylindrica grassland site in

Indonesia. Similarly, the dense shade and ability to suppress weeds may give I.

edulis a unique advantage in the pasture environment and simultaneously open a

niche for additional biodiversity. Nichols and Carpenter (2006) found that Terminalia

amazonia showed improved growth when interplanted with I. edulis. A positive effect

179

of adjacent I.edulis was observed on the survival of S. loretensis. However, the

present study only contained three incidences where rows of S. loretensis and I.

edulis were adjacent, so it was not possible to quantify this interaction. However, this

observation highlights the importance of accounting for edge effects of adjacent

treatments in sampling design. A future study could attempt to thin an existing I.

edulis plantation to introduce potential tree or shrub species such as C. pentandra

and S. loretensis that may benefit from being interplanted with I. edulis.

Prior to the pollard, I. edulis was the most effective tree species at inhibiting the

growth of the grass. However, the pollard reduced this species’ total leaf area and

subsequently I. edulis was only as effective than any other tree species at controlling

the production of grass biomass. In this experiment, the trees were pollarded as a

strategy to optimise the production of available tree foliage as fodder for cattle.

However, the results suggest that an important ecosystem function of I. edulis in

pasture reclamation is the creation of a niche via the production of shade. Further

work is needed to understand the interaction between the optimal proportion of trees

that can be pollarded for potential forage resources, and the optimal proportion of

trees that should be left without the pollard for the production of shade and

subsequent control of the B. brizantha grass.

In the present study the aim was to maximise the production of biomass for cattle

production. By 12 months all species were over 2 m tall and therefore beyond the

reach of cattle. Pollarding is therefore a useful management strategy for managers

as removal of biomass above a certain point has the dual benefit that it can be used

to increase the availability of biomass by bringing forage into the reach of animals,

180

and it can also reduce the light competition by the trees on adjacent plants. Some

evidence exists to indicate which species can survive pruning and pollarding, but

there are also species that cannot (Lang et al., 2015; Muschler et al., 1993; Garcia et

al., 2001). Care must be taken when choosing the best management approach for

each species. Separating the foliar and the branch biomass at the 18-month pollard

harvest would have improved the experiment, as it is the foliar biomass that is of

interest. In future studies, it would be of interest to measure the survival and

regrowth of trees following pollarding. This was beyond the scope of the present

study.

The failure of C. pentandra and S. loretensis to survive in the experiment was due to

an unknown cause. It is speculated the cause may have been a plant pathogen or

the decline may have resulted from competition with the pasture grass. It was

hypothesised that Leucaena leucocephala would have the greatest mortality rate of

any of the tree species, but this was not the case. This hypothesis was chosen

because L. leucocephala was reported to have failed to establish in a previous trial in

the Amazon, believed to be due to a negative effect of acidic soils and high levels of

exchangeable aluminium on this species (Salazar et al., 1993). However, L.

leucocephala has been observed to occur as a weed in the urban areas of the

Peruvian jungle cities of Puerto Maldonado and Pucallpa (personal observation). L.

leucocephala is a widely used forage tree in silvopastoral systems in tropical regions.

In this experiment, the decision to include L. leucocephala was taken because it is a

species that has existing reports of growth and biomass production and therefore it

was considered appropriate to use as a reference species by which the native

species could be compared. Yet it is also recognised as an invasive species that can

181

quickly colonise a site to the exclusion of native species. Managers of natural

systems may choose different priorities for the landscapes they devise. This study

considers the business as usual scenario in the region where cattle are produced in

association with invasive grass. The objective was to increase the agricultural

productivity and flow of ecosystem services from an already modified agricultural

landscape. While native species may be preferable, at the outset of this study it was

not known which species would be able to survive in the pasture environment, or if

any species would be able to naturally regenerate within the pasture environment.

The potential for natural regeneration of these species within pastures is an area for

future research, and of particular interest where there is a risk of invasion by L.

leucocephala. The growth and biomass production of L. leucocephala has been

shown to be effected by provenance (Prasad et al., 2011). A limitation of the present

study is it only considers a single provenance of any of the species.

It was hypothesised that the trees would positively effect soil nitrogen (N) and soil

organic matter (OM) and negatively effect soil phosphorous (P). This experiment had

limited resources and therefore it is not possible to prove or disprove this hypothesis.

There was no effect detected either between treatments or at the two distances from

the tree line for any of the soil chemical parameters considered. While no treatment

effect was detected, it is possible that there was a treatment effect that was not

detected in the present sampling regime, or that it is too soon to see measurable

effect. Therefore, the results should be treated with caution. It was not possible in

this study to conduct soil chemical analysis prior to the establishment of the

experiment. However, where resources permit this is a useful exercise. This

experiment was designed to contain five replications of each treatment with the

182

intention of minimising any differences across the site, such as soil chemistry, soil

moisture or soil compaction. Szott and Palm (1996) found that soil phosphorous

decreased following establishment of I. edulis due to the uptake of phosphorous in

the living biomass. They theorised that the phosphorous would eventually return to

the soil via leaf litter. Fisher (1995) found I. edulis increased soil nitrogen and

extractable phosphorous after 7 years. However, if the objective is production for

forage then the continuous removal of foliar biomass for fodder may result in the

exportation of nutrients held within the biomass that is removed. This could result in

a net loss of nutrients over time.

It was hypothesised that I. edulis would produce the greatest quantity of biomass

when harvested by a pollard that removed all biomass over 2 m height from the

ground. The findings of this study support this hypothesis. While this is a seminal

result for this management approach, biomass production is closely related to growth

parameters. It is possible to contextualise these findings by reviewing the results of

other studies that have experimentally grown I. edulis. I. edulis individuals in this

study were taller and thinner at breast height (DBH) than in Alegre’s (2005)

plantation in Yurimaguas, Peru. Alegre (2005) planted I. edulis at 4,444 trees ha-1 in

a 1.5 m by 1.5 m arrangement, as both a monoculture and also grown in association

with Centrosema macrocarpum. After the first year, the trees that had been grown as

a monoculture treatment reached a height of 2.5 m and a DBH of 4.4 cm. When I.

edulis was grown with C. macrocarpum, the height in the first year was 3.0 m and

the DBH 3.8 cm. The present study is not directly comparable with other studies due

to differences in planting arrangements that could be responsible for differences in

the tree morphology. The growth of I. edulis in the present study is in the same range

183

of that reported by Kanmegne (2000) in Southern Cameroon on a Ferric Acrisol soil

type. Kanmegne (2000) planted I. edulis at 3,333 trees ha-1, in 3 x 1 m rows. I. edulis

grew to 3.7 m in 12 months and by 20 months had reached a height of 7.7 m

(Kanmegne, 2000). Tilki and Fisher (1998) planted I. edulis at 1,111 trees ha-1 in a 3

x 3 m arrangement in three sites in Sarapiqui, Costa Rica. After three years, I. edulis

displayed an average DBH of 10.5 cm. Shimamoto et al. (2014) measured the

growth of naturally regenerating I. edulis in the Parana State in Southern Brazil. By

the age of 18 – 41 years they report that the average DBH of I. edulis was 24.4 cm

(± 11 SD) and an average height of 16.2 m (± 4 SD). These varied results show that

I. edulis is a species with fast growth in a number of different plantation trials.

The limitations of this experiment include the short time frame, a monospecific

planting arrangement of the five species may have favoured or prejudiced certain

species, and that only pollarded tree biomass was reported when a measure of total

tree biomass would have enabled a comparison of whether tree and grass

arrangements are more productive than grass only arrangements. It may also be of

interest to experiment with the addition of fertiliser, changing the associated grass or

tree species, and attempting the experiment in a location with a different set of

edaphic and climatic conditions. Future studies may also consider the reproduction

and establishment success for all species from seed in the invasive and allelopathic

Brachiaria brizantha cv. Marandú or how establishment is effected by exposure to

browsing. Further work would be required to understand the additional benefits of

silvopastoral systems, in particular, marketable benefits that could be used to

promote their uptake, such as the potential for carbon sequestration.

184

In the present study the aim was to maximise the production of biomass for cattle

production. By 12 months all species were over 2 m tall and therefore beyond the

reach of cattle. Pollarding is therefore a useful management strategy for managers

as removal of biomass above a certain point has the dual benefit that it can be used

to increase the availability of biomass by bringing forage into the reach of animals,

and it can also reduce the light competition by the trees on adjacent plants. Some

evidence exists to indicate which species can survive pruning and pollarding, but

there are also species that cannot (Lang et al., 2015; Muschler et al., 1993; Garcia et

al., 2001). Care must be taken when choosing the best management approach for

each species. Separating the foliar and the branch biomass at the 18-month pollard

harvest would have improved the experiment, as it was the foliar biomass that was of

interest. In future studies, it would be of interest to measure the survival and

regrowth of trees following pollarding. This was beyond the scope of the present

study.

Inga edulis, Leucaena leucocephala and Erythrina berteroana were the three

species that may be recommended as species for further assessment of their

silvopastoral potential at the study site. C. pentandra and S. loretensis may be

suitable candidates for silvopastoral systems but their relatively higher levels of

mortality shown in this study suggest they are species that are not well adapted to

the conditions of this study site. If the pasture grass inhibits the development of tree

species then planting dense lines of I. edulis and subsequent thinning may create a

niche in which other trees and shrubs can be planted. The limitations of this

experiment include the short time frame, a monospecific planting arrangement of the

five species may have favoured or prejudiced certain species, and that only

185

pollarded tree biomass was reported when a measure of total tree biomass would

have enabled a comparison of whether tree and grass arrangements are more

productive than grass only arrangements. It may also be of interest to experiment by

changes to the associated grass or tree species, and attempting the experiment in a

location with a different set of edaphic and climatic conditions. Future studies may

also consider the reproduction and establishment success for all species from seed

in the invasive and allelopathic Brachiaria brizantha cv. Marandú or how

establishment is effected by exposure to browsing. Further work would be required

to understand the additional benefits of silvopastoral systems, in particular,

marketable benefits that could be used to promote their uptake, such as the potential

for carbon sequestration.

References

Altieri, M.A., Nicholls, C.I., Henao, A., Lana, M.A., 2015. Agroecology and the design

of climate change-resilient farming systems. Agronomy for Sustainable Development

869–890. https://doi.org/10.1007/s13593-015-0285-2

Bowman, M.S., Soares-Filho, B.S., Merry, F.D., Nepstad, D.C., Rodrigues, H.,

Almeida, O.T., 2012. Persistence of cattle ranching in the Brazilian Amazon: A

spatial analysis of the rationale for beef production. Land Use Policy 29, 558–

568. https://doi.org/10.1016/j.landusepol.2011.09.009

Butterfield, R.P., 1995. Promoting biodiversity: advances in evaluating native species

for reforestation. Forest Ecology and Management 75, 111–121.

https://doi.org/10.1007/s13593-015-0285-2

https://doi.org/10.1016/j.landusepol.2011.09.009

186

Fisher, R.F., 1995. Amelioration of degraded rain forest soils by plantations of native

trees. Soil Science Society of America Journal 59, 544–

549. https://doi.org/10.2136/sssaj1995.03615995005900020039x

Ganade, G., Brown, V.K., 2002. Succession in old pastures of central Amazonia:

Role of soil fertility and plant litter. Ecology 83, 743–

754. https://doi.org/10.1890/0012-9658(2002)083[0743:SIOPOC]2.0.CO;2

Garcia, H., Nygren, P., Desfontaines, L., 2001. Dynamics of nonstructural

carbohydrates and biomass yield in a fodder legume tree at different harvest

intensities. Tree Physiology 21, 523–531. https://doi.org/10.1093/treephys/21.8.523

Garen, E.J., Saltonstall, K., Ashton, M.S., Slusser, J.L., Mathias, S., Hall, J.S., 2011.

The tree planting and protecting culture of cattle ranchers and small-scale

agriculturalists in rural Panama: Opportunities for reforestation and land restoration.

Forest Ecology and Management 261, 1684–


Gawali, A., Puri, S., Swamy, S.L., 2015. Evaluation Growth and Yield of Wheat

Varieties under Ceiba pentandra (L) Based Agrisilviculture System. Universal

Journal of Agricultural Research 3, 173–181.

Gerber, P.J., 2013. Tackling climate change through livestock : a global assessment

of emissions and mitigation opportunities. Food and Agriculture Organization of the


https://doi.org/10.1890/0012-9658(2002)083%5B0743:SIOPOC%5D2.0.CO;2

https://doi.org/10.1093/treephys/21.8.523


187

United Nations.

Grossi, G., Goglio, P., Vitali, A., Williams, A.G., 2019. Livestock and climate change:

impact of livestock on climate and mitigation strategies. Animal Frontiers 9, 69–

76. https://doi.org/10.1093/af/vfy034

Haile, S.G., Nair, V.D., Nair, P.K.R., 2010. Contribution of trees to carbon storage in

soils of silvopastoral systems in Florida, USA. Global Change Biology 16, 427–

438. https://doi.org/10.1111/j.1365-2486.2009.01981.x

Hardin, G., 1968. The tragedy of the commons. Science (New York, N.Y.) 162,

1243–8. https://doi.org/10.1126/science.162.3859.1243

Hohnwald, S., Trautwein, J., Camarão, A.P., Wollny, C.B., 2016. Relative palatability

and growth performance of capoeira species as supplementary forages in the NE-

Amazon. Agriculture, Ecosystems & Environment 218, 107–

115. https://doi.org/10.1016/J.AGEE.2015.11.015

Khurana, E., Singh, J.S., 2001. Ecology of seed and seedling growth for

conservation and restoration of tropical dry forest : a review. Environmental

Conservation 28, 39–52. https://doi.org/10.1017/S0376892901000042

Kobayashi, A., Kato-Noguchi, H., 2015. The seasonal variations of allelopathic

activity and allelopathic substances in Brachiaria brizantha. Botanical studies 56,

25. https://doi.org/10.1186/s40529-015-0105-7

https://doi.org/10.1093/af/vfy034

https://doi.org/10.1111/j.1365-2486.2009.01981.x

https://doi.org/10.1126/science.162.3859.1243


https://doi.org/10.1017/S0376892901000042

https://doi.org/10.1186/s40529-015-0105-7

188

Lambin, E.F., Meyfroidt, P., 2011. Global land use change, economic globalization,

and the looming land scarcity. Proceedings of the National Academy of Sciences of

the United States of America 108, 3465–

72. https://doi.org/10.1073/pnas.1100480108

Lang, P., Jeschke, M., Wommelsdorf, T., Backes, T., Lv, C., Zhang, X., Thomas,

F.M., 2015. Wood harvest by pollarding exerts long-term effects on Populus

euphratica stands in riparian forests at the Tarim River, NW China. Forest Ecology

and Management 353, 87–96. https://doi.org/10.1016/J.FORECO.2015.05.011

Lee, M.A., Davis, A.P., Mizeck G.G.C., Manning, P., 2017. Forage quality declines

with rising temperatures, with implications for livestock production and methane

emissions. Biogeosciences 14, 1403–1417. https://doi.org/doi:10.5194/bg-14-1403-

2017

Loker, W., Vera, R., Reátegui, K., 1997. Pasture performance and sustainability in

the Peruvian Amazon: Results of long-term on-farm research. Agricultural Systems

55, 385–408. https://doi.org/10.1016/S0308-521X(96)00102-3

Martínez, L.J., Zinck, J.A., 2004. Temporal variation of soil compaction and

deterioration of soil quality in pasture areas of Colombian Amazonia. Soil and Tillage

Research 75, 3–17. https://doi.org/10.1016/j.still.2002.12.001

McMichael, A.J., Powles, J.W., Butler, C.D., Uauy, R., 2007. Food, livestock

https://doi.org/10.1073/pnas.1100480108




https://doi.org/10.1016/S0308-521X(96)00102-3


189

production, energy, climate change, and health. The Lancet 370, 1253–

1263. https://doi.org/10.1016/S0140-6736(07)61256-2

Moraes, J.F.L. de, Volkoff, B., Cerri, C.C., Bernoux, M., 1996. Soil properties under

Amazon forest and changes due to pasture installation in Rondônia, Brazil.

Geoderma 70, 63–81.

Murgueitio, E., Calle, Z., Uribe, F., Calle, A., Solorio, B., 2011. Native trees and

shrubs for the productive rehabilitation of tropical cattle ranching lands. Forest

Ecology and Management 261, 1654–


Muschler, R.G., Nair, P.K.R., Meléndez, L., 1993. Crown development and biomass

production of pollarded Erythrina berteroana, E. fusca and Gliricidia sepium in the

humid tropical lowlands of Costa Rica. Agroforestry Systems 24, 123–

143. https://doi.org/10.1007/BF00706887

Nepstad, D.C., Stickler, C.M., Almeida, O.T., 2006. Globalization of the Amazon soy

and beef industries: Opportunities for conservation. Conservation

Biology. https://doi.org/10.1111/j.1523-1739.2006.00510.x

Nesper, M., Bünemann, E.K., Fonte, S.J., Rao, I.M., Velásquez, J.E., Ramirez, B.,

Hegglin, D., Frossard, E., Oberson, A., 2015. Pasture degradation decreases

organic P content of tropical soils due to soil structural decline. Geoderma 257-258,

https://doi.org/10.1016/S0140-6736(07)61256-2


https://doi.org/10.1007/BF00706887

https://doi.org/10.1111/j.1523-1739.2006.00510.x

190

123–133. https://doi.org/10.1016/j.geoderma.2014.10.010

Majid, N., Paudyal, B., Shebli, Z., 1998. Survival and early growth of Acacia

mangium, Ceiba pentandra and Casuarina equisetifolia on sandy tin tailings.

Pertanika Journal of Tropical Agricultural Science 21, 59–65.

OECD-FAO (Ed.), 2018. Chapter 6. Meat Market situation, in: Agricultural Outlook

2018-2027. p. 150. https://doi.org/10.1787/agr-outl-data-en

Parks, N., 2007. Livestock’s long shadow. Frontiers in Ecology and the

Environment. https://doi.org/10.1890/1540-9295(2007)5[4:D]2.0.CO;2

Pivello, V.R., Shida, C.N., Meirelles, S.T., 1999. Alien grasses in Brazilian savannas:

a threat to the biodiversity. Biodiversity and Conservation 8, 1281–

1294. https://doi.org/10.1023/A:1008933305857

Prasad, J.V.N.S., Korwar, G.R., Rao, K.V., Mandal, U.K., G.R.Rao, Srinivas, I.,

Venkateswarlu, B., Rao, S.N., Kulkarni, H.D., 2011. Optimum stand density of

Leucaena leucocephala for wood production in Andhra Pradesh, Southern India.

Biomass and Bioenergy 35, 227–

235. https://doi.org/10.1016/J.BIOMBIOE.2010.08.012

Reis, E., 2017. Opportunities and challenges to the sustainable development of

cattle raising in Brazil, 1970–2005. Economia 18, 18–

39. https://doi.org/10.1016/J.ECON.2016.07.004

https://doi.org/10.1016/j.geoderma.2014.10.010

https://doi.org/10.1787/agr-outl-data-en

https://doi.org/10.1890/1540-9295(2007)5%5B4:D%5D2.0.CO;2

https://doi.org/10.1023/A:1008933305857

https://doi.org/10.1016/J.BIOMBIOE.2010.08.012

https://doi.org/10.1016/J.ECON.2016.07.004

191

Salazar, A., Szott, L.T., Palm, C.A., 1993. Crop-tree interactions in alley cropping

systems on alluvial soils of the Upper Amazon Basin. Agroforestry Systems 22, 67–

82. https://doi.org/10.1007/BF00707472

Schiermeier, Q., 2019. Eat less meat: UN climate-change report calls for change to

human diet. Nature 2019 572:7769.

Sesto Cabral, M.E., Schabes, F.I., Sigstad, E.E., 2011. A calorimetric study of the

interaction between Brachiaria platyphylla and soil microbial activity. Thermochimica

Acta 526, 157–162. https://doi.org/10.1016/J.TCA.2011.09.015

Shimamoto, C.Y., Botosso, P.C., Marques, M.C.M., 2014. How much carbon is

sequestered during the restoration of tropical forests? Estimates from tree species in

the Brazilian Atlantic forest. Forest Ecology and Management 329, 1–

9. https://doi.org/10.1016/J.FORECO.2014.06.002

Sheil, D., Burslem, D.F. and Alder, D., 1995. The interpretation and misinterpretation

of mortality rate measures. Journal of Ecology, pp.331-333.

South, D.B., 1995. Relative Growth Rates: a Critique. South African Forestry Journal

173, 43–48. https://doi.org/10.1080/00382167.1995.9629690

Szott, L.T., Palm, C.A., 1996. Nutrient stocks in managed and natural humid tropical

fallows. Plant and Soil 186, 293–309. https://doi.org/10.1007/BF02415525

https://doi.org/10.1007/BF00707472

https://doi.org/10.1016/J.TCA.2011.09.015


https://doi.org/10.1080/00382167.1995.9629690

https://doi.org/10.1007/BF02415525

192

Tilki, F., Fisher, R.F., 1998. Tropical leguminous species for acid soils: studies on

plant form and growth in Costa Rica. Forest Ecology and Management 108, 175–

192. https://doi.org/10.1016/S0378-1127(98)00225-4

Tonucci, R.G., Nair, P.K.R., Nair, V.D., Garcia, R., Bernardino, F.S., 2011. Soil

Carbon Storage in Silvopasture and Related Land-Use Systems in the Brazilian

Cerrado. Journal of Environment Quality 40,

833. https://doi.org/10.2134/jeq2010.0162

Valentim, J.F., Andrade, C.M.S. de, 2009. Tendências e perspectivas da pecuária

bovina na amazônia brasileira. Amazônia: Ciência & Desenvolvimento 4, 7–30.

https://doi.org/10.1016/S0378-1127(98)00225-4

https://doi.org/10.2134/jeq2010.0162

193

CHAPTER IV

ALLOMETRY AND PRODUCTION OF THREE

MONOSPECIFIC SILVOPASTORAL SYSTEMS IN AMAZONIA

Abstract

The integration of trees could increase biomass production and carbon sequestration

in the cattle pastures of the Amazon rainforest. This paper evaluates the biomass

production at 24 months of three silvopastoral systems following a pollarding event

at 18 months. This study measured the root collar diameter, stump diameter,

diameter at breast height, height, crown volume, and crown diameter to predict the

biomass of total, trunk, branch and foliage components. Species and component

specific allometric equations were developed by comparison of linear, quadratic and

log-linear models using Furnival’s Index. Log-linear models were found to be

superior to linear or quadratic models for the prediction of biomass components in all

species. The results showed that a single parameter model of RCD was the best

predictor of all biomass components of Inga edulis, whereas for Erythrina berteroana

and Leucaena leucocephala a multiple model of RCD and height improved the

predictive measurements of total and trunk biomass. There was reliable prediction of

total biomass for I. edulis (R2= 0.70), E. berteroana (R2= 0.77) and L. leucocephala

(R2= 0.86). After 24 months, I. edulis had accumulated the most biomass (7.2 Mg ha-

1) when compared to E. berteroana (4.4 Mg ha-1) or L. leucocephala (4.3 Mg ha-1).

The predicted carbon sequestration of these species was greatest for I. edulis (3.4

194

Mg ha-1) when compared to E. berteroana (2.1 Mg ha-1) or L. leucocephala (2.0 Mg

ha-1). Silvopastoral systems have the potential to increase the carbon sequestration

of pastures in the Amazon rainforest.

1. Introduction

Livestock production is a major driver of global land cover change (Steinfeld et al.,

2006). The pasturelands once occupied by Amazon Rainforest are typically

unsustainable, and pasture degradation may occur within 10 years (Martínez and

Zinck, 2004). There is a need to develop a new paradigm that can reverse this trend

and enhance the provision of ecosystem services from these agricultural landscapes

(Serrão et al., 1996). It has been shown that silvopastoral systems, the inclusion of

trees and shrubs into pastures, can result in increased livestock production and also

be instrumental in landscape level sustainability (Mauricio et al., 2019). However,

during the establishment phase of silvopastoral systems, cattle may be detrimental

to the establishment of plants and for that reason may need to be excluded from

young plantations (Smit et al., 2015). Therefore, farmers that wish to transform their

land to silvopastoral systems may experience the cost of a temporary loss of land.

Farmers could be compensated for this short-term loss of land and receive payments

for the conversion of land to silvopastoral systems under schemes known as

payments for ecosystem services (Pagiola et al., 2007). Such schemes rely on an

understanding of the costs and benefits of providing an ecosystem service (Claassen

et al., 2008; Tzilivakis et al., 2019).

195

Sharrow and Ismail (2004) hypothesised that as pastures store the majority of their

carbon (C) below ground in soil organic matter whereas forests store the majority of

their C above ground that silvopasture should accumulate more C via increased

biomass production and considerable below and above ground C sequestration.

They found that after 11 years, silvopasture accumulated more C than forest (0.74

Mg ha-1 year-1) or pasture (0.52 ha-1 year-1). C sequestration per hectare is a

measurable ecosystem service for which several markets exist, and payments to

farmers for sequestered C may incentivise them to integrate trees into their pastures

(Corbera et al., 2009; Hissa et al., 2019; Blum and Lövbrand, 2019).

The International Panel on Climate Change produced guidelines in 2003 for the

estimation of C stocks via allometric equations (Penman et al., 2003). These were

intended for the measurement of C for trade mechanisms that arose from the Kyoto

protocol (for example, Reduced Emissions from Deforestation and Degradation).

Allometric equations seek to determine the potential of one measured variable to

predict another variable. A critical application of this theory is that of quantifying the

volume of tree biomass to predict C stock. For example, the diameter of a tree can

be used to predict the total biomass of that tree. A benefit of allometric equations is

that if a strong relationship can be quantified, usually via the destructive

measurement of a subsample of trees, the equation can be used to quantify the

biomass of other trees that do not have to be destroyed. Allometric relationships are

governed by site-specific conditions and management impacts. The coarse

resolution studies that attempt large scale estimates of C stocks can result in high

levels of uncertainty (Chave et al., 2014). Site and species-specific models

demonstrate improved accuracy but are of limited transferability. For example,

196

Huang et al. (2000) found that, when applied to alternative ecoregions from the one

in which they were developed, height-diameter models of white spruce Picea

glauca (Moench) Voss resulted in maximum overestimations of 29% and

underestimations of 22%.

The relationship between variables of tree growth can take many forms. The most

common models in the literature are the power model and the polynomial model

(Picard et al., 2018). When a power relationship is suspected a widely used practice

in the literature is the use of the natural log transformation on the predictor and

response variable data to achieve a linear fit (Kitteredge, 1944; Mascaro et al.,

2014). This relationship can then be described using an ordinary least squares linear

regression and then back-transformed to the arithmetic scale using a conversion

factor (e.g. Marklund, 1987).

While the power function occurs frequently in the field of tree allometry, the

predominance of authors following this approach, with little regard for alternative

methods, has been criticised as failing to detect the most appropriate model for the

distribution of the data (Packard, 2017a; Packard, 2017b). Equations can include

multiple variables but care should be taken that separate variables are not highly

correlated, as this can result in multicollinearity which is known to have a

confounding effect on the model (Sileshi, 2014). Similarly, high order polynomials

begin to result in over-fitting of the model that can result in biologically implausible

models. Sileshi (2014) consider that for each additional predictor the sample size

should be doubled, and noted that ordinary least squares regression can be useful

when few non-collinear predictors are used that have a well-understood relationship

197

to the response variable. Ultimately, the error distribution will determine which

method performs best and therefore a model cannot be selected a priori (Packard,

2014; Xiao et al., 2011; Sileshi, 2014; Mascaro et al., 2014).

Xiao et al. (2011) demonstrated that ordinary least squares linear regression was a

suitable approach for characterising data with multiplicative, heteroscedastic,

lognormal error, whereas data with additive, homoscedastic, normal error was better

suited to non-linear regression approaches. To determine the appropriate error

structure, biological reasoning or likelihood analysis by comparing the Akaike

Information Criteria (AIC) or the coefficient of determination r-squared (R2) is

commonly used (Akaike, 1974; Xiao et al., 2011). However, AIC and R2 are not

appropriate to compare between models where different transformations have been

applied to the data. In such instances, interpretation of the model residuals can be

used in to determine the most appropriate modelling approach. The Furnival’s Index

(1961) is a method that can be used to compare a variety of modelling approaches

to the same dataset, where some models use transformed data and others do not.

The theoretical basis upon which allometric equations are built has also been used

by managers of natural ecosystems who seek to predict the amount of browse

available to herbivores or the amount of browse utilized by herbivores (MacCracken

and Ballenberghe, 1993). A small body of work has investigated the experimental

removal of biomass and subsequent regrowth with the intention of maximising

production of a specific element of biomass such as coppiced sticks or livestock

forage (Harrington and Fownes, 1993; Oppong et al., 2002). There is increasing

demand for landscape level management that optimises the production of multiple

198

ecosystem service benefits, such as forage production, the formation and

maintenance of soil and the sequestration of C (MEA, 2005).

This study seeks to quantify the amount of aboveground total and foliar biomass

produced and C sequestered in the first two years of establishment of three potential

silvopastoral tree species, Inga edulis, Erythrina berteroana and Leucaena

leucocephala. It is assumed that the production of biomass by these tree species

results in additional C sequestration in the pasture when compared to a pasture with

no trees, as previous chapters have shown a negligible impact of these trees on

grass production when managed by pollarding (See Chapter III). The trees of all

three species were harvested 24 months post planting and had previously been

pollarded to a height of 2 m at 18 months. Existing allometric models for E.

berteroana, I. edulis and L. leucocephala were sought on the Internet database

Globallometree (Globallometree.org accessed July 2019). 478 existing equations

were found for L. leucocephala, but no allometric equations existed for I. edulis or E.

berteroana. As allometric equations are specific to the site in which species are

grown and the management regime, it was considered necessary to develop specific

allometric equations for these three species that considered the impact of a

management regime of pollarding at 2 m after 18 months. It was hypothesised that:

(a) When three models were compared, log-linear, linear and quadratic, the log-

linear model would display the best fit.

(b) The greatest single predictor of total biomass will be stem diameter

measurements.

(c) Multiple predictor models will improve performance over single predictor models

for the prediction of biomass components.

199

(d) Measurements of the crown will be the best predictor of total foliar biomass.

2. Methods

2.1 Measurements of Tree Biomass

Three tree species were chosen for the development of allometric models, Inga

edulis, Erythrina berteroana and Leucaena leucocephala. These species had been

grown in a Brachiaria brizantha cv. Marandú pasture in a planting arrangement of 0.5

x 7.5 m for a density of 2,667 trees ha-1 (see Chapter II for further information on the

experimental design). Five replicates of each species were available at the beginning

of the study. This experiment chose three replicates at random from each species.

Each replicate had 180 trees, for a total of 540 trees per species. All trees of all three

species had been pollarded to 2 m at 18 months. The present study occurs after 24

months of growth, and 6 months after the pollard event. Approximately 40% of the

trees in the three replicates were harvested for the development of allometric

equations. A further 30% of trees were harvested to develop estimates of C

sequestration.

In allometric studies, there is a requirement that samples should represent a range of

at least one measurement to develop models that are representative of a range of

size classes. This is because the equations can only describe the relationship

between the variables they are given. In other words, a model that was developed

based on trees of up to 2 m height is likely to display considerable error if applied to

a tree of 20 m height. Therefore, where possible, the sample should include

representatives including the maximum and minimum sizes in a range to increase

200

the accuracy of scale-dependent equations (LaBarbera, 1989). Height, age and

diameter at breast height (DBH, 1.3 m) are commonly used in selecting size classes.

However, this study considered a plantation of trees that were of the same age and

had been pollarded to 2 m. Some trees were shorter than the commonly used

parameter of DBH. Therefore, RCD was used to select size classes. For each tree

species, the RCD data were divided into 10 size classes and nine trees were

randomly drawn from each size class, to represent three trees per replicate in each

size class.

Figure 47. Diagram representing the location of measurements on tree, where CD is crown diameter, CDep is crown depth, CH is crown height, H is height, RCD is root collar diameter, SD is stump diameter and DBH is diameter at breast height.

.

Height (cm) was measured from the ground to the top of the highest meristem

(Figure 47). Crown height (cm) was measured from the ground to where foliage

volume assumes 5% of the total. Callipers were used to measure, at the widest

point, the RCD (mm), diameter at stump height (mm, SD, measured 0.3 m above

201

ground) and DBH (mm). Where trees presented branches or other features that may

have changed the diameter at this point, the decision was taken to measure at the

nearest regular point below the irregularity. In the case of the RCD, if the base of the

tree was irregularly damaged, this measurement was excluded. In trees where

multiple stems were present, each stem was measured and the mean average was

reported (Parent, 2000). A 5 m measuring tape was used to measure crown

diameter (CD, mm), taken as the mean average of three horizontal measurements of

the crown perpendicular to the ground. The crown radius (r) was calculated as the

crown diameter divided by two. Crown depth was derived using equation 4. Crown

area (CA) was assumed to be a circle through observation and was derived using

equation 5. Crown volume (CV) was assumed to equate to a cylinder, the best fit for

the pollarded trees, and was calculated using equation 6:

(Equation 4) Crown depth = height – crown height

(Equation 5) CA = πr2

(Equation 6) CV =πr2CD

Trees were felled at ground level with a chainsaw. For each tree, biomass was

divided into stem, branch and foliage. Due to resource constraints, only aboveground

biomass was considered in the present study. Each biomass component was

weighed separately using an appropriate spring-loaded scale. All biomass was

grouped according to its component and three subsamples drawn from each at the

end of every sampling period. Subsamples were dried at 80 °C until a constant dry

weight was reached. Stem samples were divided into 1 m sections and their fresh

weight volume calculated according to Newton’s equation:

202

(Equation 7) 𝑉 =𝐿(𝐴𝑏+4𝐴𝑚+𝐴𝑠)

6

Where b is diameter at the base, m is diameter at the middle and s is diameter at the

endpoint of the stem section.

2.2 Data Exploration

All analyses were carried out using the statistical program R (3.2.2, Mac OS X).

Variables were visually assessed for relationships and outliers using scatter plots.

Spearman’s correlation tests were used to determine the level of correlation between

variables as Spearman’s is a ranked approach and is therefore not sensitive to data

transformations.

The present study hypothesises that the growth presented by the species in this

study is multiplicative and therefore the best-fit models will be log-linear. This

assumption must be tested because the decision to log transform the data should

not be arbitrary (Sileshi, 2014). Three modelling approaches were chosen to test the

error distributions and fit for the formulation of allometric equations, linear, log linear

and quadratic. The residual plots and error distributions were reviewed for these

three modelling approaches. In all models both normal and kernel density distribution

models were used to plot the residuals against the predicted values to test for

normality and homogeneity. Outliers were detected by using Cook’s Distance.

203

2.3 Model Comparison and Selection

Linear, log-linear and quadratic equations were developed and compared using

Furnival’s Index for the selection of the most appropriate model (Furnival, 1961):

(Equation 8) F. I. = rMSE

GM(d(𝑓(Y))/dY)

Where rMSE is the root mean square error of the fitted equations, f(Y) is the

transformation function on the dependent variable d(f(Y))/dY is the derivative of the

transformation fitted to Y, and GM denotes the geometric mean of 1/Y (Kershaw et

al., 2016).

Furnival’s Index (F.I.) can be compared between models with transformed and

untransformed variables. A lower F.I. results from a lower error associated with the

model and therefore an improved fit. The R2 or coefficient of determination is

typically used to compare models and is interpreted as a percentage explanation of

the model fit of the data where a higher R2 represents an improved fit. The R2 is

derived from the sum of squares residual divided by the sum of squares total (the R2

can also be ‘adjusted’ for increased model parameters). However, it is inappropriate

to compare the R2 of a model with the R2 of a model that uses a transformed

predictor variable.

2.4 Model Validation

The models were cross-validated by the repeated cross validation technique. The

dataset was divided into ‘k groups’ also known as ‘k-folds’ where the training data

was defined as k = 10, which results in a manageable bias-variance trade-off, and

204

the test data was k-1 (James et al., 2013; Arlot and Celisse, 2010). This can also be

described as a random 10% subset of the data being used to test 90% of the data.

This process was repeated three times. The mean average of the root of the residual

mean squared error for three repeats is reported (rMSE). The data were

subsequently combined and the full dataset was used to generate the allometric

models.

2.5 Allometric Models

The following linear equation was used to test allometric models:

(Equation 9) y = a + b * X

Where a is the intercept coefficient, b is slope coefficient, y is a biomass component

either total biomass, trunk biomass, branch biomass, leaf biomass and X is Root

Collar Diameter (RCD, mm), Diameter at Breast Height (DBH, mm), Stump Diameter

(SD, mm), Crown Volume (CV, cm3), Crown Diameter (CD, cm) or as a function of

both RCD and Height (cm). These parameters were also modeled in the quadratic

form, where c is an additional model coefficient:

(Equation 10) y = aX2 + bX + c

Kitteredge (1944) found that Huxley's (1932) equation could be used to derive the

biomass of several species of spruce and pine trees. This is a common allometric

equation for the estimation of tree biomass:

(Equation 11) y = aXb

205

This can also be used in the logarithmic form to produce a log-linear model. The use

of the logarithmic form requires back transformation of the dependent variable. This

paper calculates the error associated with log transformation as lambda (λ) and

associated standard deviation according to the equation proposed by Marklund

(1987), where n is sample size:

(Equation 12) 𝜆 = ∑ 𝑌𝑖

𝑛𝑖=1

∑ 𝑒ln ��𝑖𝑛𝑖=1

The following log-linear allometric model was tested:

(Equation 13) ln(y) = a + b * lnX

Where ln(y) is the natural logarithmic transformation of the biomass components or

expansion factors listed above, ln(X) is the log transformation of the measurements

listed above. Where multiple predictors were present in the model the adjusted

coefficient of determination (R2) was reported.

2.6 Calculation of Carbon Stocks

The C content was estimated by multiplying the individual tree biomass (Mg ha-1) by

the IPCC default C value for trees in tropical forests of 0.47 (Eggleston et al., 2006).

The value for the C sequestration of each tree was then multiplied by the planting

density (2,667 trees ha-1). The average and standard deviation was reported. To

calculate mean annual increment (MAI) of biomass produced and C sequestered the

results of the total biomass at the 24-month destructive harvest and the pollard of

biomass at 18 months over 2 m were combined and divided by the number of years

of data. No estimate was made of the carbon present in belowground biomass as

this data was not collected in the present study due to resource constraints.

206

3. Results

3.1 Measurements of Tree Biomass

Subsamples of all biomass components showed strong relationships between wet

and dry weight (Appendix 1). The contribution of foliar biomass to total biomass was

3% in E. berteroana (Table 7), 15% in I. edulis (Table 8) and 2.4% in L. leucocephala

(Table 9). The sample size (n) of 68 for DBH of E. berteroana in Table 7 is due to not

all trees exceeding the 1.3 m at which the measurement was taken. I. edulis

produced over 60% more total biomass and 660% more foliar biomass than either E.

berteroana or L. leucocephala. I. edulis had a 223% greater crown volume than that

of E. berteroana or and a 163% greater crown volume than that of L. leucocephala.

I. edulis had 111% more branch biomass than E. berteroana and 220% more branch

biomass than L. leucocephala. Average trunk biomass of all three species were

similar 1.3 ± 0.8 kg for I. edulis, 1.0 ± 0.8 kg for E. berteroana and 1.0 ± 0.5 kg for L.

leucocephala. 6 months post pollard to 2 m, L. leucocephala had grown by 1 m to a

height of 298 cm, which represented an increase on the 242 cm height of I. edulis

and 219 cm height of E. berteroana. L. leucocephala was the species with the lowest

average diameter measurements for RCD (44 mm, compared to 67 mm of I. edulis

and 63 mm of E. berteroana) SD (35 mm, compared to 50 mm and 48 mm of I.

edulis and E. berteroana respectively) and had a DBH (22 mm) similar to that of E.

berteroana (21 mm) but lower than that of I. edulis (34 mm).

207

Table 7. Descriptive statistics of biomass parameters of individual harvested Erythrina berteroana used in constructing allometric equations, including sample size (n), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m) and diameter at breast height (DBH, 1.3 m). Mean, standard deviation (SD), minimum (min), maximum (max), lower 25th percentile (25th perc), upper 75th percentile (75th perc), as well as coefficients of skewness and kurtosis are presented.

Variable n Mean SD Min Max 25th perc 75th perc Skewness Kurtosis

Total Biomass (g) 73 1316.3 1005.1 36.4 4517 505.3 1935 1.1 0.8

Trunk Biomass (g) 73 1015.4 792.9 35.2 4087 417.9 1378 1.3 2.1

Branch Biomass (g) 72 262.6 315.8 0.6 1942 61.3 360.7 2.7 9.8

Leaf Biomass (g) 73 42 51.5 0.4 231.3 9.2 67 1.9 3.5

RCD (mm) 73 62.6 21.5 21.6 123.1 50.7 74.8 0.4 0.4

SD (mm) 73 48.2 18.3 17 105.4 33.6 61.6 0.5 0.1

DBH (mm) 68 21.4 7.6 8.4 47.7 15.8 25.4 0.8 1.0

Height (cm) 73 218.8 53.2 72 200 218.8 314.0 -0.74 0.1

Crown Volume (cm3) 73 939219.3 1028901 1341.8 4125900 156800 1390000 1.5 1.5

Crown Diameter (cm) 73 88.5 42.9 10.3 191.7 53.3 122.3 0.4 -0.7

208

Table 8. Descriptive statistics of biomass parameters of individual harvested Inga edulis used in constructing allometric equations, including sample size (n), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m) and diameter at breast height (DBH, 1.3 m). Mean, standard deviation (SD), minimum (min), maximum (max), lower 25th percentile (25th perc), upper 75th percentile (75th perc), as well as coefficients of skewness and kurtosis are presented.


Total Biomass (g) 78 2191.2 1402.6 177 6692 1269 2943 0.9 0.3

Trunk Biomass (g) 78 1327 804.7 162.5 4229 721.5 1795 1.0 1.3

Branch Biomass (g) 78 554.6 579.2 0.3 2190 71.4 835.8 1.1 0.4

Leaf Biomass (g) 78 323.1 332.7 1.6 1290 66.7 448.4 1.3 1.1

RCD (mm) 78 65.6 22.5 12.8 127.2 52.4 80.8 0.1 -0.3

SD (mm) 78 50.3 16.4 16.1 87.2 39.2 60.3 0.1 -0.5

DBH (mm) 78 34.4 14.5 12.9 89.7 25.5 39.6 1.6 3.3

Height (cm) 78 242.1 45.9 108 358 220.0 262 -0.3 2.1

Crown Volume (cm3) 78 3039987 3226258 148100 16300000 1013000 3855000 2.3 5.0

Crown Diameter (cm) 78 149.3 64.8 52.3 485.7 107.3 179.5 1.8 6.7

209

Table 9. Descriptive statistics of biomass parameters of individual harvested Leucaena leucocephala used in constructing allometric equations, including sample size (n), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m) and diameter at breast height (DBH, 1.3 m). Mean, standard deviation (SD), minimum (min), maximum (max), lower 25th percentile (25th perc), upper 75th percentile (75th perc), as well as coefficients of skewness and kurtosis are presented.


Total Biomass (g) 63 1259.5 670.1 178 2941.9 721.3 1665 0.39 -0.5

Trunk Biomass (g) 63 1056.5 546.8 174.3 2661 643.6 1378 0.44 -0.2

Branch Biomass (g) 63 172.6 215.9 0.4 1300 24.8 215.4 2.7 9.9

Leaf Biomass (g) 63 30.4 36.4 0.4 197.3 5.1 38.7 2.09 5.5

RCD (mm) 63 44 12.8 20.2 2661.3 643.6 1378 0.13 -0.5

SD (mm) 63 35 10.8 9.5 61.3 27.3 42.8 -0.1 -0.4

DBH (mm) 63 22.2 8.6 5.6 43.5 16.3 28.9 0.45 -0.5

Height (cm) 63 293.7 63.7 144 420 250 343 0.1 -0.8

Crown Volume (cm3) 63 1158571 1349790 15400 6355000 186100 1798000 1.65 2.6

Crown Diameter (cm) 63 97.8 45.5 30 220 63.5 120 0.72 -0.1

210

3.2 Data Exploration

When Spearman’s correlations were compared between biomass parameters and

tree measurements for E. berteroana, the total biomass showed the strongest

correlation with the measurement of crown volume (0.75), the trunk biomass equally

by RCD (0.71) and SD (0.71), the branch biomass by crown diameter (0.83) and the

leaf biomass by crown diameter (0.77) (Table 10). RCD and height did not have

strong correlation (0.39). For I. edulis, RCD was the strongest predictor of total

biomass (0.82), trunk biomass (0.81), and branch biomass (0.66). Crown volume

was found to have the strongest correlation with leaf biomass (0.68) (Table 11).

Height was found to have an association (0.57) with total biomass and showed a

weak correlation (0.45) with RCD. For L. leucocephala RCD was the strongest

predictor of total biomass (0.87) and trunk biomass (0.91). Branch biomass was

equally predicted by crown volume (0.67) and crown diameter (0.67). Crown

diameter was also the most correlated measurement with leaf biomass (0.78) (Table

12). In L. leucocephala, RCD had a weak correlation with height (0.45).

211

Table 10. Spearman’s correlation of E. berteroana biomass components. Total biomass (Total), trunk biomass (Trunk), branch biomass (Branch), total foliar biomass (Leaf), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m), diameter at breast height (DBH, 1.3 m), height, crown volume (CVol) and crown diameter (CD).

Total Trunk Branch Leaf RCD SD DBH Height CVol CD

Total 1 0.97 0.7 0.78 0.73 0.68 0.68 0.67 0.75 0.73

Trunk 0.97 1 0.55 0.72 0.71 0.71 0.68 0.67 0.69 0.64

Branch 0.7 0.55 1 0.7 0.63 0.43 0.51 0.43 0.72 0.83

Leaf 0.78 0.72 0.7 1 0.55 0.53 0.58 0.64 0.71 0.77

RCD 0.73 0.71 0.63 0.55 1 0.7 0.6 0.39 0.55 0.59

SD 0.68 0.71 0.43 0.53 0.7 1 0.63 0.41 0.46 0.54

DBH 0.68 0.68 0.51 0.58 0.6 0.63 1 0.55 0.56 0.58

Height 0.67 0.67 0.43 0.64 0.39 0.41 0.55 1 0.81 0.56

CVol 0.75 0.69 0.72 0.71 0.55 0.46 0.56 0.81 1 0.85

CD 0.73 0.64 0.83 0.77 0.59 0.54 0.58 0.56 0.85 1

Table 11. Spearman’s correlation of I. edulis biomass components. Total biomass (Total), trunk biomass (Trunk), branch biomass (Branch), total foliar biomass (Leaf), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m), diameter at breast height (DBH, 1.3 m), height, crown volume (CVol) and crown diameter (CD).


Total 1 0.88 0.86 0.8 0.82 0.7 0.49 0.49 0.66 0.67

Trunk 0.88 1 0.6 0.58 0.81 0.8 0.57 0.44 0.58 0.58

Branch 0.86 0.6 1 0.71 0.66 0.55 0.38 0.38 0.57 0.63

Leaf 0.8 0.58 0.71 1 0.65 0.5 0.32 0.53 0.68 0.67

RCD 0.82 0.81 0.66 0.65 1 0.78 0.62 0.47 0.67 0.66

SD 0.7 0.8 0.55 0.5 0.78 1 0.69 0.4 0.55 0.54

DBH 0.49 0.57 0.38 0.32 0.62 0.69 1 0.24 0.36 0.36

Height 0.49 0.44 0.38 0.53 0.47 0.4 0.24 1 0.63 0.4

CVol 0.66 0.58 0.57 0.68 0.67 0.55 0.36 0.63 1 0.88

CD 0.67 0.58 0.63 0.67 0.66 0.54 0.36 0.4 0.88 1

212

Table 12. Spearman’s correlation of L. leucocephala biomass components. Total biomass (Total), trunk biomass (Trunk), branch biomass (Branch), total foliar biomass (Leaf), root collar diameter (RCD, 0.05 m), stump diameter (SD, 0.3 m), diameter at breast height (DBH, 1.3 m) height, crown volume (CVol) and crown diameter (CD).


Total 1 0.94 0.76 0.82 0.87 0.77 0.53 0.57 0.69 0.71

Trunk 0.94 1 0.56 0.75 0.91 0.84 0.62 0.55 0.61 0.62

Branch 0.76 0.56 1 0.8 0.53 0.49 0.19 0.59 0.67 0.67

Leaf 0.82 0.75 0.8 1 0.65 0.54 0.33 0.73 0.75 0.78

RCD 0.87 0.91 0.53 0.65 1 0.87 0.64 0.45 0.54 0.55

SD 0.77 0.84 0.49 0.54 0.87 1 0.67 0.45 0.54 0.55

DBH 0.53 0.62 0.19 0.33 0.64 0.67 1 0.37 0.28 0.26

Height 0.57 0.55 0.59 0.73 0.45 0.45 0.37 1 0.84 0.75

CVol 0.69 0.61 0.67 0.75 0.54 0.54 0.28 0.84 1 0.92

CD 0.71 0.62 0.67 0.78 0.55 0.55 0.26 0.75 0.92 1

Appendix 2-4 show the relationship between total biomass and RCD for I. edulis

when plotted using linear, quadratic and log-linear models. The linear model fails to

fit the data points, which appear to show an exponential relationship and

multiplicative error (Appendix 2). The quadratic model fails to improve the model fit

due to being misled by amplified outliers (Appendix 3). The log-linear model appears

to be a better fit for the data and does not show any serious errors in the residual

plots (Appendix 4).

213

3.3 Model Selection

When the Furnival’s Index of all modelled approaches was compared, it was

revealed that the log-linear models are always superior to the quadratic or linear

models for all biomass parameters in all species in this study (Appendix 5 - 22). The

models with the lowest Furnival’s Index (F.I.) for each biomass parameter are

presented in Table 13. The best-fit models are plotted in Figures 48 - 51. It was only

possible to graphs with the data points for linear models (Fig 49 and Fig 50). For

species where quadratic models offered best fit the regression line has been plotted

without the data points (Fig 48 and Fig 51).

In this study, the species differ in which measurements can be used to best predict

components of biomass. The inclusion of height with RCD as a multiple predictor

resulted in the lowest F.I. for total and trunk biomass in E. berteroana and L.

leucocephala. For I. edulis the inclusion of height did not improve the prediction of

biomass and the single predictor of RCD had the lowest F.I. for all of the biomass

measurements. Predictors for the total biomass of L. leucocephala (R2=0.86) were

more effective then predictors of total biomass of either E. berteroana (R2=0.77) or I.

edulis (R2=0.70). The hypothesis that measurements of the crown would be the best

predictor of total foliar biomass was upheld for E. berteroana by the measurement of

crown diameter (R2=0.58) but rejected for L. leucocephala (RCD and height as

combined predictors with lowest F.I.) and I. edulis (RCD as single predictor with

lowest F.I.).

214

When the root mean square error (rMSE) was compared within species between

biomass parameters, the rMSE in I. edulis branch biomass was increased 282%

when compared to the rMSE present in the prediction of trunk biomass and leaf

biomass showed a 205% increase in rMSE when compared to trunk biomass. For E.

berteroana the branch biomass rMSE was increased 42% compared to trunk

biomass, and leaf biomass rMSE increased 96% compared to trunk biomass. For L.

leucocephala the branch biomass had a 495% increased rMSE than trunk biomass,

the leaf biomass rMSE showed a 320% increase when compared with the rMSE of

trunk biomass.

215

Table 13. Log-linear allometric equations demonstrated the lowest Furnival’s Index (F.I.) against linear or quadratic models for biomass components of 0.5 x 7.5 m alley rows of Erythrina berteroana, Inga edulis and Leucaena leucocephala 6 months post pollarding to a height of 2 m. Where ln(x) is the log transformation of the following measurements used as predictors of biomass: Root collar diameter (RCD mm, 0.05 m), height (cm) and crown diameter (CD, cm). ln(y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). Where b0 is the intercept coefficient, b1 and b2 are the slope coefficients, all of which are reported with standard error (S.E.) and significance (P). Lambda (λ) and standard deviation (S.D.) as back transformation according to Marklund (1987), root mean square error of model (Model rMSE) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV rMSE).

Species ln(y)

(Biomass) ln(x) b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P R2 λ S.D. F.I. Model rMSE

CV rMSE

E. berteroana Total RCD + Height -6.41 (1.02) <0.001 1.53 (0.2) <0.001 1.31 (0.25) <0.001 0.77 1.12 0.83 523.2 0.47 0.49

Trunk RCD + Height -6.91 (1.03) <0.001 1.38 (1.38) <0.001 1.47 (0.25) <0.001 0.76 1.13 0.94 409.5 0.48 0.50

Branch CD -3.99 (0.65) <0.001 2.04 (0.15) <0.001 0.73 0.25 0.19 104.8 0.68 0.74

Leaf CD -7.06 (1.07) <0.001 2.28 (0.24) <0.001 0.58 1.49 1.36 20.1 0.94 0.93

I. edulis Total RCD 0.56 (0.51) 0.28 1.68 (0.12) <0.001 0.70 1.07 0.54 759.2 0.44 0.46

Trunk RCD 0.95 (0.43) 0.03 1.47 (0.10) <0.001 0.71 1.07 0.56 409.6 0.38 0.39

Branch RCD -8.89 (1.67) <0.001 3.45 (0.40) <0.001 0.48 2.37 4.64 287.9 1.45 1.51

Leaf RCD -6.47 (1.33) <0.001 2.78 (0.32) <0.001 0.48 2.19 6.52 169.1 1.16 1.20

L. leucocephala Total RCD + Height -3.04 (0.79) <0.001 1.69 (0.12) <0.001 0.66 (0.16) <0.001 0.86 0.98 0.26 64.33 0.24 0.25

Trunk RCD + Height -2.26 (0.67) 0.001 1.71 ( 0.1) <0.001 0.47 (0.13) <0.001 0.89 1.03 0.21 39.02 0.20 0.21

Branch CD -7.16 (1.43) <0.001 2.55 (0.32) <0.001 0.51 1.86 2.33 83.7 1.19 1.23

Leaf RCD + Height -21.53 (2.75) <0.001 1.92 (0.40) <0.001 3.01 (0.55) <0.001 0.61 1.34 1.05 10.62 0.84 0.89

216

Figure 48. The log transformation of x (root collar diameter (mm) + height (cm)) derived using a linear model for the log transformation of 1) y as total biomass (g, R2 = 0.77, P <0.01) and 2) y as trunk biomass (g, R2 = 0.76, P<0.01) for Erythrina berteroana following the equations presented in Table 13.

Figure 49. The log transformation of crown diameter (cm) derived using a linear model for the log transformation of branch biomass (g, R2 = 0.73, P<0.01) and leaf

0 1 2 3 4 5 6 7

−6

−4

−2

02

4

ln(x)

ln(y

)Biomass

Total

Trunk

3.5 4.0 4.5 5.0

23

45

67

ln(CROWN DIAMETER)

ln(B

IOM

AS

S)

Biomass

Branch

Leaf

217

biomass (g, R2 = 0.58, P<0.01) for Erythrina berteroana following the equations presented in Table 13.

Figure 50. The log transformation of root collar diameter (mm) derived using a linear model for the log transformation of total biomass (g, R2 = 0.70, P<0.01), trunk biomass (g, R2 = 0.71, P<0.01), branch biomass (g, R2 = 0.48, P<0.01) and leaf biomass (g, R2 = 0.48, P<0.01) for Inga edulis following the equations presented in Table 13.

Figure 51. The log transformation of x where 1) x is root collar diameter (mm) + height (cm) derived using a linear model for the log transformation of total biomass

2.5 3.0 3.5 4.0 4.5

67

8

ln(RCD)

ln(B

IOM

AS

S)

Biomass

Total

Trunk

Branch

Leaf

0 1 2 3 4

−2

02

46

8

ln(x)

ln(y

)

Biomass

Total

Trunk

Branch

Leaf

218

(g, R2 = 0.86, P<0.01), trunk biomass (g, R2 = 0.89, P<0.01) and leaf biomass (g, R2 = 0.61, P<0.01) and 2) x is crown diameter (cm) derived using a linear model for the log transformation of branch biomass (g, R2 = 0.51, P<0.01) for Leucaena leucocephala following the equations presented in Table 13.

Total above ground biomass fraction and C stock after 24 months is presented in

Table 14. When the biomass from the destructive harvest at 24 months was

combined with the biomass from the pollard exercise at 18 months to represent total

accumulated biomass over the course of the 24 months of this study, I. edulis had

accumulated the most biomass and therefore more C (3.4 Mg C ha-1) than E.

berteroana (2.2 Mg C ha-1) and L. leucocephala (2.0 Mg C ha-1) (Table 14).

Table 14. Total above ground biomass fraction and carbon (C) stock (Mg ha-1) after 24 months, including previously removed biomass > 2 m at 18 months (Chapter III) of Erythrina berteroana, Inga edulis and Leucaena leucocephala when 2667 trees ha-1 (0.5 x 7.5 m), using the IPCC default C fraction for tropical forests of 0.47. Where n is sample size and MAI is Mean Annual Increment (Mg ha-1 year-1) (Eggleston et al., 2006).

Species n Tree Biomass

Mg ha-1

MAI Tree Biomass C Mg ha-1 MAI C

E. berterona 106 4.4 2.2 2.1 1.0

I. edulis 112 7.2 3.6 3.4 1.7

L. leucocephala 96 4.3 2.2 2.0 1.0

4. Discussion

The results of this study show that Erythrina berteroana, Inga edulis and Leucaena

leucocephala can produce substantial amounts of biomass after 24 months and

show good potential for the rapid sequestration of C in above ground biomass. This

study shows a considerable increase in C sequestration via the addition of these

trees when the tree biomass is assumed to be additional to the business-as-usual

219

carbon sequestration of monoculture pastures in the region. This assumption is

based on the trees being managed in such a way so as to have negligible impact on

grass production (see Chapter III). Further work may be carried out that examines

the interaction of C sequestration by grassland or silvopasture based on the

interaction of the tree and grass elements in such systems.

This paper alludes to an assumption in the literature that the ratio of dry weight of tree

biomass to carbon content averages around 0.47-0.49. The estimates of C in this study

were drawn from the IPCC report that reports the average value for tropical forests

(Eggleston et al., 2006). This assumption was challenged by the results of a study in

Panama that showed the ratio of biomass dry weight taken from tree cores to carbon to

vary from a proportion of 0.42–0.52 when 59 rainforest trees were sampled (Martin

and Thomas, 2011). The same study did not find any correlation between C stocks with

relative growth rate or wood density. The reliability of the estimates of C

sequestration in the present study could be improved by specific analysis of C

content in the plant material between species and biomass components and the

inclusion of sampling that considers the production of below ground biomass to

enable the C sequestered belowground to be estimated.

Site level differences such as planting arrangement, water availability, nutrient

availability, soil compaction, temperature, solar irradiation and species associations

may also limit the transferability of one study’s findings for a species biomass

220

accumulation to other regions (Sharma and Ambasht, 1991; Prasad et al., 2011).

Montagnini and Nair (2004) found that agroforestry practiced by small holders in the

tropics had a C sequestration potential of 1.5 to 3.5 Mg C ha-1 year-1 (Montagnini and

Nair, 2004). Less is known about the effects of pollarding or biomass removal on C

sequestration potential of these systems. A fodder bank in Mali, west Africa of Gliricidia

sepium, Pterocarpus lucens and P. erinaceus at 7.5 years was found to sequester

0.29 Mg C ha-1 year-1 (Takimoto et al., 2008).

The findings of the present study can be viewed in light of what other authors have

found to be true of the biomass production in different planting arrangements and

site conditions. Szott et al. (1994) planted I. edulis in Yurimaguas, Peru, in an

arrangement of 2 x 2 m and used allometric equations based on the measurement of

diameter of the trunk 0.1 m above soil (hereafter D10) for 10 harvested individuals

with a D10 between 1 and 5.5 cm to estimate the total tree biomass. They estimated

that at 17 months I. edulis had an average of 10 Mg ha-1 of biomass and by 29

months this had increased to 14.24 Mg ha-1. Shimamoto et al. (2014) measured the

biomass production of 12-year-old naturally regenerating I. edulis in the Parana

State in Southern Brazil. They sampled 9 trees and found annual accumulation of

20.73 kg (± 12.09) to 24.48 kg (± 19.71) biomass tree-1. Prasad (2011) found that in

southern India, after 51 months, a spacing of 0.8 x 5 m for L. leucocephala resulted

in a total biomass production of 47 kg tree-1 (2,500 trees ha-1) whereas 1 x 1 spacing

resulted in 38 kg tree-1, but that the per hectare biomass was higher in the 1 x 1 m

spacing (10,000 trees ha-1). López et al. (2008) planted L. leucocephala in

221

southwestern Spain at a density of 9,500 trees ha-1 and found a total biomass

production for three provenances of 23.4 Mg ha-1 (±2.9), 30.9 Mg ha-1 (±6.9) and

25.6 Mg ha-1 (±13.1) at 24 months. E. berteroana in a live fence arrangement in

Costa Rica was reported to produce 19.4 Mg ha-1 after 24 months, but the planting

arrangement was not specified (Pezo et al., 1990). Pennington (1998) planted I.

edulis at 3 x 3 m spacing at trial sites in on the Pacific Coast and the Amazon of

Ecuador. I. edulis consistently demonstrated an increased height and diameter

growth rate when compared to four other species of Inga, and produced 2.5 Mg ha-

1year-1 of woody biomass.

In this study, I. edulis had the highest rates of overall biomass production. While the

trunk biomass was relatively similar for all three species, I. edulis produced over six

times (660%) more foliar biomass than E. berteroana or L. leucocephala. Szott et al.

(1994) in Yurimaguas, Peru, found that for small I. edulis trees (D10 = 2.5 cm), trunk

accounted for 33% of total biomass, branches for 21% and leaves for 45%. In larger

trees (D10= 10 cm) the biomass was 51% trunk, 32% branches and 17% leaves. In

the present study, the contribution of biomass components to total biomass was 60%

trunk, 25% branches and 15% leaf. The differences between the present study with

that of Szott et al. (1994) may be as a result of the pollard > 2 m which may have

resulted in less trunk biomass being removed than branch and leaf components.

As expected, the log-linear models for all biomass parameters in all species had the

lowest Furnival’s Index and therefore can be considered the best models for the

222

prediction of biomass. The inclusion of multiple predictors in the model (RCD +

height) did not improve the model fit for I. edulis, but did improve the model fit for the

total and trunk biomass of E. berteroana and L. leucocephala. Lojka et al. (2005)

found that power relationships on untransformed variables were the best fit for the

prediction of I. edulis biomass. They destructively sampled 35 trees and measured

leaf litter, diameter of trunk at D10, age of trees and height. They found the highest

prediction of wood biomass (kg) by the measurement of diameter (y = 0.0466x2.3713;

R2= 0.92). D10 was also the strongest predictor of leaf biomass (kg) (y =

0.1564x1.1817; R2=0.79). Szott et al. (1994) developed allometric equations for I.

edulis based on the destructive harvest of 27 trees with a D10 of 1 – 5.5 cm and

develop allometric equations for the stem biomass (kg) (y = 0.089*(D10)2; R2=0.77),

branch biomass (kg) (y = 0.057*(D10)2; R2=0.96) and leaf biomass (kg) (y =

0.3003*(D10); R2=0.85). In the present study, the R2 of best performing models

shows poorer performance than the models present in the literature. It is likely that

the lower R2 in this study is a consequence of the difference in management

approach in this study which included pollarding all trees over 2 m at 18 months.

Managed forest plots, where trees undergo management interventions such as

coppicing or pollarding, could be expected to demonstrate different allometric

relationships when compared with trees that are free from such interference.

However, it is the effect of the management on the trees that is precisely of interest,

as the management will effect the production of key ecosystem services from the

systems. The pipe model theory asserts that the diameter of a tree is proportional to

223

the water requirements of the tree canopy and that the canopy size is determined by

the transpirational rate (Shinozaki et al., 1964). The pipe model is one of the theories

from which allometric equations are believed to derive their power.

Nygren et al. (1993) state that repeated pruning of the tree alters the development of

the trunk diameter as it changes the leaf area to trunk ratio. They propose the

formation of equations based on measurements leaf area and leaf mass related to

the diameter of branches that form following pruning. Their approach was successful

for four clones of E. berteroana, the average of which produced an R2 of 0.89 with

the equation: leaf biomass = 0.2209*CS, where CS is branch cross sectional area

measured in mm. The R2 for the leaf biomass of E. berteroana in the present study

was lower (R2= 0.58) and future studies could consider the methodological approach

of Nygren et al. (1993), which may be superior for the prediction of foliar biomass.

Lott et al. (2000) found that the approach followed by Nygren (1993) produced

reliable estimates of canopy leaf area and biomass for Grevilla robusta agroforestry

systems, but proved unsuitable for routine measurements due to the time consuming

nature of this approach. Lott et al. (2000) suggest that the measurement of trunk

cross sectional area immediately below the first branch was suitable for the

prediction of leaf area and canopy biomass in Grevilla robusta managed for poles.

There is evidence that regrowth of trees following pruning and pollarding may be in

part dependent on carbohydrate reserves, which can be depleted by coppicing or

pollarding regimes and in some cases there may also be effects of seasonality (Kays

and Canham, 1991; Raitio et al., 1994).

224

In temperate regions where plants may have well defined growing seasons, the

timing of biomass management (via, e.g. directly via pruning/pollarding/coppicing or

browsing) has been shown to have an effect on the production of biomass (Douglas,

1996; Opping, 2002). Burner (2006) studied the effect of harvest date on Robinia

pseudoacacia and found temporal variation in allometric equations post-pollard, as

foliar mass accumulated more quickly in some months than in others. Therefore,

they developed month specific equations. Seasonality would also have had a strong

effect on the trees in this study, and month specific equations may have been a more

appropriate approach but was not followed in this study. The present study could be

improved by characterising the effects of both the frequency of pollarding and also

how the timing of pollarding relative to how tropical seasonality effects foliar

production in an attempt to increase understanding of how to improve management

and production of these species.

It is a convention in forestry that DBH is the most commonly used measurement for

the development of allometric equations because it can be easily measured with high

accuracy (Kershaw et al., 2016). For all species in this experiment, when the three

heights of stem diameter measurements were compared, the RCD had the highest fit

to total biomass, followed by SD. DBH demonstrated the poorest fit. This suggests

that the RCD is a valuable parameter that could be included in future studies that

attempt to develop allometric equations for young trees in agroforestry systems. The

stem measurement of RCD was the best single predictor of total biomass for I. edulis

225

and L. leucocephala. RCD has been used when a high proportion of sample trees

are below the 1.3 m at which DBH is traditionally measured (Dutcă et al., 2018).

However, equations developed from RCD may be less accurate for trees with

buttress roots or stilts. C. pentandra was included in this experiment, and is a

species that develops buttress roots. As this tree grows the reliability of the

measurement of RCD for total biomass is likely to change. Further work would be

required to understand more about this relationship, and how RCD changes in

reliability between size classes across species. Rubilar et al. (2010) measured both

RCD and DBH for Pinus radiata and found similar results for both variables. Guedes

(2018) found that measurements of RCD were interchangeable with DBH in the

lowland miombo woodlands of Mozambique. This may be useful to estimate the

biomass of trees that have been removed from their stumps. Jagodziński et al.

(2018) developed biomass equations for Pinus sylvestris, a species they observed to

be prone to root collar swelling. Therefore, they used root collar diameter only when

trees were shorter than 0.5 m. Dutcă (2018) used root collar diameter as a

parameter on young Norway spruce trees with a maximum height of 5.5 m. Kuyah et

al. (2013) found that D10 was better at predicting below ground biomass (R2 = 0.91)

than above ground biomass (R2 = 0.88), but overestimated the biomass of most

components by 20%. They discussed that these errors were, in part, due to

uncertainties in the location of the measurement of D10. MacCracken and

Ballenberghe (1993) tested multiple models for moose browse on the Copper River

Delta, Alaska and found that mass of leaf and twig production was best predicted by

RCD.

226

Authors have suggested that the inclusion of additional variables, such as height, will

increase the predictive powers of models (Chave et al., 2005). When the additional

variable height was included in the RCD model this became the strongest predictive

model for the total biomass of L. leucocephala and E. berteroana. The inclusion of

height did not improve the model fit for I. edulis. Height and stem diameter

measurements are frequently too highly correlated for both to be included in the

same model, due to the issue of multicollinearity (Balehegn et al., 2012; Sileshi,

2014). To avoid multicollinearity, some authors have tested both height and diameter

independently and included whichever variable results in improved fit or is easiest to

measure (Pajtek, 2008). In some cases they have combined the two variables into a

single variable to improve model fit, such as the D2H in Dutcă et al. (2018) and the

DBH2H in Rubilar et al. (2010). It is possible that the inclusion of height did not

improve the model fit for I. edulis as a consequence of the pollard at 2 m. I. edulis

may have responded to the increased availability of light following pollarding by

investing more resources into the growth of lighter foliar biomass, as opposed to the

production of heavier woody biomass. Of all species, I. edulis demonstrated the

highest foliar biomass production relative to other biomass components. The foliar

biomass, which weighs less than the woody biomass, would not be represented by

the height of the tree. Therefore, it may be the case that tree species that have a

high proportion of foliar to total biomass have a greater need for species specific

models, particularly when the trees are managed by pollarding or removal of the

foliar biomass.

227

Stem-based models (RCD, SD, DBH) predicted with the greatest accuracy the stem

or total above ground biomass, and displayed poorer performance when estimating

the branch or foliar biomass. This may occur when the stem biomass, which is likely

to be highly correlated with the stem diameter, is the largest constituent of the total

biomass. That it was not the case in E. berteroana may be due to the fact that E.

berteroana had been grown from 1 m stakes, and this may have resulted in a

significantly different growth form. Manipulation of a tree via pollarding will alter both

the total biomass and the relative contribution to the total biomass of components.

Yet, pruning and pollarding interventions may not impact on measurements of stem

diameter (Kuyah et al., 2013).

Crown diameter and crown volume were strongly correlated with one another and

both displayed the greatest fit as a single predictors of the total biomass of E.

berteroana. It was hypothesised that measurements of the crown of the tree would

be the best predictor of total foliar biomass. For E. berteroana, crown volume and

crown diameter both showed marginally improved fit over the stem volume

measurements for the prediction of total biomass. This could be due to the way this

tree produces foliage, it does not rely on heavy proliferation of bifurcate branching

systems, and instead the leaves are generated from principal branches and directly

from the trunk (Nygren et al., 1993). As the measurement of the crown was taken

where 5% of the foliage began, and the foliage was observed to occur along the

entire trunk of E. berteroana, it is possible that the cylindrical measurement of crown

228

volume was able to capture the majority of the woody biomass in unison with the

foliage. This result highlights the need to consider the morphology of the tree when

choosing which elements of the tree should be measured to develop allometric

models. Where the tree has a high proportion of woody to foliar biomass, a

measurement of the trunk combined with some reference to the branching

architecture may be appropriate. Further work is needed to develop models that

consider different branching architectures of individual species.

Future studies seeking to develop transferrable estimations of biomass can include a

wide range of trees of different sizes, positioning within the stand and age (Lott et al.,

2000). Allometric equations may vary due to site specific agro ecological, climatic

and edaphic conditions (Yuen et al., 2016), altitude (Cienciala et al., 2006),

provenance (Poorter et al., 2012), input of fertiliser and the use of clones (Heinsoo et

al., 2002), spacing and stand density (Telenius and Verwijst, 1995), the composition

of mixed tree plantations (Laclau et al., 2008), the frequency and timing of biomass

removal (Oppong et al., 2002), and the age of the trees in the sample (Peichl and

Arain, 2007). Practitioners must be aware that these variables may limit the

transferability of allometric equations between sites, and therefore a process of

verification is suggested ( (Eggleston et al., 2006; Yuen et al., 2016).

The total and stem biomass of these species can be reliably predicted by the

presented equations, even when pollarding has manipulated the form of the tree.

However, the present study does not account for repeat pollarding or the temporal

229

effects of season on the trees. The log-linear transformation improved the predictive

power of allometric equations for all biomass components of all species. The rapid

accumulation of biomass of these species implies that the integration of trees into

cattle pastures in the Amazon could represent a major contribution to above ground

C storage. Further work is needed to understand the effects of management on

these species, and the models presented in this study are specific to the species and

conditions in the study area, and care should be taken when extrapolating the

results, which were developed on a homogenous group of small trees.

References

Akaike, H., 1974. A New Look at the Statistical Model Identification.

Springer, New York, NY, pp. 215–222.

https://doi.org/10.1007/978-1-4612-1694-0_16

Arlot, S., Celisse, A., 2010. A survey of cross-validation procedures

for model selection. Statistics Surveys 4, 40–79.

https://doi.org/10.1214/09-SS054

Balehegn, M., Eniang, E.A., Hassen, A., 2012. Estimation of browse

biomass of Ficus thonningii, an indigenous multipurpose fodder tree in

northern Ethiopia. African Journal of Range & Forage Science.

https://doi.org/10.2989/10220119.2012.687071

https://doi.org/10.1007/978-1-4612-1694-0_16

https://doi.org/10.1214/09-SS054

https://doi.org/10.2989/10220119.2012.687071

230

Blum, M., Lövbrand, E., 2019. The return of carbon offsetting? The

discursive legitimation of new market arrangements in the Paris

climate regime. Earth System Governance 100028.

https://doi.org/10.1016/J.ESG.2019.100028

Chave, J., Andalo, C., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus,

D., Fölster, H., Fromard, F., Higuchi, N., Kira, T., Lescure, J.-P.,

Nelson, B.W., Ogawa, H., Puig, H., Riéra, B., Yamakura, T., 2005. Tree

allometry and improved estimation of carbon stocks and balance in

tropical forests. Oecologia 145, 87–99.

https://doi.org/10.1007/s00442-005-0100-x

Chave, J., Réjou-Méchain, M., Búrquez, A., Chidumayo, E., Colgan,

M.S., Delitti, W.B.C., Duque, A., Eid, T., Fearnside, P.M., Goodman,

R.C., Henry, M., Martínez-Yrízar, A., Mugasha, W.A., Muller-Landau,

H.C., Mencuccini, M., Nelson, B.W., Ngomanda, A., Nogueira, E.M.,

Ortiz-Malavassi, E., Pélissier, R., Ploton, P., Ryan, C.M.,

Saldarriaga, J.G., Vieilledent, G., 2014. Improved allometric models

to estimate the aboveground biomass of tropical trees. Global Change

Biology 20, 3177–3190. https://doi.org/10.1111/gcb.12629

Cienciala, E., Černý, M., Tatarinov, F., Apltauer, J., Exnerová, Z.,

https://doi.org/10.1016/J.ESG.2019.100028

https://doi.org/10.1007/s00442-005-0100-x

https://doi.org/10.1111/gcb.12629

231

2006. Biomass functions applicable to Scots pine. Trees - Structure

and Function 20, 483–495. https://doi.org/10.1007/s00468-006-0064-4

Claassen, R., Cattaneo, A., Johansson, R., 2008. Cost-effective design

of agri-environmental payment programs: U.S. experience in theory and

practice. Ecological Economics 65, 737–752.

https://doi.org/10.1016/J.ECOLECON.2007.07.032

Corbera, E., Soberanis, C.G., Brown, K., 2009. Institutional

dimensions of Payments for Ecosystem Services: An analysis of Mexico’s

carbon forestry programme. Ecological Economics 68, 743–761.


Dutcă, I., 2018. Biomass data for young, planted Norway spruce (Picea

abies (L.) Karst.) trees in Eastern Carpathians of Romania. Data in

Brief 19, 2384–2392. https://doi.org/10.1016/J.DIB.2018.07.033

Dutcă, I., Mather, R., Blujdea, V.N.B., Ioraș, F., Olari, M., Abrudan,

I.V., 2018. Site-effects on biomass allometric models for early growth

plantations of Norway spruce (Picea abies (L.) Karst.). Biomass and

Bioenergy 116, 8–17. https://doi.org/10.1016/J.BIOMBIOE.2018.05.013

https://doi.org/10.1007/s00468-006-0064-4



https://doi.org/10.1016/J.DIB.2018.07.033


232

Eggleston, S., Buendia, L., Miwa, K., Ngara, T. and Tanabe, K. eds., 2006. 2006

IPCC guidelines for national greenhouse gas inventories (Vol. 5). Hayama, Japan:

Institute for Global Environmental Strategies.

Furnival, G.M., 1961. An Index for Comparing Equations Used In

Constructing Volume Tables. Forest Science 7, 337–341.

https://doi.org/10.1093/forestscience/7.4.337

Gevorkiantz, S.R., Bruce, D., Schumacher, F.X., 1997. Forest

Mensuration. Journal of the American Statistical Association 30, 780.

https://doi.org/10.2307/2277755

Heinsoo, K., Sild, E., Koppel, A., 2002. Estimation of shoot biomass

productivity in Estonian Salix plantations. Forest Ecology and

Management 170, 67–74. https://doi.org/10.1016/S0378-1127(01)00784-8

Hissa, L. de B.V., Aguiar, A.P.D., Camargo, R.R., Lima, L.S. de,

Gollnow, F., Lakes, T., 2019. Regrowing forests contribution to law

compliance and carbon storage in private properties of the Brazilian

Amazon. Land Use Policy 88, 104163.

https://doi.org/10.1016/J.LANDUSEPOL.2019.104163

Huang, S., Price, D., J. Titus, S., 2000. Development of


https://doi.org/10.2307/2277755

https://doi.org/10.1016/S0378-1127(01)00784-8

https://doi.org/10.1016/J.LANDUSEPOL.2019.104163

233

ecoregion-based height–diameter models for white spruce in boreal

forests. Forest Ecology and Management 129, 125–141.

https://doi.org/10.1016/S0378-1127(99)00151-6

Huxley, J.S., 1932. Problems of Relative Growth. Nature 129, 775–777.

https://doi.org/10.1038/129775a0

Jagodziński, A.M., Dyderski, M.K., Gęsikiewicz, K., Horodecki, P.,

Cysewska, A., Wierczyńska, S., Maciejczyk, K., 2018. How do tree stand

parameters affect young Scots pine biomass? – Allometric equations and

biomass conversion and expansion factors. Forest Ecology and

Management 409, 74–83. https://doi.org/10.1016/J.FORECO.2017.11.001

James, G., Witten, D., Hastie, T. and Tibshirani, R., 2013. An introduction to

statistical learning (Vol. 112, p. 18). New York: springer

Kays, J., Canham, C., 1991. Effects of Time and Frequency of Cutting

on Hardwood Root Reserves and Sprout Growth. Forest Science 37,

524–539. https://doi.org/10.1093/forestscience/37.2.524

Kuyah, S., Dietz, J., Muthuri, C., van Noordwijk, M., Neufeldt, H.,

2013. Allometry and partitioning of above- and below-ground biomass in

farmed eucalyptus species dominant in Western Kenyan agricultural

https://doi.org/10.1016/S0378-1127(99)00151-6

https://doi.org/10.1038/129775a0



234

landscapes. Biomass and Bioenergy 55, 276–284.


Laclau, J.-P., Bouillet, J.-P., Gonçalves, J.L.M., Silva, E.V.,

Jourdan, C., Cunha, M.C.S., Moreira, M.R., Saint-André, L., Maquère,

V., Nouvellon, Y., Ranger, J., 2008. Mixed-species plantations of

Acacia mangium and Eucalyptus grandis in Brazil: 1. Growth dynamics

and aboveground net primary production. Forest Ecology and Management

255, 3905–3917. https://doi.org/10.1016/J.FORECO.2007.10.049

Lojka, B.., Preininger, D.., Lojkova, J., Banout, J., Polesny, Z.,

2005. Biomass growth and farmer knowledge of Inga edulis in the

Peruvian Amazon. AGRICULTURA TROPICA ET SUBTROPICA 38, 44–51.

López, F., García, M.M., Yánez, R., Tapias, R., Fernández, M., Díaz,

M.J., 2008. Leucaena species valoration for biomass and paper

production in 1 and 2 year harvest. Bioresource Technology 99,

4846–4853. https://doi.org/10.1016/J.BIORTECH.2007.09.048

Lott, J.E., Howard, S.B., Black, C.R., Ong, C.K., 2000. Allometric

estimation of above-ground biomass and leaf area in managed Grevillea

robusta agroforestry systems. Agroforestry Systems 49, 1–15.

https://doi.org/10.1023/A:1006330830109



https://doi.org/10.1016/J.BIORTECH.2007.09.048

https://doi.org/10.1023/A:1006330830109

235

MacCracken, J.G., Ballenberghe, V. Van, 1993. Mass-Diameter

Regressions for Moose Browse on the Copper River Delta, Alaska.

Journal of Range Management 46, 302. https://doi.org/10.2307/4002462

Martin, A.R., Thomas, S.C., 2011. A reassessment of carbon content in tropical

trees. PLoS ONE 6. https://doi.org/10.1371/journal.pone.0023533

Mascaro, J., Litton, C.M., Hughes, R.F., Uowolo, A., Schnitzer, S.A.,

2014. Is logarithmic transformation necessary in allometry? Ten,

one-hundred, one-thousand-times yes. Biological Journal of the Linnean

Society 111, 230–233. https://doi.org/10.1111/bij.12177

Mauricio, R.M., Ribeiro, R.S., Paciullo, D.S.C., Cangussú, M.A.,

Murgueitio, E., Chará, J., Estrada, M.X.F., 2019. Silvopastoral

Systems in Latin America for Biodiversity, Environmental, and

Socioeconomic Improvements, in: Agroecosystem Diversity. Elsevier, pp.

287–297. https://doi.org/10.1016/B978-0-12-811050-8.00018-2

MEA, 2005. Millenium Ecosystem Assessment: Ecosystems and Human Well-Being

(Volume 4). Washington, DC. Island Press.

Montagnini, F., Nair, P.K.R., 2004. Carbon sequestration: An

https://doi.org/10.2307/4002462

https://doi.org/10.1111/bij.12177

https://doi.org/10.1016/B978-0-12-811050-8.00018-2

236

underexploited environmental benefit of agroforestry systems.

Springer, Dordrecht, pp. 281–295.

https://doi.org/10.1007/978-94-017-2424-1_20

Nahed-Toral, J., Valdivieso-Pérez, A., Aguilar-Jiménez, R.,

Cámara-Cordova, J., Grande-Cano, D., 2013. Silvopastoral systems with

traditional management in Southeastern Mexico: a prototype of

livestock agroforestry for cleaner production. Journal of Cleaner

Production 57, 266–279. https://doi.org/10.1016/j.jclepro.2013.06.020

Nygren, P., Rebottaro, S., Chavarra, R., 1993. Application of the

pipe model theory to non-destructive estimation of leaf biomass and

leaf area of pruned agroforestry trees. Agroforestry Systems 23,

63–77. https://doi.org/10.1007/BF00704851

Oppong, S.K., Kemp, P.D., Douglas, G.B., Foote, A.G., 2002. Effects of

season and frequency of harvest on browse yield and root carbohydrate

reserves of willows (Salix spp.) in New Zealand. New Zealand Journal

of Agricultural Research 45, 87–95.

https://doi.org/10.1080/00288233.2002.9513497

Packard, G.C., 2017a. Misconceptions about logarithmic transformation

and the traditional allometric method. Zoology 123, 115–120.

https://doi.org/10.1007/978-94-017-2424-1_20

https://doi.org/10.1016/j.jclepro.2013.06.020

https://doi.org/10.1007/BF00704851

https://doi.org/10.1080/00288233.2002.9513497

237

https://doi.org/10.1016/J.ZOOL.2017.07.005

Packard, G.C., 2017b. Why allometric variation in mammalian metabolism

is curvilinear on the logarithmic scale. Journal of Experimental

Zoology Part A: Ecological and Integrative Physiology 327, 537–541.

https://doi.org/10.1002/jez.2129

Packard, G.C., 2014. Multiplicative by nature: Logarithmic

transformation in allometry. Journal of Experimental Zoology Part B:

Molecular and Developmental Evolution 322, 202–207.

https://doi.org/10.1002/jez.b.22570

Pagiola, S., Ramírez, E., Gobbi, J., de Haan, C., Ibrahim, M.,

Murgueitio, E., Ruíz, J.P., 2007. Paying for the environmental

services of silvopastoral practices in Nicaragua. Ecological Economics

64, 374–385. https://doi.org/10.1016/j.ecolecon.2007.04.014

Parang, K., Wiebe, L., Knaus, E., 2012. Springer Texts in Statistics

An Introduction To Statistical Learning. Current Medicinal Chemistry

7, 995–1039. https://doi.org/10.2174/0929867003374372

Peichl, M., Arain, M.A., 2007. Allometry and partitioning of above-

and belowground tree biomass in an age-sequence of white pine forests.

https://doi.org/10.1016/J.ZOOL.2017.07.005

https://doi.org/10.1002/jez.2129

https://doi.org/10.1002/jez.b.22570

https://doi.org/10.1016/j.ecolecon.2007.04.014

https://doi.org/10.2174/0929867003374372

238

Forest Ecology and Management 253, 68–80.


Penman, J., Gytarsky, M., Hiraishi, T., Krug, T., Kruger, D., Pipatti,

R., Buendia, L., Miwa, K., Ngara, T., Tanabe, K., Wagner, F., 2003.

Good practice guidance for land use, land-use change and forestry.

Good practice guidance for land use, land-use change and forestry.





Picard, N., Saint-André, L. and Henry, M., 2012. Manual for building tree volume and

biomass allometric equations: from field measurement to prediction. Manual for

building tree volume and biomass allometric equations: from field measurement to

prediction, FAO; Food and Agricultural Organization of the United Nations (2012).

Poorter, H., Niklas, K.J., Reich, P.B., Oleksyn, J., Poot, P., Mommer,

L., 2012. Biomass allocation to leaves, stems and roots: meta-analyses

of interspecific variation and environmental control. New Phytologist

193, 30–50. https://doi.org/10.1111/j.1469-8137.2011.03952.x


https://doi.org/10.1111/j.1469-8137.2011.03952.x

239

Prasad, J.V.N.S., Korwar, G.R., Rao, K.V., Mandal, U.K., G.R.Rao,

Srinivas, I., Venkateswarlu, B., Rao, S.N., Kulkarni, H.D., 2011.

Optimum stand density of Leucaena leucocephala for wood production in

Andhra Pradesh, Southern India. Biomass and Bioenergy 35, 227–235.


Raitio, H., Paukkonen, K., Kauppi, A., 1994. Effects of defoliation,

nitrogen nutrition, and temperature on leafing and root carbohydrates

of birch seedlings. Canadian Journal of Forest Research 24, 1914–1920.

https://doi.org/10.1139/x94-246

Rubilar, R.A., Allen, H.L., Alvarez, J.S., Albaugh, T.J., Fox, T.R.,

Stape, J.L., 2010. Silvicultural manipulation and site effect on above

and belowground biomass equations for young Pinus radiata. Biomass and

Bioenergy 34, 1825–1837.


Serrão, E.A.S., Nepstad, D., Walker, R., 1996. Upland agricultural and

forestry development in the Amazon: sustainability, criticality and

resilience. Ecological Economics 18, 3–13.

https://doi.org/10.1016/0921-8009(95)00092-5

Sharma, E., Ambasht, R.S., 1991. Biomass, productivity and energetics


https://doi.org/10.1139/x94-246


https://doi.org/10.1016/0921-8009(95)00092-5

240

in himalayan alder plantations. Annals of Botany 67, 285–293.

https://doi.org/10.1093/oxfordjournals.aob.a088138

Sharrow, S.H., Ismail, S., 2004. Carbon and nitrogen storage in

agroforests, tree plantations, and pastures in western Oregon, USA.


https://doi.org/10.1023/B:AGFO.0000013267.87896.41

Shimamoto, C.Y., Botosso, P.C., Marques, M.C.M., 2014. How much carbon

is sequestered during the restoration of tropical forests? Estimates

from tree species in the Brazilian Atlantic forest. Forest Ecology and


Shinozaki, K., Yoda, K., Hozumi, K., Kira, T., 1964. A Quantative

Analysis of plant form - The Pipe Model Theory: Basic Analyses.

Japanese Journal Of Ecology 14, 97–105.

https://doi.org/10.18960/seitai.14.3_97

Sileshi, G.W., 2014. A critical review of forest biomass estimation

models, common mistakes and corrective measures. Forest Ecology and


Smit, C., Ruifrok, J.L., van Klink, R., Olff, H., 2015. Rewilding with

https://doi.org/10.1093/oxfordjournals.aob.a088138



https://doi.org/10.18960/seitai.14.3_97


241

large herbivores: The importance of grazing refuges for sapling

establishment and wood-pasture formation. Biological Conservation 182,

134–142. https://doi.org/10.1016/J.BIOCON.2014.11.047

Steinfeld, H., Food and Agriculture Organization of the United

Nations., Livestock, E. and D. (Firm), 2006. Livestock’s long shadow :

environmental issues and options. Food and Agriculture Organization of

the United Nations.

Szott, L.T., Palm, C.A., Davey, C.B., 1994. Biomass and litter

accumulation under managed and natural tropical fallows. Forest


https://doi.org/10.1016/0378-1127(94)90015-9

Takimoto, A., Nair, P.K.R., Nair, V.D., 2008. Carbon stock and

sequestration potential of traditional and improved agroforestry

systems in the West African Sahel. Agriculture, Ecosystems &

Environment 125, 159–166. https://doi.org/10.1016/j.agee.2007.12.010

Telenius, B., Verwijst, T., 1995. The influence of allometric

variation, vertical biomass distribution and sampling procedure on

biomass estimates in commercial short-rotation forests. Bioresource

Technology 51, 247–253. https://doi.org/10.1016/0960-8524(94)00133-L

https://doi.org/10.1016/J.BIOCON.2014.11.047

https://doi.org/10.1016/0378-1127(94)90015-9


https://doi.org/10.1016/0960-8524(94)00133-L

242

Tzilivakis, J., Warner, D.J., Holland, J.M., 2019. Developing

practical techniques for quantitative assessment of ecosystem services

on farmland. Ecological Indicators 106, 105514.

https://doi.org/10.1016/J.ECOLIND.2019.105514

Yuen, J.Q., Fung, T., Ziegler, A.D., 2016. Review of allometric

equations for major land covers in SE Asia: Uncertainty and

implications for above- and below-ground carbon estimates. Forest



https://doi.org/10.1016/J.ECOLIND.2019.105514


243

Appendices

Appendix 1. Table. The wet to dry relationship for biomass components of subsamples.taken from Inga edulis, Erythrina berteroana and Leucaena leucocephala, where R2 is the coefficient of determination, RSE is the residual square error and n is the sample size.

Species Biomass component Intercept Slope R2 RSE n

I. edulis Trunk 0.629 2.073 0.992 1.6 118

Branches -21.642 2.470 0.985 180.9 118

Leaf 7.478 2.679 0.997 49.5 118 E. berteroana Trunk 2.054 2.674 0.949 4.9 117

Branches 62.000 3.112 0.989 126.0 117

Leaf -0.480 4.918 0.966 62.1 117 L. leucocephala Trunk -0.624 1.795 0.996 1.1 115

Branches -24.161 2.246 0.998 51.2 115

Leaf 1.675 2.870 0.999 19.4 115

244

Appendix 2. Figure. Total biomass (g) and root collar diameter (mm) of Inga edulis plotted using the linear regression line (y = aX + b) showing clockwise from left (a) scatter plot and linear regression line, (b) departure from normal Q-Q plot and (c) residuals vs. leverage

20 40 60 80 100 120

01

00

02

00

03

00

04

00

05

00

06

00

0

RCD

TO

TA

LB

IOM

AS

S

−2 −1 0 1 2

−2

−1

01

23

Theoretical Quantiles

Sta

nd

ard

ized

re

sid

ua

ls

lm(TOTALBIOMASS ~ RCD)

Normal Q−Q

2

34

11

0.00 0.02 0.04 0.06 0.08 0.10

−3

−2

−1

01

23

4

Leverage

Sta

nd

ard

ized

re

sid

ua

ls

lm(TOTALBIOMASS ~ RCD)

Cook's distance 0.5

0.5

1Residuals vs Leverage

11

2

37

245

Appendix 3. Figure. Relationship between total biomass (g) and root collar diameter (mm) for Inga edulis plotted using the quadratic regression model (equation y = ax2 + bx + c) showing clockwise from left (a) scatter plot and polynomial regression line, (b) departure from normal Q-Q plot and (c) residuals vs. leverage

20 40 60 80 100 120

01

00

02

00

03

00

04

00

05

00

06

00

0

RCD

TO

TA

LB

IOM

AS

S

−2 −1 0 1 2

−2

−1

01

23

4


Sta

nd

ard

ized

re

sid

ua

ls

lm(TOTALBIOMASS ~ poly(RCD, 2))

Normal Q−Q

2

34

11

0.0 0.1 0.2 0.3 0.4

−3

−2

−1

01

23

4

Leverage

Sta

nd

ard

ized

re

sid

ua

ls

lm(TOTALBIOMASS ~ poly(RCD, 2))

Cook's distance

1

0.5

0.5

1

Residuals vs Leverage

11

2

76

246

Appendix 4. Figure. Relationship between total biomass (g) and root collar diameter (mm) for I. edulis plotted using the log-linear transformation ln(y) = a + b * lnx showing clockwise from left (a) scatter plot and regression line, (b) departure from normal Q-Q plot and (c) residuals vs. leverage

2.5 3.0 3.5 4.0 4.5

67

8

LNRCD

LN

TO

TA

LB

IOM

AS

S

−2 −1 0 1 2

−2

−1

01

23


Sta

nd

ard

ized

re

sid

ua

ls

lm(log(TOTALBIOMASS) ~ log(RCD))

Normal Q−Q

77

79

37

0.00 0.05 0.10 0.15 0.20

−2

−1

01

23

Leverage

Sta

nd

ard

ized

re

sid

ua

ls

lm(log(TOTALBIOMASS) ~ log(RCD))

Cook's distance

1

0.5

0.5

1

Residuals vs Leverage

77

76

79

247

Appendix 5. Table. Allometric equations following the linear model for biomass components of 0.5 x 7.5 m alley rows of Erythrina berteroana 6 months post pollarding to a height of 2 m. Where (x) is the measurement used as a predictor of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH (mm), 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm), (y) is the log transformation of total biomass(g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y) Biomass b0 (S.E.) P b1 (S.E.) P R2 MODEL

RMSE (F.I.) CV RMSE

RCD Total -698.4(265.4)0.01 32.2(4.0)<0.01 0.48 748.75 669.41

Trunk -335.3(234.2)0.2 21.6(3.5)<0.01 0.34 659.16 575.19

Branch -322.8(91.8)<0.01 9.3(1.4)<0.01 0.39 249.39 224.39

Leaf -45.6(15.8)<0.01 1.4(0.2)<0.01 0.33 43.09 41.24

DBH Total -136.6(303.3)0.7 71.9(13.4)<0.01 0.31 814.15 777.36

Trunk -118.3(241.2)0.6 56.1(10.6)<0.01 0.3 647.29 592.11

Branch 9.0(113.0)0.9 12.5(5.0)0.01 0.09 303.41 276.70

Leaf -27.3(16.8)0.1 3.4(0.7)<0.01 0.24 45.20 43.41

SD Total -221.7(305.0)0.5 32.4(5.7)<0.01 0.33 802.47 737.63

Trunk -236.7(237.9)0.3 26.3(4.5)<0.01 0.34 625.86 575.16

Branch 46.7(116.9)0.7 4.6(2.2)<0.05 0.06 307.62 268.01

Leaf -31.7(17.0)0.07 1.5(0.3)<0.01 0.26 44.67 42.56

CV Total 707.7(115.1)<0.01 0.0(0.0)<0.01 0.53 668.19 626.49

Trunk 637.8(107.3)<0.01 0.0(0.0)<0.01 0.35 623.07 569.37

Branch 68.6(40.7)0.1 0.0(0.0)<0.01 0.45 236.39 228.02

Leaf 1.3(4.6)0.8 0.0(0.0)<0.01 0.74 26.45 26.02

CD Total -144.2(217.0)0.5 16.7(2.1)<0.01 0.48 704.99 639.73

Trunk 127.4(199.9)0.5 10.3(2.0)<0.01 0.29 649.50 586.90

Branch -227.8(70.5)<0.01 5.4(0.7)<0.01 0.48 229.04 212.20

Leaf -43.8(10.6)<0.01 1.0(0.1)<0.01 0.56 34.39 32.69

248

Appendix 6. Table. Allometric equations following the linear model for biomass components of 0.5 x 7.5 m alley rows of Erythrina berteroana 6 months post pollarding to a height of 2 m. Where (x) is the root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass. (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g), b0 as the intercept coefficient, b1 and b2 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index F.I. for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y)

Biomass b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P R2 MODEL

RMSE (F.I.) CV RMSE

RCD + HEIGHT Total -2,138.7(464.9)<0.01 25.2(4.5)<0.01 8.4(2.1)<0.01 0.51 670.39 625.55

Trunk -1,485.3(413.3)<0.01 14.8(4.0)<0.01 7.1(1.9)<0.01 0.39 595.98 554.43

Branch -513.8(171.6)<0.01 9.3(1.7)<0.01 0.8(0.8)0.3 0.37 247.40 225.83

Leaf -139.6(26.3)<0.01 1.1(0.3)<0.01 0.5(0.1)<0.01 0.45 37.99 38.66

249

Appendix 7. Table. Allometric equations following the linear model for biomass components of 0.5 x 7.5 m alley rows of Inga edulis 6 months post pollarding to a height of 2 m. Where (x) is the measurement used a predictor of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm), (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g), b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y)

Biomass b0 (S.E.) P b1 (S.E.) P R2 MODEL

RMSE (F.I.) CV

RMSE

RCD Total -931.6(332.5)<0.01 47.7(4.8)<0.01 0.56 956.09 954.3

Trunk -415.5(185.5)0.03 26.6(2.7)<0.01 0.56 533.50 510.3

Branch -351.0(168.8)0.04 13.8(2.4)<0.01 0.29 485.25 474.5

Leaf -165.1(98.8)0.1 7.3(1.4)<0.01 0.25 283.99 279.8

DBH Total 1,100.2(398.9)<0.01 31.9(10.7)<0.01 0.1 1360.22 1348.4

Trunk 662.7(220.1)<0.01 19.3(5.9)<0.01 0.12 750.59 756.8

Branch 216.0(163.7)0.2 9.8(4.4)0.03 0.06 558.17 552.4

Leaf 221.4(95.5)0.02 2.7(2.6)0.3 0.01 325.59 317.5

SD Total -1,054.7(345.0)<0.01 64.3(6.5)<0.01 0.55 957.32 957

Trunk -571.1(180.4)<0.01 37.6(3.4)<0.01 0.61 500.51 484.4

Branch -380.8(175.4)0.03 18.5(3.3)<0.01 0.29 486.68 473.4

Leaf -102.8(107.3)0.34 8.3(2.0)<0.01 0.17 297.86 291.3

CV Total 1.529e+03 (1.940e+02)<0.01 2.162e-04 (4.332e-05)<0.01 0.24 1250.69 1257.4

Trunk 1.001e+03 (1.120e+02)<0.01 1.057e-04 (2.502e-05)<0.01 0.18 722.35 733.2

Branch 3.357e+02 (8.180e+01)<0.01 7.102e-05( 1.827e-05)<0.01 0.16 527.37 525.7

Leaf 1.923e+02 (4.679e+01)<0.01 3.951e-05(1.045e-05)<0.01 0.15 301.68 293.2

CD Total 600.5(350.5)0.09 10.7(2.2)<0.01 0.24 1251.90 1291.8

Trunk 553.2(202.7)<0.01 5.2(1.3)<0.01 0.18 724.05 750.8

Branch 26.4(147.5)0.9 3.6(0.9)<0.01 0.16 526.87 534.5

Leaf 20.9(84.4)0.9 2.0(0.5)<0.01 0.15 301.56 303.5

250

Appendix 8. Table. Allometric equations following the linear model for biomass components of 0.5 x 7.5 m alley rows of Inga edulis 6 months post pollarding to a height of 2 m. Where (x) is the root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass, (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 and b2

as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index F.I. for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y) Biomass b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P R2 MODEL

RMSE (F.I.) CV

RMSE

RCD + HEIGHT Total -2,139.6(531.7)<0.01 41.9(5.0)<0.01 6.5(2.3)<0.01 0.6 910.26 922.6

Trunk -996.0(300.6)<0.01 23.8(2.8)<0.01 3.1(1.3)0.02 0.59 514.65 506.3

Branch -718.5(278.7)0.01 12.1(2.6)<0.01 2.0(1.2)0.1 0.31 477.03 473.3

Leaf -425.1(161.8)0.01 6.1(1.5)<0.01 1.4(0.7)<0.05 0.29 276.93 276.9

251

Appendix 9. Table. Allometric equations following the linear model for biomass components of 0.5 x 7.5 m alley rows of Leucaena leucocephala 6 months post pollarding to a height of 2 m. Where (x) is the measurement used as predictor of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm), (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g), b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y) Biomass b0 (S.E.) P b1 (S.E.) P R2 MODEL RMSE (F.I.) CV RMSE

RCD Total 774.8(146.0)<0.01 46.3(3.2)<0.01 0.78 320.3 311.1

Trunk -685.9(95.9)<0.01 39.6(2.1)<0.01 0.85 210.4 198.9

Branch -55.7(94.4)0.6 5.2(2.1)0.02 0.09 207.1 178.2

Leaf .-33.250(14.4)0.02 1.4(0.3)<0.01 0.26 31.6 29.4

DBH Total 285.7(196.3)0.2 43.8(8.2)<0.01 0.32 558.6 545.8

Trunk 178.3(151.7)0.3 39.5(6.4)<0.01 0.39 431.7 415.2

Branch 109.8(76.0)0.2 2.8(3.2)0.4 0.01 216.2 189.9

Leaf -2.4(12.1)0.9 1.5(0.5)<0.01 0.12 34.4 32.3

SD Total -390.7(190.7)<0.05 47.2(5.2)<0.01 0.57 441.5 411.9

Trunk -447.3(127.1)<0.01 43.0(3.5)<0.01 0.71 294.4 279.8

Branch 76.5(93.1)0.4 2.7(2.5)0.3 0.02 215.6 189.1

Leaf -19.9(14.4)0.2 1.4(0.4)<0.01 0.18 33.2 30.5

CV Total 953.2(95.3)<0.01 0.0(0.0)<0.01 0.28 571.9 562.4

Trunk 837.7(81.2)<0.01 0.0(0.0)<0.01 0.22 487.6 471.2

Branch 104.9(33.8)<0.01 0.0(0.0)<0.01 0.13 202.6 177.2

Leaf 10.6(4.7)0.03 0.0(0.0)<0.01 0.40 28.4 25.7

CD Total 282.8(149.0)0.06 10.0(1.4)<0.01 0.46 496.0 474.3

Trunk 424.9(139.5)<0.01 6.5(1.3)<0.01 0.29 464.6 439.1

Branch -115.9(51.2)0.03 3.0(0.5)<0.01 0.39 170.3 161.7

Leaf -26.2(7.6)<0.01 0.6(0.1)<0.01 0.52 25.3 23.1

252

Appendix 10. Table. Allometric equations following the linear model for biomass components of 0.5 x 7.5 m alley rows of Leucaena leucocephala 6 months post pollarding to a height of 2 m. Where (x) is the root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass, (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 and b2 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index F.I. for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y) Biomass b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P R2 MODEL

RMSE (F.I.) CV

RMSE

RCD + HEIGHT Total -1,209.0(188.1)<0.01 41.7(3.3)<0.01 2.2(0.7)<0.01 0.81 296.9 283.9

Trunk -908.3(127.9)<0.01 37.3(2.2)<0.01 1.1(0.4)0.02 0.87 201.9 192.7

Branch -215.4(129.0)0.1 3.5(2.2)0.1 0.8(0.4)0.08 0.14 203.5 170.9

Leaf -85.2(17.8)<0.01 0.9(0.3)<0.01 0.3(0.1)<0.01 0.43 28.0 25.5

253

Appendix 11. Table. Allometric equations following a quadratic model for biomass components of 0.5 x 7.5 m alley rows of Erythrina berteroana 6 months post pollarding to a height of 2 m. Where (x) is the measurement used as a predictor of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV, cm3) and crown diameter (CD cm), (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g), b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y) Biomass b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P R2 MODEL RMSE

(F.I.) CV RMSE

RCD Total 1,405.2(15.1)<0.01 5,173.0(6.8)<0.01 158.8(0.2)0.8 0.39 748.5 726.3

Trunk 1,084.3(13.3)<0.01 3,315.9(4.9)<0.01 -251.4(-0.4)0.7 0.25 658.5 638.0

Branch 276.1(9.1)<0.01 1,620.2(6.5)<0.01 414.5(1.7)0.1 0.39 244.3 215.1

Leaf 44.9(8.4)<0.01 236.9(5.4)<0.01 -4.3(-0.1)0.9 0.29 43.1 44.5

DBH Total 1,405.2(14.2)<0.01 4,450.7(5.5)<0.01 -1,409.8(-1.7)0.09 0.32 796.0 765.4

Trunk 1,084.3(13.6)<0.01 3,471.3(5.3)<0.01 -449.7(-0.7)0.5 0.28 645.0 600.0

Branch 276.1(7.8)<0.01 771.1(2.6)0.01 -859.1(-2.9)<0.01 0.17 285.0 260.4

Leaf 44.9(8.3)<0.01 208.4(4.7)<0.01 -101.0(-2.3)0.03 0.27 43.5 42.1

SD Total 1,405.2(14.1)<0.01 4,592.7(5.6)<0.01 -245.9(-0.3)0.8 0.31 801.9 771.5

Trunk 1,084.3(14.0)<0.01 3,729.0(5.8)<0.01 124.7(0.2)0.8 0.32 625.7 604.8

Branch 276.1(7.3)<0.01 647.6(2.1)0.04 -403.7(-1.3)0.2 0.06 303.7 272.0

Leaf 44.9(8.1)<0.01 216.1(4.7)<0.01 33.1(0.7)0.5 0.26 44.5 45.5

CV Total 1,405.2(17.0)<0.01 5,875.5(8.6)<0.01 -124.6(-0.2)0.9 0.52 668.0 645.6

Trunk 1,084.2(14.1)<0.01 3,760.6(5.9)<0.01 -294.5(-0.5)0.6 0.33 622.0 574.8

Branch 276.1(9.4)<0.01 1,747.8(7.2)<0.01 97.5(0.4)0.7 0.43 236.1 246.7

Leaf 44.9(14.5)<0.01 367.1(14.4)<0.01 72.5(2.8)<0.01 0.76 24.9 24.8

254

Appendix 12. Table. Allometric equations following the quadratic model for biomass components of 0.5 x 7.5 m alley rows of Erythrina berteroana 6 months post pollarding to a height of 2 m. Where (x) is the root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass, (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 and b2 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index F.I. for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y)

Biomass b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P b3 (S.E.) P b4 (S.E.) P R2

MODEL RMSE (F.I.)

CV RMSE

RCD +

HEIGHT Total 522.7(0.8)0.4 276.1(9.3)<0.01 1,468.6(5.5)<0.01 496.5(2.0)0.05 373.8(1.4)0.2 0.52 653.8 638.5

Trunk 366.8(1.5)0.1 44.9(11.0)<0.01 160.9(4.5)<0.01 35.8(1.1)0.3 189.4(5.2)<0.01 0.37 592.4 572.9

Branch 164.3(4.9)<0.01 0.20(14.52)<0.01 0.35(3.12)<0.01 0.02(0.18)0.9 0.16(11.75)<0.01 0.41 236.8 220.5

Leaf 0.19(1.71).0.9 -0.01(-0.11)0.9 0.04(10.44)<0.01 0.14(4.95)<0.01 0.03(1.20)0.2 0.59 32.3 34.7

255

Appendix 13. Table. Allometric equations following a quadratic model for biomass components of 0.5 x 7.5 m alley rows of Inga edulis 6 months post pollarding to a height of 2 m. Where (x) is the measurement used as a predictor of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV, cm3) and crown diameter (CD cm), (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g), b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).


(F.I.) CV RMSE

RCD Total 2,195.7(107.9)<0.01 9,623.1(971.1)<0.01 -699.9(971.1)0.5 0.55 952.9 981.9

Trunk 1,327.0(60.3)<0.01 5,362.0(542.9)<0.01 -253.7(542.9)0.6 0.54 493.4 529.1

Branch 554.6(54.8)<0.01 2,786.7(493.0)<0.01 -339.3(493.0)0.5 0.28 462.7 482.1

Leaf 314.1(32.1)<0.01 1,474.4(289.2)<0.01 -106.8(289.2)0.7 0.23 275.4 288.2

DBH Total 2,195.7(137.9)<0.01 4,096.6(1,241.4)<0.01 5,446.4(1,241.4)<0.01 0.26 1218.2 1,208.0

Trunk 1,327.0(73.1)<0.01 2,484.0(657.8)<0.01 -3,447.2(657.8)<0.01 0.33 645.5 634.7

Branch 554.6(60.5)<0.01 1,266.1(544.1)0.02 -1,464.4(544.1)<0.01 0.12 533.9 543.5

Leaf 314.1(36.3)<0.01 346.5(326.2)0.3 -534.8(326.2)0.1 0.02 320.1 318.2

SD Total 2,195.7(105.6)<0.01 9,613.2(950.1)<0.01 1,954.8(950.1)0.04 0.57 932.4 938.2

Trunk 1,327.0(55.6)<0.01 5,613.7(500.0)<0.01 889.5(500.0)0.08 0.61 490.7 487.0

Branch 554.6(54.3)<0.01 2,766.4(489.0)<0.01 731.3(489.0)0.1 0.29 479.9 467.0

Leaf 314.1(33.5)<0.01 1,233.1(301.2)<0.01 334.0(301.2)0.3 0.17 295.5 296.2

CV Total 2,195.7(114.8)<0.01 6,320.2(1,033.4)<0.01 6,588.3(1,033.4)<0.01 0.49 1014.1 1,050.5

Trunk 1,327.0(70.6)<0.01 3,088.7(635.5)<0.01 -3,280.8(635.5)<0.01 0.38 623.6 634.7

Branch 554.6(54.0)<0.01 2,076.3(485.6)<0.01 -2,033.5(485.6)<0.01 0.30 476.5 484.8

Leaf 314.1(30.2)<0.01 1,155.2(271.5)<0.01 -1,274.1(271.5)<0.01 0.32 266.4 270.9

CD Total 2,195.7(119.5)<0.01 6,300.7(1,075.8)<0.01 6,056.0(1,075.8)<0.01 0.44 1055.7 1,223.3

Trunk 1,327.0(72.6)<0.01 3,056.3(653.6)<0.01 -3,023.2(653.6)<0.01 0.34 641.4 774.8

Branch 554.6(54.1)<0.01 2,086.5(487.0)<0.01 -1,997.3(487.0)<0.01 0.29 477.9 540.2

Leaf 314.1(31.6)<0.01 1,157.9(284.1)<0.01 -1,035.4(284.1)<0.01 0.26 278.8 320.3

256

Appendix 14. Table. Allometric equations following the quadratic model for biomass components of 0.5 x 7.5 m alley rows of Inga edulis 6 months post pollarding to a height of 2 m. Where (x) is the root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass, (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 and b2 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index F.I. for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y)

Biomass b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P b3 (S.E.) P b4 (S.E.) P R2

MODEL RMSE (F.I.)

CV RMSE

RCD +

HEIGHT Total 2,195.7(100.6)<0.01 8,336.5(992.2)<0.01 -853.7(906.6)0.3 2,956.2(991.8)<0.01 2,040.0(906.9)0.03 0.61 877.2 904.6

Trunk 1,327.0(56.6)<0.01 4,729.4(558.0)<0.02 -336.0(509.9)0.5 1,423.9(557.8)0.01 1,286.0(510.1)0.02 0.60 493.4 493.8

Branch 554.6(53.1)<0.01 2,375.0(523.3)<0.03 -396.2(478.2)0.4 912.1(523.2)0.09 978.3(478.4)0.04 0.32 462.7 468.0

Leaf 314.1(31.6)<0.01 1,232.1(311.5)<0.04 -121.5(284.7)0.7 620.2(311.4)0.05 620.2(311.4)0.05 0.26 275.4 299.3

257

Appendix 15. Table. Allometric equations following a quadratic model for biomass components of 0.5 x 7.5 m alley rows of Leucaena leucocephala 6 months post pollarding to a height of 2 m. Where (x) is the measurement used as a predictor of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm), (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g), b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).


(F.I.) CV RMSE

RCD Total 1,259.5(40.6)<0.01 4,645.9(322.4)<0.01 148.8(322.4)0.6 0.77 314.6 313.6

Trunk 1,056.5(26.3)<0.01 3,979.3(208.8)<0.01 289.2(208.8)0.2 0.85 203.8 199.0

Branch 172.6(26.2)<0.01 521.3(208.1)0.02 -137.1(208.1)0.5 0.07 203.1 178.8

Leaf 30.4(4.0)<0.01 145.3(31.9)<0.01 -3.3(31.9)0.9 0.23 31.1 29.7

DBH Total 1,259.5(70.7)<0.01 2,968.3(561.3)<0.01 -362.0(561.3)0.5 0.30 547.7 552.9

Trunk 1,056.5(54.2)<0.01 2,677.0(429.9)<0.01 -530.3(429.9)0.2 0.38 419.5 413.3

Branch 172.6(27.4)<0.01 191.3(217.3)0.4 136.2(217.3)0.5 -0.01 212.1 192.5

Leaf 30.4(4.3)<0.01 100.1(34.4)<0.01 32.1(34.4)0.4 0.11 33.6 33.5

SD Total 1,259.5(50.5)<0.01 3,994.2(400.9)<0.01 1,498.8(400.9)<0.01 0.64 391.2 409.7

Trunk 1,056.5(35.1)<0.01 3,639.9(278.3)<0.01 798.3(278.3)<0.01 0.74 271.6 269.2

Branch 172.6(25.1)<0.01 232.6(199.3)0.2 672.2(199.3)<0.01 0.15 194.5 204.5

Leaf 30.4(4.2)<0.01 121.7(33.3)<0.01 28.3(33.3)0.4 0.16 32.5 31.4

CV Total 1,259.4(63.9)<0.01 2,809.7(507.5)<0.01 2,119.9(507.5)<0.01 0.43 495.3 498.9

Trunk 1,056.5(57.9)<0.01 2,007.5(459.5)<0.01 1,355.8(459.5)<0.01 0.29 448.4 444.3

Branch 172.5(23.2)<0.01 620.7(183.7)<0.01 -691.8(183.7)<0.01 0.28 179.3 170.4

Leaf 30.4(3.4)<0.01 181.4(27.1)<0.01 -72.3(27.1)<0.01 0.45 26.4 24.5

CD Total 1,259.5(61.3)<0.01 3,582.4(486.4)<0.01 -902.8(486.4)0.07 0.47 474.7 470.4

Trunk 1,056.5(55.3)<0.01 2,316.9(438.8)<0.01 1,270.4(438.8)<0.01 0.36 428.2 427.6

Branch 172.6(20.9)<0.01 1,057.9(165.6)<0.01 351.8(165.6)0.04 0.41 161.7 169.9

Leaf 30.4(3.2)<0.01 207.6(25.4)<0.01 15.9(25.4)0.5 0.51 24.8 24.5

258

Appendix 16. Table. Allometric equations following the quadratic model for biomass components of 0.5 x 7.5 m alley rows of Leucaena leucocephala 6 months post pollarding to a height of 2 m. Where (x) is the root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass, (y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 and b2 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), root mean square error of model (Model RMSE, which is the same as Furnival’s Index F.I. for untransformed data) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

(x) (y) Biomass b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P b3 (S.E.) P b4 (S.E.) P R2

MODEL RMSE (F.I.)

CV RMSE

RCD +

HEIGHT Total 1,259.5(34.6)<0.01 4,300.9(12.9)<0.01 358.4(1.2)0.2 1,111.7(3.4)<0.01 561.6(1.9)0.07 0.81 314.6 284.2

Trunk 1,056.5(42.8)<0.01 3,757.6(16.7)<0.01 412.2(2.0)0.05 635.3(2.8)<0.01 214.1(1.1)0.3 0.87 188.2 194.0

Branch 172.6(6.7)<0.01 444.9(1.9)0.06 -76.1(-0.4)0.7 345.1(1.5)0.1 308.1(1.5)0.1 0.11 194.9 170.3

Leaf 30.4(8.6)<0.01 98.4(3.1)<0.01 22.3(0.8)0.4 131.3(4.1)<0.01 39.4(1.4)0.2 0.41 26.8 25.3

259

Appendix 17. Table. Allometric equations following the log-linear model for biomass components of 0.5 x 7.5 m alley rows of Erythrina berteroana 6 months post pollarding to a height of 2 m. Where ln(x) is the log transformation of the following measurements used as predictors of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm) ln(y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), lambda (λ) and standard deviation (S.D.) as back transformation according to Marklund (1987), Furnival’s Index (F.I.) according to Kershaw et al. (2016), root mean square error of model (Model RMSE) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

ln(x) ln(y)

Biomass b0 (S.E.) P b1 (S.E.) P R2 λ S.D F.I. MODEL RMSE

CV RMSE

RCD Total -2.12 (0.72) 0.0045 2.19 (0.18) <0.001 0.68 1.24 1.33 585.4 0.54 0.46

Trunk -2.09 (0.76) 0.007 2.12 (0.19) <0.001 0.65 1.28 1.66 465.3 0.55 0.39

Branch -5.87 (1.30) <0.001 2.64 (0.32) <0.001 0.50 1.42 1.23 140.0 0.92 1.51

Leaf -8.80 (1.56) <0.001 2.86 (0.38) <0.001 0.44 1.70 1.78 25.6 1.19 1.20

DBH Total 2.40 (0.58) <0.001 1.52 (0.19) <0.001 0.49 1.36 0.75 608.0 0.56 0.57

Trunk 2.27 (0.56) <0.001 1.48 (0.18) <0.001 0.49 1.21 1.03 455.9 0.54 0.55

Branch -0.18 (1.08) 0.87 1.73 (0.36) <0.001 0.26 1.57 1.47 158.2 1.03 1.07

Leaf -3.52 (1.26) 0.007 2.19 (0.42) <0.001 0.29 1.67 1.52 25.9 1.21 1.25

SD Total -0.58 (0.67) 0.39 1.95 (0.17) <0.001 0.64 1.43 0.98 596.4 0.55 0.54

Trunk -0.96 (0.63) 0.13 1.98 (0.16) <0.001 0.67 1.21 1.15 433.6 0.51 0.39

Branch -2.21 (1.28) 0.09 1.87 (0.33) <0.001 0.31 1.69 2.05 163.5 1.07 1.62

Leaf -6.84 (1.37) <0.001 2.55 (0.36) <0.001 0.42 1.89 2.48 26.2 1.22 1.25

CV Total -0.32 (0.56) 0.57 0.55 (0.04) <0.001 0.70 1.30 0.86 579.1 0.53 0.55

Trunk -0.25 (0.60) 0.68 0.52 (0.05) <0.001 0.65 1.28 1.11 471.3 0.55 0.59

Branch -4.22 (0.91) <0.001 0.70 (0.07) <0.001 0.59 1.42 1.29 130.8 0.85 0.90

Leaf -7.60 (1.08) <0.001 0.81 (0.08) <0.001 0.57 1.53 1.99 22.6 1.05 1.07

CD Total 0.57 (0.50) 0.25 1.44 (0.11) <0.001 0.70 1.18 0.23 565.1 0.52 0.55

Trunk 0.76 (0.56) 0.18 1.33 (0.13) <0.001 0.62 1.29 1.44 478.9 0.56 0.63

Branch -3.99 (0.65) <0.001 2.04 (0.15) <0.001 0.73 0.25 0.19 104.8 0.68 0.74

Leaf -7.06 (1.07) <0.001 2.28 (0.24) <0.001 0.58 1.49 1.36 20.1 0.94 0.93

260

Appendix 18. Table. Allometric equations following the log-linear model for biomass components of 0.5 x 7.5 m alley rows of Erythrina berteroana 6 months post pollarding to a height of 2 m. Where ln(x) is the log transformation of root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass. ln(y) is the log transformation of total biomass (g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), lambda (λ) and standard deviation (S.D.) as back transformation according to Marklund (1987), Furnival’s Index (F.I.) according to Kershaw et al. (2016), root mean square error of model (Model RMSE) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

ln(x) ln(y)

(Biomass) b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P R2 λ S.D F.I. Model RMSE

CV RMSE

RCD + Height Total -6.41 (1.02) <0.001 1.53 (0.2) <0.001 1.31 (0.25) <0.001 0.77 1.12 0.83 523.2 0.47 0.49

Trunk -6.91 (1.03) <0.001 1.38 (1.38) <0.001 1.47 (0.25) <0.001 0.76 1.13 0.94 409.5 0.48 0.50

Branch -7.40 (2.99) <0.001 2.4 (0.4) <0.001 0.44 (0.59) 0.46 0.41 1.66 1.47 140.5 0.91 0.96

Leaf -14.50 (2.46) <0.001 1.98 ( 0.47) <0.001 1.74 (0.6) 0.005 0.49 1.58 1.49 24.1 1.12 1.17

261

Appendix 19. Table. Allometric equations following the log-linear model for biomass components of 0.5 x 7.5 m alley rows of Inga edulis 6 months post pollarding to a height of 2 m. Where ln(x) is the log transformation of the following measurements used as predictors of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm) ln(y) is the log transformation of total biomass(g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), lambda (λ) and standard deviation (S.D.) as back transformation according to Marklund (1987), Furnival’s Index (F.I.) according to Kershaw et al. (2016), root mean square error of model (Model RMSE) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

ln(x) ln(y)

(Biomass) b0 (S.E.) P b1 (S.E.) P R2 λ S.D. F.I. Model RMSE

CV RMSE

RCD Total 0.56 (0.51) 0.28 1.68 (0.12) <0.001 0.70 1.07 0.54 759.2 0.44 0.46

Trunk 0.95 (0.43) 0.03 1.47 (0.10) <0.001 0.71 1.07 0.56 409.6 0.38 0.39

Branch -8.89 (1.67) <0.001 3.45 (0.40) <0.001 0.48 2.37 4.64 287.9 1.45 1.51

Leaf -6.47 (1.33) <0.001 2.78 (0.32) <0.001 0.48 2.19 6.52 169.1 1.16 1.20

DBH Total 3.78 (0.71) <0.001 1.06 (0.20) <0.001 0.25 1.21 0.72 1192.5 0.70 0.72

Trunk 3.39 (0.59) <0.001 1.03 (0.17) <0.001 0.32 1.20 0.66 629.3 0.58 0.60

Branch -2.31 (1.86) 0.22 2.19 (0.53) <0.001 0.18 2.65 2.83 362.3 1.82 1.87

Leaf -0.31 (1.53) 0.84 1.53 (0.44) <0.001 0.13 2.27 3.47 219.2 1.50 1.55

SD Total 0.95 (0.62) 0.13 1.68 (0.16) <0.001 0.59 2.56 1.34 888.9 0.52 0.54

Trunk 0.78 (0.45) 0.08 1.60 (0.12) <0.001 0.71 1.12 0.65 411.0 0.38 0.39

Branch -7.83 (1.86) <0.001 3.40 (0.48) <0.001 0.39 2.56 4.61 312.3 1.57 1.62

Leaf -4.88 (1.56) <0.01 2.55 (0.40) <0.001 0.34 2.61 8.16 191.4 1.31 1.37

CV Total 0.84 (0.96) 0.38 0.46 (0.06) <0.001 0.37 1.13 0.60 1092.0 0.64 0.66

Trunk 1.62 (0.86) 0.06 0.37 (0.06) <0.001 0.33 1.18 0.64 626.3 0.57 0.59

Branch -9.32 (2.53) <0.001 1.01 (0.18) <0.001 0.30 2.81 5.14 334.6 1.68 1.73

Leaf -9.30 (1.81) <0.001 0.99 (0.13) <0.001 0.44 1.93 3.83 176.1 1.21 1.26

CD Total 1.83 (0.78) 0.02 1.14 (0.16) <0.001 0.40 1.18 0.62 1070.2 0.63 0.66

Trunk 2.47 (0.70) <0.001 0.92 (0.14) <0.001 0.34 1.16 0.65 619.3 0.57 0.59

Branch -8.63 (1.93) <0.001 2.83 (0.39) <0.001 0.40 2.33 3.57 309.7 1.56 1.62

Leaf -7.14 (1.46) <0.001 2.47 (0.27) <0.001 0.47 2.04 5.29 171.8 1.18 1.24

262

Appendix 20. Table. Allometric equations following the log-linear model for biomass components of 0.5 x 7.5 m alley rows of Inga edulis 6 months post pollarding to a height of 2 m. Where ln(x) is the log transformation of root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass. ln(y) is the log transformation of total biomass(g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), lambda (λ) and standard deviation (S.D.) as back transformation according to Marklund (1987), Furnival’s Index (F.I.) according to Kershaw et al. (2016), root mean square error of model (Model RMSE) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

ln(x) ln(y)

(Biomass) b0 (S.E.) P b1 (S.E.) P b2 R2 λ S.D. F.I. Model RMSE

CV RMSE

RCD + Height Total -0.27 (1.26) 0.83 1.64 (0.13) <0.001 0.17 (0.24) 0.71 0.69 1.14 0.57 761.1 0.44 0.46

Trunk 0.49 (1.07) 0.65 1.45 (0.11) <0.001 0.1 (0.20) 0.64 0.71 1.06 0.56 409.6 0.37 0.40

Branch -8.96 (4.14) 0.03 3.45 (0.43) <0.001 0.01 (0.8) 0.99 0.47 2.41 4.72 287.9 1.45 1.52

Leaf -10.24 (3.28) <0.01 2.64 (0.34) <0.001 0.8 (0.64) 0.21 0.48 2.03 5.63 169.1 1.15 1.20

263

Appendix 21. Table. Allometric equations following the log-linear model for biomass components of 0.5 x 7.5 m alley rows of Leucaena leucocephala 6 months post pollarding to a height of 2 m. Where ln(x) is the log transformation of the following measurements used as predictors of biomass: Root collar diameter (RCD mm, 0.05 m), diameter at breast height (DBH mm, 1.3 m), stump diameter (SD mm, 0.3 m), crown volume (CV cm3) and crown diameter (CD cm) ln(y) is the log transformation of total biomass(g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), lambda (λ) and standard deviation (S.D.) as back transformation according to Marklund (1987), Furnival’s Index (F.I.) according to Kershaw et al. (2016), root mean square error of model (Model RMSE) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

ln(x) ln(y)

(Biomass) b0 (S.E.) P b1 (S.E.) P R2 λ S.D. F.I. Model RMSE

CV RMSE

RCD Total -0.19 (0.42) 0.66 1.91 (0.11) <0.001 0.82 1.05 0.30 292.6 0.27 0.28

Trunk -0.20 (0.35) 0.57 1.87 (0.09) <0.001 0.87 1.03 0.23 203.6 0.22 0.23

Branch -7.63 (2.17) <0.001 3.17 (0.58) <0.001 0.33 2.35 3.51 98.2 1.39 1.51

Leaf -8.42 (1.59) <0.001 2.96 (0.43) <0.001 0.44 1.61 1.54 14.9 1.03 1.20

DBH Total 3.89 (0.47) <0.001 1.02 (0.15) <0.001 0.42 1.12 0.56 531.8 0.50 0.51

Trunk 3.59 (0.41) <0.001 1.06 (0.14) <0.001 0.50 1.11 0.50 398.5 0.44 0.45

Branch 0.69 (1.54) 0.65 1.17 (0.51) 0.02 0.08 2.61 4.10 115.1 1.63 1.87

Leaf -1.10 (1.20) 0.36 1.25 (0.39) 0.002 0.14 1.92 2.18 18.5 1.28 1.55

SD Total 2.39 (0.59) <0.001 1.31 (0.17) <0.001 0.50 1.15 1.17 492.9 0.46 0.52

Trunk 1.89 (0.48) <0.001 1.40 (0.14) <0.001 0.63 1.10 0.57 464.0 0.38 0.39

Branch -2.26 (2.01) 0.27 1.86 (0.57) 0.002 0.15 4.80 23.82 110.8 1.57 1.62

Leaf -4.01 (1.53) 0.01 1.91 (0.42) <0.001 0.24 1.97 2.95 17.4 1.20 1.37

CV Total 3.00 (0.54) <0.001 0.30 (0.04) <0.001 0.47 1.14 0.54 507.1 0.47 0.49

Trunk 3.39 (0.55) <0.001 0.26 (0.042) <0.001 0.39 1.12 0.55 443.3 0.49 0.50

Branch -6.24 (1.40) <0.001 0.89 (0.11) <0.001 0.48 0.62 0.93 86.4 1.23 1.27

Leaf -6.43 (1.04) <0.001 0.79 (0.08) <0.001 0.56 0.42 0.51 13.3 0.92 0.95

CD Total 2.52 (0.52) <0.001 0.99 (0.12) <0.001 0.53 1.11 0.49 475.5 0.44 0.46

Trunk 3.06 (0.57) <0.001 0.84 (0.13) <0.001 0.42 1.09 0.51 431.9 0.47 0.49

Branch -7.16 (1.43) <0.001 2.55 (0.32) <0.001 0.51 1.86 2.33 83.7 1.19 1.23

Leaf -7.67 (0.98) <0.001 2.31 (0.22) <0.001 0.65 1.34 0.97 11.9 0.82 0.85

264

Appendix 22. Table. Allometric equations following the log-linear model for biomass components of 0.5 x 7.5 m alley rows of Leucaena leucocephala 6 months post pollarding to a height of 2 m. Where ln(x) is the log transformation of root collar diameter (RCD mm, 0.05 m), or RCD and Height (cm) as predictors of biomass. ln(y) is the log transformation of total biomass(g), trunk biomass (g), branch biomass (g) or leaf biomass (g). b0 as the intercept coefficient, b1 as the slope coefficient, both reported with standard error (S.E.) and significance (P). Coefficient of determination (R2), lambda (λ) and standard deviation (S.D.) as back transformation according to Marklund (1987), Furnival’s Index (F.I.) according to Kershaw et al. (2016), root mean square error of model (Model RMSE) and results of K-fold model cross validation (k=10) reported as root mean squared error (CV RMSE).

ln(x) ln(y)

(Biomass) b0 (S.E.) P b1 (S.E.) P b2 (S.E.) P R2 λ S.D F.I. Model RMSE

CV RMSE

RCD + Height Total -3.04 (0.79) <0.001 1.69 (0.12) <0.001 0.66 (0.16) <0.001 0.86 0.98 0.26 64.33 0.24 0.25

Trunk -2.26 (0.67) 0.001 1.71 ( 0.1) <0.001 0.47 (0.13) <0.001 0.89 1.03 0.21 39.02 0.20 0.21

Branch -22.3 (4.03) <0.001 2.01 (0.59) <0.001 3.36 (0.81) <0.001 0.46 2.53 6.44 109.94 1.23 1.29

Leaf -21.53 (2.75) <0.001 1.92 (0.40) <0.001 3.01 (0.55) <0.001 0.61 1.34 1.05 10.62 0.84 0.89

265

CHAPTER V

CATTLE BROWSING IMPACTS ON THREE

SILVOPASTORAL FORAGE LEGUMES IN AMAZONIA

Abstract

A lack of knowledge of locally adapted browse species limits the development of

more sustainable livestock production systems in the Amazon. Three replicates of a

control and three stands of forage legumes arranged in three alleys of 0.5 x 7.5 m

were each directly browsed by cattle to determine the forage potential of these

species. Destructive sampling of grass and tree foliage, canopy measurements and

leaf counts prior to and post browsing were used to estimate browse removal. Post-

trial, the damage of cattle on leaves and branches of trees of each species were

measured. Soil compaction was measured between treatments and prior to and

following the cattle trial. All three species were palatable to cattle and readily eaten

when fresh. Treatments containing trees and grass produced more available forage

(>2.2 Mg ha-1) for cattle than the grass only treatment (1.5 Mg ha-1). Destructive

sampling of foliar biomass below 1.6 m demonstrated that cattle consumed 99% of

Erythrina berteroana, 75% of Inga edulis and 80% of Leucaena leucocephala. When

leaves were counted from 0.5-1.5 m height the change from pre to post-trial was a

decrease of 95% for E. berteroana, 80% for I. edulis and 87% for L. leucocephala.

Cattle browsed at an average height of 1.5 m but reached a maximum height of 2.6

266

m for E. berteroana as they pulled down branches browse. This research highlights

the value of locally adapted browse species for cattle forage and the productive

potential of silvopastoral systems in the Amazon.

1. Introduction

Cattle are traditionally viewed as grazers that consume grass and other herbaceous

plants from the ground. It is often overlooked that they are also browsers and that

shrub and tree foliage can provide an important contribution to cattle diets

(Zampaligré et al., 2013; Dechnik-Vázquez et al., 2019). A number of environmental

factors are contributing to the global decline of grassland, including increased

frequency of drought, fire and changes in atmospheric CO2 levels (Morgan et al.,

2007). There is increased impetus to find browse species that can contribute to a

diverse and resilient system for livestock production in the face of climatic change

(Murgueitio et al., 2014). Under rising temperatures the forage quality of grass is

predicted to decline, whilst simultaneously methane emissions from cattle are

predicted to increase (Lee et al., 2017). In the Bragantina region of Northern Brazil,

Hohnwald et al. (2010) demonstrated that in the humid tropics, the integration of

woody components could enhance forage production, ecological stability and

sustainability when compared to grass monocultures. Historically, shrubs and trees

have been perceived as weeds that limit grass productivity and efforts have

concentrated on their elimination (Scoones, 1995). To date, little attention has been

given to developing livestock production systems in the Amazon biome that seek to

maximise the availability of tree and shrub browse to cattle, also known as

267

silvopastoral systems (Hohnwald et al., 2016). It remains to be proven in the region

that such a system could increase the production of edible biomass over traditional

models of monoculture grass pastures.

The inclusion of trees and shrubs for browse can present advantages for livestock

production and increase the provision of ecosystem services when compared with

pasture production systems. Extensive cattle production systems in the tropics lack

shade that could serve to reduce the incidence of heat stress in cattle. Heat stress in

cattle is associated with decreased feed intake (Mader and Davis, 2004),

reproductive health (Klinedinst et al., 1993), milk yield (Parsons et al., 2001) and

fitness and longevity (King et al., 2006). Certain trees and shrubs can serve to

reduce methane emissions when ingested by cattle (Pal et al., 2015). The benefits of

trees are numerous and include enhanced nutrient cycling and soil formation, the

recovery of compacted soils, the reduction of runoff, additional habitat to support

biodiversity, increased ecological connectivity and the mitigation of climate change

via the sequestration of carbon (Fisher, 1995; Montagnini and Nair, 2004; Sharrow

and Ismail, 2004; MEA, 2005; Udawatta and Kallenbach, 2011; Mauricio et al.,

2019). Farmers may experience direct benefits of planting trees, such as the

increased longevity of livestock farms. Such benefits may reduce the need for

expansion into primary forest for new pasture (Murgueitio et al., 2011).

In the Amazon, perpetual pasture degradation and abandonment has resulted in an

expanding agricultural frontier of deforestation for the continued production of cattle

268

(Martínez and Zinck, 2004). This process is viewed as the antithesis of forest

conservation. The Amazon hosts many native trees that have shown potential for

domestication in agroforestry systems, some of which are reported to be palatable to

cattle (Serrão et al., 1996; Hohnwald, 2016). A well managed browse component

could help ameliorate the chronic degradation of pastures and even mitigate some

effects of climatic change of the Amazon (Serrão et al.,1996). Furthermore, the

degraded pastureland in the Amazon represents an opportunity for afforestation

projects that could restore landscapes to productivity and simultaneously increase

the provision of ecosystem services. A key limitation to the adoption of this approach

is the lack of studies that document the impact of cattle browsing on native plants.

The development of silvopastoral systems could be an important incentive for

landowners in the Amazon to reduce reliance on extensive grass monocultures and

the use of fire in the management of pastures for cattle production (Serrão et al.,

1996; Hohnwald, 2016). Few studies have documented the forage potential of locally

adapted species. An increased understanding of which plants are palatable to cattle

and readily eaten when fresh would contribute to the development of appropriate

silvopastoral systems. The business of cattle farming seeks to increase per hectare

livestock productivity. If the inclusion of forage trees in pastures can be shown to

increase productivity over current management then there is an economic incentive

for the adoption of silvopastoral systems. A key limitation to developing such

approaches is a lack of methodologies that quantify the productive potential of

forage for cattle from trees and shrubs. In contrast to studies that consider the total

269

biomass production of trees, this study aims to identify which proportion of the total

biomass is available to cattle via a series of direct browsing experiments, and

therefore this study attempts to quantify the productive increase, as opposed to the

total increase in biomass.

This study seeks to compare novel and existing approaches to assess the forage

potential of trees and shrubs for cattle production by quantifying the browsing impact

of cattle on trees. The study compares trees destructively harvested prior to and post

cattle browsing, with alternative, non-destructive approaches. The non-destructive

approaches include measurements of the crown and leaf counts. A range of

destructive and non-destructive methodologies have been attempted here in order to

compare the consistency of the results of each method and provide a series of

experimental approaches that can be attempted in future studies with reference to

the present study. The browsing impact on the trees is considered as the diameter of

woody material consumed or broken via cattle interactions, for three species in a

random block field trial in a pasture established in the Peruvian Amazonian region of

Madre de Dios. These approaches can be used to determine the total edible

biomass available to cattle and the amount of biomass consumed, and crucially,

evaluates the accuracy of non-destructive approaches that may be used as

methodologies in on-going field trials to measure biomass removal by cattle. Such

studies can contribute to our understanding of management and productivity of

alternative silvopastoral livestock systems.

270

It was hypothesised that:

(a) Erythrina berteroana, Inga edulis and Leucaena leucocephala are palatable to

cattle and will be consumed when fresh.

(b) Tree and grass arrangements can surpass the amount of edible biomass

than that found in monoculture Brachiaria brizantha cv. Marandú pastures.

(c) The availability of browse to cattle is limited by the height of the foliage.

2. Methods

2.1 Experimental Design and Cattle Herd

A detailed description of the experimental design and establishment is available in

Chapter II. The experiment sought to evaluate the browse potential of 24-month old

Inga edulis, Erythrina berteroana and Leucaena leucocephala. These species had

previously been shown to survive and grow at the study site (Chapter II). These

three species are considered experimental treatments. The fourth experimental

treatment was a control of pasture without trees. There were five replicate plots of

each treatment available at the beginning of this study. Three replicate plots were

selected at random for the present study due to limited resources. In total, there

were 12 experimental plots, composed of three replicates of four treatments.

Eight non-lactating Bos taurus indicus heifers of approximately equal size and body

condition were selected for the browsing experiment (Edmonson et al., 1989). The

cattle were drawn from a herd that were managed via grazing on pastures and

271

browsing a 4 km roadside strip adjacent to the property. The cattle had previous

experience of browsing several different species of trees and shrubs but were

unfamiliar with the study site. The cattle were bought onto the experimental property

and into a training corral where they had a 15-day period of acclimatization on the

Brachiaria brizantha cv. Marandú grass. An electric fence was introduced to the

corral so that the cows would become accustomed to the fence.

The experiment was carried out in three eight-day phases (Table 15). Phase I began

on the 13th February 2019. The cattle were selected in pairs and each pair was

introduced into a replicate plot. After 8 days, the pairs of cattle were moved to the

next replicate plot that had been chosen at random where they remained for eight

further days. Then the cattle were moved into the last plot where they remained for a

further eight days. Phase III of the experiment concluded on the 8th March 2019.

Table 15. Pairing combinations and random rotation of cattle (8 cattle numbered 1-8) between four treatments in three, eight day phases of a cattle trial.

I. edulis L. leucocephala E. berteroana Control Grass

Phase I 1,6 7,8 2,5 3,4 Phase II 3,4 1,6 7,8 2,5 Phase III 7,8 2,5 3,4 1,6

Two cattle were allocated per treatment, for a total of eight cattle in the four

replicates at any one time (Figure 52). A one-meter wide strip was cut around all

replicates for the installation of an electric fence and to prevent cattle from browsing

adjacent plots. A single red polyester wire electric fence was installed around each

replicate plot at a height of 0.8 m. Permanent water (provided ad libitum) and 5 x 5 m

272

shade stations were installed at the edge of each plot. The area that contained the

water and shade station was excluded from all sampling.

Figure 52. Map of the cattle trial from east to west Phase I (red), Phase II (blue) and Phase III (yellow).

2.2. Browsing Detection Using Destructive Measurements

At the beginning of each eight-day sampling period the replicate number, tree

species, date, observer and subsample codes were recorded. The root collar

diameter (RCD, measured at the widest part 0.05 m above the ground) of trees

within each plot was measured and divided into ten size classes. Six trees from each

size class were randomly selected for destructive sampling; three to be harvested

prior to the introduction of cattle, and three to be harvested after the cattle had left

the replicate. Trees were harvested up to two days before the cattle were introduced,

and were harvested up to two days after they had left. The biomass was harvested in

273

two stages to consider the effect of resource height on resource intake. It was

assumed that cattle could reach a maximum height of 1.6 m. All biomass greater

than 1.6 m from ground level was harvested separately from biomass less than 1.6

m. Biomass was classified as stem, branch or foliage, and weighed using an

appropriately sized spring-loaded weight scale.


To determine whether an effect of browsing on foliar biomass could be detected, the

total foliar biomass was compared prior to and post the cattle trial. Data were first

checked for normality using the Shapiro-Wilk test for normality and the Levene’s test

for variance. In the case of I. edulis, the data were not normally distributed but had

comparable variances. Therefore, a Mann-Whitney was used to test for differences.

In the case of E. berteroana the Levene’s test demonstrated that the variance

between the results prior to and post-trial was too great to consider using the Mann-

Whitney test, therefore a Mood’s Median test was used.

To test the hypothesis that a greater proportion of foliar biomass was consumed at a

height less than 1.6 m the ratio of foliar biomass under and over 1.6 m was

compared for each tree when sampled prior to and post the cattle trial. The

proportional change was compared for each species using a proportional test. To

assess the relative amount of tree forage that had been consumed by animals

between species, the proportion of foliar biomass below 1.6 m prior to and post the

introduction of cattle was compared using a one-way Analysis of Variance (ANOVA).

274

This established the relative amount of foliage that had been consumed by the

animals between treatments and was considered as a measure of palatability.

2.3. Browsing Detection Using Non-Destructive Measurements

Within each replicate, 30 trees were selected to represent ten size classes. Tree

height was measured from the ground to the top of the highest meristem. Leaf

emergence height (LEH) was measured from the ground to where foliage volume

assumed 5% of the total. The same measurement was called canopy height when

used in the equations presented below. A 5 m measuring tape was used to measure

crown diameter (CD), taken as the mean of three horizontal measurements of the

crown perpendicular to the ground.

The crown radius (r) was calculated as the crown diameter divided by two. Crown

depth (CD) was derived using equation 1. Crown area (CA) approximates a circle

and was derived using equation 2. Crown volume (CV) was assumed to equate to a

cylinder, the best fit for the pollarded trees, and was calculated using equation 3:

(1) Crown Depth = Height – Canopy Height

(2) CA = πr2

(3) CV =πr2CD


275

To test for an impact of the cattle browsing on crown volume, the crown volumes for

each tree measured prior to and post the cattle trial were compared within a species.

The crown volume data were checked for normality using the Shapiro-Wilk test.

Crown volume was not normally distributed and therefore the homogeneity of

variances was tested using Levene’s test. The crown volumes of both I. edulis and L.

leucocephala demonstrated a homogeneous variance and therefore the Mann

Whitney U test was used with the crown volume as a continuous response variable

and prior to or post cattle trial as a categorical predictor. For E. berteroana the

Levene’s test showed that the variance within the data were not homogeneously

distributed. Therefore, the decision was taken to use the Kruskal-Wallis test to

determine if there was an effect of cattle browsing on crown volume.

To determine whether cattle had an effect on the height at which leaves were

present, the continuous response variable of LEH was tested for normality using the

Shapiro-Wilk test for each species. The LEH of both I. edulis and E. berteroana were

normally distributed and therefore paired t-tests were used to compare for each

species the LEH prior to and post the cattle trial. The LEH data for L. leucocephala

were not normally distributed and therefore a Levene’s test was employed to confirm

homogeneity of the variance. The canopy height data for L. leucocephala were of

equal variance and therefore a Mann Whitney U test was used test for an effect of

cattle browsing on LEH.

276

2.4 Browsing Detection Using Leaf Counts

Leaves were counted prior to and post the cattle trial to detect an effect of browsing

on the number of leaves. Within each species 60 branches were chosen at random

from 30 trees per replicate. Two sections of the branch were chosen. The first

section was 1 m long, and was measured from 0.5 m from the ground. Assuming the

branches were vertical, this can be described as a section 0.5 m to approximately

1.5 m above the ground. The second section was 0.5 m long and measured from 1.5

m to approximately 2 m above the ground. The replicates were exposed to cattle

within two days of sampling, and the sampling procedure was repeated the day after

the cattle left the replicates. The difference in the number of leaves prior to and post-

trial was assumed to be due to utilization by cattle.


The total number of leaves on a tree was a continuous variable. A Shapiro-Wilk test

was used to test for normality and the distribution of the variable was found to be

non-normal. A non-parametric test, the Wilcoxon signed rank test was used to

compare the total number of leaves prior to and post the cattle trial for each species.

2.5 Measurement Of Browsing Impact

Immediately following each eight-day cattle trial, 30 trees in every replicate were

visually assessed for signs of browsing or cattle impact. The height of the impact

was recorded and categorised as branch or leaf. Two damage types were reported,

either browsed or broken. The diameter of the impact was recorded with callipers.

277


The Shapiro-Wilk test was used to check the normality of the distribution of the

continuous response variables: the diameter of point of leaf browsing, diameter at

point of branch browsing, diameter of branch broken, height of leaf browsed and the

height of branch browsed. The height of leaf browsing was found to have a normal

distribution therefore a one-way ANOVA was used to determine if there was a

difference between the heights of the browsing impact between the different tree

species. Tukey’s Honest Significant Differences test was applied post-hoc to

determine which species were different. Shapiro-Wilk tests indicated the following

response variables were not normally distributed: diameter at point of branch

browsing, diameter of branch broken and the height of branch browsed. Non-

parametric tests were used for these response variables. The Kruskal-Wallis test

was applied, using species as a categorical predictor. When the result was found to

be significant, Dunn’s test was used to determine which species were different.

Dunn’s test is considered an appropriate test for groups with unequal numbers of

observations (Zar, 2010). Mortality as a result of the cattle trial was calculated as the

percentage of total trees that had died as a result of the trial.

2.6 Estimation Of Grass Forage

Prior and post-feeding trial, all ground biomass was harvested within five randomly

located 1 m2 quadrats. Sampling prior to the feeding trial subjectively considered

biomass as either an ‘edible’ or ‘nonedible’ fraction. ‘Edible fraction’ was visually

278

assessed as the leaf or stem (< 25 mm) component of the grass no less than 10 cm

from the ground. This fraction was considered to be the ‘leaf’ part of the grass and

the part most commonly favoured by cattle. The inedible fraction was the remaining

grass biomass, which was considered to be the ‘woody stem’ part of the grass and of

lower grazing value to the cattle. Each fraction was weighed separately as fresh

weight. Three subsamples were taken of each fraction and dried to a constant

weight. Post feeding trial there was no ‘edible’ fraction available to sample, therefore

sampling consisted solely of the non-edible fraction.

2.7 Effect Of Trees and Cattle on Soil Bulk Density

To determine the effect of trees on soil bulk density, samples were taken from either

within one meter of the tree line or between the rows of trees to a depth of either 0 –

10 cm or 10 – 20 cm. Five samples were taken from each of these categories, for a

total of 20 samples per replicate. A cylindrical steel corer of 5 x 9 cm was used to

give a volume of 318 cm3. The core was driven vertically into the soil and excavated

with a trowel. Once removed, the soil core was transferred to a paper bag. The core

was oven dried at 55 °C until a constant dry weight was reached and this value was

reported. Bulk density was calculated as the weight of the sample divided by the

volume of the sample and is reported in g cm3 (Swift and Bignell, 2001). To

determine the effect of cattle on soil bulk density the aforementioned sampling

procedure was repeated following the introduction of cattle. In control replicates with

no trees, sample sites were chosen randomly and no closer than 10 m to an edge of

each plot.

279


The effect of trees on soil bulk density was determined within treatments by

comparing the bulk density within the tree line with the bulk density between the

rows of trees at the 0 – 10 cm or 10 – 20 cm depth profile. The bulk density was the

continuous response variable to the categorical treatment of species. Within each

species and depth, the bulk density was tested for normality by using Shapiro-Wilk

tests. The data were found to be normally distributed. Paired t-tests were used for

each species and depth to compare the compaction of the soils within the tree line

with the compaction of the soils between the rows of trees.

Subsequent to the cattle trial the sampling methodology was repeated to test for an

effect of cattle on soil compaction. Using each treatment as a predictor variable, the

bulk density post-cattle trial at two depths was tested for normality by using Shapiro-

Wilk tests. Where the data were normally distributed, paired t-tests were used to

compare the bulk density prior to or post the introduction of cattle. The data were not

normally distributed for L. leucocephala at 5-10 cm depth, or for E. berteroana at

either the 0-10 cm or 10-20 cm depths. In these cases Kruskal-Wallis tests were

used. If an effect of cattle on soil bulk density was found, it was explored via t-tests

to determine whether there was an effect of trees on soil bulk density by comparing

the soil bulk density from within the tree line with that taken from between the rows of

trees.

280

3. Results

Cattle browsing reduced the amount of available foliage for all species (Figures 53-

56). This effect was detected by all methodologies. When foliar biomass was

destructively sampled below 1.6 m, the results show foliar biomass reductions for all

three tree species following the introduction of the cattle. A greater proportion of

biomass was lost below 1.6 m than over 1.6 m for E. berteroana (P< 0.05), I. edulis

(P <0.01) and L. leucocephala (P< 0.01). Destructively sampled L. leucocephala

showed an increase in total foliar biomass of 20% (P<0.01) post-browsing; this result

is discussed further in the discussion section. When foliar biomass below 1.6 m was

compared prior to and post the cattle trial, E. berteroana showed a 99% decrease, I.

edulis a 75% decrease and L. leucocephala an 80% decrease (Table 16). When the

change in foliar biomass below 1.6 m was compared prior to and post cattle trial

there was no effect of species detected. When total biomass (i.e. below and above

1.6 m) was considered there was a reduction in total foliar biomass of E. berteroana

of 92% (P<0.01) and of I. edulis by 30% (P<0.05).

Table 16. Dry foliar biomass for individual trees (dry matter g tree-1), standard deviation (S.D.) and number of individuals sampled (n) prior to and post cattle browsing either mean total tree foliar biomass or foliar biomass harvested <1.6 m on potential browse species Erythrina berteroana, Inga edulis and Leucaena leucocephala. Significant differences (P<0.05) from prior value to post value following cattle browsing denoted by an asterisk (*).

Prior or Post Foliar biomass Foliar biomass <1.6 m

Species Cattle Trial n g tree-1 S.D. g tree-1 S.D.

E. berteroana Prior 87 45.3 66.2 15.5 66.2

Post 90 3.8* 11.8 0.2* 11.8

I. edulis Prior 88 297.1 347.7 141.7 347.7

Post 90 204.5* 348.2 34.9* 348.2

L. leucocephala Prior 87 60.4 193.0 6.6 193.0

281

Post 90 72.3* 277.9 1.3* 277.9

Figure 53. Cattle browsing Inga edulis

282

Figure 54. Browsing line where cattle consumed all the foliage they could reach.

Figure 55. Post-browsing, an Erythrina berteroana row is missing several trees

283

Figure 56. Leucaena leucocephala defoliated on low branches

E. berteroana demonstrated the greatest loss of crown volume due to browsing, a

67% decrease (P <0.01), followed by I. edulis (51% decrease, P <0. 01) (Table 17).

For the species L. leucocephala there was a 31% reduction but this was a

statistically weaker effect of browsing on crown volume (P = 0.06). Post browsing,

the leaf emergence height increased in E. berteroana by 49% (P<0.01), in I. edulis

by 50% (P <0.001) and in L. leucocephala by 20% (P <0.01) during the browsing

period.

284

Table 17. Mean average (Av) and standard deviation (S.D.) of change per individual tree in crown volume (M3) and leaf emergence height (LEH, cm) prior to and post a cattle browsing experiment on individuals (n) of potential browse species, Erythrina berteroana, Inga edulis and Leucaena leucocephala. Significant differences between sampling period prior to and post (P<0.05) denoted by an asterix (*).

Species

Prior to cattle trial Post cattle trial

Parameter Av S.D. n Av SD n

Crown Volume (M3) E. berteroana 0.94 1.15 88 0.31* 0.51 50

I. edulis 2.01 2.05 48 1.02* 1.65 46

L. leucocephala 1.54 2.6 83 1.06 2.05 75

LEH (cm) E. berteroana 112.6 40.1 88 167.8* 74.2 50

I. edulis 101.6 33.3 48 152.5* 35.1 46

L. leucocephala 191 45.9 83 230.7* 53.5 75

The mean number of leaves decreased among all species as a result of cattle

browsing (Table 18). E. berteroana lost the greatest proportion of leaves in both the

0.5 – 1.5 m and the 1.5 – 2.0 m category (95% and 91% respectively). I. edulis and

L. leucocephala both lost more leaves under 1.5 m (80%, 87%) than over 1.5 m

(48%, 80%). When the methodologies used to measure browsing impact prior to and

post cattle trial are compared between species they show comparable results

between species, as shown for E. berteroana in Figure 57, I. edulis in Figure 58 and

L. leucocephala in Figure 59.

285

Table 18. Mean leaf count per tree (number of leaves tree-1) and standard deviation (S.D.) prior to and post a cattle-browsing trial on individuals (n) of potential browse species Erythrina berteroana, Inga edulis and Leucaena leucocephala. Significant differences (P<0.05) between leaves counted prior to and post browsing are denoted by an asterix (*).

Sampling Range Species

Prior/ Post n

Number of leaves tree-1 S.D.

Mean reduction (%)

1.5 m - 2 m E. berteroana Prior 65 26.3 18.6

Post 65 1.4* 3.5 -95

I. edulis Prior 70 17.6 18.8

Post 70 9.2* 14.4 -48

L. leucocephala Prior 37 16.1 14.4

Post 37 3.9* 7.2 -76

0.5 m - 1.5 m E. berteroana Prior 132 29.3 23.6

Post 132 2.6* 4.6 -91

I. edulis Prior 163 24.6 23.2

Post 163 5.0* 6.6 -80

L. leucocephala Prior 99 16.2 13.6

Post 99 2.1* 5.0 -87

Figure 57. Percentage change of Erythrina berteroana biomass for four assessment methodologies used to detect biomass removal following cattle browsing.

-100

-80

-60

-40

-20

0

20

40

60

80

100

Destructive <160cm

Crown Volume Leaf Count 50-150 cm

Leaf EmergenceHeight

Pro

po

rtio

na

l ch

an

ge

286

Figure 58. Percentage change of Inga edulis biomass components between methodologies used to detect biomass removal following a cattle-browsing experiment.

Figure 59. Percentage change of Leucaena leucocephala biomass components between methodologies used to detect biomass removal following a cattle-browsing experiment.

Browsing damage was recorded on branches for all species (Table 19). The

browsed branches of E. berteroana showed an increased diameter compared with

-120

-100

-80

-60

-40

-20

0

20

40

60

Destructive <160cm



Pro

po

rtio

na

l ch

an

ge

-120

-100

-80

-60

-40

-20

0

20

40

Destructive <160cm



Pro

po

rtio

na

l ch

an

ge

287

other species. The branches of E. berteroana had a 35% greater diameter than I.

edulis and 65% greater than L. leucocephala. There was no effect of species on

maximum height of branches browsed. E. berteroana had the highest rate of

mortality when measured immediately post browsing (12.5%) that was six times

greater than that of the species with the lowest mortality, I. edulis (2.1%). L.

leucocephala had a mortality of 4.8%.

Table 19. Impacts measured following a cattle trial on three potential browse species Erythrina berteroana, Inga edulis and Leucaena leucocephala. Impact types were classified as browsing-type damage (Browsed) or breakage-type damage (Broken) of two categories, branches (all species) and leaves (I. edulis). Average diameter (mm), average height (cm) and maximum height (cm) and respective standard deviations (S.D.) reported on a number (n) of individual trees. Significant differences between species (P<0.05) denoted by asterisk (*).

Impact Category

Impact Type

Diameter (mm) Height (cm)

Species Average S.D. Average Max S.D. n

E. berteroana Branch Broken 12.8* 7.2 153 240 41.7 126

Branch Browsed 5.0* 1.8 156 256 42.7 222

I. edulis Branch Broken 9.0* 4.7 134 195 37.0 39

Branch Browsed 3.9* 1.4 150 246 36.3 181

Leaf Browsed 2.5 0.7 144 197 55.3 245

L. leucocephala Branch Broken 6.5* 3.8 135 260 44.1 41

Branch Browsed 2.8* 0.9 154 225 36.2 113

There was no effect of treatment on the amount of edible or non-edible biomass

available prior to the cattle trial (Table 20). When grass biomass was compared prior

to and post the cattle trial, 100% of the edible fraction was consumed in all

treatments. If the edible fraction of grass in each treatment was combined with the

results of the destructive foliage estimates <1.6 m as a measure of total available

edible foliage, the no-tree control treatment would produce 1.5 Mg ha-1 less available

edible forage than the tree treatments of E. berteroana (2.3 Mg ha-1), I. edulis (2.3

Mg ha-1) or L. leucocephala (2.2 Mg ha-1).

289

Table 20. Edible and non-edible fractions of mean average dry weight grass biomass (Mg ha-1) produced in four treatments, reported for sampling carried out prior to and post the introduction of cattle.

Prior (n = 15) Post (n = 15)

Treatment Fraction Average (Mg ha-1) S.D.

Average (Mg ha-1) SD

Control Edible 1.5 0.6 0

Non-edible 30.2 9.7 38.8 9.1

E. berteroana Edible 2.3 0.8 0

Non-edible 29.4 10.9 29.8 10.0

I. edulis Edible 1.9 0.5 0

Non-edible 22.7 8.5 30.7 11.8

L. leucocephala Edible 2.2 0.9 0

Non-edible 30.2 12.9 31.4 7.8

Prior to the cattle trial there were no differences in the soil bulk density within the line

of trees or between the rows of trees of any of the species tested. Following the

cattle trial, the soils of the I. edulis replicates had become compacted at both the 0-

10 cm (P <0.01) and the 10-20 cm horizon (P <0.01). This effect was found to occur

within the tree line at both the 0-10 cm (P <0.05) and 10-20 cm horizon (P <0.01),

and between the rows of trees at the 10-20 cm horizon (P <0.05). No effect of the

introduction of cattle on soil compaction was found in any of the other treatments.

4. Discussion

This study sought to identify the forage potential of three tree species. The results

show that the three tree species are palatable when fresh and readily eaten by

cattle. When the biomass of traditional monoculture was compared with the available

biomass of the silvopastoral system the results showed that the inclusion of these

tree species would increase the forage available to cattle. Therefore, the

290

silvopastoral system represents an increase in per hectare productivity and has the

potential to support more cattle per hectare than the pasture-only conventional

system. The reported increase in production per hectare is a conservative estimate

as this figure only considers the foliar biomass up to a height of 1.6 m. Over the

course of the experiment, it was found that cattle were able to access browse up to a

height of 2.6 m. The estimation only considered foliar biomass as the sole

component of edible biomass. In practice, it was found that cattle also browsed

branches. The trees therefore provided a greater amount of edible biomass to the

cattle than that reported in this study.

Destructive sampling was able to describe and characterise the impact of browsing

on the foliar biomass of all species in this experiment. Although samples were drawn

from the same population there was a degree of natural deviation between the

samples of trees destructively harvested prior to and post the introduction of cattle.

An example of this in the present study was that destructive sampling detected an

increase in total foliar biomass for L. leucocephala. This result is a methodological

artefact and implies that a sample size must be chosen that is proportional to the

variation within the population and the predicted browsing impact. Often this is

difficult to establish in advance and the present study detected an increase in

biomass despite a reasonable sample size. A drawback to this approach is that there

is a high input of labour for the results.

291

Destructive sampling of foliar biomass does not have to destroy the tree. It would be

possible to strip the trees of foliar biomass prior to an experiment and compare it

with the foliar biomass harvested subsequent to a browsing experiment. This is also

known as simulated browsing. The advantage of this is the tree can be measured

over time, a possibility that is removed when the tree is destructively harvested. A

disadvantage is that simulated browsing may not accurately represent the height of

browsing which may be greater or lower than expected, or the animal’s interaction

with the browse, for example, an animal’s use of woody branches or ability to

manipulate branches by pulling on them. Destructive sampling may be a necessary

approach to determine the total biomass of trees and this approach is often used to

calibrate or evaluate alternative, non-destructive techniques (Barnes, 1976).

Sanon et al. (2007) found that the mean height reached by browsing cattle was 1.47

m and the maximum was 1.9 m. Schoenbaum et al. (2018) found in Israel that cattle

(described as local breed x Hereford) could reach a maximum grazing height of 2.2

m. Ouédraogo-Koné et al. (2006) found that cattle could browse to a height of 1.1 m.

In the present study, cattle were found to browse on average between 1.3 and 1.6 m.

However, the maximum heights reported to have reached cattle in this study

exceeded 2.5 m. It was only possible to identify the maximum diameter of leaf

browsing impact in the species I. edulis due to partial consumption of the leaf by

cattle. When cattle browsed E. berteroana or L. leucocephala leaves the entire leaf

was removed. Therefore, there was no visible post-trial impact to record. In

anticipation of the effect of browsing height on resource availability, the destructive

292

sampling had considered all above ground biomass in two height categories, over

1.6 m and below 1.6 m. A greater proportion of biomass was consumed below 1.6 m.

If the total amount of foliar biomass were considered as the estimation of total

available forage, the results would have underestimated the proportion of available

forage consumed. If the resource is to be directly browsed a study must seek to

establish the effect of resource height on direct browsing by the animal as this will

influence estimations of the availability of forage (Renaud et al., 2003). This is

because when large trees are considered, only a fraction of the total foliar biomass is

actually within reach and therefore available to the cattle. Alternatively, forage

systems such as cut and carry or lopping systems may aim to determine the edible

fraction of biomass available to animals with little regard for the browsing height

(Balehegn et al., 2012).

The non-destructive approach to measuring a browsing impact using crown volume

prior to and post cattle trial is limited by the fact that crown volume does not consider

whether biomass is within reach of the cattle and an effect of light browsing may not

be detected. The impact of browsing was detectable in the present study,

presumably due to the trees having a low crown due to pollarding. Trees in cattle

pastures may exhibit browse lines; this term describes a visible line of defoliation

caused by cattle consuming all foliage within their reach. This study introduces the

concept of leaf emergence height (LEH), which was conceptualised as attempt to

capture the occurrence and height of browse lines on the tree species in the study.

The LEH increased for all species post cattle trial, which suggests the trees were

293

consumed by the cattle. However, the LEH does not indicate how much forage was

consumed. Furthermore, trees that may offer some supplemental forage value but

are not palatable enough to be completely defoliated may be overlooked by this

technique. The LEH could be of use when determining an appropriate pollard height

for species.

The number of leaves lost as a result of the cattle trial was strikingly similar to the

results of the destructive harvest of trees below 1.6 m. The methodology similar to

that described by Rees (1974). The present study initially planned to employ the leaf

counting technique only in the 0.5-1.5 m range. However, after the first round of the

cattle trial it became apparent that in certain instances the cattle were browsing

above this range. Therefore, an additional category of 1.5 – 2 m was specified. A

greater proportion of leaves were lost in the 0.5 – 1.5 m range for all species, as this

range was the easiest for cattle to reach. Rees (1974) did not establish a uniform

height for sampling within the experiment. This is possibly due to different

management approaches between their study and this study. In Rees’ (1974) study

the plants were pollarded to 1 m height in the season in which the study was carried

out and it can be assumed that all tree forage was accessible to cattle.

The specific management of this silvopastoral system is related to the quantities of

biomass produced by both the tree and grass elements. For instance, this study was

limited to one planting arrangement of 0.5 x 7.5 m, equating to 2667 trees ha-1. Six

months prior to the cattle trial, all trees had been pollarded to 2 m height to reduce

294

competition with grass (Chapter III). It was found that the pollarding reduced the

negative effects of trees on the grass (presumed to be due to light competition).

However, the pollarding may also have impacted the tree’s production and

distribution of foliar biomass and these effects were not studied. The browsing

resulted in a substantial reduction in foliage on the trees, and this may have

increased the light available to the grass. Further work is needed to understand how

the tree and grass elements interact in silvopastoral systems. For example, the

planting arrangement, the selection of tree and grass companion species, the use of

fertilisers and pruning or pollarding management can all effect the production per

hectare and were not evaluated in the present experiment.

A limitation of this study was that it did not consider how the availability of grass

affected the practice of tree browsing. While it was observed that the trees were

consumed at the same time as the grass there was no data collected to suggest that

consumption of browse was as a result of the depletion of the edible grass biomass.

The results of this study indicate a substantial portion of inedible biomass. The

Brachiaria brizantha cv. Marandú pasture had been sown by hand in March 2017

and it was first grazed 24 months later at the time of the present experiment. This

meant that a large fraction of the grass biomass had become inedible, as the mature

stems of grass are fibrous and unpalatable to cattle. This study employed a

subjective estimation of edible biomass fraction. The total edible fraction was

completely devoured by the completion of the trial, which suggests the methodology

underestimated the total edible fraction of grass. Future studies may consider

295

simulating the grazing of the grass by repeatedly cutting the grass, an approach that

was beyond the resources of this study.

The foliage of all tree species was consumed by cattle and there were no differences

detected between species in the amount of foliage consumed. This suggests that all

species were palatable and that relative palatability was not detected in the present

study. I. edulis produced the most overall biomass and the most accessible and

edible foliage for the cattle of all the tested species. The results of this study confirm

that I. edulis is palatable to cattle and readily eaten when available. It observed that

the cattle preferred to be in the shade of I. edulis than in the sun. I. edulis

experienced branch browsing and damage to branches as a result of cattle

interaction. However, I. edulis demonstrated the lowest immediate mortality following

cattle interaction. This may be because stems of I. edulis were of an increased

diameter when compared with those of E. berteroana or L. leucocephala (Chapter

III). It is possible that stem size and mortality following the cattle trial were related.

While beyond the scope of the present study, this could be of importance in

determining an appropriate stage at which silvopastoral systems could be exposed

to cattle.

Hohnwald et al., (2015) reported that I. edulis was a high producer of woody and

leafy biomass, a frequently occurring tree in Capoeira, a nitrogen fixing legume and

palatable for cattle. Hohnwald (2016) reported that I. edulis had ‘bad performance’

on a scale of palatability and biomass, which they described as a ‘positive value’ for

296

forage quality and ‘intermediate’ values for mortality. Hohnwald (2016) found that I.

edulis suffered from the overgrown grass sward, but when established it was robust.

He also mentions that the small trees may have not been obvious to browsing cattle

amongst the tall grass and may have been overlooked, ultimately resulting in a

negative palatability value. The present study differs from Hohnwald (2016) in the

experimental design, establishment and management of trees and finds contrary to

Hohnwald (2016) that I. edulis is a suitable candidate for further investigation for

forage potential. Dechnik-Vázquez et al. (2019) rated I. edulis 11th of 268 species on

the amount of times the species appeared browsed. They found tannin content of

0.31% but no presence of toxic cyanogenic glycosides.

E. berteroana consistently exhibited the greatest proportion of foliar biomass lost, the

greatest diameter of branches consumed and broken and the highest mortality

following the cattle trial when compared with the other species. Unforeseen in the

experimental design was that cattle would employ unique browsing strategies for

each species. It was observed that the cattle displayed a browsing habit for E.

berteroana of pulling down branches on which to feed. The branches would

frequently break during this process and often sections of the broken branches

would also be chewed and consumed. Post-trial, almost all the branches of E.

berteroana trees had been broken, in some cases entire plants were missing,

presumed destroyed by the cattle. This phenomenon was also reflected in the post-

impact monitoring, where E. berteroana reported the highest average and maximum

297

point of branch browsing, and the greatest increase in leaf emergence height

following the experiment.

This study was limited to the planting of 1 m stakes of E. berteroana. This may have

contributed to a substantially different growth form from than that observed when E.

berteroana is grown from seed, where it tends to form a more robust and resistant

tree (personal observation). The 1 m stakes produced adventitious shoots formed

from the callus tissue between the bark and the wood of the cut surface that may

have resulted in weaker branches than a natural growth form of E. berteroana. This

may have made the trees of this species susceptible to cattle damage. Future

studies could consider propagation from 15 cm cuttings or from seed.

Of the three tree species, L. leucocephala produced the least available foliar

biomass. It was observed that this was due to L. leucocephala growing out of reach

of the cattle as the pollard at 2 m, which favoured the production of adventitious buds

that grew at a height of 2 m. This resulted in a minimal contribution to available

forage for cattle. If the tree is produced with the intention of being directly browsed

by cattle, a pollarding height under 2 m may be considered to encourage leaf

production at accessible heights. Guevarra et al. (1978) compared three harvest

frequencies of 2.5, 3 and 4 months, three planting arrangements of 15 x 30 cm, 30 x

50 cm or 45 x 50 cm and two accessions of L. leucocephala in North Kohala, Hawaii.

They found that for optimal forage production, a planting arrangement of 15 x 50 cm

and pollarding when the plants were approximately 1 m in height were desirable

298

management practices to increase forage yield. The present study reports the ability

of this species to grow in the Amazon in association with B. brizantha cv. Marandú.

L. leucocephala has been widely documented as a tropical forage tree but it has not

yet been reported to grow successfully in the Amazon. Szott et al. (1991) speculated

that L. leucocephala failed to establish due to high levels of exchangeable aluminium

at their experimental site in Yurimaguas, Peru. L. leucocephala is considered a

highly invasive tree (Richardson, 1998). However, no incidences of natural

regeneration were observed in the invasive B. brizantha cv. Marandú pasture.

Prior to the cattle trial there were no within species differences in soil bulk density

between or within the rows of trees. This may be due to the relatively undisturbed

nature of the site at the beginning of the study, as cattle had not been present on the

land for 30 years and no machinery was used in the preparation of the experiment.

After the cattle trial the soils had a higher bulk density in the I. edulis parcels. This

corresponds with the observation that the cattle were spending their time in the

shade of I. edulis.

Research has shown that in traditional pasture models of the Amazon the highest

nutrient concentrations occur in the most utilized and therefore compacted areas, for

example, surrounding the water trough. This may result in a net loss of nutrients over

time as the land’s nutrients are essentially exported and subsequently leached

(Boddey et al., 2004). If the cattle are spending time in the shade of I. edulis, these

nutrients may be circulated back into the tree and become available to the animals

299

via consumption of the foliage. Furthermore, over 5 years the roots of I. edulis have

been shown to reverse the compaction of soils (Fisher, 1995). A relevant extension

of this work would be to attempt to capture the effects of on-going compaction by

cattle on growth and mortality of I. edulis. For example, planting trees in alley rows

may result in different patterns of cattle movement and behaviour, which may

influence the distribution of soil compaction and nutrients across pastures. A logical

extension of the present work is a critical assessment of the chemical composition of

these foliages and their impact on animal health.

In conclusion, the inclusion of trees into pastures can provide an increase in

available forage for cattle. There are multiple techniques, both destructive and non-

destructive that can be used to estimate the utilisation of browse plants by cattle.

Cattle are generalist browsers and may include leaves and twigs into their diet.

Where resource availability is limited by resource height, cattle may find ways to

manipulate browse trees to access forage out of their reach. Locally adapted species

may be able to provide palatable forage for cattle. Further work is needed to

understand the impacts of these forage species on animal health.

References

Balehegn, M., Eniang, E.A., Hassen, A., 2012. Estimation of browse

biomass of Ficus thonningii, an indigenous multipurpose fodder tree in

northern Ethiopia. African Journal of Range & Forage Science.

https://doi.org/10.2989/10220119.2012.687071

https://doi.org/10.2989/10220119.2012.687071

300

Barnes, D.L., 1976. A review of plant‐based methods of estimating food

consumption, percentage utilisation, species preferences and feeding

patterns of grazing and browsing animals. Proceedings of the Annual

Congresses of the Grassland Society of Southern Africa 11, 65–71.

https://doi.org/10.1080/00725560.1976.9648781

Boddey, R.M., MacEdo, R., Tarré, R.M., Ferreira, E., De Oliveira,

O.C., De, C., Cantarutti, R.B., Pereira, J.M., Alves, B.J.R.,

Urquiaga, S., 2004. Nitrogen cycling in Brachiaria pastures: The key

to understanding the process of pasture decline. Agriculture,

Ecosystems and Environment 103, 389–403.


Dechnik-Vázquez, Y.A., García-Barrios, L., Ramírez-Marcial, N., van

Noordwijk, M., Alayón-Gamboa, A., 2019. Assessment of browsed plants

in a sub-tropical forest frontier by means of fuzzy inference. Journal

of Environmental Management 236, 163–181.


Edmonson, A.J., Lean, I.J., Weaver, L.D., Farver, T., Webster, G.,

1989. A Body Condition Scoring Chart for Holstein Dairy Cows. Journal

of Dairy Science 72, 68–78.

https://doi.org/10.1080/00725560.1976.9648781



301

https://doi.org/10.3168/JDS.S0022-0302(89)79081-0

Fisher, R.F., 1995. Amelioration of Degraded Rain Forest Soils by

Plantations of Native Trees. Soil Science Society of America Journal

59, 544. https://doi.org/10.2136/sssaj1995.03615995005900020039x

Guevarra, A.B., Whitney, A.S., Thompson, J.R., 1978. Influence of

Intra-Row Spacing and Cutting Regimes on the Growth and Yield of

Leucaena1. Agronomy Journal 70, 1033.


Hohnwald, S., Rischkowsky, B., King, J.M., Camarão, A.P., Rodrigues

Filho, J.A., Zeppenfeld, T., 2015. Intensive cattle browsing did not

prevent fallow recuperation on smallholder grass-capoeira pastures in

the NE-Amazon. Agroforestry Systems 89, 813–828.

https://doi.org/10.1007/s10457-015-9815-9

Hohnwald, S., Trautwein, J., Camarão, A.P., Wollny, C.B., 2016.

Relative palatability and growth performance of capoeira species as

supplementary forages in the NE-Amazon. Agriculture, Ecosystems &

Environment 218, 107–115. https://doi.org/10.1016/J.AGEE.2015.11.015

King, J.M., Parsons, D.J., Turnpenny, J.R., Nyangaga, J., Bakari, P.,

https://doi.org/10.3168/JDS.S0022-0302(89)79081-0



https://doi.org/10.1007/s10457-015-9815-9


302

Wathes, C.M., 2006. Modelling energy metabolism of Friesians in Kenya

smallholdings shows how heat stress and energy deficit constrain milk

yield and cow replacement rate. Animal Science 82, 705–716.

https://doi.org/10.1079/ASC200689

Klinedinst, P.L., Wilhite, D.A., Hahn, G.L., Hubbard, K.G., 1993. The

potential effects of climate change on summer season dairy cattle milk

production and reproduction. Climatic Change 23, 21–36.

https://doi.org/10.1007/BF01092679





Mader, T.L., Davis, M.S., 2004. Effect of management strategies on

reducing heat stress of feedlot cattle: Feed and water intake.

Journal of Animal Science 82, 3077–3087.

https://doi.org/10.2527/2004.82103077x

Martínez, L.J., Zinck, J.A., 2004. Temporal variation of soil

compaction and deterioration of soil quality in pasture areas of

Colombian Amazonia. Soil and Tillage Research 75, 3–17.

https://doi.org/10.1079/ASC200689

https://doi.org/10.1007/BF01092679


https://doi.org/10.2527/2004.82103077x

303


Morgan, J.A., Milchunas, D.G., LeCain, D.R., West, M., Mosier, A.R.,

2007. Carbon dioxide enrichment alters plant community structure and

accelerates shrub growth in the shortgrass steppe. Proceedings of the

National Academy of Sciences 104, 14724–14729.


Murgueitio, E., Calle, Z., Uribe, F., Calle, A., Solorio, B., 2011.




Murgueitio, E.R., Chará, J.O., Cuartas, C.C., Naranjo, J.R., Barahona,

R.R., 2014. Intensive Silvopastoral Systems (ISPS), mitigation and

adaptation tool to climate change. Tropical and Subtropical

Agroecosystems 17, 501–507.

Ouédraogo-Koné, S., Kaboré-Zoungrana, C.Y., Ledin, I., 2006. Behaviour

of goats, sheep and cattle on natural pasture in the sub-humid zone of

West Africa. Livestock Science 105, 244–252.

https://doi.org/10.1016/J.LIVSCI.2006.06.010





304

Pal, K., Patra, A.K., Sahoo, A., Kumawat, P.K., 2015. Evaluation of

several tropical tree leaves for methane production potential,

degradability and rumen fermentation in vitro. Livestock Science 180,

98–105. https://doi.org/10.1016/J.LIVSCI.2015.07.011

Parsons, D.J., Armstrong, A.C., Turnpenny, J.R., Matthews, A.M.,

Cooper, K., Clark, J.A., 2001. Integrated models of livestock systems

for climate change studies. 1. Grazing systems. Global Change Biology

7, 93–112. https://doi.org/10.1046/j.1365-2486.2001.00392.x

Rees, W.A., 1974. Preliminary Studies into Bush Utilization by Cattle

in Zambia, Journal of Applied Ecology 11 (1), 207-214

Renaud, P.C., Verheyden-Tixier, H., Dumont, B., 2003. Damage to

saplings by red deer (Cervus elaphus): effect of foliage height and

structure. Forest Ecology and Management 181, 31–37.

https://doi.org/10.1016/S0378-1127(03)00126-9

Richardson, D.M., 1998. Forestry Trees as Invasive Aliens Article in

Conservation Biology. https://doi.org/10.1046/j.1523-1739.1998.96392.x

Sanon, H.O., Kaboré-Zoungrana, C., Ledin, I., 2007. Behaviour of

goats, sheep and cattle and their selection of browse species on


https://doi.org/10.1046/j.1365-2486.2001.00392.x

https://doi.org/10.1016/S0378-1127(03)00126-9

https://doi.org/10.1046/j.1523-1739.1998.96392.x

305

natural pasture in a Sahelian area. Small Ruminant Research 67, 64–74.

https://doi.org/10.1016/J.SMALLRUMRES.2005.09.025

Schoenbaum, I., Henkin, Z., Yehuda, Y., Voet, H., Kigel, J., 2018.

Cattle foraging in Mediterranean oak woodlands – Effects of management

practices on the woody vegetation. Forest Ecology and Management

419-420, 160–169. https://doi.org/10.1016/J.FORECO.2018.03.017

Scoones, I., 1995. Exploiting heterogeneity:habitat use by cattle in

dryland Zimbabwe. Journal of Arid Environments 29, 221–237.

https://doi.org/10.1016/S0140-1963(05)80092-8




https://doi.org/10.1016/0921-8009(95)00092-5





Szott, L.T., Palm, C.A., Sanchez, P.A., 1991. Agroforestry In Acid

https://doi.org/10.1016/J.SMALLRUMRES.2005.09.025


https://doi.org/10.1016/S0140-1963(05)80092-8

https://doi.org/10.1016/0921-8009(95)00092-5


306

Soils Of The Humid Tropics. Advances in Agronomy 45, 275–301.

https://doi.org/10.1016/S0065-2113(08)60043-7

Udawatta, R.P., Kallenbach, R.L., 2011. Agroforestry Buffers for

Nonpoint Source Pollution Reductions from Agricultural Watersheds

Influence of Cover Crop Managements on X-Ray CT-Measured Pore

Parameters and Hydraulic Properties of Soil View project

Parameterization of the APEX model View project. Article in Journal of

Environmental Quality. https://doi.org/10.2134/jeq2010.0168

Zampaligré, N., Dossa, L.H., Schlecht, E., 2013. Contribution of

browse to ruminant nutrition across three agro-ecological zones of

Burkina Faso. Journal of Arid Environments 95, 55–64.

https://doi.org/10.1016/J.JARIDENV.2013.03.011

https://doi.org/10.1016/S0065-2113(08)60043-7

https://doi.org/10.2134/jeq2010.0168


307

CHAPTER VI

CONCLUSIONS

1. Summary of Research Findings

Chapter I highlighted some of the advances and challenges in the current research

into silvopastoral systems. Chapter II outlined a framework of decision-making

processes for the design of agroforestry systems and used the contextual problem of

livestock production induced chronic pasture degradation to describe the

development of a novel silvopastoral system for the Amazon. The objective of the

silvopastoral system was to reduce the degradation of pastures via the inclusion of

locally adapted and native trees, and to offer sustainable improvements in

productivity when compared with the traditional model of extensive cattle farming in

the Amazon.

Chapter III followed the growth of five locally adapted species for 18 months from

when they were planted in a Brachiaria brizantha cv. Marandú pasture. It was found

that the tree species responded differently to the pasture environment. Erythrina

berteroana, Inga edulis and Leucaena leucocephala demonstrated positive increase

in height, diameter at breast height and root collar diameter when measured in all 6-

308

month intervals. After 12 months, Senegalia loretensis and Ceiba pentandra began

to decrease in height as mortality increased for these two species. At 18 months, I.

edulis was 4.4 m tall and was found to halve grass production. All species were

pollarded, and I. edulis produced more pollard biomass (0.42 Mg ha-1) than E.

berteroana (0.19 Mg ha-1) or L. leucocephala (0.27 Mg ha-1). Six months later, all

species but C. pentandra were found to negatively effect grass production. However,

the magnitude of this negative effect was less for I. edulis than in the previous

monitoring period.

It was concluded that some species are better adapted to the environmental

conditions of the study site than others. A shorter, 12-month experiment may have

incorrectly concluded that S. loretensis and C. pentandra were well adapted, but

these species had declined within 18 months. It should also be taken into

consideration that it is unknown how E. berteroana, I. edulis and L. leucocephala will

respond to the pasture environment after 18 months. In-situ field experiments can

assist in the evaluation of tree species in the context of the pests and pathogens

present at the site. It was reported that at the present site all species suffered bark

damage by rodent species, and that C. pentandra was susceptible to attacks from

ants of the Atta genus. These findings demonstrated some of the advantages of in-

situ long-term field experiments.

Rising levels of carbon in the atmosphere are driving global climatic changes and

this has resulted in a market for services and processes that can remove

309

atmospheric carbon, such as afforestation. The potential carbon sequestration that

could be provided by silvopastoral systems is an important market-benefit that could

enable payments to farmers for the adoption of more sustainable practices. A

commonly used non-destructive approach to determine the biomass of a tree, and by

extension the carbon sequestration of a tree, is the use of allometric equations to

determine the relationship between an easily measurable element of the tree and it’s

total biomass.

Chapter IV developed species and component specific allometric equations and

compared the predictive power of a variety of tree measurements against biomass.

The potential combinations were assessed for their fit to the data using linear,

quadratic and log-linear models and the models were compared using the Furnival’s

Index. Log-linear models were found to be superior to linear or quadratic models for

the prediction of biomass components in all species. The results showed that a

single parameter model of root collar diameter (RCD) was the best predictor of all

biomass components of I. edulis, whereas for E. berteroana and L. leucocephala a

multiple model that included RCD and height improved the predictive measurements

of total and trunk biomass. The models for the prediction of total biomass showed

the best fit for the species of L. leucocephala (R2= 0.86) followed by E. berteroana

(R2= 0.77) and I. edulis (R2= 0.70). After 24 months, I. edulis had accumulated the

most biomass (7.2 Mg ha-1) when compared to E. berteroana (4.4 Mg ha-1) or L.

leucocephala (4.3 Mg ha-1). The predicted carbon sequestration of these species

was greatest for I. edulis (3.4 Mg ha-1) when compared to E. berteroana (2.1 Mg ha-

310

1) or L. leucocephala (2.0 Mg ha-1). Chapter IV concludes that silvopasture could

offer increased sequestration of carbon in the form of trees that increase above

ground biomass. This finding is based on the assumption of active management of

trees to reduce negative impacts on the growth, and therefore the carbon

sequestration, of associated grasses. The rapid accumulation of above ground

biomass in I. edulis in the pasture environment and its resilience to pollarding makes

this species an attractive option for planting arrangements to maximize carbon

sequestration.

All three tree species were palatable to cattle and readily eaten when fresh.

Destructive sampling of foliar biomass below 1.6 m suggested that cattle consumed

99% of E. berteroana, 75% of I. edulis and 80% of L. leucocephala. When leaves

were counted from 0.5-1.5 m the change pre to post-trial was a decrease of 95% for

E. berteroana, 80% for I. edulis and 87% for L. leucocephala. Cattle browsed at an

average height of 1.5 m but reached a maximum height of 2.6 m for E. berteroana as

they pulled down branches browse. Cattle consumed the tips of branches and

leaves. E. berteroana trees had been developed from cuttings and they were found

to be the most damaged by exposure to cattle.

Treatments containing trees and grass were found to produce more available forage

(>2.2 Mg ha-1) than the grass only treatment (1.5 Mg ha-1). It is important to consider

the effect of pollarding on these results. After 18 months, there was a negative effect

of I. edulis on the production of pasture grass. This occurred when the I. edulis had

311

an average height of 4.4 m. Subsequently, pollarding of I. edulis at 18 months 2 m

reduced this negative effect. The increased production of available forage biomass

and was greater than the decreased production of the pasture grass biomass.

Pollarding is a useful strategy for managers of silvopastoral systems, as competition

for resources such as light can negatively affect plant productivity.

Practical recommendations for those seeking to plant a silvopastoral system for

cattle production in the Brachiaria pastures of the Amazon are as follows. Establish

80% I. edulis and 20% alternative species in a planting arrangement of 0.5 x 7.5-15

m. The first pollard should occur at the onset of a rainy season when 70% of all trees

are greater than 3 m height. The first pollard should apply to 60% of all I. edulis and

be to a height of 2.5 m. The rationale behind this decision is to make some I. edulis

foliage available as forage for cattle, also to leave some I. edulis to provide shade for

the months where the rain is less frequent and contribute to the continued provision

of ecosystem services. In the absence of cattle the pollard is still a useful way of

thinning the tree line and maximizing the production of litter to contribute to healthy

topsoil. Leaving a considerable proportion of trees will also ensure there are I. edulis

fruit available to replenish the I. edulis plant stocks and expand the planted area.

2. Comparison with Previous Research

This research contributes to a growing body of work that catalogues silvopastoral

systems as an approach to livestock production that can meet the challenge of

sustainably feeding a growing population. Murgueitio et al. (2011) assert that

312

silvopastoral systems are instrumental for the productive rehabilitation of cattle

production and biodiversity conservation in agricultural landscapes. They state that

native trees and shrubs can be integrated into intensive silvopastoral systems and

that this is an important approach for climate change adaptation and mitigation.

Dagang and Nair (2003) reviewed two decades of research on silvopastoral systems

in Central America and find that such systems were not widely implemented by

farmers in spite of the potential for silvopasture to improve pasture productivity and

buffer deforestation. They suggest future researchers must understand the mind-set

of farmers when managing their resources, which is to maximize perceived gains

within the constraints of the system dynamics. They assert that silvopastoral

research must seek to produce fast results, be of low risk for farmers to adopt and

have a favorable impact on farming systems (Dagang and Nair, 2003).

This study relied on an understanding of the forage potential of locally adapted

plants. Serrão et al. (1996) advocate the domestication of native Amazonian plants

for agroforestry systems and conclude that more research is needed to identify

optimal configuration of land use systems. The importance of locally adapted plants

for the development of silvopastoral systems has been highlighted by many authors

(including but not limited to Butterfield, 1995; Ibrahim et al., 2006; Guimarães-Beelen

et al., 2006; Murgueitio et al., 2011; Nunes et al., 2016; Dechnik-Vázquez et al.,

2019). However, research on the forage potential of trees that are adapted to

Amazon pastures is scarce and to date has concentrated on the North Eastern

Amazon in Brazil (Hohnwald et al., 2010; Hohnwald, 2016). The present study is in

313

the South Western Amazon, Peru, 1700 miles from Hohnwald’s (2016) study site. I.

edulis, a species described in Hohnwald’s (2016) experiment was also present in the

region of the present study site and therefore selected for the present experiment as

a parallel with Hohnwald’s (2016) study. However, a different planting design was

used to that of Hohnwald (2016). The planting design of the present silvopastoral

system was taken from the densest rows of alley planted I. edulis (0.5 m) found in

the literature (Alegre and Rao, 1996; Leblanc et al., 2006).

Devendra and Ibrahim (1999) consider that one of the greatest challenges facing the

development of silvopastoral systems is the need to improve integrated resource

movement to increase productivity on whole farm systems that are driven by market-

orientated access. The present study found an increase in two key benefits, that of

increased forage production, which is a benefit that accrues to the farmer, and that of

enhanced carbon sequestration which is experienced as a global good but has an emerging

market. Numerous studies have shown the potential of trees to produce high quality

fodder for animals (Haque et al., 1986; Pezo et al., 1990; Douglas et al., 1996; Ainalis and

Tsiouvaras, 1998; Rengsirikul et al., 2011). The majority of this work has concentrated on

the production of forage in fodder bank arrangements where the production of forage is

measured relative to a lopping, coppicing or pollarding approach (Muschler et al., 1993;

Larbi et al., 2005; Ortega-Vargas et al., 2013; Casanova-Lugo et al., 2014; Khedri et al.,

2018).

314

When trees are grown in rows they are known as alleys. Alleys are commonly used to

grow food crops where prunings from the rows of trees can be left between the rows for

nutrient inputs or weed control or the pruning can be cut and carried to livestock and

used as a fodder bank (Duguma et al., 1988; Leblanc et al., 2006; Ghezehei et al., 2015). In

the present study it was envisaged that the cattle could directly browse the alleys of trees.

The planting design of this study shares certain design characteristics (east to west

planting lines, objective of maximizing production) with the ‘intensive’ silvopastoral

systems described by Murgueitio et al. (2011). Murgueitio et al. (2011) suggest that in an

intensive silvopastoral system a planting density may be upwards of 10,000 plants ha-1,

whereas the present study planted forage trees at a density of 2667 plants ha-1.

The present study reports an increase in carbon sequestration due to enhanced

production of above ground biomass. The carbon storage of trees is directly related to the

production of total biomass, therefore this study may underestimate total carbon

sequestration (Nair et al., 2010). Sharrow and Ismail (2004) hypothesized that as grasses

store the majority of their carbon below ground and trees and shrubs store the majority of

their carbon above ground, combinations of trees and grasses would be able to provide

increased sequestration of carbon. The potential for and uncertainties of silvopastoral

systems to sequester carbon are well documented (see Nair et al., 2010; Feliciano et

315

al., 2018). In the present study it is not known what effect the pollarding may have on

the production of biomass of these species or how this may effect the carbon

sequestration in this system. Ultimately, managers must find a balance between the

benefits they wish to accrue with a silvopastoral system.

The present study only considers the above ground biomass of trees in relation to

potential for carbon sequestration and does not consider the below ground biomass or the

effect of trees in the present study on soil carbon. Feliciano et al., (2018) reviewed 88

studies on soil carbon. They found that the greatest soil carbon sequestration results

from the transition of a grassland system to a silvopastoral system (4.38 Mg

C ha−1 yr−1). Higher soil organic content is related to tree density (Saha et al., 2009).

Soils proximate to trees, compared to soils at a greater distance from trees, store

increased carbon (Howlett, 2009; Takimoto et al., 2009). Haile et al. (2010) found on

Spodosols and Ultisols in Florida that silvopastoral systems had greater potential to

store more stable carbon in the soil at all depths compared with the treeless system.

3. Theoretical Implications of Findings

Wilderness continues to be converted to agriculture to meet the food needs of a

growing global population. Livestock is a major driver of land cover change and while

awareness is of the impacts of meat consumption on the environment is growing in

western societies, the global demand for meat and dairy products is projected to

have increased by 15% in 2027 (OECD‐ FAO, 2018). As the agricultural frontier

316

expands into neotropical rainforest areas, it leaves in its a wake a mosaic of

pasturelands that have been degraded by poor management practices such as fire

regimes and overgrazing (Costa and Rehman, 1999). There is an urgent need to

develop new technology for the productive rehabilitation of these lands if wilderness

areas are to remain while simultaneously feeding a growing global population

(Murgueitio et al., 2011). It is theorised that intensification of existing cattle systems

will reduce the incentive to deforest additional land, in practice this is likely to depend

on political and economic circumstances, such as market saturation (Latawiec et al.,

2014; Müller-Hansen et al., 2019). However, silvopastoral systems would result in

net afforestation that would provide increased ecological connectivity in an

increasingly fragmented landscape (Cadavid-Florez et al., 2019; Vaca et al., 2019).

The practice of silvopasture has been described as a “triple win” land use (Chará,

2019). Silvopastoral systems have the potential to increase cattle stocking densities

by providing an increase in forage that can therefore intensify production in a way

that simultaneously increases carbon sequestration. The adoption of silvopastoral

systems may be a crucial strategy of farmer adaptation to projected climate impacts

which include a decline in the forage quality of grasses under increased

temperatures (Lee et al., 2017), and increased frequency and intensity of extreme

weather patterns, such as drought (Jat et al., 2016; Lawson et al., 2019). The

present study chose tree five species to assess their silvopastoral potential, of which

three were considered to be well adapted to the present study site. There are many

more species in the Amazon that represent good potential for forage (Hohnwald,

317

2016; Dechnik-Vázquez et al., 2019). Different species combinations may also be

more successful than others. The present study alludes to the potential of Inga edulis

as a nurse tree to assist in the establishment of silvopastoral systems in the

Brachiaria brizantha cv. Marandú pasture grasses, as it is able to rapidly accumulate

foliar biomass and open up a niche which due to a reduced presence of grass may

have a reduced influence of the allelopathic effects of the grass.

The findings of this study imply that there is an excellent potential for further

development of silvopastoral systems in the Amazon. Silvopastoral systems in the

Amazon could provide rapid and substantial increases in available forage biomass

and carbon sequestration over monoculture grass pastures. This combination of

benefits accruing at both the farm and global level supports the need to develop

payment-based incentives for landowners to adopt silvopastoral systems.

4. Limitations of the Study

The design of the present study was ambitious. The positive outcome was that it was

able to answer some general questions about the potential for silvopastoral systems

in the Amazon. However, a key limitation of this approach is that often the fine scale

interactions that are interesting design elements of such a system were overlooked.

For example, there was a single planting arrangement of 0.5 x 7.5 m. This meant

that the findings were less comparable with those of the literature, which commonly

focus on tree growth and biomass production in 2 x 2 m or 3 x 3 m arrangements.

318

The results were drawn from a single site, which means the results cannot be

transferred to other areas. Species were arranged in monospecific stands, whereas

in practice mixed stands could increase biodiversity, reduce disease and exploit

more resources by inherent variation such as branching pattern, leaf area, niche

occupancy or rooting depth. All species were drawn from single accessions when in

reality there are multiple accessions of species some of which may have been more

successful than others. The experiment was only carried out once, and therefore is

temporally limited to the specific prevailing conditions of the time at which it was

carried out.

There were six dominant species in this experiment, consisting of five tree species

and one pasture grass. Ideally, there would have been a much more thorough

revision of potential species to use in the experiment but this was not possible due to

the time sensitive nature of the current project. However, there are many more tree

species with excellent potential not considered in the study. No consideration was

taken for the chemical constituents of the foliage, or potential toxicity. The decision to

use Brachiaria brizantha cv. Marandú was taken because it is one of the most

productive grasses available in the region and also presents some of the most

challenging conditions for tree establishment (Sobanski and Marques, 2014). The

decision was taken to arrange the trees in monospecific stands. All management

decisions that excluded an alternative approach can be considered as impact

limitations of the present study as they further narrow the scope of the present study.

319

The trees were pollarded to 2 m at 18 months. This was considered an essential

management strategy in keeping with the objective of optimal forage production, but

also had the effect of limiting the comparisons that could be drawn between the

present study and other studies in the literature. The pollarded biomass was weighed

as a total biomass. In retrospect, it would have been valuable to separate the stems,

branches and foliage. This would have been informative to consider the amount of

forage or fuel wood that could be gathered by the pollard. The original research

design allowed for repeat pollarding and manual defoliation at varying frequencies

and intensities. Due to unforeseen changes to the project timeline this experiment

was not possible.

In Chapter III there were two sampling approaches to measuring the effect of trees

on the regrowth of grass. The first approach considered a 30-day regrowth of grass

that had been cut to 10 cm. It was originally intended to measure consecutive 30-day

regrowth intervals. This approach would have also captured a seasonal effect on

grass growth. The second consideration of the grass biomass occurs after the

pollard event and considers the regrowth following a 6-month interval. Care was

taken to consider the two results separately and this occurrence was an

unintentional artefact of field sampling.

In Chapter V there was no treatment effect on the total fractions of grass biomass

prior to and post the cattle trial. Five 1 m2 grass quadrats were taken in each

replicate prior to the cattle trial and five post the cattle trial. This resulted in n=15 for

320

each treatment prior to the cattle trial and n=15 post the cattle trial. It is possible that

the sample sizes were too small to detect an effect of treatment. When a

retrospective power analysis was carried out it was found hat the number of samples

that would have been required to detect a treatment effect would have exceeded

0.09 ha, the size of each replicate. Therefore, this was not a feasible approach for

the present study. This occurred because the grass had been left to grow for 24

months and this may have resulted in a greater inter-sample variation than that

which may have been found in an active pasture. In anticipation of this, the study

design included a subjective visual assessment of ‘edible biomass’.

The estimation of the contribution of tree biomass to carbon stocks only considered

the above ground biomass and did not consider the below ground components of the

trees. The carbon fraction was derived from the literature. However, as detailed

biomass estimates were made for the three species it would have improved the

study to have component specific estimates of carbon for each tree species. This

was beyond the scope of the present study. No consideration was given to the soil

carbon potential of the silvopastoral system, which could have yielded interesting

results.

321

5. Future Research

5.1 Direct Extensions of the Study

In the present study, the species Erythrina berteroana, Inga edulis and Leucaena

leucocephala were able to establish in a pasture environment when planted as

saplings. This was a costly approach as each tree was planted as a seed in an

individual bag of soil and transported to the planting location. In the present study,

this was feasible due to an on-site nursery. If this technology were to be widely

adopted then a useful study would compare the cost and labour of the present

approach with direct seeding and on-going management of the surrounding pasture

grass.

5.2 Broader Issues for Future Work

A lack of studies document the forage potential of trees native to the Amazon.

Despite the relatively recent history of cattle farming in the Amazon, there is local

ecological knowledge of which species are preferentially browsed by cattle. Once

species have been identified, cafeteria type feeding trials could be used to evaluate

intake and effect on animal health. There is also an effect of the chemicals found in

milk following the consumption of certain forages that should be assessed.

The present study is primarily concerned with the production of forage and

sequestration of carbon. Future research could seek to increase on-farm revenue.

Incorporating a timber element into the silvopastoral system could generate added

322

economic incentives. Fast-growing leguminous timber trees with a low leaf area

index could be incorporated into the rows of trees and simultaneously enhance the

sequestration of carbon. Fruit trees could be an additional form of on-farm income.

Fertilisers could be used to account for the nutrients exported from the site in the

form of timber or fruits. The increased biodiversity of such stands could increase the

farm productivity and resilience by incorporating a more diverse selection of tree

products.

More work is required to understand the costs of implementing silvopastoral systems

and farmers perceptions of such systems. Outreach and education may be required

to explain the system to landowners (Latawiec et al., 2014). It is likely that some form

of credit or payment would be required to assist in the transition to silvopastoral

systems. Some examples of these approaches exist in the literature from other

regions and could be usefully applied in the study region of Madre de Dios, Peru

(Calle et al., 2009; Garbach et al., 2012; Cranford and Mourato, 2014).

5.3 Practical Next Steps

The growing demand for rehabilitation and reforestation projects could be met by the

extension of this research into real world scenarios. The following discussion

summarises some ideas and concepts that occurred as a result of this study and

may be of use when considering how to translate these studies into action. In order

to extend the interpretation of the results of this research it is necessary to replicate

silvopastoral systems to many more sites, which would allow a sample of

323

silvopastoral systems to be directly compared, both between silvopastoral sites with

common and contrasting site conditions and with alternative land uses. This could be

done with the participation and local ecological knowledge of cattle farmers. A survey

of farms could evaluate a farmer’s local ecological knowledge and their willingness

and conditions to engage with the development of such trial silvopastoral systems.

The present study has described numerous benefits of silvopastoral systems

accruing at the farm level. It is likely that landowners will require some assistance to

initially adopt the system and over time will see the benefits and be incentivised by

their own experience to continue to expand the system. Crucially, they would be

involved in every stage, including the production of the plants from seed. Such an

approach would ensure that in the future, once they are able to enjoy the benefits of

the system they have the experience of all stages of the process to continue

expanding the silvopastoral system on their farm and adapting the system to meet

their own needs. The next step is transferring this technology to the farmers and

supporting them to create the silvopastoral system that best fits their needs, and

simultaneously create a baseline from which the system can be measured over time,

in different contexts and thereby attract greater interest and investment.

7. Conclusion

Silvopastoral farming systems can make livestock farming more economically and

environmentally sustainable when compared with traditional models of cattle farming

in the Amazon. The use of robust and well-replicated on-farm experiments such as

the one recommended by the framework in this thesis can be used to evaluate the

considerable potential of locally adapted species for use in silvopastoral systems.

324

When compared with conventional models of farming in the region, the silvopastoral

system presented in this study provided gains in productivity, accruing as increased

total biomass, an increase in available forage for cattle and the sequestration of

carbon. Widespread adoption of silvopastoral systems on existing cleared land in the

Amazon could reduce the need for further deforestation, reverse the trend of pasture

degradation and soil erosion, increase biodiversity in agricultural landscapes and

contribute to the mitigation of climatic change.

References

Ainalis, A.B., Tsiouvaras, C.N., 1998. Forage production of woody

fodder species and herbaceous vegetation in a silvopastoral system in

northern Greece. Agroforestry Systems 42, 1–11.

https://doi.org/10.1023/A:1006112330453

Alegre, J.C., Rao, M.R., 1996. Soil and water conservation by contour

hedging in the humid tropics of Peru. Agriculture, Ecosystems &

Environment 57, 17–25. https://doi.org/10.1016/0167-8809(95)01012-2

Butterfield, R.P., 1995. Promoting biodiversity: advances in

evaluating native species for reforestation. Forest Ecology and


Cadavid-Florez, L., Laborde, J., Zahawi, R.A., 2019. Using landscape

https://doi.org/10.1023/A:1006112330453

https://doi.org/10.1016/0167-8809(95)01012-2

325

composition and configuration metrics as indicators of woody

vegetation attributes in tropical pastures. Ecological Indicators 101,

679–691. https://doi.org/10.1016/J.ECOLIND.2019.01.072

Calle, A., Montagnini, F., Andrés, F., 2009. Farmer’s perceptions of

silvopastoral system promotion in Quindío, Colombia. Bois et Forets

des Tropiques 300, 79 – 93.

Chará J., Reyes E., Peri P., Otte J., Arce E., S.F., 2019.

Silvopastoral Systems and their Contribution to Improved Resource Use

and Sustainable Development Goals (SDG): Evidence from Latin America.

Rome, Italy.

Costa, F.P., Rehman, T., 1999. Exploring the link between farmers’

objectives and the phenomenon of pasture degradation in the beef

production systems of Central Brazil. Agricultural Systems 61,

135–146. https://doi.org/10.1016/S0308-521X(99)00043-8

Cranford, M., Mourato, S., 2014. Credit-Based Payments for Ecosystem

Services: Evidence from a Choice Experiment in Ecuador. World

Development 64, 503–520.

https://doi.org/10.1016/j.worlddev.2014.06.019


https://doi.org/10.1016/S0308-521X(99)00043-8

https://doi.org/10.1016/j.worlddev.2014.06.019

326

Dagang, A.B.K., Nair, P.K.R., 2003. Silvopastoral research and

adoption in Central America: Recent findings and recommendations for

future directions, in: Agroforestry Systems. Kluwer Academic

Publishers, pp. 149–155. https://doi.org/10.1023/A:1026394019808

Dechnik-Vázquez, Y.A., García-Barrios, L., Ramírez-Marcial, N., van

Noordwijk, M., Alayón-Gamboa, A., 2019. Assessment of browsed plants

in a sub-tropical forest frontier by means of fuzzy inference. Journal

of Environmental Management 236, 163–181.


Devendra, C., Ibrahim, M., 1999. Silvopastoral systems as a strategy

for diversification and productivity enhancement from livestock in the

tropics. Centro Agronomico Tropical de Investigación y Enseñanza

(CATIE ) 2020, 8–22.

Douglas, G.B., Bulloch, B.T., Foote, A.G., 1996. Cutting management of

willows ( Salix spp.) and leguminous shrubs for forage during summer.

New Zealand Journal of Agricultural Research 39, 175–184.

https://doi.org/10.1080/00288233.1996.9513176

Duguma, B., Kang, B.T., Okali, D.U.U., 1988. Effect of pruning

intensities of three woody leguminous species grown in alley cropping

https://doi.org/10.1023/A:1026394019808


https://doi.org/10.1080/00288233.1996.9513176

327

with maize and cowpea on an alfisol. Agroforestry Systems 6, 19–35.

https://doi.org/10.1007/BF02220106

Garbach, K., Lubell, M., DeClerck, F. a J., 2012. Payment for

Ecosystem Services: The roles of positive incentives and information

sharing in stimulating adoption of silvopastoral conservation

practices. Agriculture, Ecosystems and Environment 156, 27–36.


Ghezehei, S.B., Everson, C.S., Annandale, J.G., 2015. Can productivity

and post-pruning growth of Jatropha curcas in silvopastoral systems be

regulated by manipulating tree spacing/arrangement without changing

tree density? Biomass and Bioenergy 74, 233–243.

https://doi.org/10.1016/j.biombioe.2015.01.007

Grieser, D., 2013. Mathematisches Problemlösen und Beweisen : Eine

Entdeckungsreise in die Mathematik.

https://doi.org/10.1016/S0065-2113(10)08005-3

Guimarães-Beelen, P.M., Berchielli, T.T., Beelen, R., Araújo Filho,

J., Oliveira, S.G. de, 2006. Characterization of condensed tannins

from native legumes of the Brazilian northeastern semi-arid. Scientia

Agricola 63, 522–528. https://doi.org/10.1590/S0103-90162006000600002

https://doi.org/10.1007/BF02220106


https://doi.org/10.1016/j.biombioe.2015.01.007

https://doi.org/10.1016/S0065-2113(10)08005-3

https://doi.org/10.1590/S0103-90162006000600002

328

Haile, S.G., Nair, V.D., Nair, P.K.R., 2010. Contribution of trees to

carbon storage in soils of silvopastoral systems in Florida, USA.

Global Change Biology 16, 427–438.

https://doi.org/10.1111/j.1365-2486.2009.01981.x

Haque, I., Jutzi, S., Neate, P.J.H., 1986. Potentials of forage

legumes in farming systems of sub-Saharan Africa, in: Proceedings of a

Workshopheld at ILCA, Addis Ababa, Ethiopia, 16-19 September 1985.

International Livestock Centre for Africa, p. 583.

Hohnwald, S., 2016. Relative palatability and growth performance of

capoeira species as supplementary forages in the NE-Amazonia.

Agriculture, Ecosystems & Environment 218, 107–115.


Hohnwald, S., Trautwein, J., Camarão, A.P., Gerold, G., Wollny,

C.B.A., 2010. Resource Management and Rural Development Biomass

Production and Relative Palatability of Possible Supplementary Forage

Plants of the Northeastern Amazon.

Ibrahim, M., Villanueva, C., Casasola, F., Rojas, J., 2006. Sistemas

silvopastoriles como una herramienta para el mejoramiento de la

https://doi.org/10.1111/j.1365-2486.2009.01981.x


329

productividad y restauración de la integridad ecológica de paisajes

ganaderos. Pastos y Forrajes 29, 383–419.

Jat, M.L., Dagar, J.C., Sapkota, T.B., Yadvinder-Singh, Govaerts, B.,

Ridaura, S.L., Saharawat, Y.S., Sharma, R.K., Tetarwal, J.P., Jat,

R.K., Hobbs, H., Stirling, C., 2016. Climate Change and Agriculture:

Adaptation Strategies and Mitigation Opportunities for Food Security

in South Asia and Latin America. pp. 127–235.

https://doi.org/10.1016/bs.agron.2015.12.005

Khedri, L., Ghahramany, L., Ghazanfari, H., Pulido, F., 2018. A

quantitative study of pollarding process in silvopastoral systems of

Northern Zagros, Iran. Forest Systems 26, e018.

https://doi.org/10.5424/fs/2017263-11433

Larbi, A., Anyanwu, N.J., Oji, U.I., Etela, I., Gbaraneh, L.D.,

Ladipo, D.O., 2005. Fodder Yield and Nutritive Value of Browse Species

in West African Humid Tropics: Response to Age of Coppice Regrowth.


https://doi.org/10.1007/s10457-005-0922-x

Latawiec, A.E., Strassburg, B.B.N., Valentim, J.F., Ramos, F.,

Alves-Pinto, H.N., 2014. Intensification of cattle ranching production

https://doi.org/10.1016/bs.agron.2015.12.005

https://doi.org/10.5424/fs/2017263-11433

https://doi.org/10.1007/s10457-005-0922-x

330

systems: socioeconomic and environmental synergies and risks in

Brazil. Animal 8, 1255–1263. https://doi.org/10.1017/S1751731114001566

Lawson, G., Dupraz, C., Watté, J., 2019. Can Silvoarable Systems

Maintain Yield, Resilience, and Diversity in the Face of Changing

Environments? Agroecosystem Diversity 145–168.

https://doi.org/10.1016/B978-0-12-811050-8.00009-1

Leblanc, H.A., Nygren, P., McGraw, R.L., 2006. Green mulch

decomposition and nitrogen release from leaves of two Inga spp. in an

organic alley-cropping practice in the humid tropics. Soil Biology and

Biochemistry 38, 349–358.






Müller-Hansen, F., Heitzig, J., Donges, J.F., Cardoso, M.F.,

Dalla-Nora, E.L., Andrade, P., Kurths, J., Thonicke, K., 2019. Can

Intensification of Cattle Ranching Reduce Deforestation in the Amazon?

Insights From an Agent-based Social-Ecological Model. Ecological

https://doi.org/10.1017/S1751731114001566

https://doi.org/10.1016/B978-0-12-811050-8.00009-1



331

Economics 159, 198–211. https://doi.org/10.1016/J.ECOLECON.2018.12.025

Murgueitio, E., Calle, Z., Uribe, F., Calle, A., Solorio, B., 2011a.




Muschler, R.G., Nair, P.K.R., Meléndez, L., 1993. Crown development

and biomass production of pollarded Erythrina berteroana, E. fusca and

Gliricidia sepium in the humid tropical lowlands of Costa Rica.

Agroforestry Systems 24, 123–143. https://doi.org/10.1007/BF00706887

Nunes, A.T., Cabral, D.L.V., Amorim, E.L.C., Santos, M.V.F. dos,

Albuquerque, U.P., 2016. Plants used to feed ruminants in semi-arid

Brazil: A study of nutritional composition guided by local ecological

knowledge. Journal of Arid Environments 135, 96–103.


OECD-FAO (Ed.), 2018. Chapter 6. Meat Market situation, in:

Agricultural Outlook 2018-2027. p. 150.


Ortega-Vargas, E., López-Ortiz, S., Burgueño-Ferreira, J.A., Campbell,



https://doi.org/10.1007/BF00706887



332

W.B., Jarillo-Rodríguez, J., 2013. Date of pruning of Guazuma

ulmifolia during the rainy season affects the availability,

productivity and nutritional quality of forage during the dry season.


https://doi.org/10.1007/s10457-013-9608-y





Rengsirikul, K., Kanjanakuha, A., Ishii, Y., Kangvansaichol, K.,

Sripichitt, P., Punsuvon, V., Vaithanomsat, P., Nakamanee, G., Tudsri,

S., 2011. Potential forage and biomass production of newly introduced

varieties of leucaena (Leucaena leucocephala (Lam.) de Wit.) in

Thailand. Grassland Science 57, 94–100.

https://doi.org/10.1111/j.1744-697X.2011.00213.x




https://doi.org/10.1016/0921-8009(95)00092-5

https://doi.org/10.1007/s10457-013-9608-y

https://doi.org/10.1111/j.1744-697X.2011.00213.x

https://doi.org/10.1016/0921-8009(95)00092-5

333





Sobanski, N., Marques, M.C.M., 2014. Effects of soil characteristics

and exotic grass cover on the forest restoration of the Atlantic

Forest region. Journal for Nature Conservation 22, 217–222.

https://doi.org/10.1016/J.JNC.2014.01.001

Vaca, R.A., Rodiles-Hernández, R., Ochoa-Gaona, S., Taylor-Aquino,

N.E., Obregón-Viloria, R., Díaz-García, D.A., Navarrete-Gutiérrez,

D.A., 2019. Evaluating and supporting conservation action in

agricultural landscapes of the Usumacinta River Basin. Journal of

Environmental Management 230, 392–404.



https://doi.org/10.1016/J.JNC.2014.01.001


designing silvopastoral systems for the amazon: a framework ...

Documents

Transcript of designing silvopastoral systems for the amazon: a framework ...