Adequate control treatments for long term multiple

objective agroforestry research

C.Dupraz

INRA,Lepse,Place Viala,34060Montpellier,France

Tel.+ 33 4 99 61 23 39

Fax + 33 4 67 52 21 16

Email :Dupraz@ensam.inra.fr

Key words:experiment design,resource capture,competition,farming

system,weeds,tree density

Abstract

Adequate control treatments for long term multiple objective

agroforestry research

Agroforestry experiments usually include control plots of either

pure crop or pure tree stands. A clear distinction should be made

between intensively managed biophysical controls and farming

system controls with realistic labour input and management regimes.

Trying to draw biophysical conclusions from farming system controls

(or the reverse) is often not justifiable.The design and management

of these elusive control plots is a complicated issue which is often

overlooked. Many factors beyond the control of the experiment

manager can disturb long term field agroforestry experiments.Some

examples from French agroforestry experiments illustrate how

uncontrolled factors may bias the results,including the proportion

of harvested to planted trees, the weeding regimes, and the use of

tree-shelters. The analysis of agroforestry data could be more

efficient when considering a continuum of tree - crop mixture

management options between the agroforestry plot and the non-

agroforestry plot.The concept of biophysical control plots becomes

then less essential.A relevant modelling approach of interactions

between trees and crops should 1)perform correctly for any tree/crop2

proportion and even for pure stands,when setting the parameters of

the other component to zero, 2) provide for the inclusion of new,

uncontrolled factors that could emerge over time. The biological

efficiency of agroforestry systems may however be a subordinate

criterion for agroforestry adoption, as observed at the moment in

France.Agroforestry systems with poor biological outcomes can even

be very attractive in some ecological or sociological conditions,

and only farming system controls may allow to highlight this aspect.

3

Introduction

Agroforestry systems aim at taking advantage of tree - crop

interactions in order either to improve land productivity (Fukai and

Trenbath, 1993) or to tackle environmental (Price, 1995) or

sociological (Dupraz, 1994) issues. Under a specified management

regime, agroforestry field experiments consequently intend to

identify and quantify interactions between trees and crops and/or

animals. Usually both competitive (negative) and facilitative

(positive) interactions are described, but their evaluation always

implies the comparison of intraspecific and interspecific

relationships (Ranganathan and de Wit,1996).Control plots of either

pure trees (forestry control plot) or pure crops (agricultural

control plot) are therefore essential.Interferences between plots

(Ong,1996)and various edge effects (Langton,1990,Van Noordwijk and

Ong, 1996) have been recognised as sources of error when assessing

tree-crop interactions and should always be considered when

designing factorial experiments in agroforestry.These aspects will

not be discussed here.We will focus on some overlooked aspects of the

design and management of control plots in agroforestry experiments:

the objective assigned to the control plots, the appropriate

structure of the control plots, the need to avoid long term4

unexpected confounding shifts,the intensity of tree management in

forestry control plots.Most of the biophysical literature on tree-

soil-crop interactions ignores these issues, and thus may give the

incorrect impression that results are of universal value. Some

examples taken from our experiments in temperate and Mediterranean

France will illustrate these pitfalls.Confronted to these problems,

a modelling strategy is put forward to address the biophysical

interactions in agroforestry, and the case for farming system

controls is questioned.

5

Biophysical and/or farming system controls in

agroforestry experiments?

A clear distinction should be made between intensively managed

biophysical controls and farming system controls with realistic and

practicable labour input and management regimes. The former are

entirely managed by the researchers (or a farmer strictly following

researchers’ instructions),while the latter are entirely under the

farmer's management rule. Trying to draw biophysical conclusions

from farming system controls (or the reverse) is often not

justifiable,as will be illustrated in the following sections.

The distinction between biophysical and farming system controls may

be questioned :most experimenters look for intermediate solutions

in cases where the pure biophysical control is too unrealistic or

costly.It may not be easy to plead and raise funds for a biophysical

control with no real-world meaning such as trees spaced and managed

as in an agroforestry plot but without intercrops. This is most

relevant to agroforestry research as control plots necessarily have

large sizes. Moreover, a purely biophysical control may also

implicate unrealistic effects which make the validity of the control

doubtful. This would be the case in a plot of isolated trees with a

6

perfectly clean (weeded) soil in-between. In most environments,

unprotected soil with no litter input would loose organic matter,

mineralised nutrients would be leached,the soil would acidify.These

processes would not occur with the same intensity in the

agroforestry plots and the control plots, and would disturb the

interpretation.

Farming system controls may also be quite unrealistic when the

farmers have no know-how in agroforestry management, and rely on

researchers' instructions. However, in our on-farm experiments, we

observed that the farmers very rapidly become independent and take

initiatives in managing the plots.Farming system controls often do

not allow to quantify biophysical processes,but provide very useful

qualitative indications on the nature of tree-crops interactions.An

experiment that would include both biophysical and farming system

control plots would gain much information.

The appropriate structure of biophysical control

plots in agroforestry experiments

Most agroforestry experiments use the corpus of intercropping as

defined for annual crops (Mead and Willey,1980,Hiebsch and McCollum,

1987).But the design and management of control plots in agroforestry7

experiments have to address some specific aspects. A useful

distinction may be made between systems where the tree component

reaches a steady state (such as regularly pruned trees in alley-

cropping)and systems where the tree component is allowed to develop

a full canopy (as in most temperate agroforestry systems).In the

first case,trees and crops rapidly reach a stationary state,making

the design of control plots relatively easy. However, tree-crop

interactions in alley cropping systems may also evolve over years

due to the root growth of the trees (Fernandes et al, 1993) or the

changes in soil organic matter and nutrient availability (Haggar et

al, 1993).In the second case, the trees are not cut back, no steady

state phase can be observed and the interactions between trees and

intercrops evolve continuously over time.Typically,in such a tree-

crop association, young tree seedlings are very sensitive to crop

water competition,while mature trees will dominate the intercrops

through light capture. The design of control plots should allow to

bring these different phases in the tree-crop relationship to the

forefront.

8

Both additive and substitutive control plots are necessary

Should control plots be derived from the agroforestry scheme using

an additive,substitutive or intermediate design (Vandermeer,1989)?

We illustrate this aspect for a temperate silvoarable system. When

growing high quality hardwoods ( Prunus avium at 200 stems/ha)with a

wheat intercrop,what non-agroforestry control treatments should be

used? For an additive design of the control plots,we would set up a

forest plantation of Prunus avium at the same tree stocking rate as

the agroforestry plantation (200 trees/ha)and a field of wheat with

uncropped bays similar to those in an agroforestry plot. For a

substitutive design of the control plots, we would set up a forest

plantation of Prunus avium at 800 trees/ha (today’s advised stocking)

and a field of wheat. But we could also consider an intermediate

design between these two extremes. The substitutive design is

usually adopted,but would not allow a full assessment of tree-crop

interactions.For example,the additive control consisting of a tree

stand at an agroforestry density but with no intercrop would be

useful to separate the impact of the low tree density from the impact

of the crop competition on tree growth. Substitutive designs offer

the possibility to calculate the Land Equivalent Ratio (LER,Mead and

Willey, 1980), which is a common and well adopted measure of

9

intercropping efficiency. For a comprehensive investigation of

tree-crop interactions, a design including both additive and

substitutive control plots is therefore advisable, as adopted in

most recent silvoarable experiments in France.

Both pure tree and pure crop controls are necessary

Cannell et al (1996)revisited the central agroforestry hypothesis.

They indicate that in a successful agroforest,the tree must acquire

resources that the crop would not otherwise acquire.This approach is

«  tree oriented  ».This is the consequence of the implicit choice of a

sole herbaceous crop as a comparative basis.If a pure tree stand had

been taken as a reference,the hypothesis would have been formulated

as: the crop must acquire resources that the trees would not

otherwise acquire.And if a combination of both pure crops and pure

tree stands was the reference, the approach would no longer be

apposite, as stated by Vandermeer (1989).But the resource capture

approach could be further developed. Separating « harvested  » and

« recycled  » resources,as suggested by Cannell et al (1996),implies

that any « harvested  » resource captured by the plants is equivalent,

regardless of its final destination.This may be questioned,and will

be illustrated by the impact of different planted/harvested tree

10

ratios in agroforestry and forestry plots on the optimum resource

allocation to high value final products

In most tree plantations, some of the trees will be culled before

reaching any economic value. The proportion of such unmarketable

thinned trees is usually higher in a forestry plantation than in an

agroforestry plantation. In an agroforestry stand, the resources

captured at the expense of the crop to feed these trees have a higher

marginal cost than the resources captured by marketable final trees.

Therefore agroforestry designs with low planted/harvested tree

ratios will favour intercrop productivity,but increase the risk of

obtaining an incomplete mature tree stand.In a forestry stand,the

resources captured by unmarketable trees at the expense of the final

trees are also wasted. Different forestry control treatments may

then be designed,with a variable cost of establishing,tending and

culling trees with no commercial value.Therefore,the optimisation

of agroforestry designs implies an understanding of the risks

associated with low density plantations, and leads to define

Conservative, Prudent, Risky and Daring tree plantation designs in

agroforestry (Table 1).The design of both agroforestry plots and

forestry controls should therefore allow to assess the probability

of tree failure as affected by tree density.Multiple controls with

11

different tree plantation densities would therefore be required.In

agroforestry plots,the distance between tree lines is often imposed

by mechanisation of the intercrop,and higher densities are achieved

through decreasing the spacing between trees on the plantation

lines.In most French agroforestry schemes,no productive thinnings

are expected,and the final trees will be selected as soon as they can

be identified, probably before age 10, as suggested by Dupraz and

Lagacherie (1990).Depending on the cost of tree establishment, and

the probability of tree failure, the optimum strategy may then be

determined. A full reasoning implies that the potential impact of

higher initial tree densities on the intercrop productivity should

be taken into account.

The need to avoid long term unexpected confounding shifts in

agroforestry experiments

Simplifying field agroforestry experiments often results in high

levels of confounding of treatment variables (Huxley,1990).But even

if these aspects have been carefully managed at the time of setting

up the experiment, they may pose a problem later on. Agroforestry

experiments are set up for many years, usually many decades in

temperate areas,in order to obtain a time-integrated evaluation over

12

the whole rotation.Unpredictable events or uncontrolled factors may

alter the very nature of the control plots in time. Biophysical

control plots may become useless if a new action is required in the

control plot but not in the agroforestry plot (or the reverse),

resulting in an intractable confounding effect.A possible solution

would be to design large plots that allow a splitting design when

such a new action has to be included in later years.Three examples

from temperate agroforestry experiments will illustrate this

aspect.

Weeds at the tree -crop interface

As pointed out by Vandermeer (1989),weeds may alter the comparison of

pure crops and associations as they prevent the use of the LER

criterion.This is most relevant to agroforestry experiments,where

agroforestry stands very seldom consist of only trees and crops.

Weeds are often a compulsory partner in the interface between trees

and crops,as in Imperata infested rubber plantations in Asia (Bagnall-

Oakeley et al, 1997). They are usually not taken into account in

tropical experiments, perhaps assuming perfect low-cost hand

weeding of the interface, but were found to be essential when

considered (Schroth and Zech,1995).These weeds may add an unexpected

13

dimension to an agroforestry plantation.Weeding the trees in a non

agroforestry control treatment and in an agroforestry plot in

identical fashion is a difficult issue.

Weeds interfere in the allocation of resources to final harvested

products (either trees or crops),and may alter the comparison by

affecting the agroforestry stand and the non-agroforestry control

treatment in different ways and with different intensities.Here are

some of the reasons observed in our experimental plots for such a

bias:

The use of different weed killers: trees in agroforestry

plantations are often protected by shelters,allowing for the use

of contact herbicides banned in forest plantations.

Different weed species thrive in agroforestry and forestry

plantations, as a consequence of the tilling regime, the crop

phenology,and the weeding operations (often mechanical in forest

plantation, and chemical in agroforestry plantations).This was

observed in 1996 at the Restinclières estate, which is the most

extensive agroforestry experiment running in France.In a walnut

( Juglans nigra x regia ) - wheat association experiment, the 4 m wide

strips of uncropped land along the tree lines were invaded in

14

summer by red poppies ( Papaver Rhoeas ), resulting in high water

competition with the trees, while in the forestry control

treatment the trees were choked in spring with cruciferous species

( Sinapis arvensis ) that the farmer could not destroy in due time,

neither with weedicides nor mechanically.

The possibility to cross-till the forestry control plots and the

agroforestry plots depends on the tree density on the plantation

line. Usually, forestry stands are not planted on a rectangular

pattern and cannot be cross-tilled,resulting in an infestation of

perennial weeds year after year. Agroforestry stands could be

cross-tilled, but this is not often performed as it imposes

numerous driving manoeuvres along the short side of the plot.

Therefore, the repeated ploughing of narrow cropped bays in

agroforestry plantations leads to deep edge furrows that may

complicate the tending of the uncropped strips along the tree

lines. In some experiments, as in the Restinclières estate, this

resulted in the forsaking of the ploughshares,replaced by plough

discs. As the agricultural control plots remained tilled with

conventional ploughshares,an unforeseen bias followed.

15

Pest propagation and control

Weeds are not an isolated example of a disturbing factor that could

appear with time.Other factors,such as pests or pathogens may lead

to further discrepancies between agroforestry treatments and

control treatments. A simple example was observed in an experiment

where the tree owner was alerted by the intercrop farmer that aphids

were invading some trees. He immediately sprayed the contaminated

trees, stopping the aphid propagation. The control forestry

treatment was not watched over by the farmer, and was not easily

accessible due to the amount of weeds and shrubs:it was not sprayed,

resulting in a much higher infestation level.

Tree-shelters in temperate silvopastoral stands

In a silvopastoral system, young seedlings need to be protected

against browsing animals. In Europe, tree-shelters are widely used

for this purpose (Dupraz et al, 1997).Forestry control plantations

are not grazed,and usually trees need not be individually protected

with shelters.Shelters strongly influence the growth of young trees.

It could be argued that an experiment with two forestry controls (one

with sheltered trees and one with unsheltered trees) and with two

agroforestry treatments (one with sheltered trees and one with

16

unsheltered but protected trees,provided that a system allowing the

safe grazing of animals next to the tree could be designed) would

enable experimenters to isolate the impact of the shelter and of the

grass or weed competition on tree growth. But tree-shelters have

other management impacts :they allow easy mechanical control of the

weed regrowth close to the tree,help monitor diseases through visual

separation of the tree and its environment,permit the use of contact

weedicides without danger for the tree,speed up the opening of buds

in the spring,motivate the owner,who is impressed by tree emergence

out of the shelter,etc.A few unexpected elements have more recently

been observed: birds use the shelters as roost, and fertilise

generously the young trees, sheep weed the trees while feasting on

fresh weeds in the weeded area around the shelter,and incidentally

bring more manure to the trees. In a given experiment, any of these

events may take place,thereby further distancing the two treatments,

rendering them no longer representative of the simple initial

factorial design.

The intensity of tree management in forestry control plots

What intensity of tree management is appropriate in forestry control

plots? The intensity of plant management is often much higher in

17

agriculture than in forestry,and this poses a major problem for the

choice of controls in agroforestry.The performance of trees,soils

and crops depends on their management;as the management of soils and

trees in forestry differs substantially from that in agriculture,

there is no unequivocal choice of management regime for agroforestry

which allows meaningful comparisons with both forestry and

agricultural alternatives at the same time. The intensity of the

weeding regime of the forestry control plot will be used to

illustrate this aspect.

Tending a young forestry stand is demanding. Weeds are aggressive.

The early growth of trees will primarily be influenced by the

intensity of weeding.What is the best forestry control treatment? A

biophysical control consisting of a perfectly weeded tree stand is

often used to assess the intercrop influence on tree growth.

Alternatively,a farming system control may be included in an on-farm

experiment and consists of a forestry plantation managed by the

farmer. Two Prunus avium agroforestry experiments in Languedoc-

Roussillon (France) were set up in the early 1990’s using a

biophysical control in Notre-Dame de Londres (NDL) and a farming

system control in Pomy. These experiments have been described by

Dupraz et al (1995)for NDL and Balandier and Dupraz (1998)for Pomy.

18

While the growth of agroforestry trees was similar on both sites,the

difference in growth of the forestry control treatment was

impressive:perfectly weeded trees in the biophysical control at NDL

grew fast,while poorly weeded trees in the farming system control at

Pomy grew slowly. As a result, opposite conclusions concerning the

efficiency of the agroforestry scheme could erroneously be drawn

from these results (Figure 1) if one would ignore the management

intensities used.

The significance of these two control treatments is entirely

different.On one hand the biophysical forestry control is both time

consuming and expensive,and should be applied for a long period (up

to 20 or 30 years);it does not make sense in most farming systems.This

was unfortunately made clear when such an on-farm biophysical

control plot was unintentionally killed by a weedicide application

by the farmer in one of our experiments: keeping the plot free of

weeds is impracticable in most real farming situations.On the other

hand,the management of the farming system forestry control depends

on the agreement of the farmer with this scheme.The cost of tending

the forestry plantation should not be underestimated,with no short

term revenues. Most farmers logically decided to manage their

forestry control plots reducing expenses to a minimum,resulting in

19

poor tree growth. Such experiments were strongly debated by

foresters who argued that the forestry control plot was not

representative of standard management, and that the resulting

overwhelming advantage of agroforestry was unfair.But these control

plots had demonstrated that in some situations, standard forestry

management is rejected by farmers. The on-farm control plot was

effective as a farming system control,and helped to assess tree-crop

interactions at a high level of integration,including many aspects

that a biophysical control plot would have ignored.To date,none of

the 9 on-farm agroforestry experiments in Languedoc-Roussillon has

received proper application of standard forestry weeding schemes on

the forestry control plots (Table 2).

A strategy to minimise the importance of

biophysical control plots: modelling a continuum

from an agroforest to a monocrop?

Long-term factorial experiments may therefore not be adapted to

agroforestry research,and process-based modelling may be part of the

answer to such difficulties (Sanchez,1995,Balandier et al,1998).The

structure of the model should be able to take into account the

unexpected factors that make the agroforestry and control plots20

increasingly different over time. With such a modelling approach,

biophysical control treatments may help both calibration and

validation of the models, but are no longer essential. However,

farming system controls will remain essential as farmers’ behaviour

will usually involve too many unknowns and will hardly be

predictable by models. We therefore advocate the inclusion of

various control treatments in agroforestry experiments, provided

that an « open  » modelling approach is used for the interpretation of

the results,and that a clear distinction is made between biophysical

controls and farming system controls.

The « open  » modelling approach should serve :

to simulate any intermediate design between the intensive

association and the monocrops, including the sole crop or

pure trees control plots when setting the parameters of the

absent element to zero;

to use the farmer’s actions as command variables (such as the

evolution of tree -crop distances);

to simulate tree and crop characteristics over the entire

life of the trees, as this is the only way to get a complete

diagnosis of the system.Annual Land Equivalent Ratios must

21

be integrated over the whole tree rotation,as illustrated in

Figure 2,where the first 6 years are actual data from a Prunus

avium - Festuca arundinacea experiment in Notre-Dame de Londres,

30 km north of Montpellier (Dupraz et al,1995).The curves have

been extrapolated (Dupraz and Newman, 1997) using the

following assumptions: no impact of the trees on the sward

yield until year 10;sward yield decreases to a 20% relative

yield at year 30;this 20% relative yield is maintained until

tree harvest at age 60; the decrease results from an

additional 3m wide strip not being sown every 6 years from

year 10 to year 30; 100 final trees are harvested in the

agroforestry stand versus 125 in the forestry stand; final

trees have the same per tree timber volume at harvest time;

thinning products are not included in the calculation;timber

annual yields are computed as timber volume increments.LER

in year n is calculated with the standing timber volume,

integrating the timber increments from year 1 to year n.

Therefore,LER in year n is not equal to the sum of the tree and

sward partial yields (PY) in year n. The value of the LER

integrated over time is very sensitive to the tree - crop

performance,as illustrated by the impact of the duration of

22

the intercrop (Figure 3),and by the possibility to obtain a

residual pasture production under the trees after the end of

the intercropping phase as in option 1 of Figure 3.

to include any new factor arising from the management history

of agroforestry and control plots. It may be assumed that

object-oriented models are more versatile in this aspect.

Such an approach was adopted in developing the ALWAYS

silvopastoral stand model (Bergez and Msika,1996,Balandier

et al,1998).

Forsaking factorial experiments and focusing only on studying

processes and interactions in « promising  » tree/crop associations,

as advocated by some authors (Sanchez, 1995), may not result in

operative know-how,as the integration of processes is usually out of

reach of the available models. Therefore a combined approach with

basic interactive processes elucidated by specific experiments,and

on-farm field experiments to validate or integrate these processes

in a farming system perspective seems advisable.

23

The plus of farming system controls: when

biological interactions are subordinate

Price (1995)stressed a number of situations where agroforestry can

give greater economic benefits than monocrops with no biological

benefits from the plant association. Only farming system controls

may draw attention to the necessity of rejecting biophysically

successful agroforestry systems,such as some alley cropping schemes

(Ong,1994)or the adoption of poorly efficient agroforestry systems.

The success of agroforestry schemes among farmers and landowners in

France for the last three years is modest but real: without any

extension scheme running,more than 300 ha of agroforests have been

planted.The reasons behind the adoption of such practices are often

not related to biological interactions between trees and crops.Two

key motives may be mentioned in Languedoc-Roussillon (Dupraz,1994):

agroforestry plantations are almost free of fire hazards and

agroforestry plantations help to compromise between landowners and

farmers.

Such motives are not easy to put into a factorial experiment design.

Poor agronomic LERs may be an acceptable price to pay to achieve such

targets. Economic LERs taking into account the probability of

24

partial failure (or even destruction) of both the non-agroforestry

control treatment and the agroforestry treatment would be more

realistic. LERs that would include a term for the externalities

(resulting utilities not directly related to the production of

goods)generated by the different options should also be considered

(Etienne and Rapey, 1998). In some agroforestry systems, poor

biological productivity of the agroforestry stand would not

discourage the decision makers: the reasons quoted above are of

greater importance.In such circumstances,biological interactions

may not be a key issue.

Conclusions

Many agroforestry experiments do not include all the treatments that

would be required for explaining observed results. Sometimes the

pure tree (forestry) stand is not available (such as in many alley

cropping experiments, or in many temperate agroforestry

experiments),and almost always the low density forest stand (with a

tree density similar to the agroforestry treatment)is not available.

In addition control treatments often have an erratic fate that

introduces many unexpected aspects and disqualifies them as

controls,according to the initial protocol.

25

A clear distinction should be made between:

biophysical controls, aimed at elucidating mechanisms of

tree-crop interactions; they may be implemented both in

controlled conditions and in real farms,if the farmer agrees

to follow a strict protocol;a complete design should include

both additive and substitutive control plots;

farming system controls, aimed at appraising the

acceptability and agronomic efficiency of agroforestry and

forestry schemes in a given farming system; these controls

should always be implemented in a real farm.

Both controls may be used as demonstration plots, given that

appropriate comments are provided to the visitors. From the

scientific point of view, trying to draw biophysical conclusions

from farming system controls (or the reverse) is generally a

rejection criterion for papers submitted to scientific journals.An

« open  » modelling approach that could include the shifts in control

treatments may help to take advantage of these situations.

Biologically poorly efficient agroforestry systems (as

demonstrated by biophysical controls) may however be readily

adopted by farmers (as evidenced by farming system controls)if they

26

meet a critical need,such as providing fire protection or avoiding

land tenure conflicts. Biophysical and farming system controls

should therefore be set up concurrently when possible.

Acknowledgements

We express our thanks to Götz Schroth,Meine van Noordwijk and Daniel

Auclair for contributing to the improvement of the manuscript with

thorough and constructive suggestions.Special thanks are due to the

farmers and landowners involved in our experimental agroforestry

network: they contributed mostly to these ideas through their

commitment in the experiments with a critical mind.

References

Bagnall-Oakeley H, Conroy C, Faiz A, Gunawan A, Gouyon A, Penot E,

Liangsutthissagon S,Nguyen HD and Anwar C (1997).Imperata management

strategies used in smallholder rubber-based farming systems.

Agroforestry Systems 36:83-104

Bergez JE and Msika B (1996) A silvopastoral model for the E.U.In:

Western European Silvopastoral Systems, Etienne M (Ed.), INRA

Editions:207-220

27

Balandier P and Dupraz C (1998) The growth of widely spaced forest

trees :a case study from recent temperate agroforestry plantations

in France.Agroforestry Systems This issue

Balandier P,Bergez JE and Etienne M (1998)Use of the ALWAYS computer

model to test management practices on silvopastoral plots.

Agroforestry Systems This issue

Cannell MGR, Van Noordwijk M and Ong CK (1996) The central

agroforestry hypothesis:the trees must acquire resources that the

crop would not otherwise acquire.Agroforestry Systems 34:27-31

Dupraz C (1994) Prospects for easing land tenure conflicts with

agroforestry in Mediterranean France: a research approach for

intercropped timber orchards. Agroforestry Systems 25:181-192

Dupraz C and Lagacherie M (1990)Culture de feuillus à bois précieux

en vergers pâturés sur des terres agricoles du Languedoc-Roussillon:

le réseau expérimental APPEL.Forêt Méditerranéenne 12 (4):447-457

Dupraz C,Dauzat M,Girardin N and Olivier A (1995). Root extension of

young wide-spaced wild cherry trees in an agroforest as deduced from

the water budget. In :Ehrenreich JH, Ehrenreich DL and Lee HW (eds)

Growing a sustainable Future,pp 46-50.Proceedings of the 4th North-

American Agroforestry Conference,Boise,Idaho,USA

28

Dupraz C, Bergez JE and Balandier P (1997)Improved ventilated tree

shelters as a key tool for innovative agroforestry practices in

Europe.In:Proceedings of the Montpellier International Conference

on Agroforestry for Sustainable Land-use,June 1997,pp 275-280.Inra-

Cirad editors,Montpellier,France

Dupraz C and Newman SM (1997) Temperate Agroforestry: The European

Way.In:Gordon A and Newman S (eds)Temperate Agroforestry Systems,pp

181-236.CAB International,Wallingford,Oxon,UK

Etienne M and Rapey H (1998)Simulating integration of agroforestry

into livestock farmers’ projects in France. Agroforestry Systems

This issue.

Fernandes ECM,Davey CB and Nelson LA (1993)Alley cropping on an acid

soil in the upper Amazon:mulch,fertiliser,and hedgerow root pruning

effects.In:Technologies for sustainable agriculture in the tropics.

ASA Special publication 56, pp 77-96. Madison, American Society of

Agronomy

Fukai S and Trenbath BR (1993) Processes determining intercrop

productivity and yields of component crops.Field Crops Research 34:

274-271

29

Haggar JP,Tanner EVJ,Beer JW and Kass DCL (1993)Nitrogen dynamics of

tropical agroforestry and annual cropping systems.Soil Biol Biochem

25:1363-1378

Hiebsch CK and McCollum RE (1987) Area x time equivalency ratio: A

method for evaluating the productivity of intercrops.Agron.J.79:15-

22

Huxley PA (1990) Experimental agroforestry. In: MacDicken KG and

Vergara NT (eds)Agroforestry Classification and Management,pp 332-

353.John Wiley and Sons,New York,USA

Langton S (1990)Avoiding edge effects in agroforestry experiments;

the use of neighbour-balanced designs and guard areas.Agroforestry

Systems 12:173-185

Mead R and Willey RW (1980)The concept of a "Land Equivalent Ratio"and

advantages in yields from intercropping.Expl.Agric.16:217-228

Ong CK (1996)A framework for quantifying the various effects of tree-

crop interactions. In: Ong CK and Huxley P (eds) Tree-Crop

Interactions: A Physiological Approach, pp 1-23. CAB International,

Wallingford,UK

Ong C (1994)Alley farming:pies in the sky? Agroforestry Today 6(3):8-

11

30

Price C (1995) Economic evaluation of financial and non-financial

costs and benefits in agroforestry development and the value of

sustainability.Agroforestry Systems 30:75-86

Ranganathan R and de Wit CT (1996)Mixed cropping of annuals and woody

perennials: an analytical approach to productivity and management.

In:Ong CK and Huxley P (eds)Tree-Crop Interactions:A Physiological

Approach,pp 25-50.CAB International,Wallingford,UK

Sanchez PA (1995)Science in agroforestry.Agroforestry Systems 30:5-

55

Schroth G and Zech W (1995)Root length dynamics in agroforestry with

Gliricidia sepium as compared to sole cropping in the semi-deciduous

rainforest zone of West Africa.Plant Soil 170:297-306

Van Noordwijk M and Ong C (1996)Lateral resource flow and capture -

the key to scaling up agroforestry results.Agroforestry Forum,7(3):

29-31

Vandermeer J (1989) The Ecology of Intercropping. Cambridge

University Press,237 pp

31

Captions to Figures

Figure 1: Agroforestry to forestry control height and diameter

growth ratios of Prunus avium trees in two agroforestry experiments

in Southern France.Intensively (NDL)managed and extensively (POMY)

managed forestry control plots led to opposite conclusions

regarding the efficiency of agroforestry, while the growth of

agroforestry trees was similar at the two sites. Dissymetrical

confidence intervals of AgroForestry (AF) to Forestry Control (FC)

ratios are indicated (P=0.05)

Figure 2: Time change of Land Equivalent Ratios (LER) in an

agroforestry stand.( Prunus avium trees in a Festuca arundinacea sward).

As the trees grow older,the integrated LER is not equal to the sum of

the annual partial yields (PY)for intercrop and trees.

Figure 3:LER of an agroforestry stand as influenced by the duration

of the intercrop. Two options are compared regarding the pasture

production under the trees after the removal of the intercrop:a 20%

relative yield in option 1, and no pasture production in option 2.

(Calculations based on the same data and hypotheses as in figure 2).

32

33

Table 1: Planting and tending strategies to tackle tree

failure risk in agroforestry plantations

Agroforest

scheme

Plantation/

final tree

ratio

Beating up

policy

Intensity of tree

management

Conservativ

e

6-8 first year low (forestry like)

Prudent 4 2 first years intermediate

Risky 2 5 first years intensive

Daring 1 permanent very intensive

(horticulture like)

34

Table 2: A dismal record of the fate of the forestry control

treatment in Languedoc-Roussillon on-farm agroforestry

experiments

35

Forestry control treatment as expected for ?

Site Weeding Protection against wild

animals

Unexpected event

Campagne NO Never weeded NO Browsed by deer

No tree growth

Cassagnas NO Never weeded NO Browsed by red

deer

All dead,forestry control treatment lost

Corneilla del Vercol

NO Trees destroyed by

mechanicalweeding

NO Browsed by cows

Poor pruning scheme,forestry control treatment lost

Notre-Dame de Londres

NO Full (excessive)

weeding during the

first 4 years, then nothing

YES Flood,agricultural control treatment lost

Pomy NO Weeded onlyonce

YES

Portes NO Weeded onlytwice

NO Browsed by deer

All dead,forestry control treatment lost

Restinclières

NO Date of operation

veryvariable

YES

Valmanya NO No weeding during the

first 3 years

YES

Vézénobres NO Not weeded YES First year with an intercrop in the forestry control and no

intercrop in the agroforestryplot

The weeding protocol included spot weeding during the first 5 years,and the

protection of trees against browsing was to be secured by a fence.

36

0.2

0.6

1

1.4

1.8

2.2

-1 0 1 2 3 4 5 6

YEARS AFTER PLANTATION

TREE HEIGHT

POM YNDL

0.2

0.6

1

1.4

1.8

2.2

-1 0 1 2 3 4 5 6

YEARS AFTER PLANTATION

TREE DIAMETER

37

0.9

1

1.1

1.2

1.3

1.4

1.5

Rela

tive

val

ues

0 10 20 30 40 50 60 Years

LER PY(crop) + PY (tree)

0.8

0.9

1

1.1

1.2

1.3

1.4

LER

10 20 30 40 50 60 70 80 90 100 Duration of intercrop (% of tree life)

Option 1 Option 2