G

A

UA

MKIIa

b

c

d

e

f

g

h

a

ARRAA

KCNCASS

1

mgp

d

0h

ARTICLE IN PRESS Model

GEE-4500; No. of Pages 16

Agriculture, Ecosystems and Environment xxx (2013) xxx– xxx

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment

jo ur nal ho me page: www.elsev ier .com/ locate /agee

nderstanding the impact and adoption of conservation agriculture infrica: A multi-scale analysis

arc Corbeelsa,∗, Jan de Graaffb, Tim Hycenth Ndahc, Eric Penota, Frederic Baudrond,rishna Naudina, Nadine Andrieua, Guillaume Chirata, Johannes Schulerc,

saiah Nyagumboe, Leonard Rusinamhodzib, Karim Traore f, Hamisi Dulla Mzobag,van Solomon Adolwah

French Agricultural Research Centre for International Development (CIRAD), Montpellier, FranceWageningen University (WUR), Wageningen, The NetherlandsLeibniz-Centre for Agricultural Landscape Research (ZALF) e.V., Eberswalder Strasse 84, 15374 Müncheberg, GermanyInternational Maize and Wheat Improvement Center (CIMMYT), Addis Ababa, EthiopiaInternational Maize and Wheat Improvement Center (CIMMYT), Harare, ZimbabweInstitute of Environment and Agricultural Research (INERA), Bobo-Dioulasso, Burkina FasoAfrican Conservation Tillage Network (ACT), Nairobi, KenyaInternational Center for Tropical Agriculture (CIAT), Nairobi, Kenya

r t i c l e i n f o

rticle history:eceived 18 February 2013eceived in revised form 28 August 2013ccepted 21 October 2013vailable online xxx

eywords:onservation agricultureo-tillagerop residuesdoptionmallholdersub Saharan Africa

a b s t r a c t

Conservation agriculture (CA) is increasingly promoted in Africa as an alternative for coping with the needto increase food production on the basis of more sustainable farming practices. Success with adoptingCA on farms in Africa has been limited, despite more than two decades of research and developmentinvestments. Through analyzing past and on-going CA experiences in a set of case studies, this paperseeks to better understand the reasons for the limited adoption of CA and to assess where, when and forwhom CA works best. CA is analyzed and understood within a framework that distinguishes the followingscales of analysis: field, farm, village and region. CA has a potential to increase crop yields in the fields,especially under conditions of erratic rainfall and over the long-term as a result of a gradual increase ofoverall soil quality. The impact on farm income with the practice of CA on some fields of the farm is farless evident, and depends on the type of farm. The lack of an immediate increase in farm income with CAexplains in many cases the non-adoption of CA. Smallholders have often short-term time horizons: futurebenefits do not adequately outweigh their immediate needs. Another key factor that explains the limitedCA adoption in mixed crop-livestock farming systems is the fact that crop harvest residues are preferablyused as fodder for livestock, preventing their use as soil cover. Finally, in most case studies good markets

for purchase of inputs and sale of produce – a key prerequisite condition for adoption of new technologies– were lacking. The case studies show clear evidence for the need to target end users (not all farmers arepotential end user of CA) and adapt CA systems to the local circumstances of the farmers, considering inparticular the farmer’s investment capacity in the practice of CA and the compatibility of CA with his/herproduction objectives and existing farming activities. The identification of situations where, when andfor whom CA works will help future development agents to better target their investments with CA.

. Introduction

Conservation agriculture (CA) has increasingly been pro-

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

oted by international organizations, donors, farm and non-overnmental organizations as a means to overcome continuingoor profitability, food insecurity and soil degradation on

∗ Corresponding author at: UR Annual Cropping Systems, CIRAD, Avenue’Agropolis, 34398 Montpellier Cedex 5, France. Tel.: +33 556191974061.

E-mail address: corbeels@cirad.fr (M. Corbeels).

167-8809/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.agee.2013.10.011

© 2013 Elsevier B.V. All rights reserved.

smallholder farms in Africa. The key principles on which CA isbased, namely no or minimum tillage, soil cover with crop residues,and the use of crop rotations or associations are attractive from anagronomic point of view (e.g. Hobbs, 2007; Thierfelder et al., 2013).Retention of crop residues under CA is expected to increase soilcarbon, compared to conventional, tillage-based cropping whereresidues are taken from the field. This is seen as an important pro-

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

cess explaining the increased soil productivity over time under CAcompared to the conventional systems (e.g. Hobbs, 2007). Besides,based on experimental evidence of increased water productiv-ity under sub-optimal rainfall conditions, CA has been attributed

ING Model

A

2 tems a

tca

hdtascspf

GmmfiSfditatd

pe(cteeaseick(iGet

tiaBZCnaiftmioctiwsb

region (Fig. 1). We examined all scales and their interactions withemphasis on the most relevant factors to explain CA adoption orrefusal. Each scale has its own analytical approaches and/or models.The performance of CA at field scale was assessed through analyzing

Fig. 1. Conceptual representation of the determinants of adoption of conservationagriculture. Adoption (A) is conditioned by its technical performance (P), subject to

ARTICLEGEE-4500; No. of Pages 16

M. Corbeels et al. / Agriculture, Ecosys

he potential for mitigating negative effects from future climatehange, when rainfall is projected to decrease and be more unreli-ble (e.g. Thierfelder and Wall, 2010).

Adoption rates of CA by smallholder farmers in Africa remain,owever, low (Kassam et al., 2009) and this, despite more than twoecades of research and development investments. For example,he proportion of the total cropland area under CA in Zambia, Kenyand Zimbabwe is lower than 1%. This is in huge contrast with theituation, for example, in South America where about 50% of theropped area is cultivated without tillage in CA systems. However, ithould be pointed out that adoption of CA in that part of the world isredominantly successful on mechanized, medium- to large-scalearms, and by far less on smallholder farms.

In their review paper on CA and smallholder farming in Africa,iller et al. (2009) questioned whether CA should be so widely pro-oted for smallholder farmers in Sub-Saharan Africa (SSA). Theain concluding argument was that CA systems probably do not

t within the majority of current smallholder farming systems inSA. They argued that CA can offer substantial benefits for certainarmers in certain locations at certain times, recognizing the wideiversity of farmers in terms of resource endowments and farm-

ng systems. Therefore, a challenge that demands for research iso identify where and how particular CA practices may best fit,nd which farmers in any given community are likely to benefithe most (Giller et al., 2011). Answering these questions will helpirecting the investment efforts with CA dissemination.

Although the three CA principles are common to CA systems,utting them into practice can vary substantially across agro-nvironments and farmers. The specific components of a CA systemseeding method, plant density, fertilization, weed control, soilover and type of crops in the rotation or in association) canhus be very different across broad locations and types of farm-rs (e.g. Erenstein et al., 2012; Scopel et al., 2013). This means thatxisting CA systems from elsewhere in the world clearly requiredaptation to the local circumstances and specific contexts ofmallholders in SSA before they can be widely disseminated andxpected to be adopted. Aspects that hereby need to be considerednclude: farmers’ production objectives and constraints, expectedosts (requirements in terms of inputs, equipment, labour but alsonowledge and confidence about the performance of CA), benefitsespecially in the short term), production and financial risk of CA,nput supply and marketing, and technical advice (Erenstein, 2003;iller et al., 2009; Mazvimavi and Twomlow, 2009). A poor consid-ration of the above aspects is in many CA projects accountable forhe ineffective dissemination and for refusal of CA.

The purpose of this paper is to gain a better understanding ofhe reasons for the limited adoption of CA on smallholder farmsn Africa by analyzing past and on-going CA experiences through

number of case studies that were identified across SSA, i.e. inurkina Faso, Kenya, Malawi, Tanzania, Madagascar, Zambia andimbabwe. A better comprehension of why, where, and for whomA works best is a prerequisite for knowing where (and whereot) and for whom (and whom not) to promote large-scale CAdoption in Africa. Through the case studies we have assessed thempact of CA practices and their feasibility for local smallholderarmers at different scales, i.e. from the single field plot level tohe regional scale. To do this, different analytical approaches and

odeling tools were employed with the overall aim to gain betternsights in the major factors and processes that determine whetherr not CA is a viable option for the local smallholder farmers in thease study regions. At field scale, we focus on the biophysical factorshat explain short- and long-term yield responses to CA. These are

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

mportant as they help in identifying the agro-ecological settingshere CA can be successfully promoted. At farm-scale, we assess

hort- to medium-term profitability of CA. Immediate economicenefits are often seen as a major driver behind CA adoption by

PRESSnd Environment xxx (2013) xxx– xxx

smallholder farmers, who often have short-term needs to feed theirfamily. Then, our focus is on trade-offs in the use of crop residuesat this scale and that of the village, since they determine to a largeextent whether or not CA practices fit in the mixed crop-livestockfarming systems of Africa. Lastly, at the scale of the region, we focuson the pre-conditions for a widespread adoption of CA in SSA.

2. Approaches and methods

2.1. Case studies

We selected a number of past and on-going projects on research,development and dissemination of CA practices with smallholderfarmers in SSA, and analyzed each project as a case study. Allprojects were situated in regions where one or more cereals (maize,sorghum, millet or rice) are the main crops grown in mostly mixedcrop-livestock farming systems. The projects were chosen on thebasis of a set of criteria that included the availability of documenta-tion and data on the CA research and dissemination activities in theproject, the interest and impact of the project for the local stake-holders, and the in-depth local knowledge of the project and itsdynamics by the researchers and development agents involved. Thecase studies were considered as specific examples to illustrate moregeneral determinants and principles of CA adoption or refusal bysmallholders in SSA. In total 10 case studies were examined. A briefdescription of the different case studies and the related CA projectsis given in Table 1.

2.2. Overall approach of the multi-scale analysis

For the assessment of the performance of the CA practices andtheir potential for widespread adoption we used a frameworkthat distinguishes the following scales: field, farm and village, and

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

the opportunities and tradeoffs (T) that operate at farm and village scales and con-strained by different aspects of the context (C) in which the farming system operatesincluding market, socio-economic, institutional and policy conditions defining theinnovation system and the variability inherent to the physical environment (e.g.climate change).

Please cite

this

article in

press

as: C

orbeels, M

., et

al., U

nd

erstand

ing

the

imp

act an

d ad

option

of con

servation agricu

lture

in A

frica: A

mu

lti-scalean

alysis. A

gric. Ecosyst.

Environ

. (2013),

http

://dx.d

oi.org/10.1016/j.agee.2013.10.011

AR

TIC

LE

IN P

RE

SS

G M

odel

AG

EE-4500;

No.

of Pages

16

M.

Corbeels et

al. /

Agriculture,

Ecosystems

and Environm

ent xxx (2013) xxx– xxx

3

Table 1Description of the case studies and related projects on conservation agriculture.

Case study Project Project activities Estimated number offarmers involved

Rainfall Principal cropping system

South-western Burkina Faso PRODS/PAIA pilot project, funded byFAO and implemented by INERA

Introduction, adaptation, anddissemination of CA systems throughfarmer field schools

2000 Uni-modal, 800–1200 mm Maize, cotton, sorghum, millet

Lake Aloatra, Madagascar BVLac project, 2003–2013, funded byAgence Franc ais de développement(AFD)

Research, development anddissemination of CA systems throughfarm organisations

3000 Unimodal, 800–1200 mm Rice-based and maize, cassavaand groundnuts

Vakinankaratra, Madagascar BVPI project, 2006–2012, funded byAgence Franc ais de développement(AFD)

Research, development anddissemination of CA systems throughfarm organizations

1000 Unimodal, 1200–1400 mm Rice-based and maize,vegetables

Central Malawi CIMMYT project, and Total Land Careproject, 2005–2013, funded by IFADand BMZ, Germany

Research, development anddissemination of CA systems (dibblestick, jab planter and planting basins)via lead farmer approach

15000 Unimodal 700–1100 mm Maize

Bungoma, Kenya CA-SARD project, 2004–2011, fundedby German Trust Fund andimplemented by ACT and FAO

Adaptation, development anddissemination of CA systems (directseeder and, jab planter) via the farmerfield school approach

500 Bimodal, 500–1000 mm during1st rains and 430–800 mmduring 2nd rains

Maize-based and beans

Karatu, Tanzania CA-SARD project, 2004–2011, fundedby German Trust Fund andimplemented by ACT and FAO

Adaptation, development anddissemination of CA systems (directseeder and jab planter) via the farmerfield school approach

500 Bimodal, 400–1000 mm perseason

Maize-based and beans

Kafue, Zambia Conservation farming unit, fundedthrough a group of donors includingNorway, Sweden and Finland, startedin the 1996

Adaptation and scaling out of CA(principally planting basins andMagoye Ripper)

1000 Unimodal, 800–1200 mm Maize-based, cotton, sorghum,millet, cassava

Monze, Zambia CIMMYT project, funded by IFAD Research, development anddissemination of CA systems via leadfarmer approach

100 Unimodal, 750 mm Maize, cotton

North-eastern Zimbabwe CIMMYT project, 2005–2013, fundedby IFAD and BMZ, Germany andvarious NGO initiatives

Research, development anddissemination of CA systems (directseeder, ripper, planting basins and jabplanter) via lead farmer approach

2000 Uni-modal, 750–1000 mm Maize

Mid-Zambezi Valley, Zimbabwe CIRAD project (2007–2010) funded bythe EU

Introduction and adaptation of CAsystems (direct seeder and ripper)

200 Unimodal, 450–650 mm Sorghum, cotton

ARTICLE IN PRESSG Model

AGEE-4500; No. of Pages 16

4 M. Corbeels et al. / Agriculture, Ecosystems and Environment xxx (2013) xxx– xxx

Table 2Modeling tools with their objectives at each scale and the case studies.

Modeling tool Reference Scale Objective Case studies

DSSAT Jones et al. (2003) Field To quantify crop yield responses from CA North-eastern Zimbabwe and Monze, ZambiaOlympe Attonaty et al. (2005) Farm To analyse the effect of CA practice on farm

household net incomeVakinankaratra, Madagascar; Bungoma, Kenya; Karatu,Tanzania; Central Malawi; North-eastern Zimbabwe andLake Aloatra, Madagascar

CLIF Baudron et al. (2013) Farm To quantify trade-offs between CAcropping and livestock production

Mid-Zambezi Valley, Zimbabwe

GANESH Naudin et al. (2013) Farm To optimize the allocation of fields for CAand quantify trade-offs between crop andlivestock production

Lake Aloatra, Madagascar

Flow model Andrieu et al. (2013) Village To quantify trade-offs in the use of cropresidues by different types of farmers

South-western Burkina Faso

QAToCA Ndah et al. (2013) Region To analyse the regional conditions andadopt

South-western Burkina Faso; Bungoma, Kenya; Karatu,

ccetrfifttuttCadiasaiseaistsmsitAbfdaeN

stT

2

t(

factors that affect CA

rop yield data from conventional experimentation and using theropping system simulation model DSSAT (Jones et al., 2003). How-ver, it is clear that misleading conclusions can be drawn abouthe attractiveness of CA for farmers by only analyzing crop yieldesponses at the field plot level. Other factors at the scale of thearm or above intervene. Given the fact that short-term profitabil-ty is a prime factor determining the relative advance smallholderarmers perceive with any new technology over their current prac-ices, an analysis of the farm-scale economics of CA can help assesshe potential for adoption of CA. This type of analysis was donesing the bio-economic farm model Olympe (Attonaty et al., 2005)hat takes into account the trade-offs that may exist in the alloca-ion of available resources (e.g. cash, labour, land and nutrients) toA, which may in turn affect the performance and income of otherctivities on the farm. In general, these trade-offs are important inetermining how a new technology may fit into a given farm. Since

n many farming systems in Africa, especially those of semi-aridnd sub-humid regions, the availability of crop residues is limited,trong competing uses exist at the farm, but also at the village-scalemong different types of farmers (e.g. crop farmers versus pastoral-sts). The traditional common right of free grazing in many farmingystems in SSA makes crop residues non-private products for farm-rs. Keeping crop harvest residues on the field as soil cover with CAnd not feeding them to livestock, results in significant trade-offsn livestock production. These trade-offs, were analyzed at farm-cale using simple models of crop-livestock systems that capturehe main interactions between the crop and livestock productionubsystems of a farm. These latter models must be seen as ad hocodels (Affholder et al., 2012) i.e. models that are designed for a

pecific modeling problem. Each of the trade-off models respondedn this way to the specificities of a given case study with respect tohe cropping and livestock subsystems and their main interactions.t village-scale, trade-offs were analyzed using a simple model ofiomass flows. Finally, when the context or external environmentor farming (regional scale) is considered, policy and institutionalimensions are seen as important for successful dissemination anddoption of CA. These factors were analyzed with a qualitativexpert-based assessment tool of CA adoption in Africa (QAToCA,dah et al., 2013).

We describe the methods and approaches used at the differentcales of analysis below. An overview of the modeling tools withheir objectives at each scale for selected case studies is given inable 2.

.3. Field-scale analysis: crop yield responses to CA

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

At field-scale we analyzed datasets on crop yield responseso CA and used the cropping system modeling platform DSSATJones et al., 2003). DSSAT incorporates process-based crop growth

ion Tanzania; Central Malawi; North-eastern Zimbabwe andKafue, Zambia

models for more than twenty different crops that can be linked tosoil models and to databases that describe weather, genotype, soiland experimental conditions and measurements for applying themodels to different agro-ecological situations (Jones et al., 2003).In our study, we used DSSAT version 4.5 with CERES-Maize as thecrop model, the soil water balance model as developed by Ritchie(1998), and CENTURY to simulate soil carbon and nitrogen dynam-ics (Parton et al., 1987; Porter et al., 2010).

To simulate tillage effects we assumed that the following foursoil properties vary with tillage in the model (Andales et al., 2000):(1) bulk density; (2) saturated hydraulic conductivity; (3) the soilrunoff curve number, and (4) soil water content at saturation. Thesesoil properties after a tillage event are input and they change backto a settled value, following an exponential curve that is a functionof cumulative rainfall kinetic energy since the last tillage opera-tion. Mulching effects on the soil water balance were simulated bythe DSSAT model through three soil water-related processes: (1)rainfall interception by the mulch; (2) reduction of soil evapora-tion rates, and (3) reduction of surface water runoff (Porter et al.,2010).

With DSSAT we addressed two specific questions that contributeto a better understanding of the crop yield responses to CA: (1)in how far can CA practices mitigate negative effects from futureclimate change, when rainfall is predicted to be lower and moreerratic, and (2) is increased soil carbon under CA the main fac-tor that explains increased crop productivity in the long term?Crop growth simulation models, such as DSSAT, have been usedto simulate and predict long-term effects on yields under CA (e.g.Sommer et al., 2007; MacCarthy et al., 2009). Whether these mod-els capture all the mechanisms involved remains, however, an openquestion.

For the first question, DSSAT was calibrated and tested usingdata from a CA experiment conducted by CIMMYT (Thierfelder andWall, 2012) at the Henderson Research Station (17◦35′S, 30◦38′E,1136 m.a.s.l.) near Harare in Zimbabwe. The site is characterizedby a sub-humid subtropical climate with an average annual rain-fall of about 880 mm. Rain falls during summer from Novemberuntil early April, but the occurrence of prolonged dry spells thatmay coincide with critical stages of crop growth is common. In theregion, rainfall is projected to decline by an estimated 10% by 2030(Lobell et al., 2008). Average annual temperature is about 22 ◦C.The site has a slope of about 5–7% and the soil was classified as adystric Arenosol. For the modeling exercise, two tillage treatmentswere considered: (1) the conventional farmer‘s practice of plough-ing the soil to a shallow depth (10–15 cm) without retention of cropresidues (CT); (2) the no-tillage practice using a direct seeder with

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

retention of crop residues (2 ton DM ha−1) on the soil surface (CA).We ran the model to simulate maize yields for water-limited condi-tions under the present climate using 45 years of daily climatic data

ING Model

A

tems a

(vraDnw(g(acd

fWTemNdebrdameotw(1

2

2

mafnwaoada(uowmtitpslf

tiils

ARTICLEGEE-4500; No. of Pages 16

M. Corbeels et al. / Agriculture, Ecosys

baseline scenario, BS) from Harare (source: Meteorological Ser-ices Department of Zimbabwe) and under three plausible futureainfall scenarios for the region. These were: (1) a 15% decrease innnual rainfall, RS; (2) a 15% increase in the duration of dry spells,S; and (3) the combination of scenarios 1 and 2, RDS. Each sce-ario also included a temperature increase of 1.1 ◦C. The scenariosere constructed using the stochastic weather generator LARS-WG

Semenov and Barrow, 1998). We predicted water-limited maizerain yield for the Henderson site under the 4 weather scenariosincluding the baseline climate) and for the 2 tillage treatments (CTnd CA). The model was run for each year separately, thus withoutonsidering possible long-term effects from CA. Planting date wasuring the last week of October.

For the second question, DSSAT was run against observed datarom a 6-years experiment conducted by CIMMYT (Thierfelder and

all, 2009; Thierfelder et al., 2013) on a Lixisol at the Monze Farmerraining Centre (16◦24′S, 27◦44′E, 1103 m.a.s.l.) in Zambia. Thexperimental site is characterized by a sub-humid subtropical cli-ate with an average annual rainfall of about 750 mm. Rains start inovember and end in April. The occurrence of prolonged dry spellsuring the rainy season is common. Two tillage treatments from thexperiment were considered: (1) the conventional tillage (mould-oard plough) treatment (CT) with removal of the crop harvestesidues, and (2) the CA treatment with the use of an animal tractionirect seeder and crop residue mulching. We first calibrated DSSATgainst observed data from the CT treatment and then ran theodel for the CA treatment, considering the no-tillage and mulch

ffects as described above and including the effects of mulchingn the soil organic matter dynamics as simulated by CENTURY. Wehen ran the model for 40 consecutive years for the two treatmentsith generated weather data that were based on observed data

1978–2007) from the Magoye weather station (16◦00′S, 27◦36′E,027 m.a.s.l, source: Zambia Meteorological Department).

.4. Farm- and village-scale analyses

.4.1. The farm-scale economics of CAFor the analysis of farm-scale economics of CA a distinction was

ade in the case studies (see Table 2) between farmers who havedopted (or are at least are experimenting with) CA on part of theirarm and were still using or experimenting with it, and farmers whoever tried it. In the case study of Vakinankaratra in Madagascar,e also included a sample of farmers who had used CA, but had

bandoned it again. In each case study, detailed household surveysn the general characteristics of the households, on the croppingnd livestock systems (activities, inputs, outputs), labour calen-ars and cash flows, including off-farm income were undertakennd the data were subsequently analyzed with the Olympe modelAttonaty et al., 2005). This model consists of a database and a sim-lation tool. The database is structured into several modules basedn general concepts of farming systems. The main modules dealith: (1) categories of inputs (fertilizers, seeds, labour, and invest-ents) and outputs (crop and livestock products); (2) activities on

he farm (crop and livestock) and off-farm; and (3) farm character-stics (land, capital, equipment, available labour). The simulationool calculates gross and net income of the farm household and theroduction subsystems (crop and livestock) from the householdurvey data. It allows assessing the impact of scenarios of techno-ogical change and/or prices on the economic performance of thearm household and its subsystems.

For the case study of Lake Alaotra in Madagascar we exploredhe impact of the practice of CA on total net household income

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

n the medium-term (10 years). The main crop in this study areas rice, grown in paddy fields or on dryland (non-irrigated low-and or hillside fields). Maize, cassava and groundnuts are theecondary crops grown in fields on the hillsides. The introduction

PRESSnd Environment xxx (2013) xxx– xxx 5

and dissemination of CA in the region started in 2003 through alarge-scale project (Bassin Versant du Lac Alaotra, BVLac project).An ex-post assessment of the impact of CA on farm householdincome over the medium-term using the model Olympe can give usbetter insights in the potential economic benefits of CA, consistentwith the production objectives of farmers. In total, 37 detailed farmhousehold surveys were conducted to construct the input databasefor Olympe. In the modeling exercise, three farm types were con-sidered: (1) Type C, medium-sized farms with 1–3 ha of paddy riceand less than 3 ha non-irrigated lowland or hillside fields; theyhave small-scale livestock (zebu cattle, pigs, poultry) activities andoff-farm activities to generate extra income; (2) Type D, medium-sized farms with less than 1 ha of paddy rice and 2–3 ha hillsideor non-irrigated lowland fields with small-scale livestock activitiesand some off-farm activities; (3) Type E, small-sized farms withless than 0.5 ha of paddy rice fields and less than 1 ha lowland orhillside fields. These farmers sell their labour to other farms. Wecompared the farm-level economics between the current conven-tional cropping practices based on tillage and CA cropping includingcrop rotations with legumes as proposed by the BVLac project. Theanalysis considers the implementation of CA on the non-irrigatedlowland and hillside fields. The structure of the farm and the paddyrice fields were assumed to remain invariable, while the input(fertilizer, seeds, pesticides, labor) prices and the selling prices ofproducts were kept constant over the 10 years simulation period.Climatic effects are taken into account by considering yield fluc-tuations according to the last 5 climatic years: one good year, twoaverage years, one very good year and one very bad year, repeatedtwice over 10 years. Based on recorded yield data, we assumed thatthe crop yields of CA systems increased over time at 4% per year andwere less sensitive to rainfall variability (15% versus 50% yield dropduring dry years for CA and CT, respectively).

2.4.2. Quantifying trade-offs in the use of crop residues at farmscale

We have analyzed the trade-offs in the use of crop residues atfarm-scale for two case studies (see Table 2). The first case studydeals with mixed crop-livestock smallholder farming systems inthe semi-arid Zambezi Valley in northern Zimbabwe, a region thatis characterized by low rainfall (450-650 mm) with severe dry spellsduring the growing season, resulting in low crop biomass produc-tion levels and high pressure on the crop residues. Sorghum, maizeand cotton are the main crops grown on the farms in the region. TheCrop-Livestock Interaction at Farm-scale (CLIF) model (Baudronet al., 2013) was built to analyze the trade-offs and possible syn-ergies that exist between crop and livestock production. Field andfarm data were collected from surveying 176 farms in the studyregion as input for the model. The interactions between the cropand livestock subsystems of the farms in the study region that wereconsidered in CLIF were: (1) cattle feeds during the dry season onsorghum harvest residues not retained as soil cover on the fields;(2) cattle provides manure for increased sorghum production, and(3) cattle provides traction for land preparation and weeding, i.e.the area of cropland of a farm is a function of the number of cattle.

The second case study explores the trade-offs around the useof crop residues on smallholder farms that raise dairy cows formilk production at Lake Alaotra in Madagascar. In the study, threefarm types were considered: Types C, D and E (see above). An opti-mization whole-farm model, GANESH (Goals oriented Approachto use No till for a better Economic and environmental sustaina-bility for SmallHolders), was built for the trade-off analysis (Naudinet al., 2013). Data for model input were collected from household

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

surveys on farms that participated in the BVLac project. Total netfarm income over three years was optimized with the model. Weexplored the impact of livestock intensification (increased numberof milking cows) on the optimal level of CA practice on the farm

IN PRESSG Model

A

6 tems and Environment xxx (2013) xxx– xxx

u(iofsoc

2s

alzlwwoiactcgoa3ffddcfslf

2

aTapataaaa(nitdosaawo

io

Fig. 2. Comparison of maize grain yields under conventional tillage (CT) and conser-

ally improve biological, chemical and physical properties of the soil(e.g. Thierfelder et al., 2013). We tested this hypothesis through

ARTICLEGEE-4500; No. of Pages 16

M. Corbeels et al. / Agriculture, Ecosys

nder scenarios of (1) altered soil cover with crop residues, and2) altered fodder prices. The number of cows on a farm was var-ed between 0 and 12, with the possibility to purchase forage fromutside the farm. The model considers that cows can be fed withodder produced on the farm – during the rainy season on the hill-ides (Brachiaria sp. and Stylosanthes guianensis) and during theff-season in the paddy fields (Vicia villosa or Dolichos lablab.), withrop residues, and with fodder bought on the market.

.4.3. Quantifying trade-offs in the use of crop residues at villagecale

The trade-offs in the use of crop residues at village scale werenalyzed for Koumbia, a village of 9000 ha and 5311 inhabitantsocated in the sub-humid (800–1200 mm of rainfall) agro-pastoralone of Burkina Faso. Three types of farmers co-habit in the vil-age with strong competing uses of crop residues: (1) crop farmers,

ho grow cotton for sale and cereals for home consumption, andho keep a small number of draught animals (83% of the farmers

f the village); (2) livestock farmers who own large herds of cattlen a more or less transhumant way for milk or meat production,nd grow cereals exclusively for home consumption (10%); and (3)rop-livestock farmers ‘who emerged from one of the two previousypes (7%). They are former crop farmers who began investing theirotton revenue in cattle fattening, or livestock farmers who startedrowing crops (cotton or cereals) for sale. About 36% of the surfacef the village is occupied with cropland, 32% is savannah grasslandnd 30% is natural protected area. Farm surveys were conducted on0 farms in the village (four livestock farmers, five crop-livestockarmers and 21 crop farmers). The structural characteristics of theirarms their cereal crop residue management practices, mode of pro-uction of organic fertilizer, and their use of mineral fertilizers wereocumented through individual interviews. A simple model wasonstructed that calculates the flows of biomass between the dif-erent types of farmers (Andrieu et al., 2013). The village fodderelf-sufficiency was calculated from the amount of crop residueseft on the fields for free grazing and collected as fodder, and theeed requirement of the entire village herd.

.5. Analysis of ‘regional’ scale factors

To analyze the regional conditions and factors that affect CAdoption, we developed a Qualitative expert-based Assessmentool of CA adoption in Africa (QAToCA, Ndah et al., 2013). The toolssesses the relative CA adoption potential in a given region (orroject) and diagnoses the supporting and hindering factors to CAdoption. The tool was built based on conceptual models of innova-ion systems, diffusion theories and relevant literature. The factorsre grouped under seven thematic areas: (1) characteristics of CA asn object of adoption; (2) capacity of promoting organizations; (3)ttributes of diffusion strategy; (4) institutional frame conditionst regional level; (5) institutional frame conditions at village level;6) market conditions at village and regional level, and (7) commu-ity’s perception at village and regional level. Each thematic area

s further declined in a series of operational questions that addresshe particular factors. QAToCA is meant as a self-assessment toolirected to regional experts, research teams and managers of devel-pment projects enabling them to assess their CA project along aystematic list of questions and criteria, to reflect on their CA relatedctivities and to eventually adjust or redesign them on the basis of

more explicit understanding of the problems and opportunitiesith the development and dissemination of CA. It gives a quick

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

verview of information on the CA status and adoption potential.We have used the tool to analyze CA adoption in six case stud-

es across Africa (see Table 2). For each of these case studies, ane-day workshop was organized during which the QAToCA tool

vation agriculture (CA) system. Data are from selected sites in Zimbabwe, Malawi,Zambia, Kenya and Tanzania.

was applied with multi-stakeholders who are involved in the CAdevelopment and dissemination activities of the related projects.

3. Main findings and key lessons learned

In this section of the paper we present the main findings andinsights gained from the multi-scale analysis on the performance ofCA throughout the different case studies, and which are importantfor understanding the conditions and determinants of adoption ornon-adoption of CA by smallholder farmers in SSA.

3.1. Field-scale: CA and its effects on crop yields

3.1.1. Regular crop yield benefits from CA take time to occurWe have analyzed the short-term effect (less than 3 years) of CA

on maize yields for a set of locations in Africa. The data are storedin a database.1 Although the observed short-term crop responsesto CA tend to be positive, they do vary, and can be neutral or neg-ative (Fig. 2). In general, it is difficult to determine precisely theunderlying causes of the variable crop responses to CA, becausethey are the result of complex interacting crop and soil processesthat are modified under CA. Often, increased soil water availabilityis the principal factor that is responsible for a short-term positiveresponse of crop yields to CA. Examples in the literature (Mkogaet al., 2010; Mupangwa et al., 2012; Thierfelder and Wall, 2012)show that short-term CA benefits on yields are common under con-ditions of water-limited crop growth. On the other hand, increasedweed competition and problems with seed germination are seenas main factors that cause crop yields to be lower under CA thanunder conventional tillage practices during the first years of imple-mentation (Mashingaidze et al., 2012). Skillfulness of the no-tillageplanting technique is critical for good crop establishment. The factthat immediate crop responses to CA are variable, and not alwayspositive, is a bottleneck for CA adoption by resource-poor farmers,since they demand immediate returns with their investments inCA.

While short-term yield effects of CA are variable, benefits areexpected to accumulate over time, because CA is known to gradu-

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

meta-analysis by combining results from different long-term CA

1 http://ca2africa.ciat.cgiar.org.

ARTICLE IN PRESSG Model

AGEE-4500; No. of Pages 16

M. Corbeels et al. / Agriculture, Ecosystems and Environment xxx (2013) xxx– xxx 7

Duration (years)

403020100

Wei

gh

ted m

ean d

iffe

rence

s (t

ha-1

)

-2

-1

0

1

2

3n = 134

(a)

403020100

Wei

gh

ted m

ean d

iffe

ren

ces

(t h

a-1)

-2

-1

0

1

2

3 n = 29 7

(b)

Fig. 3. Weighted mean differences in maize grain yield over time between (a)continuous no-tillage and continuous conventional tillage and (b) between CA(including rotation) and continuous conventional tillage. Data are from publishedexperiments under semi-arid and sub-humid conditions. Bars show 95% confidenceintervals (adapted from Rusinamhodzi et al., 2011).

Table 3Effect of climate change scenarios on maize grain yield (kg ha−1) as simulated byDSSAT under conventional tillage (CT) and conservation agriculture (CA) for theHenderson site nearby Harare, Zimbabwe. Variation coefficient in parenthesis.

BS RS DS RDS

CT 3107 (0.39) 2607 (0.35) 2577 (0.41) 2254 (0.43)CA 3830 (0.35) 3166 (0.34) 3328 (0.37) 2832 (0.40)

BS: base-line scenario, historical weather.RS: a 15% decrease in annual rainfall.DS: a 15% increase in the duration of dry spells.R

mhteHyv

3r

rguaPaa

Fig. 4. Cumulative probability functions of maize grain yield as simulated by theDSSAT model for the historical (baseline, BS) climate and the climate change scenario(CC) with 15% decrease in annual rainfall and 15% increase in the duration of dryspells, under conventional (CT) and conservation agriculture (CA) systems at the

DS: RS and DS combined.

aize experiments that were conducted in semi-arid and sub-umid regions (Rusinamhodzi et al., 2011). The analysis showshat crop yields under CA tend to accumulate over the long term,specially when maize is grown in rotation with a legume (Fig. 3).owever, for benefits on crop yields to occur it may take up to 15ears. Again, these results are highly site specific, as shown by theariability in the data.

.1.2. CA can mitigate the negative yield effects from more erraticainfall with climate change

The DSSAT simulation study on maize production in Zimbabweevealed that under the baseline scenario of current climate maizerain yield was on average about 720 kg ha−1 higher under CA thannder CT (Table 3). This was mainly due to increased water avail-

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

bility as a result of decreased runoff under CA compared to CT.redicted yields varied broadly, from a minimum of 1003 kg ha−1 to

maximum of 6483 kg ha−1 depending on seasonal rainfall amountnd distribution. As expected, average simulated grain yields were

Henderson Research Station, Zimbabwe.

for both cropping practices lower under future climate scenarios(Table 3). The simulation results indicated that the impact of a 15%increase in the duration of seasonal dry spells (DS scenario) is atleast as large as that of a 15% decrease in annual rainfall (RS sce-nario). Under the RDS scenario of decreased rainfall with longerdry spells, the model predictions suggested a decrease in maizegrain yields of about 25 to 30%, which is in agreement with thevalue (30%) projected for southern Africa in a broad-scale analy-sis by Lobell et al. (2008). The cumulative probability distributionfunctions of simulated maize grain yield for the BS and RDS climatescenarios under CT and CA are presented in Fig. 4. Under the currentclimate the probability of producing at least 3000 kg ha−1 grains is41% and 67% for respectively CT and CA. Under future climate, dueto water stress the probability drops to respectively 15% and 43%.The results suggest that in the ‘normal’ years the negative impactof climate change can be mitigated by adopting CA. However, inthe high- and low-rainfall years there will be a higher risk of loweryields.

3.1.3. Causes of long-term yield benefits are multifaceted and notwell captured in simulation crop growth models

The long-term simulated soil carbon and maize grain yieldsfor the experimental site in Zambia are shown in Fig. 5. UnderCA soil carbon levels remained more or less constant during theinitial years, while under the CT treatment there was a signifi-cant decrease, which is in agreement with the observations thatshowed significantly higher soil carbon levels in the CA treatmentfrom the third year onwards (Thierfelder et al., 2013). According tothe model predictions, soil carbon was after 40 years more than5 tons ha−1 lower under CT compared to CA (Fig. 5b). However,this differentiation in soil carbon levels had no long-term effecton the simulated grain yields. Grain yields were principally deter-mined by the rainfall amounts and distribution, with constantlyhigher simulated yields under CA compared to CT, principally as aresult of improved soil moisture under CA (Fig. 5a). Observed datashowed, however, improved maize grain yields over time underCA compared to CT, i.e. in the fifth and sixth year of the experi-

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

ment (Thierfelder et al., 2013). Thus, from these results we mayconclude that – at least for the conditions of the present study –the mechanisms of increased soil carbon and associated supply ofnitrogen represented in the model did not explain the observed

ARTICLE ING Model

AGEE-4500; No. of Pages 16

8 M. Corbeels et al. / Agriculture, Ecosystems a

Fig. 5. Simulated (DSSAT model) maize grain yields (a) and soil carbon (b) underconventional (CT) and conservation agriculture systems (CA) at the Monze FarmerT

ycuubttsiaSc

3

3

ossslati

raining Centre experimental site in Zambia.

ield increases with time. With the DSSAT model we only suc-eeded to approximate the observed increases in maize grain yieldsnder CA versus CT, if we assumed a restricted root developmentnder CT resulting in lower water uptake. This assumption wasased on the observed root-hampering plough pan under the CTreatment, which disappeared over time under CA. This suggestshat CA induces over time more complex changes in soil properties,uch as improved soil structure or increased soil biological activ-ty, which may affect long-term crop yield responses but whichre not simulated by common crop growth models such as DSSAT.imulating the long-term effects of CA on crop yields remainshallenging.

.2. The farm-scale economics of CA

.2.1. Comparison of CA and non-CA farmsThe results of the farm-scale analysis comparing the economics

f farmers who have adopted CA and farmers who did not are pre-ented in Table 4. In general the size of farms that are using CA islightly higher than that of non-CA farms, but that is probably notignificant. Typically, CA farms have less than about 30–40% of their

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

and under CA, presumably since farmers are still cautious about itnd have land (soils) and crops less suitable for CA. Family size (andhus labour availability) seems also not to play a very important rolen the adoption process, and during the surveys there were both

PRESSnd Environment xxx (2013) xxx– xxx

farmers who indicated that CA saves labour, and those who find thatit increases labour. In all case studies it appeared that CA farmershad more cattle than non-CA farmers. This may be surprising, giventhe competition for harvest residues between mulching and live-stock feed (see below). In the Tanzanian case study, CA farmers hadhigher livestock earnings than non-CA farmers; the opposite wastrue in the Kenyan case study. Off-farm earnings were either simi-lar or somewhat higher for non-CA farmers. Earnings from croppingwere higher for CA farmers in the Kenyan and Malawi/Zimbabwecase studies, but lower in the Tanzanian case study comparedwith the non-CA farmers. In the Tanzanian case study, the distinc-tion between fields under CA and under conventional practice wasnot always that clear, since some so-called CA farmers used con-ventional tillage during the survey year because of drought anddifficulties to plant in no-tilled soil. Some farmers also claimed thatwith tillage they enhanced rainfall infiltration into the soils; oth-ers tilled their soils to protect the crop residues from being eatenby grazing animals. A closer look at the earnings from croppingshows that CA fields gave in all case studies higher incomes perha than non-CA fields from farms that are not practicing CA; CAfarmers had also a higher income per ha from their CA fields thanfrom their non-CA fields, with exception of the Malawi/Zimbabwecase study. Crop yields were generally higher with CA than non-CA(data not shown) and this was for farmers also the most impor-tant reason for adopting/experimenting with CA. However, thesehigher yields were often obtained because of higher inputs in termsof fertilizer, herbicides and labour on CA fields compared to non-CA fields. For example, in Malawi/Zimbabwe the mean maize yieldwas 2097 kg ha−1 on CA fields, compared to 1038 kg ha−1 on non-CA fields, but farmers applied on average 10% more fertilizer andspent 45% more labour time on CA fields. Finally, the results ofour analysis show that CA farmers do not seem to have system-atically higher levels of net income than non-CA farmers. This is toa great extent a consequence of the various trade-offs that existswith the use of available resources (land, labour and cash). It isalso important to point out that in all case studies farmers arestill in an early stage of adoption and some of them may still turnback to their traditional way of cropping, as some have done in theMadagascar case. From the farmers who have abandoned CA in theMadagascar case study, 70% mentioned lower income as the mainreason.

3.2.2. Medium-term impact of CA on farm household incomeResults on simulated total net farm household income over the

medium-term (10 years) comparing farms with and without CAusing the Olympe model for different farm types at the Lake Alao-tra site in Madagascar are shown in Fig. 6. Total net farm householdincome is the sum of all net margins from all agricultural activi-ties on the farm plus off-farm income. The results illustrate thatthe effect on total income of introducing CA into the farm stronglydepends on the type of farm. On the type C farms that are orientedtoward paddy rice production, total income is after 10 years about9% higher with practicing CA. This improvement is not very sub-stantial, because the overall income of this type of farms is largelygenerated by the rice production on the paddy fields. It means thatthese farms probably do not have a particular interest to adoptthe CA systems on their drylands. On type D farms, the practiceof CA increased total income with 19% compared to the conven-tional system. This increase is more significant than on type Cfarms, because of the lower proportion of paddy fields. It meansthat for this farm type CA systems contribute to securing income,especially during the dry years. On the smaller type E farms, CA

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

improved farm household income by 23% compared to the con-ventional system after 10 years, for the same reasons as type Dfarms. CA systems can secure income on this type of farms as well.However, farms of type D and E have a much smaller cash balance

ARTICLE IN PRESSG Model

AGEE-4500; No. of Pages 16

M. Corbeels et al. / Agriculture, Ecosystems and Environment xxx (2013) xxx– xxx 9

Table 4Farm economic data from the agro-economic surveys held in the respective study areas in 2010–2012 and analyzed with the Olympe model (values in Euro in the respectiveyears).

Country Madagascar Kenya Tanzania Malawi/ZimbabweRegion Vakinankaratra Bungoma Karatu (various)Exchange rate 1 D 2600 Ariary 105 KES 2000 TShs 215 MWK/1.31 USDYear 2010 2011 2011 2012

Farm category CA Aband. Never CA Non-CA CA Non-CA CA Non-CASample size No 21 17 22 25 25 25 25 25 13

Average farm size ha 0.88 1.81 1.05 1.03 0.93 1.82 1.30 2.13 1.83Of which CA % n/a 0 0 31 0 29 0 26 0Family size No 5 5 6 7 6.8 6.5 5.8 n/a n/aAverage no of cattle No 1.8 1.2 1.2 3.4 2.9 3.6 2.9 3.04 2.4Annual crop income D /yr n/a n/a n/a 813 615 478 563 423 323Annual livestock income D /yr n/a n/a n/a 165 207 275 237 n/a n/aCrop and livestock income D /yr 290 264 272 978 822 753 800 n/a n/aOff-farm income D /yr 341 330 332 496 514 187 360 353 526Total farm income D /yr 631 594 604 1473 1336 940 1160 776 849Family expense D /yr n/a n/a n/a 1149 1013 1099 956 1526 960

Value crop production D /ha n/a n/a n/a 1456 1017 523 551 199 176CA crop production D /ha n/a – – 1526 – 598 – 185 –Non-CA crop production D /ha n/a n/a n/a 1256 1017 419 551 203 176

Fig. 6. Calculated (Olympe model) net farm income for three farm types, without(CT) and with (CA) the practice of conservation agriculture at Lake Aloatra, Madagas-car. Type C farms have 1–3 ha of paddy rice and less than 3 ha non-irrigated lowlandor hillside fields; Type D farms have less than 1 ha of paddy rice and 2–3 ha hillsideor non-irrigated lowland fields and Type E farms have less than 0.5 ha of paddy ricefields and less than 1 ha lowland or hillside fields. Values of income are expressedi

tebtfiCatpTi9gtytai

Fig. 7. Calculated (Olympe model) gross margin for the conven-tional (CT)—maize–maize rotation and conservation agriculture

effects on crop yields in the long term, as the application of theavailable manure from cattle. When considering also the role of

n Malagasy ariary (1kAr ≈ 0.350 Euro).

hen those of type C, which limits their capacity to invest consid-rably in upland fields. The difference in farm household incomeetween the CA and conventional system was mainly related tohe effect of CA practiced on upland hillside fields (Fig. 7). Theelds of the lowlands were small (<0.1 ha) on all farm types. TheA systems on the hillside fields consist of a rotation that includes

legume (groundnut) and upland rice, while the conventional sys-em is continuous maize cropping. With CA the gross margin fromroducing on hillside fields increased by about 40% over 10 years.he increase was mainly the result of the gradual increase over timen yields of upland rice and maize under CA with a peak in year

that correspond to the harvest of upland rice that has a higherross margin than maize or groundnuts. In the conventional sys-ems gross margin dropped in year 4 and 9 because of the reducedields during dry years. Based on the above results, we may expecthat the more a farm is oriented towards rainfed cropping, the more

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

doption of CA is interesting in terms of improving farm householdncome.

(CA)—maize–rice–maize–groundnut rotation—systems on hillside fields atLake Aloatra, Madagascar. Values of gross margin expressed in Malagasy ariary(1kAr ≈ 0.350 Euro).

3.3. Farm and village scale: crop residues for feeding the soil orthe cows?

3.3.1. Trade-offs and synergies between CA and livestock atfarm-scale

Results from the trade-off analyses in two case studies, i.e. Zam-bezi Valley, Zimbabwe and Lake Aloatra in Madagascar, illustratethe site-specificity of the interactions between crop and animalproduction. In the mixed-crop livestock systems in Zimbabwewhere cereal crop residues constitute an important source of feedfor livestock during the dry season, the number of cattle that canbe kept on a farm per unit area of sorghum is strongly and neg-atively correlated with the fraction of sorghum residues retainedas soil cover on the field (Fig. 8a). Since crop growth is limited bynitrogen, fertilization with nitrogen has an effect on the relation-ship. On the other hand, the density of cattle grazing on a field hada small effect on the sorghum yields per unit area (Fig. 8b). Accord-ing to the model, mulching the field with crop residues had similar

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

traction that cattle plays, a positive relationship appeared betweencattle number and total crop production of the farm (Fig. 8c and

ARTICLE IN PRESSG Model

AGEE-4500; No. of Pages 16

10 M. Corbeels et al. / Agriculture, Ecosystems and Environment xxx (2013) xxx– xxx

Fig. 8. (a) Predicted (CLIF model) long-term number of cattle that can be sustained per surface area of sorghum, (b) predicted long-term sorghum grain yield as a function ofthe long-term number of cattle sustained per surface area of sorghum, (c) predicted long-term farm production of cereals as a function of the predicted long-term numberof cattle owned by the farm, and (d) predicted long-term farm production of cotton as a function of the predicted long-term number of cattle owned by the farm, case studyZambezi Valley, Zimbabwe (adapted from Baudron et al., 2013).

dagagcatshsm

gccssdnpbi8hoop

). For example, the model predicts that a farm produces an aver-ge of 3.2 tons of grains (no cotton) with no cattle, 2.9 tons ofrains and 4.7 tons of cotton with two cattle, and 3.5 tons of grainsnd 7.9 tons of cotton with four cattle in the case of low nitro-en fertilization. These results illustrate the key importance ofattle traction for crop production in the study area. Cultivatingn area as large as possible is an important risk adverse strategyhat farmers adopt in this region, where farming is more con-traint by labour than by land. It is clear that in this context, croparvest residues are in the first place fed to cattle, impeding large-cale dissemination and adoption of CA practices with crop residueulching.The results of the trade-off analysis in the case study in Mada-

ascar are for three farm types shown in Fig. 9. Overall, the resultslearly show that synergies may exist as well as trade-offs betweenrop and livestock production on these farms, depending on thepecific farming context. Farmers only start practicing CA on hill-ide fields, when the number of cows becomes larger than 6–8,epending on the farm type, soil cover and forage price. Below theseumbers, farmers are not interested in CA because they prefer toroduce the most profitable crops, such as groundnut and cassava,ut that are difficult to grow under CA. Under the scenario of min-

mum soil cover (30%) on the CA fields, all farms with more than (small farms) or 10 (medium–sized farms) cows will decide to

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

ave about 60–80% of their hillside fields under CA, irrespectivef the fodder price. On medium-sized farms, this high percentagef CA fields is even possible with a soil cover of 95%, if the fodderrice is low, because farmers are able to purchase the feed for their

cows from outside the farm. However, in the case of a high fodderprice, when feed for the cows has to come from the farm, the com-petition for crop residues makes that the fields under CA with 95%soil cover are no longer preferable as soon as the number of cowsexceeds 11. On small farms, CA with full soil cover is less of anoption, because the competition for crop residues is higher. One ofthe external factors that can strongly influence the synergies/trade-offs is the market price of fodder. If fodder is produced locally andavailable on the market, the pressure on the crop harvest residuesis less, which opens opportunities for CA dissemination. This con-stitutes a main difference with the case study in Zimbabwe, wherethe pressure on crop residues is high, given the limited availabil-ity of feed for livestock. In Lake Alaotra, CA systems are a realoption for farmers with dairy cattle, provided that fodder cropsare integrated into the systems as an extra source of feed for thecattle.

3.3.2. Conflicts between free-grazing and CA at village-scale: theneed for territory arrangements

The trade-offs in the use of crop residues at village scale areillustrated for the case study in Burkina Faso. Private utilization ofcrop residues by the different types of farmers in the Koumbia vil-lage represents less than 20% of the crop residue biomass producedon the farm (Fig. 10). The bulk of crop residues are left on the field

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

and consequently available for free grazing by livestock from thevillage and from outside. At village scale, self-sufficiency in live-stock feed during the dry season is estimated to be around 60% ofthe nutritional requirements of the entire village herd, indicating

ARTICLE ING Model

AGEE-4500; No. of Pages 16

M. Corbeels et al. / Agriculture, Ecosystems a

Fig. 9. Predicted (GANESH model) percentage of hillside fields under CA as a func-tion of the number of cows on the farm for three farm types in the Alaotra Lake region,Madagascar (a) medium-sized Type D farms with mainly hillside fields (2–3 ha) andless than 1 ha of paddy rice; (b) medium-sized Type C farms with 1–3 ha of irrigatedpaddy rice fields and less than 1 ha hillside fields, and (c) small-sized Type E farmswith less than 0.5 ha of paddy rice fields and less than 1 ha hillside fields, and asa function of soil cover (30 and 95%) and fodder price (0.1 and 0.3 kAr kg−1 DM).1kAr ≈ 0.350 Euro.

aupt–fwap

(Table 5). It has been argued that the number of practices thatare required to be changed with CA at the same time necessi-tates a major transformation in crop and soil management practices

high pressure on the crop residues (data not shown). Privatese of cereal crop residues for soil cover or composting has theotential to increase the maize yields on the individual farms andotal maize production at village scale, but logically – as a trade-off

would also mean an increased pressure on the feed resourcesor livestock during the dry season at village scale (Fig. 10). Itould result in an increased grazing pressure on the savannah

rea and may exacerbate conflicts between crop farmers and

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

astoralists.

PRESSnd Environment xxx (2013) xxx– xxx 11

3.4. Regional scale: pre-conditions and constraints to adoption ofCA

The overall results of the QAToCA exercise in terms of rate ofpotential CA adoption are shown in Table 5. It has to be notedthat a high CA adoption potential for some case studies takes intoaccount the likelihood of adoption of the three CA principles, butinclusive with the chance of partial adoption of one or two princi-ples of CA. Most often, farmers experiment and tend to adopt one ortwo of the CA principles as an eventual entry point to full adoptiononce benefits are perceived for the enhancement of their personalgoals (Triomphe et al., 2007). Farmers go through a learning processbefore full adoption (Pannell et al., 2006).

The most outstanding observation from the QAToCA analysisin the six case studies is the recurrent assessment that marketconditions for inputs and outputs are not in place for the adop-tion of CA to happen. This has been found in five out of the sixcase studies. Only in the Malawian case study good market con-ditions are considered to be fulfilled for potential adoption of CA.Probably, this explains to a large extent the success of CA in theMalawian project. Estimates show that Total Land Care (the mainimplementing organization of the project) has reached out to about32,000 farmers who are now practicing CA on a total surface areaof 12,830 ha (NCATF, 2012). Market conditions scored especiallylow in the Zimbabwean case study, which obviously is related tothe current fragile economic situation in the country. In general,good market conditions should essentially been seen as prerequi-site conditions for adoption as they are mostly outside the controlor influence of the project. Unless these prerequisite conditions aremet, there can be no prior expectation of CA adoption. We haveobserved that in many projects that promote CA, the ‘higher-scale’conditions for adoption, such as markets, are often poorly consid-ered. In fact, most projects create their own enabling environmentfor the implementation of CA practices by providing financial andtechnical support (e.g. the purchase of inputs for farmers by theproject), but once this stops the majority of farmers revert to theirformer crop management practices. Probably related to this, thecapacity of the promoting organization to develop and promote CAand attributes of CA dissemination strategies received high scoresin most case studies (Table 5). The positive appraisal of the CAdissemination strategy can presumably also be attributed to theuse of participatory learning and extension approaches such asthe Farmer Field Schools (Kenya, Tanzania) or the Lead Farmerapproach (Malawi), and that were considered as effective for thedissemination of CA by the experts.

More surprisingly, the institutional (policy) frame conditionsat regional (and village) level were evaluated as rather good inseveral case studies. In most of the study regions, CA is endorsedas a sustainable cropping practice by national and local insti-tutes. In particular, the national governments of Kenya, Tanzania,Zimbabwe, Malawi and Zambia have incorporated CA in theirstrategic plans for the development of the agricultural sector. Thequestion, however, remains how effective are these institutes andpolicy that are put in place in promoting CA. More research isneeded on the question of public policies and institutional arrange-ments and factors that support or hinder the diffusion and adoptionof CA.

Looking into the specific factors that may explain adoption orrefusal of CA, the QAtoCA analysis revealed that, while some arerecurrent, others are specific to the region or project. The ‘com-plexity of CA as a practice’ came up in three case studies (Kenya,Zambia, Zimbabwe) as a main hindering factor to adoption of CA

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

(Erenstein, 2002). CA is a knowledge-intensive cropping practice

ARTICLE IN PRESSG Model

AGEE-4500; No. of Pages 16

12 M. Corbeels et al. / Agriculture, Ecosystems and Environment xxx (2013) xxx– xxx

F lage fa

tOatapTgaftewwtoctprroL

ig. 10. (a) Crop residue left on the field versus crop residue produced, and (b) vilmendment at Koumbia, Burkina Faso (adapted from Andrieu et al., 2013).

hat needs capacity building with farmers and extension services.ther recurrent constraints to CA adoption were the availabilitynd accessibility (cost) of markets for CA inputs (specialized no-illage implements, (legume) seeds, and herbicides), the limitedvailability of social networks for interacting on CA and the com-etition for crop residues with its use as livestock feed (see above).he latter was clearly brought in relation with the practice of freerazing, and was by many experts seen as the bottleneck for CAdoption by smallholders in Africa. The existence and strength ofarmers’ social networks and local organizations have been showno be positively related to adoption in a number of studies (Pannellt al., 2006). The availability of basic infrastructure for marketing,hich is linked to the overall poor market conditions (see above),as seen as a main hindering factor in the Kenyan case study. In

he Tanzanian, Zimbabwean and Burkina Faso case study, the lackf quality control structures (certification) was evaluated as a mainonstraint for CA adoption. It was related to difficulty to differen-iate as to which farmers practice ‘full’ CA and which ones onlyartially implement CA or are just involved in some kind of CA-

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

elated activities. The identification of this constraint is probablyelated to the awareness by the experts that there is a need forptimal management in order to obtain the full benefits from CA.astly, the level of administrative set up was seen as a hindering

odder self-sufficiency as a function of the proportion of crop residues used as soil

factor in the Zambian case study, while land access, ownership anduse, was identified as a main hindering factor for the case studiesboth in Malawi and Zambia. There is ample evidence that secureland use rights promote investments in land, such as with CA orother conservation practices in more general (e.g. Neef, 2001).

4. The need to tailor CA interventions to the end users

From the multi-scale analysis in the above sections, there isclear evidence that CA practices need to be tailored to local cir-cumstances of the farmers. This agrees with the conclusions ofKnowler and Bradshaw (2007), Giller et al. (2009) and Erensteinet al. (2012). From the QAToCA analysis, it was suggested that themarket conditions in the majority of the case studies (Kenya, Tan-zania, Zimbabwe, Zambia and Burkina Faso, Table 4) are hinderingthe widespread adoption of CA. This is generally true as well forother agricultural technologies: there is a general lack of effectivesupport for smallholder agriculture in much of SSA, such that thereare actually economic disincentives to investment in agriculture

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

(Ehui and Pender, 2005). Private sector support is often uncertain,because only a small part of farm output is marketed. However, therecent food crisis put market regulations and production incentivesback on the world agenda, with a particular focus on Africa where

ARTICLE IN PRESSG Model

AGEE-4500; No. of Pages 16

M. Corbeels et al. / Agriculture, Ecosystems and Environment xxx (2013) xxx– xxx 13

Table 5Drivers and constraints for adaption of conservation agriculture in six case studies in sub Saharan Africa—as assessed by the QAToCA tool.

Case study Adoption potential Thematic constraints Thematic drivers Specific hindering factors

Bungoma, Kenya 82% Market conditions Attributes of disseminationstrategyPolitical and institutionalframework at regional level

Complexity of CAResidue and seed requirementversus availabilityAvailability of basic marketinfrastructure

Karatu, Tanzania 83% Market conditions Capacity of promoting institutionsAttributes of disseminationstrategy

Availability of quality controlstructures

Kafue, Zambia 75% Market conditions institutionalframe conditions at villagelevel

Capacity of promoting institutionsAttributes of disseminationstrategyPolitical and institutionalframework at regional level

Cost of CA and liquidity issueComplexity of CAAvailability of social networksState/level of administrative set-upLand access and ownershipHousehold spatial distributionAvailability of market for outputsAccessibility of market for outputsAcceptability of CA by youngfarmers

South-eastern Zimbabwe 62% Market conditions Capacity of promoting institutionsPolitical and institutionalframework at regional level

Cost of CA and liquidity issueComplexity of CAAvailability of social networksResidue and seed requirementversus availabilityMachinery requirement versusavailabilityFlexibility/adaptabilityLack of communication channelsAvailability of quality controlstructuresAvailability of market for outputsAccessibility of market for outputsAcceptability of CA by youngfarmers

Central Malawi 87% Institutional frame conditionsat village level

Capacity of promoting institutionsAttributes of disseminationstrategyPolitical and institutionalframework at regional levelMarket conditions

Land access and ownership

South-western BurkinaFaso

82% Market conditions Capacity of promoting institutionsPolitical and institutional

Availability of quality controlstructures

yaaoi2peadnsgoititfib

ield gaps and hence the perspectives of production increasesre the largest (Tittonell and Giller, 2013). Many CA options thatre inappropriate for subsistence-oriented farms might becomepportunities for market-oriented farms, as suggested by studiesn emerging countries (e.g. in Brazil and India, Erenstein et al.,012). As argued by Sumberg (2005), the market, institutional andolicy contexts, in which a new technology is promoted, shouldssentially be seen as prerequisite conditions. Markets and policyre often outside the control or influence of the CA development-issemination process. This does not mean that they are to beeglected; they have to be identified at the early stage of the projecto to be able to make the necessary adaptations of the CA systems. Aoodness-of- fit between the CA innovation and the targeted groupf farmers is to a large extent determined by the short-term prof-tability of adopting CA. Smallholders in SSA often have short-termime horizons and immediate needs. From our analysis it seems that

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

n the majority of cases farmers in SSA do not substantially increaseheir farm household income through the practice of CA on someelds of their farm. This is mainly because the short-term yieldenefits from CA are small or highly variable. On the other hand,

framework at regional and villagelevel

CA can increase income over the medium-term (10 years), becauseof the expected yield benefits over time, but these increases willdepend on the type of farms, their production assets and objec-tives. CA investments should in the first place target situationswhere short-term crop responses are expected to be positive. Sincethese primarily take place through increased soil moisture conser-vation and crop water productivity as a result of mulching, targetregions are those where crop production is primarily determinedby limited soil moisture supply (under current or future climate).Unfortunately, these are also the regions where the pressure oncrop residues is highest because of the low crop biomass productionand traditionally large livestock populations. The competition forcrop residues with livestock is a key issue that has to be consideredwhen promoting CA. Territory arrangements between villagers, butalso with pastoralist outside the village can offer solutions for suffi-cient residue retention on some fields of the village in areas where

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

free grazing is practiced. This implies establishing new contractualrules between crop farmers, agro-pastoralists and pastoralists, thatbypass the traditional rule of free grazing. Institutional arrange-ments governing grazing privileges are not trivial and often difficult

ING Model

A

1 tems a

tontafwoitfAtstmwpfemssaiittnniopshhbrtlpfcCptbAby

5

pibHmocmapm

ARTICLEGEE-4500; No. of Pages 16

4 M. Corbeels et al. / Agriculture, Ecosys

o establish; they depend on the cultural, socio-economic andrganizational specificities of the agro-pastoral farming commu-ities. For example, in the Soil Conservation project PCS-ESA2 inhe Tupuri de Sirlawé village in Northern Cameroon new spatialrrangements were constructed by the local villagers that allowedor the practice of CA cropping (Dongmo et al., 2012). CA fieldsere established in some areas of the village with the agreement

f all villagers not to let their cattle freely roam in that area dur-ng the dry season. To counterbalance the resulting reduced accesso feed resources, the project also introduced the production ofodder crops in the village. As shown in the case study at Lakelaotra in Madagascar, synergies between livestock keeping and

he practices of CA may arise if fodder crops are part of the CAystems. In many situations the incorporation of fodder crops canherefore be seen as an effective local adaptation of CA for many

ixed crop-livestock farming systems in SSA. Feeding livestockith crop residues is certainly favored in situations where land islentiful and livestock is of primary importance as animal tractionor extending the area under cultivation. In these situations, farm-rs opt to spread their resources over a large area as a strategy toitigate risk, rather than concentrating labour and cash inputs on

mall areas (Baudron et al., 2012). Consequently, investments in CAhould preferably target situations where land is relatively scarce,nd where opportunities exist to direct farm development towardsntensification. In general, the practice of CA needs to be compat-ble with other farm and non-farm activities. The example, withhe paddy rice-growers in the Lake Aloatra case study illustrateshat farmers who have their main income from activities that areot well-matched with the practice of CA (paddy rice fields) mayot be interested in CA, and prefer to invest in those other activ-

ties that generate their income. Their neighbor farmers that relyn income from dryland cropping may perceive CA as a profitableractice. Dissemination efforts of CA should therefore carefully con-ider the production objectives of the target farmers. On the otherand, it seems likely that CA will be most rapidly adopted by small-older farmers with adequate resources of cash and labour, and noty the most resource-constrained groups. For example, mediumesourced farmers at Lake Aloatra in Madagascar who have the casho buy livestock feed from the local markets, have the capacity toeave their cereal crop residues as a soil cover on their fields. Theoorer farmers do not have the means to do this, especially whenodder prices are high. Targeting and adapting CA practices shouldlearly consider the farmers’ investment capacity in the practice ofA. In general, CA requires relatively high inputs of nutrients, sim-ly to produce enough biomass for soil cover. It has been shownhat CA with high fertilizer rates gives better comparative yieldenefits than CA with low fertilization (Rusinamhodzi et al., 2011).nother component of CA systems that requires cash are the her-icides. When herbicides are not used under no-tillage, lower cropields are often observed with increased labour requirements.

. Conclusions

The analysis of CA research, development and disseminationrojects in SSA at different scales: field, farm, village and region can

nform about the determinants of CA adoption and non-adoptiony smallholders. Conservation agriculture can increase crop yields.owever, immediate yield benefits are highly variable, and areost likely to occur when crops are drought-stressed. The practice

f CA has the potential to conserve soil moisture through the soilover of crop residues, which makes it an effective technology for

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

itigating the negative effects from less and more erratic rainfalls a result of climate change. With CA, crops yields are expected torogressively increase in time, as a result of the gradual improve-ent of soil quality. Yield increases are, however, difficult to predict

PRESSnd Environment xxx (2013) xxx– xxx

because they are location-specific and current crop growth modelsdo not capture all the mechanisms involved. Eventhough small-holders may understand that crop yields will increase with thepractice of CA, the future benefits often do not compensate fortheir immediate needs to provide for their family. With the absenceof immediate positive yield responses, CA is unlikely to result inimmediate increases in farm income, which is a major constraintfor rapid adoption of CA. Although the economic benefits of CAare still difficult to quantify and are often confounded by loca-tion specificity, type of farming system and/or seasonal variability,the case studies suggest that farmers who are practicing CA onsome fields of their farms do not increase total farm householdincome compared to their neighbors who are not practicing CA.The case studies analyzed in this paper also reiterate the impor-tance of good markets of input supply and sale of produce as aprerequisite condition for the widespread adoption of CA. As sug-gested by studies in emerging-markets countries such as Brazil andIndia, many CA practices that are inappropriate for subsistence-oriented farms may become viable options for market-orientedfarms. Farmers adapt and implement CA technologies with theirown understanding of the principles, their aspiration and possibil-ities to integrate them into their farming systems, and their actualaccess to knowledge, advice and resources. The ex-ante identi-fication of opportune situations for adapting and implementingCA is a challenge that demands active research and develop-ment from a multi-stakeholder, multi-scale, and interdisciplinaryperspective. It must consider the multiple scales at stake as repre-sented in Fig. 1, in which technical performance (i.e., crop yieldsat the field plot level) is but one of the determinants of adop-tion. At each scale, difficulties might emerge that impede, slowdown or even reverse the adoption process of CA. Too often CAprojects tend to focus heavily on agronomic, field-scale matters,often to the detriment of dealing properly with issues arising atother scales or being of a different nature. Priority is often givento demonstrating CA rather than to adapting it in a participa-tory manner to the local context, even though the use of localgroup-based learning approaches such as ‘farmer field schools’ and‘lead farmer to farmers-extension’ is increasing. Given the broadrange of stakeholders involved in the development and diffusionof CA, a multi-stakeholder approach through a so called innova-tion network is probably the best approach for adapting CA tothe local conditions of farmers. Such a local innovation networkof farmers, extension agents, researchers, input suppliers, equip-ment manufacturers, service providers, traders, and policymakersshould foster dynamic interactions and synergies for joint learn-ing and experimenting with CA to develop viable CA practices. Arole for research today is to help improve the policy and institu-tional context for innovation, which may imply shifting researchresources toward institutional innovation and away from technol-ogy development.

Acknowledgements

This work was carried out as part of the European Union (EU)funded project ‘Conservation Agriculture in AFRICA: Analysing andFoReseeing its Impact—Comprehending its Adoption ‘(CA2Africa,www.ca2africa.eu; grant agreement no. 245347). The opinionsexpressed herein are the sole responsibility of the authors.

Appendix A.

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

Table A1.Table A2.Table A3.Table A4.

ARTICLE IN PRESSG Model

AGEE-4500; No. of Pages 16

M. Corbeels et al. / Agriculture, Ecosystems and Environment xxx (2013) xxx– xxx 15

Table A1Soil model input parameters for the DSSAT simulations at the Henderson Research Station near Harare in Zimbabwe.

Limit inf. of layer(cm)

Albedo (estimatedfrom colour)

Potential soil evaporation(estimated from colour andtexture) (mm/d)

Drainage rate (estimatedfrom texture) (mm/d)

Runoff curve number (estimatedfrom texture and slope)

0 0.15 6 0.75 80

Limit inf. of layer (cm) (Measured) (Estimated from texture and orga. C)

Clay (%) Silt (%) Organic C (%) pH inwater

CEC(cmol/kg)

Bulk density(g/cm3)

Lowerwaterlimit(cm3/cm3)

Drainedupper limit(cm3/cm3)

Saturatedupper limit(cm3/cm3)

Hydraulic conductivity(cm/h)

−12 8 17 1.0 5.1 7.0 1.34 0.09 0.18 0.464 2.6−28 11 15 0.3 5.3 5.4 1.42 0.088 0.162 0.439 2.6−45 10 23 0.4 5.9 7.4 1.45 0.086 0.177 0.427 2.6−60 3 12 0.4 6.3 4.4 1.40 0.044 0.103 0.448 6.1−87 3 12 0.4 6.3 4.7 1.40 0.08 0.103 0.448 6.1

−100 4 13 0.4 6.4 0.8 1.41 0.09 0.107 0.444 6.1−120 4 13 0.4 6.4 1.2 1.55 0.047 0.108 0.394 6.1

Table A2Soil model input parameters for the DSSAT simulations at the Monze Farmer Training Centre in Zambia.

Limit inf. of layer(cm)

Albedo (estimatedfrom colour)

Potential soil evaporation(estimated from colour andtexture) (mm/d)

Drainage rate (estimatedfrom texture) (mm/d)

Runoff curve number (estimatedfrom texture and slope)

0 0.14 3 0.75 84

Limit inf. of layer (cm) (Measured) (Estimated from texture and orga. C)

Clay (%) Silt (%) Organic C (%) pH inwater

CEC(cmol/kg)

Bulk density(g/cm3)

Lowerwaterlimit(cm3/cm3)

Drainedupper limit(cm3/cm3)

Saturatedupper limit(cm3/cm3)

Hydraulic conductivity(cm/h)

−22 16 5 0.6 4.4 3.1 1.67 0.122 0.187 0.38 2.6−56 37 8 0.3 5.1 4.0 1.48 0.228* 0.314 0.40 0.4−80 38 8 0.01* 5.3 5.3 1.48 0.264* 0.309 0.39 0.1

−107 44 7 0.01* 5.6 5.4 1.33 0.296* 0.331 0.39 0.1>−107 43 7 0.01* 5.7 6.7 1.48 0.304* 0.324 0.39 0.1

Table A3Maize genotype-specific input parameters for the DSSAT simulations at the Hender-son Research Station near Harare in Zimbabwe.

Cultivar P1 P2 P5 G2 G3 PHINT

SC513 240 0.045 770 550 9 55MRI624 260 0.045 500 950 9 65

P1: Thermal time from seedling emergence to the end of the juvenile phase(expressed in ◦C day, above a base temperature of 8 ◦C) during which the plant is notresponsive to changes in photoperiod. P2: Extent to which development (expressedas days) is delayed for each hour increase in photoperiod above the longest pho-toperiod at which development proceeds at a maximum rate (which is consideredto be 12.5 h). P5: Thermal time from silking to physiological maturity (expressed in◦C day above a base temperature of 8 ◦C). G2: Maximum possible number of kernelsper plant. G3: Kernel filling rate during the linear grain filling stage and under opti-mum conditions (mg day−1). PHINT: Phyllochron interval, i.e. the interval in thermaltime (◦C day) between successive leaf tip appearances.

Table A4Maize genotype-specific input parameters for the DSSAT simulations at the MonzeFarmer Training Centre in Zambia.

Cultivar P1 P2 P5 G2 G3 PHINT

CIM627 177 1.598 789 741 7.5 53CIM635 100 0.799 778 280 4.9 77ZSS261 210 0.496 625 632 6.6 57

P1: Thermal time from seedling emergence to the end of the juvenile phase(expressed in ◦C day, above a base temperature of 8 ◦C) during which the plant is notresponsive to changes in photoperiod. P2: Extent to which development (expressedas days) is delayed for each hour increase in photoperiod above the longest pho-toperiod at which development proceeds at a maximum rate (which is consideredto be 12.5 h). P5: Thermal time from silking to physiological maturity (expressed in◦C day above a base temperature of 8 ◦C). G2: Maximum possible number of kernelsper plant. G3: Kernel filling rate during the linear grain filling stage and under opti-mum conditions (mg day−1). PHINT: Phyllochron interval, i.e. the interval in thermaltime (◦C day) between successive leaf tip appearances.

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

ING Model

A

1 tems a

R

A

A

A

A

B

B

D

E

E

E

E

G

G

H

J

K

K

L

M

M

M

M

M

76–90.Triomphe, B., Kienzle, J., Bwalya, M., Damgaard-Larsen, S., 2007. Case-study project

ARTICLEGEE-4500; No. of Pages 16

6 M. Corbeels et al. / Agriculture, Ecosys

eferences

ffholder, F., Tittonell, P., Corbeels, M., Roux, S., Motisi, N., Tixier, P., Wery, J., 2012.Ad hoc modeling in agronomy: what have we learned in the last 15 years? Agron.J. 104, 735–748.

ndales, A.A., Batchelor, W.D., Anderson, C.E., Farnham, D.E., Whigham, D.K., 2000.Incorporating tillage effects into a soybean model. Agr. Syst. 66, 69–98.

ndrieu, N., Vayssières, J., Corbeels, M., Blanchard, M., Vall, E., Tittonell, P., 2013.Synergies at farm scale and trade-offs at village scale: biomass uses in an agro-pastoral system of the Sudano-Sahelian zone of Burkina Faso. Agr. Syst., underreview.

ttonaty, J.M., Le Bars, M., Allaya, M., Le Grusse, P., 2005. OLYMPE, Manueld‘utilisation. CIHEAM-IAM, Montpellier.

audron, F., Andersson, J.A., Corbeels, M., Giller, K.E., 2012. Failing to yield? Ploughs,conservation agriculture and the problem of agricultural intensification. Anexample from the Zambezi Valley, Zimbabwe. J. Dev. Stud. 48, 393–412.

audron, F., Delmotte, S., Corbeels, M., Herrera, J.M., Tittonell, P., 2013. Multi-scaletrade-off analysis of cereal residue use for livestock feeding vs. soil mulching inthe Mid-Zambezi Valley, Zimbabwe. Agr. Syst., under review.

ongmo, A.L., Vall, E., Dugué, P., Njoya, A., Lossouarn, J., 2012. Designing a processof co-management of crop residues for forage and soil conservation in Sudano-Sahel. J. Sustain. Agr. 36, 106–126.

hui, S., Pender, J., 2005. Resource degradation, low agricultural productivity, andpoverty in sub-Saharan Africa: pathways out of the spiral. Agr. Econ. 32,225–242.

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

renstein, O., 2003. Smallholder conservation farming in the tropics and subtropics:a guide to the development and dissemination of mulching with crop residuesand cover crops. Agr. Ecosyst. Environ. 100, 17–37.

renstein, O., Sayre, K., Wall, P., Hellin, J., Dixon, J., 2012. Conservation agriculturein maize and wheat-based systems in the (sub)tropics: lessons from adapta-tion initiatives in South Asia, Mexico, and Southern Africa. J. Sustain. Agr. 36,180–206.

iller, K.E., Witter, E., Corbeels, M., Tittonell, P., 2009. Conservation agriculture andsmallholder farming in Africa. The heretics‘view. Field Crop. Res. 114, 23–34.

iller, K.E., Corbeels, M., Nyamangara, J., Triomphe, B., Affholder, F., Scopel, E., Tit-tonell, P., 2011. A research agenda to explore the role of conservation agriculturein African smallholder farming systems. Field Crop Res. 124, 468–472.

obbs, P.R., 2007. Conservation agriculture: what is it and why is it important forfuture sustainable food production? J. Agr. Sci. 145, 127–137.

ones, J.W., Hoogenboom, G., Porter, C.H., Boote, K.J., Batchelor, W.D., Hunt, L.A.,Wilkens, P.W., Singh, U., Gijsman, A.J., Ritchie, J.T., 2003. The DSSAT croppingsystem model. Eur. J. Agron. 18, 235–265.

assam, A., Friedrich, T., Shaxson, F., Pretty, J., 2009. The spread of conservation agri-culture: justification, sustainability and uptake. Int. J. Agric. Sustain. 7, 292–320.

nowler, D., Bradshaw, B., 2007. Farmers’ adoption of conservation agriculture: areview and synthesis of recent research. Food Pol. 32, 25–48.

obell, D.B., Burke, M.B., Tebaldi, C., Mastrandrea, M.D., Falcon, W.P., Naylor, R.L.,2008. Prioritizing climate change adaptation needs for food security in 2030.Science 319, 607–610.

acCarthy, D.S., Sommer, R., Vlek, P.L.G., 2009. Modeling the impacts of contrastingnutrient and residue management practices on grain yield of sorghum (Sorghumbicolor (L.) Moench) in a semi-arid region of Ghana using APSIM. Field Crop Res.113, 105–115.

ashingaidze, N., Madakadze, C., Twomlow, S., Nyamangara, J., Hove, L., 2012. Cropyield and weed growth under conservation agriculture in semi-arid Zimbabwe.Soil Till. Res. 124, 102–110.

azvimavi, K., Twomlow, S., 2009. Socioeconomic and institutional factors influenc-ing adoption of conservation agriculture by vulnerable households in Zimbabwe.Agr. Syst. 101, 20–29.

Please cite this article in press as: Corbeels, M., et al., Understanding the imanalysis. Agric. Ecosyst. Environ. (2013), http://dx.doi.org/10.1016/j.agee.2

koga, Z.J., Tumbo, S.D., Kihupi, N., Semoka, J., 2010. Extrapolating effects of conser-vation tillage on yield, soil moisture and dry spell mitigation using simulationmodeling. Phys. Chem. Earth, 686–698.

upangwa, W., Twomlow, S., Walker, S., 2012. Reduced tillage, mulching and rota-tional effects on maize (Zea mays L.), cowpea (Vigna unguiculata (Walp) L.) and

PRESSnd Environment xxx (2013) xxx– xxx

sorghum (Sorghum bicolor L. (Moench)) yields under semi-arid conditions. FieldCrop Res. 132, 139–148.

Naudin, K., Bruelle, G., Salgado, P., Penot, E., Scopel, E., Lubbers, M., de Ridder, N.,Giller, K.E., 2013. Trade-offs around use of biomass for livestock feed and soilcover at farm level in the Alaotra lake region of Madagascar. Agr. Syst., underreview.

NCATF, 2012. Approaches to the Implementation of Conservation Agriculture AmongPromoters in Malawi, Base Line Study. IFPRI and Min. of Agriculture, Irrigationand Water Development, Lilongwe.

Ndah, H.T., Schuler, J., Uthes, S., Zander, P., Triomphe, B., Mkomwa, S., Corbeels,M., 2013. Adoption potential of Conservation Agriculture in Africa: a newlydeveloped assessment approach (QAToCA) applied in Kenya and Tanzania. LandDegrad. Dev., http://dx.doi.org/10.1002/ldr.2191.

Neef, A., 2001. Land tenure and soil conservation practices: evidence from WestAfrica and Southeast Asia. In: Mohtar, R.H., Steinhardt, G.C. (Eds.), Sustainingthe Global Farm. Selected papers from the 10th International Soil ConservationOrganization meeting held May 24–29 1999 at Purdue University and the USDA-ARS National Soil Erosion Research Laboratory. , pp. 125–130.

Pannell, D.J., Marshall, G.R., Barr, N., Curtis, A., Vanclay, F., Wilkinson, R., 2006. Under-standing and promoting adoption of conservation practices by rural landholders.Aust. J. Exp. Agr. 46, 1407–1424.

Parton, W.J., Schimel, D.S., Cole, C.V., Ojima, D.S., 1987. Analysis of factors control-ling soil organic matter levels in Great Plains grasslands. Soil Sci. Soc. Am. J. 51,1173–1179.

Porter, C., Jones, J., Adiku, S., Gijsman, A., Gargiulo, O., Naab, J., 2010. Modelingorganic carbon and carbon-mediated soil processes in DSSAT v4.5. Oper. Res.10, 247–278.

Ritchie, J.T., 1998. Soil water balance and plant stress. In: Tsuji, G.Y., Hoogenboom, G.,Thornton, P.K. (Eds.), Understanding Options for Agricultural Production. KluwerAcademic Publishers, Dordrecht, The Netherlands, pp. 41–54.

Rusinamhodzi, L., Corbeels, M., van Wijk, M., Rufino, M.C., Nyamangara, J., Giller, K.E.,2011. Long-term effects of conservation agriculture practices on maize yieldsunder rain-fed conditions: lessons for southern Africa. Agron. Sustain. Dev. 31,657–673.

Scopel, E., Triomphe, B., Affholder, F., Da Silva, F.A.M., Corbeels, M., Xavier, J.H.V.,Lahmar, R., Recous, S., Bernoux, M., Blanchart, E., De Carvalho Mendes, I., DeTourdonnet, S., 2013. Conservation agriculture cropping systems in temperateand tropical conditions, performances and impacts. A review. Agron. Sustain.Dev. 33, 113–130.

Semenov, M.A., Barrow, E.M., 1998. Use of a stochastic weather generator in thedevelopment of climate change scenarios. Clim. Change 35, 397–414.

Sommer, R., Wall, P.C., Govaerts, B., 2007. Model-based assessment of maize crop-ping under conventional and conservation agriculture in highland Mexico. SoilTill. Res. 94, 10–83.

Sumberg, J., 2005. Constraints to the adoption of agricultural innovations. Is it timefor a re-think? Outlook Agr. 34, 7–10.

Thierfelder, C., Wall, P.C., 2009. Effects of conservation agriculture techniques oninfiltration and soil water content in Zambia and Zimbabwe. Soil Till. Res. 105,217–227.

Thierfelder, C., Wall, P.C., 2010. Investigating conservation agriculture (CA) systemsin Zambia and Zimbabwe to mitigate future effects of climate change. J. CropImprov. 24, 113–121.

Thierfelder, C., Wall, P.C., 2012. Effects of conservation agriculture on soil quality andproductivity in contrasting agro-ecological environments of Zimbabwe. Soil UseManage. 28, 209–220.

Thierfelder, C., Mwila, M., Rusinamhodzi, L., 2013. Conservation agriculture in east-ern and southern provinces of Zambia: long-term effects on soil quality andmaize productivity. Soil Till. Res. 126, 246–258.

Tittonell, P., Giller, K.E., 2013. When yield gaps are poverty traps: the paradigm ofecological intensification in African smallholder agriculture. Field Crop Res. 143,

pact and adoption of conservation agriculture in Africa: A multi-scale013.10.011

background and method. In: Boahen, P., Dartey, B.A., Dogbe, G.D., BoadiF A.D.,Triomphe, B., Damgaard-Larsen, S., Ashburner, J. (Eds.), Conservation Agricultureas Practised in Ghana. ACT, CIRAD, FAO, Nairobi.