Sms

Ma

b

c

a

ARRA

KBRTECN

1

Da1b

h0

Field Crops Research 171 (2015) 1–10

Contents lists available at ScienceDirect

Field Crops Research

jou rn al hom epage: www.elsev ier .com/ locate / fc r

hort- and medium-term impact of manual tillage and no-tillage withulching on banana roots and yields in banana-bean intercropping

ystems in the East African Highlands

.T. Mulielea,c, C.L. Bieldersa,∗, P.J.A. van Astenb

Earth and Life Institute, Université catholique de Louvain, Croix du Sud 2, L7.05.02, B-1348, Louvain-la-Neuve, BelgiumInternational Institute of Tropical Agriculture, P.O. Box 7878, Kampala, UgandaInstitut National pour l’Etude et la Recherche Agronomiques, B.P 2037, Kinshasa/Gombe, Democratic Republic of the Congo

r t i c l e i n f o

rticle history:eceived 1 April 2014eceived in revised form 24 October 2014ccepted 25 October 2014

eywords:ananaootsillageast African Highlandsonservation tillageematode

a b s t r a c t

Banana-bean intercropping systems are common in the bimodal rainfall areas of the East African High-lands and are characterized by low banana productivity. In these systems, the soil is tilled manually twicea year before bean planting with potentially damaging effects to the shallow banana root system. No-tillage with mulching (NT + M) may constitute an interesting alternative to conventional manual tillage(CMT) to avoid such root damage and improve banana productivity. The objectives of this study weretherefore (i) to assess tillage-induced damage to the banana rooting system and its subsequent recovery,and (ii) to evaluate the impact of three NT + M systems vs. CMT on banana root distribution and bananabunch weight. At two sites in the D.R. Congo, the cord root length density (RLD) and fresh weight (FW)were monitored monthly in the top 0.2 m of the soil over a 5 to 6-month period following manual tillage,and compared with NT + M plots. Immediately after tillage and on average over the two sites, cord RLDand FW in the top 0.1 m of the soil were reduced on average to 15% and 16%, respectively, of the levelsobserved under NT + M. At 0.3 m from the rhizome, cord roots needed 2–4 months to recover to a levelsimilar to the one observed prior to tillage. On average over the two sites, direct root damage by tillagecaused the loss of 47% and 63% of the RLD and FW observed in NT + M plots, respectively. The remainingrooting deficit (38% of RLD and 21% of FW) was hypothesized to originate from differences in root growthconditions unrelated to immediate mechanical root damage. There was no evidence that the mechanicaldamage of roots by tillage affected banana growth in the short term. The medium-term effect of CMT andNT + M treatments was evaluated at three sites (two in D.R. Congo and one in Rwanda) 30 months afterbanana planting. At two sites out of three, root density profiles indicated lower rooting densities in the top

0.1 m of the soil in CMT plots compared with NT + M plots. Banana bunch weight was consistently lowerin CMT plots compared with NT + M plots. Compared with NT + M, CMT appears to affect banana rootingand bunch weight in the medium term under the pedo-climatic conditions of the East African Highlands.No-tillage with mulching may constitute an alternative to manual tillage to enhance the sustainabilityof these systems but its impact on the whole intercropping system’s productivity must be verified.

. Introduction

In the East African Highlands (Uganda, Rwanda, Burundi, eastern.R. Congo, north-western Tanzania and western Kenya), bananas

re an essential commodity for 30 million people (Karamura et al.,998). The East African Highland Bananas (Musa AAA-EA) dominateanana production. They are predominantly grown on small-holder

∗ Corresponding author. Tel.: +32 10 473714.E-mail address: charles.bielders@uclouvain.be (C.L. Bielders).

ttp://dx.doi.org/10.1016/j.fcr.2014.10.015378-4290/© 2014 Elsevier B.V. All rights reserved.

© 2014 Elsevier B.V. All rights reserved.

farms, occupy 30% of the land (van Asten et al., 2004) and sup-ply on average between 16% and 31% of the population’s calorierequirement (Abele et al., 2007). Bananas are also an importantsource of income for farmers, transporters, traders and processorsof banana and banana by-products (Karamura and Karamura, 1994;Eledu et al., 2004; Jagwe et al., 2008).

Two main banana-based cropping systems can be identified insmall-holder agriculture in the East African Highlands: sole crop-

ping of bananas, and intercropping systems. Bananas are frequentlyintercropped with annual crops, of which bush beans (Phaseolusvulgaris) are the most common (Dowiya et al., 2009). Before plant-ing of the beans, the entire space in between banana plants is tilled

2 rops R

maD0moh

ansttwaatBr(c

nmcf(phreRa

ptal0Sdmrepsfteossptrthwpebo(vet

M.T. Muliele et al. / Field C

anually using a hand hoe to prepare the seedbed. This is done bi-nnually (approx. September and February) in Rwanda, Burundi,.R. Congo and Uganda. The soil is typically tilled to a depth of.15–0.2 m (Dowiya et al., 2009), up to 10 cm from the bananaat. All residues present at the soil surface are generally taken

ff the field to facilitate tillage. This soil management practice isenceforth referred to as conventional manual tillage (CMT).

In banana-bean intercropping systems, tillage is clearly directedt the bean and not at the banana. Indeed, sole banana crops areot tilled, and weed control is achieved through (self-) mulchingupplemented with occasional superficial weeding. According tohe farmers, tillage in banana–bean intercropping systems favorshe beans’ performance, presumably by reducing competition witheeds and by providingfavorable seedbed conditions. Tillage may

lso reduce below ground competition with banana roots for waternd nutrients, bananas having been shown to be more competitivehan beans in the intercropping system (Wortmann et al., 1992).y improving the contact between the soil matrix and organicesidues, tillage may also accelerate organic matter decompositionBalesdent et al., 2000), thereby liberating additional nutrients forrop production.

Besides these positive effects, tillage may also have a number ofegative effects. Tillage has been associated with low soil organicatter (SOM) contents (Lal and Kimble, 1997). Decreased SOM

ontent may lead to reduced aggregate stability, increased sur-ace sealing and hence an increased risk of runoff and soil erosionGlenn and Welker, 1989; Gosai et al., 2009). When surface mulch isresent, which is always the case in banana sole cropping in small-older farms, banana plots are not prone to erosion. However, theemoval of the mulch and frequent tillage in CMT may stronglynhance the risk of runoff and erosion (Rishirumuhirwa, 1993; Leoux et al., 2005). The removal of crop residues prior to tillage willlso enhance nutrient depletion, and accelerate soil degradation.

When carried out in established banana plantations rather thanrior to banana planting, tillage may also damage the banana roo-ing system, as reported for other perennial crops (e.g., Lipeckind Berbec, 1997). Indeed, bananas are known to have a shal-ow root system, with about 45–85% of the root mass in the top.3 m (Blomme, 2000; Draye et al., 2005; Robinson and Galánaúco, 2010). Hence, the bi-annual tillage is likely to repeatedlyamage an important fraction of the banana root system in theost nutrient-rich topsoil. This practice, which is akin to shallow

oot pruning, may reduce the plants’ ability for water and nutri-nt uptake, which is especially relevant in much of the bananaroducing areas of eastern Africa given the generally low fertilityoils (Delvaux, 1995; van Asten et al., 2004) and sub-optimal rain-all conditions (rainfall < 2000 mm yr−1 and not well distributedhroughout the year; Robinson and Galán Saúco, 2010; van Astent al., 2011). Robinson and Alberts (1989) reported that up to 80%f water taken up by bananas originated from the top 0.3 m of theoil, highlighting the functional importance of roots in the top-oil. Root injury by tillage may possibly also increase the risk oflant root infection. Furthermore, regeneration of the rooting sys-em following tillage-induced damage will also require additionalesources on behalf of the plant and may restrict plant produc-ivity, as reported for other plants (Schroth, 1999). Many studiesave documented significant positive correlations between buncheight (or banana yield) and below ground biomass for bananalants (e.g., Lassoudière, 1978; Serrano and Marín, 1998; McIntyret al., 2000; Blomme et al., 2002), implying that any decrease in rootiomass may lead to decreased production. Furthermore, the vigorf the main pseudostem strongly determines the vigor of suckers

Blomme, 2000), such that better root systems leading to moreigorous main pseudostems would lead to more vigorous suck-rs. Having more vigorous suckers is an advantage, as this reduceshe duration between two successive banana harvests from the

esearch 171 (2015) 1–10

same mat, and hence could increase plantation productivity perunit space and time (Blomme, 2000).

As early as in the beginning of the 20th century, Higgins (1904)noted that the replacement of banana roots damaged by tillage“must make very considerable demands upon the stores of foodin the corm”. He thus advised that “plowing should not be doneat a time when all the supplies of stored food are required for thedeveloping of the flower-bud or of the fruit”. This recommendationis still being issued by some researchers today (Robinson and GalánSaúco, 2010; Lassoudière, 2012). However, it seems to be basedmore on common sense (in view of the shallow banana rooting sys-tem) than on actual scientific evidence of direct negative impactsof tillage on the banana rooting system and banana productivity.Like Higgins (1904) and Lassoudière (2012) recently acknowledgedthat the practice of tillage in existing banana plantations remains adebated issue.

No-tillage with mulching (NT + M) has been much promoted asan alternative to conventional tillage in many temperate and trop-ical cropping systems. Benefits of NT systems include preservingsoil structure; savings of time, energy and water; water erosioncontrol; and maintaining biological activity and diversity (Derpschet al., 2010). In addition, NT avoids disturbing the rooting system ofperennial crops. Mulching with organic residue provides weed con-trol (Lipecki and Berbec, 1997), stimulates soil biological activity,helps maintain SOM levels and conserve water, supplies nutrients,and further contributes to improving soil physical and physico-chemical properties (Erenstein, 2002). By helping maintain morefavorable moisture conditions, mulching may also favor denser rootexploration near the soil surface. NT + M may thus constitute aninteresting alternative to CMT in banana-bean intercropping sys-tems.

Hence, the objectives of this study were two-fold: (i) to assessthe short-term, tillage-induced damage to the rooting system ofEast African Highland banana plants and its subsequent recovery,and (ii) to evaluate the impact of NT + M systems vs. CMT on bananaroot parameters, plant growth and banana bunch weight.

2. Materials and methods

Two types of experiments were carried out. (1) Field experi-ments at two different locations specifically aiming at evaluatingbanana rooting dynamics over a 5–6-month period followingtillage. These are further referred to as ‘banana root growth exper-iments’. (2) Field experiments at three different locations aimingat assessing the effect of CMT and three different NT + M soil man-agement practices on banana root distribution. These are furtherreferred to as ‘NT + M experiments’.

2.1. Study sites

Banana root growth experiments were carried out at twolocations: (1) at Kabamba (South-Kivu, D.R. Congo) in three long-established (>30 years), small-holder banana plantations, and (2) atMulungu (South Kivu), in selected treatments of the NT + M exper-iment described below.

The NT + M experiments were carried out at three sites, namelyat Mulungu and Walungu (South Kivu), and Rubona (Rwanda). AtMulungu and Rubona, the researcher-managed experiments werecarried out on the research stations of the national research institu-tions (Institut National pour l’Etude et la Recherche Agronomiquesand Institut des Sciences Agronomiques du Rwanda, respectively).

At Walungu, the researcher-managed experiment was laid out ina farmer’s field. In the previous season, the land had been grownwith sweet potato at Mulungu, and an association of annual crops(sorghum, taro, sweet potato, bean) at Walungu and Rubona.

M.T. Muliele et al. / Field Crops Research 171 (2015) 1–10 3

Table 1Main characteristics of the study sites.

Site and soil characteristics Experimental site/country

Mulungu/DR Congo Walungu/DR Congo Rubona/Rwanda Kabamba/DR Congo

Latitude 2.335◦S 2.441◦S 2.486◦S 2.11◦SLongitude 28.788◦E 28.411◦E 29.772◦E 28.51◦EAltitude (m) 1699 1638 1630 1570Annual rainfall (mm/an) 1700 1650 1170 1650Mean annual air temperature (◦C) 19.1 19.0 19.2 19.1Soil-Parent material Volcanic ash Basalt Granite Volcanic ash-Classification (FAO) Nitisols Ferralsols Acrisols Nitisols/Ferralsols

* Clay

*

mtaAlmKl

2

2

ttbstssptbtbaotego

2

mt(wmmma1tmbwd

-Texture (0–0.4 m) Clay

Source: Gaidashova (2009), Bouwmeester et al. (2009), this study.

All sites are located at similar altitudes and have comparableean annual air temperatures (Table 1). Rainfall has a bimodal dis-

ribution at all sites, with rainfall maxima in October–Novembernd March–April. A marked dry season occurs from June tougust–September. Mean annual rainfall at Rubona is substantially

ower than at the three other sites. The climate is Aw3 (tropical cli-ate with 3 months of dry season) at all sites according to the

öppen classification. Soils are clayey at Mulungu and Walungu,oamy clay at Kabamba and sandy clay loam at Rubona (Table 1).

.2. Experimental layout and management strategies

.2.1. Banana root growth experimentsIn each of the three farms at Kabamba, one plot was selected

hat had previously been cultivated with a banana-bean associa-ion, with tillage (manual hoeing to approx. 0.15 m depth) beforeean planting. Each plot was divided into two sub-plots. The firstub-plot was kept under the traditional banana management sys-em (CMT with export of crop residues), whereas the secondub-plot was converted into a no-tillage with banana self-mulchystem (NT + SM; banana residue recycled as mulch). Each sub-lot consisted of 25 plants spaced between 2 and 3 m. The NT + SMreatment was implemented in February 2009. The entire space in-etween banana plants was planted with beans, up to 10 cm fromhe pseudostem. In the CMT treatment, tillage occurred 7–10 daysefore bean planting. In the NT + M treatments, beans were plantedfter making a small hole in the soil with a stick. No fertilizers,rganic manure or pesticides were applied. At Mulungu, observa-ions were made in the CMT and NT + SM treatments of the NT + Mxperiment (see below). At the start of the observations, plots hadone through three (Kabamba) and five (Mulungu) cropping cyclesf beans.

.3. NT + M experiments

All experiments were setup in April–May 2008. The experi-ental layout was a randomized complete block design with four

reatments and four replications. CMT with export of crop residuessee banana root growth experiments for details) was comparedith three alternative NT + M treatments: NT with banana self-ulch (NT + SM), NT with self-mulch + Hyparhenia diplandra grassulch (NT + HM) and NT with self-mulch + Tripsacum laxuum grassulch (NT + TM). External mulches (Hyparhenia or Tripsacum) were

pplied at the rate of 25 t ha−1 dry matter (DM) in the first year, and2.5 t DM ha−1 yr−1 from the second year onward. A single applica-ion of banana residue mulch was applied at planting in NT + SM to

imic the quantity of self-mulch usually present in an establishedanana plantation. The quantity of banana mulch applied initiallyas based on the average annual smallholder banana mulch pro-uction in each study area (22, 18, and 10 t DM ha−1 at Mulungu,

Sandy clay loam Clay loam

Walungu, and Rubona, respectively; Bouwmeester et al., 2009).Subsequent self-mulching consisted in leaving crop residues, non-functional banana leaves (pruned monthly; lamina surface less than50% green) and pseudostems (after banana harvest) in the field.

Previously disinfected, vigorous sword suckers of banana cul-tivar ‘Injangi’ (AAA-EA cooking banana) at Rubona and ‘Ndundu’(AAA-EA beer banana) at the other sites were planted at a 2 m × 2 mspacing (2500 plants/ha). Each plot comprised 35 banana plants.Bush beans of the local ‘Kirundu’ variety were sown in all treat-ments at a spacing of 0.4 m × 0.2 m, with two seeds per hole(250,000 plants/ha). In the NT + M treatments, beans were plantedafter making a small hole in the soil with a stick. The entire spacein-between banana plants was planted with beans, up to 10 cmfrom the pseudostem. No mineral fertilizers, organic manure orpesticides were applied. De-suckering was performed monthly toeliminate any extra suckers beyond the permitted one plant percycle per mat. The male bud was removed after the complete open-ing of female flowers. Given the abundance of weeds in CMT plots,weeding was done superficially using a hand hoe whereas NT + Mplots were hand-weeded by pulling.

2.4. Measurements

2.4.1. Banana root growth experimentsIn each plot, 4 (Mulungu) and 7 (Kabamba) plants approx. 1 m

high were selected which were more than 6 months from floweremergence. This is because roots are produced continuously untilflowering, after which the rate of root production drops sharply(Beugnon and Champion, 1966; Lavigne, 1987). Before starting thebanana root growth experiments, banana growth indicators (pseu-dostem circumference at soil level (base) and 1 m aboveground,and plant height) were subjected to statistical analysis (Student’st-test), and no significant differences were found between the twotreatments.

The banana’s root system consists of adventitious roots. Primaryroots, also called cord roots, arise from the corm. On these cordroots, secondary and occasionally tertiary lateral roots are formed(Riopel and Steeves, 1964; Draye et al., 2005). Cord roots are mostlyresponsible for anchorage and transport of water and nutrients,whereas lateral roots contribute effectively to the absorption ofwater and nutrients. In the present study, cord roots were sam-pled. According to Robinson and Galán Saúco (2010, p. 51), theoverall effectiveness of a banana plant’s water and nutrient uptakepotential depends directly on the number of cord roots presentand the vigor of root extension through the soil. Hence, cord rootsare a reasonable proxy for the full banana root system. In addition,

given their large size, their determination is much less subject toexperimental error than finer roots.

Observations were done during 5 (Mulungu) or 6 months(Kabamba), at monthly intervals, starting on the day of tillage

4 rops R

(lsgyfias2aeospsttt

aRblaerha

doud

2

ctat

ttmvsaIa4tdc(tprauofiht

M.T. Muliele et al. / Field C

August) and until the next tillage. The banana root system (rootength density and fresh weight) was assessed by means of coreampling (0.3 m diameter; Blomme, 2000). This method provides aood estimate (>80%) of the banana root system (Blomme, 2000),et is less time consuming and destructive than digging up theull root system. Immediately after tillage, banana roots cut dur-ng tillage were carefully collected from the entire tillage depthnd removed from the plot in order to avoid interference withubsequent observations. Eight angles (0◦, 45◦, 90◦, 135◦, 180◦,25◦, 270◦, and 315◦) were defined clockwise in a horizontal planeround each mat. The 0◦ angle was defined in the direction of themergence of the daughter sucker of the selected mat. For a givenbservation date and mat one of the eight possible angles waselected randomly, while avoiding sampling at a previously sam-led location. The center of the metal cylinder was placed at theelected angle and at 0.3 m from the pseudostem, and driven intohe ground in 0.1-m increments up to 0.2 m depth. For each layer,he cord roots were extracted by washing and allowed to drain inhe shade to remove excess water.

Cord root fresh weight and length were measured using a bal-nce (precision: 0.01 g) and a ruler (precision: 1 mm), respectively.oot necrosis was evaluated according to the method describedy Speijer and De Waele (1997). Five randomly selected, 0.1-m

ong functional cord root segments were selected for each matnd depth, and cut in half lengthwise. The root necrosis index wasxpressed as the percentage of the total root length with necroticoot cortex. Banana growth was monitored by measuring planteight as well as the circumference of the pseudostem at its basend 1 m above ground.

Because of non-normally distributed data at certain observationates, Wilcoxon’s rank score test was applied to assess the effectf treatments on the various root parameters. Student’s t-test wassed to assess treatment effects on plant growth. All analyses wereone using SAS 9.2 Enterprise Guide 4.2 software.

.5. NT + M experiment

All observations were made 30 months after banana planting,orresponding to 2 months after the last tillage in the CMT plots. Athe time of measurements, CMT plots had been tilled four (Rubona)nd five (Mulungu and Walungu) times. All measurements pertaino the top 0.4 m of the soil profile.

Banana root distribution was characterized on one represen-ative plant per plot. The representative plant was selected so aso have a pseudostem base circumference equal to or close to the

ean base circumference of all the plants of the plot. A 2-m wideertical soil profile was dug at 0.4 m from the pseudostem of theelected plant. Roots were then carefully exposed and countedfter overlaying a 0.1 m × 0.1 m grid (Delvaux and Guyot, 1989).n the upper 0.2 m of the profile, the diameter of the roots waslso recorded, distinguishing between roots smaller or greater than

mm (Delvaux and Guyot, 1989). For each 0.1-m depth increment,he total number of roots was determined, which was used toerive (1) the average root density per layer (roots m−2), (2) theumulative number of roots down to a given depth (roots m−1),3) the percentage roots present in a given layer compared withhe total number of roots in the profile, and (4) the cumulativeercentage of roots up to a given depth. The use of percentagesather than absolute values facilitates comparison of treatmentsnd sites in terms of shape of the root distribution, whereas these of absolute values is important to highlight differences basedn the number of roots. Because root densities are quite variable

rom one profile to another and because the 0.1 m layer thicknesss somewhat arbitrary in comparison to the thickness of the soilorizons, cumulative data with depth might in some cases bet-er reveal treatment effects than layer-by-layer data. Hence, we

esearch 171 (2015) 1–10

opted for an analysis of both layer-by-layer data and cumulativevalues.

Penetration resistance was measured vertically at 0.1 m incre-ments using a hand penetrometer (type IB, #06.06.10 EijkelkampAgrisearch, 0.5 cm2 cone area, cone angle: 30◦). To minimize theeffect of differences in water content on penetration resistance,three 0.2 × 0.2 m holes were dug to the desired depth, and the soilin the hole was wetted using 1 L of water. Five measurements weremade in each hole, one hour after applying the water. In addition,three 100 cm3 undisturbed soil cores were sampled at 0.1-m depthincrements to determine the soil bulk density.

Banana bunches (1st crop cycle) were collected at Mulungu,Walungu and Rubona, and were used to determine the mean freshbunch weight per treatment. Bunch weight data was not collectedat Kabamba.

Data analysis was carried out using SAS 9.2 Enterprise Guide 4.2Software. Analysis of variance (ANOVA) was performed for each siteand soil depth separately. Significantly different means (P < 0.05)were separated using the Student–Newman–Keuls test.

3. Results

3.1. Banana root growth experiments

3.1.1. Root fresh weight and root length densityIn order to smooth out temporal fluctuations, root length den-

sity and fresh weight in the CMT treatment were expressed as apercentage of root length density or fresh weight in the NT + SMtreatment. At Mulungu, immediately after tillage, the mean cordroot fresh weight and length density in the top 0.1 m of the soilin the CMT plots corresponded respectively to 6% and 5% of themean fresh weight and length density in NT + SM plots (Fig. 1). Atthis depth, cord root fresh weight and length density increasedup to the third month after tillage, after which both root lengthdensity and fresh weight remained fairly constant. Throughout theobservation period, mean cord root fresh weight in the top 0.1 mwas significantly lower (P < 0.05) in CMT plots than in the NT + SMplots, whereas cord root length density in the CMT treatment dif-fered significantly from the NT + SM treatment only on the firsttwo observation dates as well as on the last date (5th month). At adepth of 0.1–0.2 m, there was a general tendency for the cord rootlength density and fresh weight to increase throughout the obser-vation period relative to the NT + SM treatment (Fig. 1). However,no significant difference was found between the two treatments.

At Kabamba, immediately after tillage, the mean cord root freshweight and length density in the top 0.1 m of the soil in the CMTplots corresponded respectively to 26% and 27% of the mean freshweight and length density in NT + SM plots (Fig. 2). A gradualincrease in cord root fresh weight and length density in CMT plotswas observed until the 5th month after tillage, but significant dif-ferences between treatments were observed only at the start (forroot length density and fresh weight) and 1 and 3 months after thestart of the observations (for root fresh weight only). As opposedto Mulungu, the effect of tillage at Kabamba extended into the0.1–0.2 m soil layer. Tillage significantly reduced the mean freshweight (−41%) and length density (−49%) of the cord roots imme-diately after tillage (Fig. 2). For this soil layer, the cord roots inCMT plots recovered to levels similar to the NT + SM plots one (rootlength density) and two (root fresh weight) months after tillage.

At neither site did the banana’s rooting parameters (0–0.1 m)in the CMT treatments recover to the same level as in the

NT + SM treatments in between two successive tillage operations(Figs. 1 and 2). On average after 4 to 5 months, i.e., upon reachinga near constant value, cord root fresh weight and root length den-sity in CMT at Mulungu reached only 45% and 67%, respectively, of

M.T. Muliele et al. / Field Crops Research 171 (2015) 1–10 5

-100

-50

0

50

100

150

200

543210

Months after tillage

Cord

root fr

esh w

eig

th (%

)

CMT 0-10 cm CMT 10-20 cm

(a)

-10 0

-50

0

50

100

150

200

543210

Months after tillage

Co

rd r

oo

t le

ng

th d

en

sity (

%)

CMT 0-10 cm CMT 10 -20 cm

(b)

Fig. 1. Evolution of cord root fresh weight (a) and length density (b) in the con-ventional manual tillage treatment as a percentage of the fresh weight or lengthdb

tK

3

dst

3

iadgttb(

3

3

id

0

50

100

150

200

6543210

Months after tillage

Co

rd r

oo

t fr

esh

we

igh

t (%

)

CMT 0-10 cm CMT 10 -20 cm

(a)

0

50

100

150

200

6543210

Months after tillage

CMT 0-10 cm CMT 10 -20 cm

(b)

Co

rd r

oo

t le

ng

th d

en

sity (

%)

Fig. 2. Evolution of cord root fresh weight (a) and length density (b) in the con-ventional manual tillage treatment as a percentage of the fresh weight or lengthdensity of cord roots in the no-tillage with self-mulch treatment at Kabamba. Error

ensity of cord roots in the no-tillage with self-mulch treatment at Mulungu. Errorars = standard error.

he fresh weight and length density observed in NT + SM (Fig. 3). Atabamba, these values were 79% and 90%, respectively.

.2. Root necrosis

At both sites and for all treatments, depths and observationates, the root necrosis index was low (6.0% ± 0.56; n = 52). Noignificant differences were found between the CMT and NT + SMreatments (data not shown).

.3. Banana growth

Pseudostem circumference and height were used for calculat-ng the pseudostem volume as indicator of plant performance. Atll sites, one month after tillage, the mean volume of the pseu-ostems was similar in both treatments (Fig. 4). At Mulungu, plantsrew faster in the NT + SM plots than in the CMT plots (Fig. 4a) andhe pseudostem volumes started differing significantly betweenhe two treatments as from four months after tillage. In contrast,anana plant growth at Kabamba was similar for both treatmentsFig. 4b).

.4. NT + M experiment

.4.1. Banana root distribution profilesBanana rooting patterns varied widely across study sites, both

n terms of absolute root densities (number dm−2) and rootistribution with depth (Fig. 5). For all sites and treatments, most

bars = standard error.

roots (62–99%) were confined within the top 0.3 m of the soilprofile.

At Mulungu (Fig. 5M), the maximum root density was observedin the 0.1–0.2 m soil layer irrespective of the treatment, averagingbetween 3.0 and 4.2 roots dm−2 depending on the treatment. A sig-nificant treatment effect was observed only for the 0–0.1 m depth(P < 0.01). In the CMT treatment, 9% of the roots were present in thetop 0.1 m, as opposed to 18–26% for the NT + M treatments.

At Walungu, one observes a three-fold increase in root densityfrom the first to the second layer and a sharp decrease in densitybelow 0.3 m (Fig. 5W). Depending on the treatments, the maxi-mum root density varied between 1.6 and 3.2 roots dm−2. In theupper 0.1 m of the soil, root density was not affected by treatments.Below 0.1 m depth, significant treatment effects were occasion-ally observed, but these effects differed depending on the depthand rooting parameter (root density or cumulative root number,expressed in absolute terms or percentages). In none of the casesdid the treatment contrasts mark a clear distinction between theCMT treatment on the one hand and the three NT + M treatments(NT + SM, NT + HM and NT + TM) on the other hand. Hence they arenot further discussed.

At Rubona, banana root density increased gradually in the top0.3 m and then remained near constant (Fig. 5R). Root densitiesnever exceeded 2.0 roots dm−2. Compared with the three NT + Mtreatments, the CMT treatment had a significantly lower (P < 0.05)root density (Fig. 5R) and percentage of roots (4.3% vs. 13–15%)in the upper 0.1 m of the soil. The cumulative percentage of rootsin NT + M treatments (40–41%) remained significantly higher thanin the CMT treatment (21%) up to 0.2 m depth. The percentage of

roots present at the 0.2–0.3 m depth was significantly higher in theCMT treatment (40%) compared with the three other treatments(27–29%).

6 M.T. Muliele et al. / Field Crops Research 171 (2015) 1–10

0%

20%

40%

60%

80%

100%

Roo t FW Roo t LD Roo t FW Root LD

Mulung u Mulungu Kabamb a Ka bamba

roo�ng deficit roots damage d by �lla geremaining live roots

(a)

0

20

40

60

80

100

0 2 4 6 8 10 12 14

Root

leng

th o

r fre

sh w

eigh

t in

CMT

rela

�ve

to N

T+SM

(%

)

Month s

Roo�ng defici t

Remaini ng li ve roots

Rootsdamaged by �lla ge

(b)

Fig. 3. (a) Distribution of banana cord root fresh weight (FW) or length density(LD) in conventional manual tillage (CMT) treatment in the top 0.1 m of the soilimmediately after tillage, relative to the fresh weight or length density of roots inthe no-tillage with self-mulch treatment (NT + SM, = 100%). (b) Schematic concep-tualization of short-term root dynamics following tillage. ‘Remaining live roots’ arethe live roots still present in the soil immediately after tillage. ‘Roots damaged bytillage’ corresponds to the fraction of the roots that were lost by direct mechanicaldamage during tillage but that regenerated in between two successive tillage oper-ations. ‘Rooting deficit’ corresponds to the difference between the roots observed inthe NT + SM treatment and the roots observed in the CMT treatment immediatelyb

oMti

3

sltarbdbrr0b

np

0

2

4

6

8

10

12

6543210

Months after tillage

Pse

ud

oste

m v

olu

me

(d

m³)

CMTNT+SM

(a)

0

2

4

6

8

10

12

76543210

Months after t illageP

se

ud

oste

m v

olu

me

(d

m³)

CMT

NT+SM

(b)

Fig. 4. Growth rate of banana pseudostem in conventional manual tillage (CMT) and

efore the next tillage operation.

A higher proportion of roots ≤4 mm in diameter was generallybserved in CMT plots compared with the NT + M treatments atulungu and Walungu, but not at Rubona (Table 2). On average,

here were 19% and 31% more fine roots in the CMT treatments thann the NT + M treatments at Mulungu and Walungu, respectively.

.5. Soil physical properties

The study sites differed strongly in terms of their bulk den-ity and penetration resistance profiles. Bulk density values wereowest at Mulungu and highest at Rubona (Fig. 6). On the con-rary, penetration resistance values were lowest overall at Rubonand highest at Walungu (Fig. 7). For a given site, the penetrationesistance and bulk density profiles showed similar trends. Bothulk density and penetration resistance tended to increase withepth at Mulungu. At Walungu, bulk density decreased slightlyetween 0–0.1 and 0.1–0.2 m depth, then increased rapidly. Theapid increase below 0.2 m depth was also observed for penetrationesistance at this site. At Rubona, bulk density was lowest between.1 and 0.3 m depth, whereas penetration resistance was lowest

etween 0.1 and 0.2 m depth.

Irrespective of the sites and depths, treatments did not sig-ificantly affect the bulk density of the soil. On the contrary,enetration resistance in the 0–0.1 m layer was significantly

no-tillage with self-mulch (NT + SM) plots (a) at Mulungu and (b) at Kabamba. Errorbars = standard error.

affected by treatments at all three sites (Fig. 7). At this depth, pen-etration resistance was significantly lower in CMT than in NT + SMand NT + HM at Mulungu and Rubona, and NT + SM and NT + TMat Walungu. Below 0.1 m soil depth, no significant differences inpenetration resistance were found between treatments.

3.6. Banana bunch weight

On average, banana bunch weight was similar at Rubona(12.6 ± 6.6 kg) and Mulungu (13.2 ± 3.1 kg) and lower at Walungu(8.1 ± 1.8 kg). There were significant treatment effects on bananabunch weight (P ≤ 0.001) at all three sites. At each site, bunchweight in the CMT treatment was significantly lower than in allthree NT + M treatments, with the exception of Walungu whereCMT and NT + HM were not significantly different. At Walungu andRubona, banana bunch weight tended to increase with increasingroot density (Fig. 8). At Mulungu, this trend was least marked.

4. Discussion

Both sets of experiments revealed that treatments mostlyaffected roots in the top 0.1 m of the soil (Figs. 1, 2 and 5). Thisis consistent with the fact that manual tillage was used. Hence thediscussion will focus on that layer.

The field measurements revealed a strong negative impact oftillage on the cord root length density and fresh weight (Figs. 1–3).Root damage caused by tillage has been reported in fruit crops otherthan banana. For instance, Glenn and Welker (1989) reported areduction in root density of peach trees (Prunus persica) within thesoil depth affected by tillage. In fruit tree plantations, root pruningis a practice whose effects on the root system may be compara-ble to those of tillage. For apple trees, Ferree (1994) reported that

the density of fine roots (<1 mm) had decreased by 20% and coarseroots (>1 mm) by 40% in the 0–0.3 m soil layer after 9 years of prun-ing. These losses are smaller than in the present study (Fig. 3),which may relate to differences in rooting characteristics, tillage

M.T. Muliele et al. / Field Crops Research 171 (2015) 1–10 7

0

1

2

3

4

5

0-10 10-20 20-30 30-40

Depth (cm)

Roo

ts/d

m2

CMT NT+SMNT+HM NT+TM

a

bcb

c

M

0

1

2

3

4

5

0-10 10-20 20 -30 30 -40

Depth (cm)

Roo

ts/d

m2

CMT NT+SMNT+HM NT+TM

aab

bb

W

b

abab

0

1

2

3

4

5

0-10 10 -20 20 -30 30 -40

Depth (c m)

Roo

ts/d

m2

CMT NT+SMNT+HM NT+TM

b

ab b

R

Fig. 5. Banana cord root distribution in the top 0.4 m of the soil profile at Mulungu(M), Walungu (W), and Rubona (R) as a function of treatment. CMT = conventionalmanual tillage. NT + SM, NT + HM and NT + TM refer to no-tillage with self-mulch,Hyparhenia mulch and Tripsacum mulch, respectively. Measurements were done 30months after planting, and 2 months after last tillage in CMT. Error bars = standardea

itbs

TPWmmNw

Ft

0

10

20

30

40

1.51.41.31.21.110.90.8

Dep

th (c

m)

Bulk density (g cm-3)

Mulungu

Walung u

Rubona

Fig. 6. Soil bulk density profiles at Mulungu, Walungu and Rubona. Curves are aver-

rror. For a given depth, means with the same letters are not significantly differentt = 0.05 level. If no letters are shown, this means there was no treatment effect.

mplement, depth of tillage and depth of characterization. In addi-ion, tree crops have a permanent root framework, whereas the

anana root system is regenerated with each new sucker produced,uch that the relative impact of tillage may vary over time.

able 2ercentage of roots ≤4 mm in diameter in the top 0.2 m of the soil at Mulungu,alungu and Rubona, as a function of the treatment. Measurements were done 30onths after planting, and 2 months after the last tillage operation in the CMT treat-ent. Values = means ± standard error. CMT = conventional manual tillage. NT + SM,T + HM and NT + TM refer to no-tillage with self-mulch, with Hyparhenia mulch andith Tripsacum mulch, respectively.

Percentage of roots ≤4 mm in diameter

Site CMT NT + SM NT + HM NT + TM P-value

Mulungu 70 ± 8a 50 ± 7b 47 ± 8b 55 ± 7ab 0.0048Walungu 92 ± 4a 64 ± 7b 65 ± 5b 54 ± 9b 0.0040Rubona 90 ± 4a 84 ± 6a 83 ± 4a 91 ± 3a 0.4323

or a given site, treatments with the same letters are not significantly different athe 0.05 probability level.

ages over all treatments. Measurements were done 30 months after planting. Errorbars = standard error.

4.1. Short-term effects of mechanical damage

Conceptually, the reduction in root length density and freshweight in CMT vs. NT + SM plots immediately after tillage can resultfrom immediate mechanical damage to roots on the one hand(‘roots damaged by tillage’), and medium-term (18–30 months)impacts of the CMT vs. NT + M treatment on root growth on theother hand (i.e., ‘rooting deficit’; Fig. 3b). According to this concep-tual representation of root dynamics in CMT, one may infer from theexperimental data that the damage caused by tillage on the rootingsystem (0–0.1 m) in the CMT plots at Mulungu resulted in the lossof 41 and 62% of the root fresh weight and length density observedin NT + SM, respectively (Fig. 3a). At Kabamba, the correspondingfigures are 53% and 63%, similar to the figures of Mulungu. Basedupon the roots present in CMT plots at the end of a tillage cycle(‘roots damaged by tillage’ + ‘remaining live roots’; Fig. 3a and b),91% and 93% (Mulungu) and 67% and 70% (Kabamba) of, respec-tively, the root fresh weight and root length density in CMT arebrutally destroyed by the bi-annual hoeing.

The rate of root recovery was faster in terms of root lengthdensity than in terms of root fresh weight (Figs. 1 and 2). Accordingto Lecompte et al. (2002), the diameter of cord roots decreasestowards the apex, such that younger, shorter roots would havesmaller diameters than older, longer roots. Hence, the roots recol-onizing the surface horizon near the rhizome in CMT plots may atfirst have been finer than mature roots, explaining why soil recolo-nization occured faster in terms of length than in terms of weight.Though root diameter was not measured in the banana root growthexperiments, this statement is supported by evidence from theNT + M experiments. Indeed, 2 months after tillage, cord roots hada smaller diameter in the CMT treatments than in the NT + M treat-ments at two out of three sites (Table 3). The temporal dynamics inspecific root length (root length per unit root fresh weight) providesome additional insight (data not shown). Immediately after tillage,specific root length was similar or only slightly higher in the CMTand NT + SM treatments, reflecting the fact that the roots remainingin the CMT treatment immediately after tillage were of similardiameter or only slightly thinner than those present in the NT + SMtreatment. Thereafter, specific root length first increased duringthe first 2 to 3 months, then decreased again towards values similarto those observed in CMT plots immediately after tillage. The initialincrease in specific root length is due to young thin cord rootscolonizing the soil after tillage. The thickening of these roots overtime thereafter leads to a gradual decrease in specific root length.

At both sites, recovery of the rooting system in the top 0.1 mof the soil at 0.3 m from the pseudostem proceeded over a periodof 2–4 months (Figs. 1 and 2). At greater distances from the

8 M.T. Muliele et al. / Field Crops Research 171 (2015) 1–10

0

10

20

30

40

0 5 10 15 20 25 30 35

Penetration resistan ce (kg/cm2)

De

pth

(cm

)

CMT

NT+SM

NT+HM

NT+TM

M

b baba

0

10

20

30

40

0 5 10 15 20 25 30 35

Penetration resistance (kg/cm2)

De

pth

(cm

)

CMT

NT+SM

NT+HM

NT+TM

W

a bbab

0

10

20

30

40

0 5 10 15 20 25 30 35

Penetration resistance (kg/cm2)

De

pth

(cm

)

CMT

NT+SM

NT+HM

NT+HTM

R

ababa

Fig. 7. Penetration resistance profiles at Mulungu (M), Walungu (W), and Rubona (R)as a function of treatments. CMT = conventional manual tillage. NT + SM, NT + HM andNT + TM refer to no-tillage with self-mulch, Hyparhenia mulch and Tripsacum mulch,respectively. Measurements were done 30 months after planting, and 2 months afterlst

pr(fc(re

illfHm

0

5

10

15

20

25

0 0.5 1 1.5 2 2.5 3

Bana

na b

unch

wei

ght (

kg)

Mean root density (roots dm-2)

Walungu CMT

Walungu NT+M

Mulungu CMT

Mulungu NT+M

Rubona CMT

Rubona NT+M

Fig. 8. Relationship between mean banana bunch weight and mean root density

in place for 1.5 years, corresponding to three bean crop cycles. TheNT + SM treatment did not significantly modify soil chemical prop-erties (Table 3) and penetration resistance (Fig. 7) compared withCMT. Although there is some indication that roots did not recover to

Table 3Selected properties of the topsoil (0–0.2 m) at Mulungu (n = 8) and Kabamba (n = 6).Values = mean ± standard error.

Soil properties Mulungu* Kabamba

Ntot (%) 0.42 ± 0.01 0.47 ± 0.06Ctot (%) 5.15 ± 0.15 6.47 ± 0.62Exch. Ca (cmolc kg−1) 18.98 ± 0.75 11.04 ± 0.96Exch. K (cmolc kg−1) 1.23 ± 0.16 2.52 ± 0.36Exch. Mg (cmol kg−1) 4.47 ± 0.19 6.45 ± 0.55

ast tillage in CMT. Error bars = standard error. For a given depth, means with theame letters are not significantly different at = 0.05 level. If no letters are shown,his means there was no treatment effect.

seudostem, the recovery is likely to take longer. Banana cordoots have been reported to grow at rates up to 3.5 cm day−1

Lassoudière, 1978), and cord roots may extend for several metersrom the pseudostem (Beugnon and Champion, 1966). Hence,olonization of the soil at mid-distance between banana plants1–1.5 m) would start after 1–1.5 month in the best of cases. Theecovery rates reported here therefore correspond to the highernd values of the soil’s re-colonization rates by banana roots.

The dramatic decrease in rooting in the top 0.1 m of the soil andts slow recovery are expected to hamper the plant’s ability to col-ect nutrients and water from the topsoil for several months. This

oss of roots would be particularly harmful during dry periods, andor the low fertility soils (i.e., Ferralsols, Acrisols) of the East Africanighlands whose fertility is largely dependent on the soil organicatter content of the topsoil (Delstanche, 2011). These harmful

(0–0.4 m depth) per treatment at Mulungu and Walungu. CMT: conventional manualtillage, NT + M: no tillage with mulching (self-mulch, Tripsacum mulch or Hyparheniamulch). Error bars = standard error.

effects would be even stronger if tillage coincided with flowering,since root emission drops sharply after flowering (Beugnon andChampion, 1966; Lavigne, 1987). The loss of roots by tillage wouldalso require a certain investment on behalf of the plant to regener-ate the rooting system (Rodrigues and Cabanas, 2009). Finally, froma functional perspective, roots from the topsoil may contributemore than proportionally to the banana plant’s performances sincesecondary roots tend to be absent from cord roots growing atdepth (Blomme, 2000). For all these reasons, tillage-induced dam-age to roots may lead to losses in banana productivity, as has beenreported for other crops (P. persica: Glenn and Welker, 1989; Malusdomesticus: Schupp and Ferree, 1988, 1990; Olea europaea: Tisdall,1989; Vitis venifera: Pickering et al., 2005).

Based on the results of the present study, the extent of thebanana productivity loss from direct root damage must be put inperspective. Indeed, banana root profiles, averaged over all treat-ments at Mulungu, Walungu, and Rubona, indicate that the numberof roots in the top 0.1 m of the soil profile represents only 12–18%of the roots in the top 0.4 m (Fig. 5), and even less knowing thatroots may extend up to 1 m depth at some sites (data not shown).Hence, only a small proportion of the total number of cord roots isactually damaged by tillage. In addition, root biomass commonlyrepresents approx. 20% of the total plant biomass (Blomme, 2000),such that the replacement of the roots in the top 0.1 m of the soil isunlikely to place a large demand on the banana plants.

The experiment at Kabamba is well suited to assess the short-term impact of mechanical damage. At this site, old (>30 years)small-holder banana plantations managed under CMT were con-verted to NT + SM. At the time of observations, treatments had been

c

P available (mg kg−1) 86.39 ± 4.24 54.01 ± 7.93pHH2O 6.3 ± 0.04 6.5 ± 0.05

*Source: S. Bizimana, personal communication. Ntot: total nitrogen, Ctot: total organiccarbon. Ca, Mg, K and P were extracted using Mehlich-3 solution.

rops R

trtpNddtbn

4

t(me3t

mcmcasdwbrr(wobntRpmdrd

tpssfCrhsNttci

4

rab

M.T. Muliele et al. / Field C

he full extent of what is observed in NT + SM plots (Figs. 2 and 3),oot length density and fresh weight in CMT plots at the end ofhe observation period did not differ significantly from the NT + SMlots. Hence, at this site, most of the difference in rooting betweenT + SM and CMT immediately after tillage can be attributed toirect mechanical damage of the roots. Yet, plant growth did notiffer significantly across treatments (Fig. 4). This seems to indicatehat the abandonment of tillage did not procure any short-termenefit to the bananas, i.e., mechanical damage was insufficient toegatively impact plant growth.

.2. Medium-term effects of CMT and NT + M treatments

Especially at Mulungu, mechanical damage alone cannot explainhe total loss in root length density or fresh weight in the top 0.1 mFig. 1). The additional rooting deficit thus appears to relate to

edium-term (30 months) effects of the treatments (Fig. 3). Thexperiments at Mulungu, Walungu, and Rubona, which lasted for0 months, provide some insight into the medium-term impact ofhe treatments on rooting and plant productivity.

At Mulungu, Walungu, and Rubona, the CMT and NT + M treat-ents were implemented at the same time. The land had been

ultivated prior to the experiment, but not with bananas. Measure-ents were carried out after four or five bean crop cycles. In the

ase of Mulungu, where root growth dynamics were monitored, 55% and 33% deficit in root fresh weight and root length den-ity, respectively, could not be explained by immediate mechanicalamage (Figs. 1 and 3). This loss in root length density and fresheight in the top 0.1 m also cannot be explained by the limited time

etween two successive tillage operations (6 months), as the rootecovery curves seem to indicate that rooting characteristics hadeached a fairly constant value by the end of the observation periodsFigs. 1 and 2). Contrary to Kabamba where most of the root lossas related to short-term mechanical damage and no impact was

bserved on plant growth, plant growth was negatively affectedy tillage at Mulungu (Fig. 4). In addition, bunch weight was sig-ificantly lower in the CMT treatment compared with the NT + Mreatments (Fig. 8). The latter was also observed at Walungu andubona. By comparison with Kabamba, it therefore appears that theoorer performance of banana plants at Mulungu largely relates toedium-term effects of the treatments rather than from the imme-

iate mechanical damage of roots. It seems likely that the sameeasoning applies to Walungu and Rubona, even though rootingynamics were not specifically studied at these two sites.

Although there was no measurable evidence that soil proper-ies were less favorable for root growth in CMT than in NT + Mlots at Mulungu (Table 3 and Figs. 6 and 7), better water con-ervation, stronger buffering of soil temperature or better nutrientupply from the decomposing mulches may, for instance, haveavored root growth in the NT + M treatments as compared withMT without measurably altering soil properties. In addition, theepeated mechanical damage over a 30-month period may alsoave weakened the main pseudostems, and consequently theuckers (Blomme, 2000), in CMT plots. Both positive factors inT + M plots and negative factors in CMT plots may thus explain

he observed medium-term changes in rooting between the tworeatments. Further studies will, however, be required to better dis-riminate among the various mechanisms that affect root growthn the CMT and NT + M treatments.

.3. Root necrosis

Both at Kabamba and Mulungu, the extent of necrosis of bananaoots was low, irrespective of depth and treatment. This may reflect

low nematode pressure at these sites, even at Kabamba whereananas had been cultivated for more than three decades on the

esearch 171 (2015) 1–10 9

same plot. For necrosis values ≤50%, Gaidashova (2009) did notfind a significant negative impact of nematodes (Pratylenchus good-eyi) on the East African Highland banana growth and bunch weightat Kibuye, Rwanda. Similar results were reported by Delstanche(2011) and Wairegi (2010). Root damage by nematodes thereforedoes not seem to be a major constraint in banana production at theMulungu and Kabamba sites and it appears not to be enhancedby tillage. This result agrees with earlier observations made byBouwmeester et al. (2009) and Delstanche (2011) who reportedthat nematodes are not a major constraint to banana productionin many areas of the African Great Lakes Region (Rwanda, Uganda,Burundi, Sud-Kivu).

4.4. Comparison of NT + M treatments

Because there were no consistent differences in root develop-ment across the NT + M treatments in the top 0.1 m (Fig. 5), and inview of the lack of significant differences among NT + M treatmentsregarding bunch weight, it appears that the addition of externalmulches was unable to substantially modify the banana rootingsystem and productivity compared with self-mulching within thetimeframe of the experiments. This may be because insufficientquantities of mulch were applied to have a measurable impact,because of insufficient quality of the mulch, or because of slowdegradation of the mulch which was surface applied rather thanincorporated into the soil.

Both sets of experiments indicate that tillage tends to reduce therooting parameters of banana plants immediately after tillage andfor several months thereafter. The data further indicate that, overthe medium term, the rooting system in CMT plots may not fullyrecover to the levels observed in NT + M plots. In addition, bananabunch weight was lower in CMT than in NT + M and there was a ten-dency for bunch weight to increase with increasing root density attwo of the three NT + M experimental sites (Fig. 8). The evidencetherefore points towards a weak short-term impact of tillage, butin the medium or long term the abandonment of tillage and itsreplacement by a NT + M system may be profitable to the bananaplants’ productivity. However, in small-holder farms, the short-term benefits of tillage in banana-bean intercropping systems (i.e.,improved seedbed structure for bean planting and reduced belowground competition) rather than the possible longer-term nega-tive effects may very well drive farmers’ decision making regardingcropping practices. Hence, further studies are needed to investigatethe productivity of the whole intercropping system, and not justthe banana component. In addition, the issue of tillage in bananaintercropping systems should be investigated over a wider range ofagro-pedological conditions, over a longer timeframe and for differ-ent stages of crop development in order to better evaluate whethertillage threatens the sustainability of banana-bean intercroppingsystems.

5. Conclusions

This study demonstrates that tillage causes severe damage to therooting system of banana plants compared with NT + M systems. Inthe case of manual tillage, the damage is mostly restricted to thetop 0.1 m and appears not to favor root necrosis. Recovery of theroot system, both in terms of length density and fresh weight, tookseveral months in the vicinity of the pseudostem, and is likely totake longer at greater distances from the pseudostem. During thisperiod, the plants may be exposed to an increased risk of water and

nutrient deficiency.

In absolute terms, the actual damage to roots at the presentstudy sites remains limited, since only a fairly small percentage ofthe total cord roots were present in the top 0.1 m. When mechanical

1 rops R

daawlba

caotarnbwatttst

A

fsA

R

A

v

v

B

B

B

B

B

D

D

D

D

D

D

0 M.T. Muliele et al. / Field C

amage explained most of the difference in rooting between CMTnd NT + SM, as was the case at Kabamba, no immediate neg-tive impact on banana growth could be evidenced. However,hen the rooting system was not able to recover to levels simi-

ar as those observed in NT + M systems, as observed at Mulungu,anana growth and productivity (bunch weight) appear negativelyffected.

In the medium term, NT + M appears to favor root developmentompared with CMT plots, and may thus provide more favor-ble conditions for banana development. Within the time framef the experiment, there were no clear benefits in terms of roo-ing and productivity from combining banana residue mulch withn external mulch as compared with self-mulching with bananaesidues. Considering the cost of external mulch (labor, opportu-ity cost), NT + M may thus be the preferred alternative to tillage inanana–bean intercropping systems. However, the present studyould have to be complemented with in-depth evaluations of

boveground productivity of both crops in the intercropping sys-em over a longer period of time. In addition, it is recommended thathe study be extended to a wider range of soil and climatic condi-ions, as the (dis-)advantages of (no-)till may very well depend onoil and climatic conditions as well as on the degree of intensifica-ion of the banana production system.

cknowledgments

The authors gratefully thank the Belgian Directorate Generalor Development Cooperation (DGDC) who, through CIALCA (Con-ortium for Improving Agriculture-based Livelihoods in Centralfrica), funded this study.

eferences

bele, S., Twine, E., Legg, C., 2007. Food Security in Eastern Africa and the Great Lakes.Crop Crisis Control Project, Kampala (CRS/IITA).

an Asten, P.J.A., Gold, C.S., Okech, S.H., Gaidashova, S.V., Tushemereirwe, W.K., DeWaele, D., 2004. Soil quality problems in East African banana systems and theirrelation with other yield loss factors. InfoMusa 13, 20–25.

an Asten, P.J.A., Fermont, A.M., Taulya, G., 2011. Drought is a major yield loss factorfor rainfed East African highland banana. Agr. Water Manag. 98, 541–552.

alesdent, J., Chenu, C., Balabane, M., 2000. Relationship of soil organic matterdynamics to physical protection and tillage. Soil Tillage Res. 53, 215–230.

eugnon, M., Champion, J., 1966. Etude sur les racines du bananier. Fruits 21,309–327.

lomme, G., 2000. The Interdependence of Root and Shoot Development in Banana(Musa spp.) under Field Conditions and the Influence of Different BiophysicalFactors on this Relationship. Katholieke Universiteit Leuven, Leuven, Belgium,pp. 183 (Ph.D. dissertation).

lomme, G., Swennen, R., Tenkouano, A., 2002. The effect of soil bulk density on rootand overall plant development in six banana varieties. InfoMusa 11, 38–40.

ouwmeester H., van Asten P. and Ouma E. 2009. Mapping key variables of bananabased cropping systems in the Great Lakes Region: partial outcomes of thebaseline and diagnostic surveys. CIALCA Technical Report 11.

elstanche, S., 2011. Drivers of Soil Fertility in Smallholder Banana Systems in theAfrican Great Lakes Region. Université catholique de Louvain, Louvain-la-Neuve,Belgium, pp. 458 (Ph.D. dissertation).

elvaux, B., 1995. Soils. In: Gowen, S. (Ed.), Banana and Plantains. Chapmans Pub-lishers, London, UK, pp. 230–257.

elvaux, B., Guyot, P., 1989. Caractérisation de l‘enracinement du bananier auchamp. Incidences sur les relations sol-plante dans les bananeraies intensivesde la Martinique. Fruits 44, 633–647.

erpsch, R., Friedrich, T., Kassam, A., Hongwen, L., 2010. Current status of adoptionof no-till farming in the world and some of its main benefits. Int. J. Agric. Biol.Eng. 3, 1–25.

owiya, N.B., Rweyemamu, C.L., Maerere, A.P., 2009. Banana (Musa spp. Colla)

cropping systems, production constraints and cultivar preferences in easternDemocratic Republic of Congo. J. Anim. Plant Sci. 4, 341–356.

raye X., Lecompte F., Pages L., 2005. Distribution of banana roots in time andspace: new tools for an old science. In: Turner D.W., Rosales F.E. (Eds.). BananaRoot System: Towards a Better Understanding for its Productive Management.

esearch 171 (2015) 1–10

Proceeding of an International Network for the Improvement of Banana andPlantain (INIBAP) symposium, INIBAP. Costa Rica, Montpellier, France, pp.58–74.

Eledu, C., Karamura, E., Tushemereirwe, W.K., 2004. Agroecological distribution ofbanana systems in the Great Lakes Region. Afr. Crop Sci. J. 12, 33–42.

Erenstein, O., 2002. Crop residue mulching in tropical and semi-tropical countries:an evaluation of residue availability and other technological implications. SoilTillage Res. 67, 115–133.

Ferree, D.C., 1994. Root distribution following 9 years of root-pruning“Me1rose”/M.26 Apple Trees. HortScience 29, 76–78.

Gaidashova, S., 2009. Effect of Plant Parasitic Nematodes and Arbuscular MycorrizalFungi on Banana (Musa spp.) in the East African Highland Cropping Systems.Université catholique de Louvain, Louvain-la-Neuve, Belgium, pp. 270 (Ph.D.dissertation).

Glenn, D., Welker, W., 1989. Cultural practices for enhanced growth of young peachtrees. Am. J. Altern. Agr. 4, 8–11.

Gosai, K., Arunachalam, A., Dutta, B.K., 2009. Influence of conservation tillage on soilphysicochemical properties in a tropical rainfed agricultural system of northeastIndia. Soil Tillage Res. 105, 63–71.

Higgins, J.E. 1904. The banana in Hawai. Hawaï Agr. Exp. Stat. Bull. 7, Honolulu, 53p.

Jagwe J., van Asten P., Steffen A., 2008. Banana Marketing in Rwanda, Burundi andSouth-Kivu. Technical Report 9. CIALCA Project Survey.

Karamura, E., Frison, E.A., Karamura, D.A., Sharrock, S., 1998. Banana productionsystems in eastern and southern Africa. In: Picq, C., Fouré, E., Frison, E.A. (Eds.),Bananas and Food Security. INIBAP, Montpellier, pp. 401–412.

Karamura D., Karamura E., 1994. A provisional checklist of bananas cultivars inUganda. National agricultural Research organization (NARO). International Net-work for the Improvement of Banana and Plantain. INIBAP, Montpellier, France,28 p.

Lal, R., Kimble, J.M., 1997. Conservation tillage for carbon sequestration. Nutr. Cycl.Agroecosyst. 49, 243–253.

Lassoudière, A., 1978. Quelques aspects de la croissance et du développement dubananier ‘Poyo’ en Côte d‘Ivoire Le système radical. Fruits 33, 314–338.

Lassoudière, A., 2012. Le bananier: Un siècle d’innovations techniques. EditionsQuae, Versailles, France, pp. 352.

Lavigne, C., 1987. Contribution à l’étude du système racinaire du bananier: mise aupoint de rhizotrons et premiers résultats. Fruits 42, 265–271.

Lecompte, F., Vaucelle, A., Ozier-Lafontaine, H., Pagès, L., 2002. Number, position,diameter and initial direction of growth of primary roots in Musa. Ann. Bot. 90,43–51.

Lipecki, J., Berbec, S., 1997. Soil management in perennial crops: orchards and hopgardens. Soil Tillage Res. 43, 169–184.

Le Roux, J.J., Sumner, P.D., Rughooputh, S.D.D.V., 2005. Erosion modelling and soilloss prediction under changing land use for a catchment on Mauritius. S. Afr.Geogr. J. 87, 129–140.

McIntyre, B.D., Speijer, P.R., Riha, P.R., Kizito, F., 2000. Effects of mulching on biomass,nutrients and soil water in banana inoculated with nematodes. Agron. J. 92,1081–1085.

Pickering A.H., Warrington I.J., Woolley D.J., 2005. Root pruning of Cabernet Sauvi-gnon (Vitis vinifera L.) decreases bunch stem necrosis. Limiting factors forprimary production. June, 2005, Lincoln, New Zealand (Conference Presenta-tion). 13 p.

Riopel, J., Steeves, T., 1964. Studies on the roots of Musa acuminata ‘Gros Michel’ 1.The anatomy and development of main roots. Ann. Bot. 28, 475–494.

Rishirumuhirwa, T., 1993. Facteurs anthropiques de l‘érosion dans les régions desmontagnes et hauts plateaux de la C.E.P.G.L (Burundi–Rwanda–Zaïre). RéseauEros.—Bull. 13, 53–62.

Robinson, J.C., Alberts, A.J., 1989. Seasonal variation in the crop water-use coefficientof banana (cultivar ‘Williams’) in the subtropics. Sci. Hortic. 40, 212–225.

Robinson, J.C., Galán Saúco, V., 2010. Bananas and Plantains. In: Crop ProductionScience in Horticulture Series 19, 2nd Edition. CABI, Wallinford, UK, pp. 311.

Rodrigues, M.A., Cabanas, I.E., 2009. Manutenc ão do solo. In: Rodrigues, M.A., Correia,C.M. (Eds.), Manual da Safra econtra safra do olival. IPB, Braganc a, pp. 41–57.

Schroth, G., 1999. A review of belowground interactions in agroforestry, focussingon mechanisms and management options. Agroforest. Syst. 43, 5–34.

Schupp, J.R., Ferree, D.C., 1988. Effect of root pruning at flour levels of severityon growth and yield of ‘Me1rose/26 Apple Trees. J. Am. Soc. Hortic. Sci. 113,194–198.

Schupp, J.R., Ferree, D.C., 1990. Influence of time of root pruning on growth, mineralnutrition, net photosynthesis and transpiration of young apple trees. Sci. Hortic.42, 299–306.

Serrano, E., Marín, D.H., 1998. Disminución de la productividad bananera en CostaRica. Corbana 23, 85–96.

Speijer P.R., De Waele D., 1997. Screening of Musa germoplasm for resistance andtolerance to nematodes. INIBAP Technical Guidelines 1. INIBAP, Montpellier,France, pp. 47.

Tisdall, I.M., 1989. Soil Manag. Acta Hortic. 240, 161–168.

Wairegi, L., 2010. Management Practices and Opportunities in the East African High-

land Banana (Musa AAA-EA) Production in Uganda. Wageningen University, TheNetherlands, pp. 180 (PhD dissertation).

Wortmann, C.S., Sengooba, T., Kyamanywa, S., 1992. Banana and bean intercropping:factors affecting bean yield and land use efficiency. Exp. Agr. 28, 287–294.