doubled-up legume rotations improve soil fertility and maintain … · doubled-up legume rotations...

11
Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based cropping systems in Malawi Alex Smith, Sieglinde Snapp , John Dimes, Chiwimbo Gwenambira, Regis Chikowo Department of Plant, Soil and Microbial Sciences, Michigan State University, 1066 Bogue St., Rm. A286, East Lansing, MI 48824, United States abstract article info Article history: Received 11 December 2015 Received in revised form 5 March 2016 Accepted 13 March 2016 Available online 2 April 2016 Smallholder farmers in Malawi must cope with small farm size, low soil fertility and production risks associated with rainfed agriculture. Integration of legumes into maize-based cropping systems is advocated as a means to increase production of diverse nutrient-dense grains and improve soil fertility. It is difcult to achieve both aims simultaneously, however. Short-duration grain legumes rarely produce enough biomass to appreciatively improve soils, and long duration pigeonpea, commonly grown in Malawi as a dual purpose crop, produces little or no edible grain as a consequence of grain-lling into the dry season. A novel technology is the doubled-up le- gume rotation (DLR) system in which two legumes with complementary phenology are intercropped and grown in rotation with maize. Initial performance from on-farm research is favorable; however, it is crucial to under- stand competition for resources in mixed cropping systems under variable soil and climate conditions. We used soil and crop yield data from farmer participatory trials to parameterize the Agricultural Production Systems Simulator (APSIM) and evaluate its performance in simulating observed treatments at three locations in central Malawi. We used the calibrated APSIM model to investigate the performance of DLR and other maize-based sys- tems across 26 growing seasons (19792005) in the three agroecologies. We simulated two DLR systems (maize rotated with a groundnut/pigeonpea or soybean/pigeonpea intercrop), maize rotated with groundnut or soy- bean, maize intercropped with pigeonpea, and continuous maize under a range of N fertilizer inputs. We extend- ed ndings to the household level by determining calorie and protein yields of these systems, and calculating the chance that an average household could meet their food requirements by dedicating all available farmland to a given system. Simulated maize grain yields in DLR and maize-grain legume rotations were essentially equivalent, and exceeded yields in maize/pigeonpea intercrop and sole cropped maize receiving comparable fertility inputs. All rotation systems were more likely to meet household calorie and protein needs than other systems receiving equivalent inputs. DLR systems accumulated higher total soil C and N over time than traditional rotation systems in areas where pigeonpea performed well. However, the effects of improved soil fertility on maize yields were counterbalanced by factors including N immobilization and water availability. We conclude that where growing conditions allow, DLR can harness the complementary phenology of pigeonpea to build soil quality for the future without reducing maize yields or compromising household food production in the immediate term. © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Keywords: APSIM Soil fertility Climate variability Legume rotation Pigeonpea intercrop Agroecology 1. Introduction 1.1. Background Smallholder farmers practicing rain fed agriculture in sub-Saharan Africa (SSA) face recurring episodes of food insecurity, and food demand is expected to increase in coming decades (Thornton et al., 2011). At the same time, climate change will alter patterns of temperature and rain- fall in SSA that may cause many areas to develop climate regimes with no present-day analog. With the threat of climate change and increasing demands placed on agroecosystems, farmers will need to adapt to new conditions and the unrelenting requirement of fertility inputs to im- prove food production. Incorporation of legumes into cereal based cropping systems has frequently been advocated as a means of increas- ing soil fertility and agroecological resilience for farmers with limited access to nutrient resources (Snapp et al., 1998; Thierfelder et al., 2012). However, many legumes utilized to improve soil fertility do not provide edible grain, while smallholder farmers are often forced to pri- oritize food production and crop sales over potential soil fertility bene- ts (Snapp et al., 2002). As a result, recent farmer-participatory research efforts have focused on incorporating soil fertility building legumes into Agricultural Systems 145 (2016) 139149 Corresponding author. E-mail address: [email protected] (S. Snapp). http://dx.doi.org/10.1016/j.agsy.2016.03.008 0308-521X/© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Contents lists available at ScienceDirect Agricultural Systems journal homepage: www.elsevier.com/locate/agsy

Upload: others

Post on 27-Mar-2020

7 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

Agricultural Systems 145 (2016) 139–149

Contents lists available at ScienceDirect

Agricultural Systems

j ourna l homepage: www.e lsev ie r .com/ locate /agsy

Doubled-up legume rotations improve soil fertility and maintainproductivity under variable conditions in maize-based croppingsystems in Malawi

Alex Smith, Sieglinde Snapp ⁎, John Dimes, Chiwimbo Gwenambira, Regis ChikowoDepartment of Plant, Soil and Microbial Sciences, Michigan State University, 1066 Bogue St., Rm. A286, East Lansing, MI 48824, United States

⁎ Corresponding author.E-mail address: [email protected] (S. Snapp).

http://dx.doi.org/10.1016/j.agsy.2016.03.0080308-521X/© 2016 The Authors. Published by Elsevier Ltd

a b s t r a c t

a r t i c l e i n f o

Article history:Received 11 December 2015Received in revised form 5 March 2016Accepted 13 March 2016Available online 2 April 2016

Smallholder farmers in Malawi must cope with small farm size, low soil fertility and production risks associatedwith rainfed agriculture. Integration of legumes into maize-based cropping systems is advocated as a means toincrease production of diverse nutrient-dense grains and improve soil fertility. It is difficult to achieve bothaims simultaneously, however. Short-duration grain legumes rarely produce enough biomass to appreciativelyimprove soils, and long duration pigeonpea, commonly grown in Malawi as a dual purpose crop, produces littleor no edible grain as a consequence of grain-filling into the dry season. A novel technology is the doubled-up le-gume rotation (DLR) system inwhich two legumeswith complementary phenology are intercropped and grownin rotation with maize. Initial performance from on-farm research is favorable; however, it is crucial to under-stand competition for resources in mixed cropping systems under variable soil and climate conditions. Weused soil and crop yield data from farmer participatory trials to parameterize the Agricultural Production SystemsSimulator (APSIM) and evaluate its performance in simulating observed treatments at three locations in centralMalawi. We used the calibrated APSIMmodel to investigate the performance of DLR and othermaize-based sys-tems across 26 growing seasons (1979–2005) in the three agroecologies.We simulated two DLR systems (maizerotated with a groundnut/pigeonpea or soybean/pigeonpea intercrop), maize rotated with groundnut or soy-bean, maize intercroppedwith pigeonpea, and continuousmaize under a range of N fertilizer inputs.We extend-ed findings to the household level by determining calorie and protein yields of these systems, and calculating thechance that an average household could meet their food requirements by dedicating all available farmland to agiven system. Simulatedmaize grain yields in DLR andmaize-grain legume rotationswere essentially equivalent,and exceeded yields in maize/pigeonpea intercrop and sole croppedmaize receiving comparable fertility inputs.All rotation systems weremore likely to meet household calorie and protein needs than other systems receivingequivalent inputs. DLR systems accumulated higher total soil C and N over time than traditional rotation systemsin areas where pigeonpea performed well. However, the effects of improved soil fertility on maize yields werecounterbalanced by factors including N immobilization and water availability. We conclude that where growingconditions allow, DLR can harness the complementary phenology of pigeonpea to build soil quality for the futurewithout reducing maize yields or compromising household food production in the immediate term.

© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords:APSIMSoil fertilityClimate variabilityLegume rotationPigeonpea intercropAgroecology

1. Introduction

1.1. Background

Smallholder farmers practicing rain fed agriculture in sub-SaharanAfrica (SSA) face recurring episodes of food insecurity, and food demandis expected to increase in coming decades (Thornton et al., 2011). At thesame time, climate change will alter patterns of temperature and rain-fall in SSA that may cause many areas to develop climate regimes with

. This is an open access article under

nopresent-day analog.With the threat of climate change and increasingdemands placed on agroecosystems, farmers will need to adapt to newconditions and the unrelenting requirement of fertility inputs to im-prove food production. Incorporation of legumes into cereal basedcropping systems has frequently been advocated as a means of increas-ing soil fertility and agroecological resilience for farmers with limitedaccess to nutrient resources (Snapp et al., 1998; Thierfelder et al.,2012). However, many legumes utilized to improve soil fertility do notprovide edible grain, while smallholder farmers are often forced to pri-oritize food production and crop sales over potential soil fertility bene-fits (Snapp et al., 2002). As a result, recent farmer-participatory researchefforts have focused on incorporating soil fertility building legumes into

the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Page 2: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

Table 3Soil properties measured in experimental trials used to parameterize models for scenarioanalysis. Plant available water capacity (PAWC) is calculated as soil water at the drainedupper limit minus soil water at the 15 bar lower limit (Burk and Dalgliesh, 2008).

Depth(cm)

% Sand % Silt % Clay % C % N pH Bray P(ppm)

PAWC(mm)

Lintipe0–15 22 29 48 2.3 0.13 4.9 13.5 19.215–30 19 26 55 2.2 0.12 4.9 12.5 18.030–60 18 24 58 2.0 0.11 5.0 8.5 34.860–90 6 16 78 0.7 0.05 5.1 2.5 34.590–120 5 19 76 0.4 0.04 5.2 2.5 34.8

Kandeu0–15 75 17 9 1.0 0.09 5.0 72.5 10.715–30 74 13 13 0.7 0.06 5.4 42.5 10.530–60 68 17 16 0.8 0.06 5.2 27.5 24.960–90 40 20 40 1.5 0.12 4.9 9.5 36.6

Golomoti0–15 52 22 26 0.9 0.06 4.5 56.5 16.215–30 43 20 37 0.5 0.05 4.2 20 14.630–60 34 19 47 0.1 0.04 4.9 12 18.360–90 27 22 51 0.2 0.05 4.3 16.5 16.890–120 29 21 50 0.2 0.02 4.2 12 11.7

Table 1Seasonal rainfall at the three study locations, in mm.

Site 2012/2013 2013/2014 Mean 1979–2005 CV 1979–2005

Lintipe 1018 926 937 0.19Kandeu 888 864 855 0.23Golomoti 794 933 834 0.27

140 A. Smith et al. / Agricultural Systems 145 (2016) 139–149

maize-based cropping systems using approaches that do not compro-mise food crop production (Snapp et al., 2010).

Pigeonpea (Cajanus cajan) is a woody legume which can contributeN to agricultural soils through the production of N rich biomass whilealso producing edible grain (Chikowo et al., 2004). Pigeonpea has typi-cally been integrated into African cropping systems as an intercropwitha staple cereal crop, generally maize (Myaka et al., 2006; Snapp et al.,2010; Waddington et al., 2007). The practice of intercropping canincrease the yield stability and resource use efficiency of agriculturalsystems (Trenbath, 1999), but under resource-limited conditions yieldsin intercropmay be suppressed due to competitive interactions (Morrisand Garrity, 1993). When pigeonpea is intercropped directly withmaize, maize has often been found to achieve yields equivalent to solecrop (Chirwa et al., 2003; Myaka et al., 2006), and in some cases maizeyields have increased over time as pigeonpea adds fertility to the soil(Rusinamhodzi et al., 2012). However, in a long-term study spanning 10growing seasons, Waddington et al. (2007) found that maize wassuppressed in intercrop with pigeonpea during some growing seasons.

In order to harness the soil-improving properties of pigeonpeawhile maintaining maize yields, researchers and farmers have begunexperimenting with novel cropping systems now known as doubled-up legume rotations (DLR) (Snapp et al., 2002). In DLR, maize isgrown in rotation with an intercrop of two legumes which typicallyinclude one grain legume grown as an understory to a longer-durationlegume intended to produce N-rich biomass. As of 2016 the MalawiMinistry of Agriculture has released DLR systems for general use,allowing extensionists to actively promote these technologies(Ngwira, personal communication). To date, large-scale on-farm studiesof these DLR systems in Malawi have shown comparable grain yields,increased protein yields, and increased fertilizer efficiency when com-pared with maize monocrops (Snapp et al., 2010). There is evidencethat across much of Malawi the temporal complementary of a short-lived annual grain legume such as soybean (Glycine max) or groundnut(Arachis hypogaea) avoids or minimizes competition for water andother resources relative to the longer-duration, initially slow growingpigeonpea (Snapp and Silim, 2002). However, given that farmers mustcontendwith a highly variable climate, studies which span a large num-ber of growing seasons are needed to understand the resilience of DLRsystems in the face of climatic variability.

The Agricultural Production Systems Simulator (APSIM) is a process-based crop simulation model which has been extensively tested andused in sub-Saharan Africa (Keating et al., 2003; Robertson et al.,

Table 2Cropping systems examined in this study, with abbreviations. Bold cropping system abbreviatiicized abbreviations indicate systems used in scenario analysis, while italicized systemswere nolegume density and bracketed numbers indicate pigeonpea density.

Abbrev. System description Maize kg N/ha

Mz0 Continuous maize 0Mz35 Continuous maize 35Mz69 Continuous maize 69MzPp35 Maize pigeonpea intercrop 35MzPp69 Maize pigeonpea intercrop 69GnRot Maize groundnut rotation 35SbRot Maize soybean rotation 35GnDLR Maize groundnut pigeonpea DLR 35SbDLR Maize soybean pigeonpea DLR 35

2005). Together with its maize module which is based on the CERESmaize model, APSIM has been expanded to simulate a wide range ofcrops and cropping systems (Carberry et al., 1989). APSIM can be runusing long-term climate data, and simulations have been used to under-stand climate risk in agriculture (Ollenburger and Snapp, 2014; Rurindaet al., 2015). APSIM is relatively unique among process based crop sim-ulation models in its capacity to simulate intercrop systems (Malezieuxet al., 2009). APSIM employs an arbitrator routine to partition resourcesincluding light, water, and N between intercrop components (Carberryet al., 1996). The ability of APSIM to simulate maize legume intercropsand legume-cereal rotations has been tested using field data from sub-Saharan Africa (Adiku et al., 1995; Robertson et al., 2005). While thepigeonpea module for APSIM was initially developed using field datafrom India (Robertson et al., 2001) it has since been used to simulatemaize/pigeonpea cropping systems in Malawi (Ollenburger andSnapp, 2014).

1.2. Objectives

(I) Use APSIM to evaluate the performance of maize in DLR systemsacross 26 growing seasons in three agroecologies of centralMalawi, and contrast this performance to maize grown in rota-tion with a single grain legume, maize pigeonpea intercrop, andcontinuous maize.

(II) Examine simulated soil C and N dynamics to explore sustainabil-ity implications of DLR systems compared with other maize-based systems.

ons indicate systems that were grown in the field and used in model calibration. Non-ital-t used in scenario analysis. Under legume plants/m2, un-bracketed numbers indicate grain

Legume kg N/ha Maize plants/m2 Legume plants/m2

NA 4.2 NANA 4.2 NANA 4.2 NA12 3.8 [3.2]12 3.8 [3.2]12 4.2 7.612 4.2 18.212 4.2 7.4 [3.2]12 4.2 17.4 [3.2]

Page 3: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

Fig. 1. Results of model calibration for maize, groundnut and soybean grain yields at a) LintipeCrops are abbreviated asMz=maize, Gn= groundnut, and Sb= soybean. Bars represent obsemodeled yields. Bars with a single dot include yields from only a single growing season, while bCropping system abbreviations are given in Table 1.

Table 4Parameters manipulated during model calibration, and the final values employed inscenario analysis.

Kandeu Lintipe Golomoti

Fbiom (fraction organic C in microbialbiomass: 0–15, 15–30, 30–60, 60–90and 90–120 cm)

0.02, 0.01,0.01, 0.01,0.01

0.02, 0.01,0.01, 0.01,0.01

0.02, 0.01,0.01, 0.01,0.01

Finert (fraction inert organic C: 0–15,15–30, 30–60, 60–90, 90–120 cm)

0.55, 0.65,0.85, 0.99,0.99

0.7, 0.85,0.95, 0.99,0.99

0.4, 0.6,0.85, 0.90,0.99

Crop residue removal in dry season 70% 70% 70%Groundnut variety Custom Custom CustomGroundnut planting depth (mm) 75 75 75Soybean variety MG_3 MG_3 MG_3Soybean planting depth (mm) 50 50 50Pigeonpea variety Custom Custom CustomPigeonpea planting depth (mm) 30 30 30

141A. Smith et al. / Agricultural Systems 145 (2016) 139–149

(III) Determine whether DLR systems are capable of consistentlymeeting the calorie and protein needs of average farming house-holds based on household size and landholding.

2. Materials and methods

2.1. Research context

This study was conducted as part of the Africa RISING (Research InSustainable Intensification for the Next Generation) initiative, exploringstrategies for sustainable intensification of maize-based croppingsystems in central Malawi based on long-term farmer participatory re-search in the country (Snapp et al., 1998). Our modeling investigationinvolved three extension planning areas (EPAs) in central Malawi se-lected to represent three distinct agroecological zones: high elevationwith high water availability (Lintipe EPA), mid elevation with interme-diate water availability (Kandeu EPA), and low elevation with highevaporative demand and moderate water availability (Golomoti EPA;

, b) Kandeu and c) Golomoti, as well as d) pigeonpea biomass yields at all three locations.rved yields from field experiments, with error bars showing standard error. Dots representars with two dots include yields for both the 2012–2013 and 2013–2014 growing seasons.

Page 4: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

Table 5Household composition and landholding based on surveys of farming households at the study locations. Calorie and protein production per hectare required tomeet household needs areindicated by “Calories req.” and “Protein req.”

Children under 15 Adults 15–69 Seniors 70 and over Landholding (ha) Calories req. (Mcal/ha/yr) Protein req. (kg/ha/yr)

Lintipe 2.40 2.65 0.11 0.71 6097 114Golomoti 2.48 2.54 0.09 0.83 5121 95Kandeu 2.21 2.80 0.16 0.89 5005 94

142 A. Smith et al. / Agricultural Systems 145 (2016) 139–149

Table 1). One experimental trial at each site was utilized in this study,and these trials were managed collaboratively by researchers andfarmers (Snapp et al., 2002).

2.2. Household data collection

We collected data from households at each study location through asurvey conducted in June and July 2013, described previously (Hockett,2014). Respondents included 40 randomly selected households partici-pating in the Africa RISING project and 40 randomly selected householdsat each location not involved in the project (control households). Infor-mationprovidedby survey respondents includedhousehold composition,the amount of land being actively farmed, crops grown and agriculturalmanagement practices.

2.3. Agronomic trials

Yield data from the three experimental trials described above fromthe 2012–2013 through 2013–2014 growing seasons were used in thisstudy. Each trial included seven treatments, replicated three times

Fig. 2. Stability analysis of modeled maize yields at a) Lintipe, b) Kandeu and c) Golomoti. MaiYields for each system in each year are plotted against an environmental index consisting oLinear regressions of yield by environmental index are shown in these plots. See Table 2 fosupplementary data (Table S2).

each in 5 m × 5 m plots and arranged in a randomized complete blockdesign. (Table 2, also see supplementary data Table S1). Treatments re-flect amix of current farmer practice, alongwith two DLR systems. Cropvarieties used in each trial were recommended by the Malawi govern-ment as appropriate for the local environment.

All crops were planted in December. Crops were planted on ridgesformed by hand-hoe, following local practice. Intercrops used a semi-additive design, in which total plant density was highest in intercrop,and each intercrop partner was planted at a slightly lower densitythan in monocrop. Fertilizer applications in treatments receiving35 kg N/ha or 69 kg N/ha were split between a basal application atplanting and a top-dress application. Crop grain yields were measuredin each trial for all crops except for pigeonpea in the 2012–2013 and2013–2014 growing seasons. All plants within the three planting ridgesin the center of each five by five meter plot were harvested. Harvestedgrains were sun-dried for at least four weeks until grain moisture haddropped to between 14 and 16%, and then weighed. Grain moisturewas determined using amoisturemeter and all grain yields are reportedat 12.5%moisture content. Pigeonpea biomass wasmeasured in 2014 atthe Lintipe and Kandeu locations. Final plant densities for maize were

ze yields in rotation systems are presented for the response year, and are not annualized.f the mean maize yield for all cropping systems at a given location and in a given year.r cropping system abbreviations. R2 values for these regressions are presented in the

Page 5: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

143A. Smith et al. / Agricultural Systems 145 (2016) 139–149

measured in the 2013–2014 and 2014–2015 growing seasons, whilelegume final plant densities were measured only in the 2014–2015growing season.

2.4. Soil sampling and analysis

We collected soil profile data fromall trials after the 2013maize har-vest (Table 3). One soil pit was dug to a depth of 120 cm at each studylocation. From these pits, soil samples were taken at depths of 0–15 cm, 15–30 cm, 30–60 cm, 60–90 cm, and 90–120 cm. At Kandeu,the pit could only be dug to a depth of 90 cm due to impenetrableclay subsoil, and 90–120 cm samples were not taken. Additionalsamples were taken from three random points in all experimentaltrials. Samples were taken with a hand hoe at three depths: 0–15,15–30 and 30–60 cm. Samples from each depth were composited,resulting in one 0–15 cm, one 15–30 cm, and one 30–60 cm samplefrom each trial.

All soil samples were air dried and passed through a two mm sieve,and rocks and large pieces of organicmatterwere discarded. Soil texturewas determined from sieved soil using the hydrometer method(Kellogg Biological Station LTER, 2008). Soil pH in distilled water at a1:2 soil to water ratio was determined (Hendershot et al., 1993), usinga Metler Toledo SevenEasy S20 pH meter. A subsample of each soilwas pulverized using a shatter mill for chemical analysis. Total carbonand total nitrogen were determined from pulverized soil using a CarloErba NA1500 SeriesII Combustion Analyzer (Kellogg Biological StationLTER, 2003). Available P was determined from soil ground to pass a1 mm sieve and analyzed by the Michigan State University Soil andPlant Nutrition Laboratory using the Bray P extraction and colorimetricmethod (Bray and Kurtz, 1945).

Fig. 3. Stability analysis ofmodeled legume yields at a) Lintipe, b) Kandeu and c)Golomoti. Yieldthe mean groundnut or soybean yield for all cropping systems at a given location and in a givTable 2 for cropping system abbreviations. R2 values for these regressions are presented in the

2.5. Simulation analysis

APSIM version 7.7 was calibrated using field data from the 2012–2013 and 2013–2014 growing seasons. Scenario analysis utilized thecalibrated model to simulate cropping system performance across aperiod from 1979 to 2005.

2.5.1. Model parameterizationAPSIM was parameterized using soil data collected from the three

experimental trials. For all experimental trials, data on soil C, soil N,pH, and texture from the hand hoe samples at depths of 0–15, 15–30and 30–60 cm were used. In the 60–90, and 90–120 cm layers, datafrom the soil pit were used (Table 3). At Kandeu, data from the60–90 cm layer were used to parameterize both the 60–90 and the90–120 cm soil layers (Mabapa et al., 2010).

Soil water at saturation, field capacity, and the 15-bar lower limitwas determined based on % sand, % clay and % soil organic matterusing the SPAW model (Saxton and Willey, 2006) (Table 3). We deter-mined soil organic matter as a function of total soil carbon using a con-version factor of 1.724 (Kerven et al., 2000). SPAW also provided anestimate of bulk density, which was used as a model input. The soilwater at air dry parameter was set slightly below the 15 bar lowerlimit in the top two soil layers, and equal to the 15-bar lower limit indeeper layers. Crop lower limits for extractable soil water were setequal to the 15-bar lower limit. Additional soil parameters includingcrop kl and XF were selected from a library of generic soil profiles forAPSIM that includes soil descriptions from Malawi (APSIM Initiative,2013).

Daily rainfall data obtained using rain gauges at each study locationbetween 2012 and 2014 were used for purposes of model calibration.Temperature and solar radiation data were obtained from the National

s for each system in each year are plotted against an environmental index (EI) consisting ofen year. Linear regressions of yield by environmental index are shown in these plots. Seesupplementary data (Table S2).

Page 6: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

144 A. Smith et al. / Agricultural Systems 145 (2016) 139–149

Centers for Environmental Protection (NCEP) Climate Forecast SystemVersion Two, which includes daily meteorological data at all points onthe globe at a resolution of 0.25° latitude and longitude (Saha et al.,2014). Daily maximum temperatures were increased by 1.1 °C anddaily minimum temperatures by 1.7 °C based on a comparison ofNCEP temperature data with nearby meteorological stations. In orderto prevent excessive carryover of soil water from one season to thenext, soil water was reset on October 1st of each year.

Plant densities in our calibrationmodels were set based on densitiesmeasured in the field (Table 2). Maize plant densities measured in the2013–2014 season were used, along with legume densities measuredin the 2014–2015 season.

Two maize cultivars were used in our crop simulations. The SC401cultivar description was used to simulate medium duration varietiesSC403 and PAN53 at Linthipe and Golomoti. The PAN6479 cultivar de-scription was used to simulate the shorter duration varieties DK8033and DK8053 at Kandeu and Golomoti. Appropriate legume cultivar de-scriptions were determined through the calibration process. All cultivardescriptions used in the simulation offield trials are listed in the supple-mentary data (Table S1).

Predicted maize, groundnut and soybean grain yields from the2012–2013 and 2013–2014 seasons were compared with measuredyields in the two seasons. In addition, pigeonpea stover biomass yieldswere compared with yields obtained in 2013–2014. Soil parameters(Fbiom, Finert) and crop residue management were first calibrated toaccurately simulate unfertilizedmaize grain yields at each site. Simulat-ed yields of legumes were calibrated by adjusting cultivar descriptionsand planting depth as shown in Table 4 until the modeled yields fellas close as possible to measured yields (Fig. 1). Adjustment of cultivar

Fig. 4. Total soil C as a percentage of soil dry mass in the top 15 cm of soil at a) Linti

descriptions affected the energy allocation and phenology of the crops,while the adjustment of planting depth had a minor effect on emer-gence time.

Groundnut and pigeonpea cultivar descriptions included in APSIMwere unable to satisfactorily predict grain and biomass yields in our tri-als and therefore custom groundnut and pigeonpea cultivar descrip-tions were created. The custom groundnut cultivar was based on the“Chico” cultivar with the maximum harvest index adjusted from 0.45to 0.35. The lower harvest index is consistent with a spreading ratherthan erect canopy structure of the test cultivar. The custom pigeonpeacultivar was based on the “medium duration” cultivar descriptionincluded in APSIM, with biomass yields increased by adjusting themax-imum harvest index from 0.2 to 0.1, based on data from detailed bio-mass characterization of pigeonpea in these trials (Gwenambira, 2015).

2.5.2. Scenario analysisScenario analyseswere run fromNovember 1, 1979 through June 30,

2005 using soil and crop parameters from the calibrated models at eachstudy location. For these analyses maize cultivar sc401 was used for allsites. Rainfall data for long-term simulations were obtained fromMalawi Meteorological Service stations that were located nearby andat similar elevation to the study sites. The best station to representeach study locationwas selected by comparing rainfall valuesmeasuredby the extension planning area office at each study location for the years2010–2013 with nearby meteorological stations. Solar radiation andtemperature data were obtained from the NCEP Climate Forecast Sys-tem reanalysis, which includes daily meteorological data at all pointson the globe at a resolution of 0.3° latitude and longitude (Saha et al.,2010). Daily maximum and minimum temperatures were adjusted as

pe, b) Kandeu, and c) Golomoti. See Table 2 for cropping system abbreviations.

Page 7: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

145A. Smith et al. / Agricultural Systems 145 (2016) 139–149

in calibration, and soil water was reset on October 1st of each year.Crops were planted in a window from November 1st through January31st, when 20 mm of rain had accumulated over the course of 3 days(Kassie et al., 2013).

At each site, all cropping systems listed in Table 2 in non-italicizedfont were simulated. For rotational systems, two staggered simulationswere run so that each phase of the rotation was present in every grow-ing season. We set planting densities for each sole crop equal to the av-erage planting density for that sole crop across all treatments in all fieldtrials. Similarly, we set planting densities for each intercrop equal to theaverage planting density for that crop in intercrop across all treatmentsin all field trials (Table 2). Stability analysis was used to evaluate thesimulated yields of different cropping systems across the full range ofgrowing seasons from 1979 through 2005 (Lightfoot et al., 1987). Fora given species, grain yields from each cropping system were plottedagainst the average species yield across all cropping systems in a givenseason. The index yield is an estimate of the seasonal conditions forcrop growth, providing a basis for comparing crop yields in differentcropping systems across seasons, ranging from the least favorable tothe most favorable seasons.

We also determined the chance that average farming householdscouldmeet their calorie and protein needs by growing the cropping sys-tems examined in this study. For smallholder farming households, thechance of producing enough food to meet household needs has beenused as a measure of production risk, as well as a partial indicator ofhousehold food security (Snapp et al., 2014). The calorie and proteinproduction per hectare required to meet the nutritional needs of a typ-ical farming household were calculated based on mean household sizeand landholding at each study location (Table 5). We obtained averagedaily calorie and protein requirements for children 0–14 years of age,adults 15–69 years engaged in moderate to heavy activity, and seniors70 years and older, from the Food and Agriculture Organization

Fig. 5. Total soil N as a percentage of soil dry mass in the top 15 cm of soil at a) Linti

(2004) and World Health Organization (2007). We determined the av-erage number of children 14 and under, adults 15–69, and seniors 70and over, and used these data to calculate household protein and calorierequirements. Calorie and protein productionwas determined based onaverage household landholding, simulated grain yields, and standardvalues for the calorie and protein contents of grain crops in Malawi(Drost, un published result; Snapp et al., 2010). For the purposes ofthis analysis, we assumed that all of the household's arable land wasdedicated to a given cropping system. In the case of rotational systems,half was dedicated to one stage of the rotation and half to the other.

3. Results

3.1. Predictive performance

Simulated maize and legume grain yields generally approximatedthe observed yields from the 2012–2013 and 2013–2014 growing sea-sons (Fig. 1). While maize yields were sometimes under-predicted,the model captured the direction and relative magnitude of maizeyield responses to N fertilization and legume introduction at all sites.In most cases, the model also captured the response of groundnut andsoybean to intercropping with pigeonpea. The sole exception concernssoybean at Lintipe where field observations show suppression of soy-bean when intercropped with pigeonpea, while the model does notshow suppression of soybean.

Simulated pigeonpea biomass yields also approximated measuredbiomass yields in the 2013–2014 season (Fig. 1). There were several ex-ceptions, however. Biomass yields of pigeonpea grown in intercropwithgroundnut at Lintipe were greatly under-predicted, while biomassyields of pigeonpea intercropped with maize at Kandeu were greatlyover-predicted.

pe, b) Kandeu, and c) Golomoti. See Table 2 for cropping system abbreviations.

Page 8: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

146 A. Smith et al. / Agricultural Systems 145 (2016) 139–149

3.2. Scenario analysis

Maize yields at all sites were highly responsive to fertilizer in morefavorable growing seasons. In less favorable seasons at Kandeu andGolomoti, the two driest sites, fertilizer input had a limited effect onmaize yield. Both DLR and traditional rotation systems increasedmaize grain yields when compared with monoculture maize or maizepigeonpea intercrops receiving the same fertilizer rate (35 kg N/ha;Fig. 2). At all sites, this increase in maize yield was approximatelyequal to the increase achieved when an additional 34 kg N/ha wasadded to continuousmaize systems. Increases inmaize yields due to ro-tation were most marked in more favorable growing seasons. In poorergrowing seasons at Kandeu and Golomoti, maize grown in rotationachieved yields only slightly above continuous maize.

DLR systems generally had an impact onmaize yields that was com-parable to traditional rotation systems. The sole exception was at

Fig. 6. Probability that average households will meet calorie and protein needs with their owcropping system will meet at least 100%, at least 200% and at least 300% of household calorie a

Lintipe, where maize yields in DLR systems were lower than traditionalrotations in better growing seasons.

Grain yields of soybean and groundnut were unaffected bypigeonpea intercrops across the full range of sites and season (Fig. 3).

Stover production was much higher in DLR systems and maize/pigeonpea intercrops than in other cropping systems (see supplemen-tary data Figure S1). The difference was particularly marked at Lintipeand Kandeu, where pigeonpea biomass yields were relatively high,and less pronounced at Golomoti where pigeonpea yields were lower(Fig. 1).

3.3. Simulated soil conditions

Total C and N in the top 15 cm of soil decreased over the course ofour study period (1979–2005) for all sole cropped maize systems inall study locations (Figs. 4 and 5). The magnitude of the decrease was

n production at a) Lintipe, b) Kandeu and c) Golomoti. Bars show the chance that eachnd protein needs in any given year. See Table 2 for cropping system abbreviations.

Page 9: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

147A. Smith et al. / Agricultural Systems 145 (2016) 139–149

inversely related to the amount of N fertilizer added to the system. Ro-tation of maize with a grain legume slightly reduced the magnitude ofthis decrease in soil C and N. However, when pigeonpea was added tothe cropping system (as an intercrop with either grain legumes ormaize) higher levels of total soil C and N were achieved, causing totalC and N to remain approximately constant (Lintipe, Golomoti) or in-crease (Kandeu). Pigeonpea had the greatest impact on total soil C andN atKandeu and Lintipe, where pigeonpea biomass yieldswere relative-ly high. Effects of pigonpea on soil C and N at Golomoti were muted bypoorer pigeonpea performance.

In contrast to total soil N, NO3 at the time of maize plantingwas con-sistently higher in traditional rotations than in other cropping systemsincluding DLR systems (see supplementary data Figure S2). This patternwas strongest at the Lintipe andKandeu locations, and somewhatweak-er at Golomoti. We also observed that available soil water at maizeplanting was higher in traditional rotations than in DLR systems atLintipe (see supplementary data Figure S3).

3.4. Chance of meeting food needs

At all locations, both traditional rotations andDLR systemswere ableto meet household calorie and protein needs in themajority of growingseasons based on average household size and landholding. The chancesof traditional rotations and DLR systemsmeeting household calorie andprotein needs were equal to or greater than those produced by solecropped maize receiving the same fertilizer rate of 35 kg N/ha (Fig. 6).DLR systems did not produce any consistent differences in the chanceof meeting household calorie and protein needs when compared withtraditional rotations.

4. Discussion

The accurate predictions of observed maize yields (Fig. 1) confirmprior observations that APSIM is able to predict maize response to inor-ganic and organic N inputs (Carberry et al., 1989; Robertson et al., 2005).Our study has also shown that APSIM is able to effectively simulatecompetitive dynamics in both maize/legume and legume/legumeintercrops.

DLR systems and traditional rotations produced maize yields thatwere higher than continuous maize or maize pigeonpea intercropsystems receiving equivalent fertility inputs (Fig. 2). In fact, the yieldboost due to rotation was approximately equal to the effects of addingan additional 34 kg N/ha in the recommended fertilizer treatment. Theyield advantage achieved inmaize legume rotations relative to continu-ousmaizewas not always uniform fromone growing season to the next,however. At Kandeu and Golomoti, the two drier locations, the advan-tages of rotation depended onwater availability. In poorer growing sea-sons when crop growth was water-limited, rotation did not greatlyimprovemaize yields.Most of the improvement came inmore favorableseasons when crop growth was less limited by water. This pattern didnot hold at Lintipe, the wettest location. There, the maize yield advan-tage gained through rotation was relatively uniform from season toseason.

DLR systems improved soil fertility over time when compared withtraditional rotations, as indicated by increasing total C and N levels(Figs. 4 and 5). This raises the question of whyDLR systems did not pro-duce highermaize yields than traditional rotations, and in fact producedlower maize yields in the best growing seasons at the most productivestudy location. DLR systems produce far higher levels of non-grainaboveground biomass (or stover) than traditional rotations due to thepresence of pigeonpea. In our models, 30% of all stover is incorporatedinto the soil after harvest. The immediate effect of incorporating largeamounts of stover into the soil appears to be immobilization of a portionof the plant available N at the time of maize planting. This pattern is es-pecially apparent at the Lintipe and Kandeu study locations, where soilNO3 levels are much lower in DLR systems than in traditional rotations

at the time of maize planting. The effects of this N immobilization ap-pear to counter-balance any positive effects that increasing levels oftotal soil C and N could otherwise have on maize yields in DLR systemsat these locations. We also explored the possibility that maize yields inDLR systems may be suppressed through increased soil water draw-down by pigeonpea in the previous growing season. An examinationof available soil water at maize sowing shows little effect of pigeonpeaon water available to maize in the next growing season at the twodrier sites, Kandeu and Golomoti (see supplementary data figure S3).At Lintipe, more water was available to maize following a grain legumethan following a grain legume pigeonpea intercrop. This may alsopartially account for the fact that maize yields in DLR systems weresometimes lower than yields in traditional rotations at Lintipe.

At the Golomoti study location we observed far less improvement insoil fertility over time in DLR systems compared with other croppingsystems, much lower biomass production in DLR systems, and limitedevidence of N immobilization. These differences are attributable topoorer pigeonpea performance at Golomoti than at the other locations.Golomoti stands apart as the most water limited study location. Sincegrain production by soybean and peanut at Golomoti were not sup-pressed in DLR systems (Fig. 3), it appears that in legume-legume inter-crops the faster-establishing grain legumes consumed the majority ofavailable water resources before later-emerging pigeonpea could be-come established, resulting in low pigeonpea yields. This supportsprior observations of heightened competitive dynamics in intercropsunder water-limited conditions (Morris and Garrity, 1993). However,it also supports the farmer-relevance and temporal compatibility ofthese grain legume combinations, one being short and the otherlonger-duration where the shorter-duration soybean and groundnutare prioritized by most farmers (Snapp and Silim, 2002). The observa-tion that these short-lived species had first access to soil moisture inwater-limited environments is supportive of farmer acceptance of theDLR systems.

Chances of meeting household calorie and protein needs were con-sistently greater in rotational systems than in either maize pigeonpeaintercrop or continuous maize receiving 35 kg N/ha. This pattern wasapparent for both DLR systems and traditional rotations. The patternalso held true across all growing seasons, from the most favorable tothe least favorable. At the same time, rotation systems received lowerN inputs than continuous maize, as legume crops grown in this studywere fertilized with only 12 kg N/ha. Therefore, both DLR systems andtraditional rotations ensure food production at least equal to continuousmaize systems, while requiring lesser investments in fertilizer. It is im-portant to note that this analysis is based on a hypothetical situation inwhich farminghouseholds choose to dedicate all of their arable land to asingle cropping system. In reality, chances of meeting household foodneeds would depend on farmers' land allocation decisions as well as ahost of other factors.

In this study, we chose to take a conservative approach to quantify-ing the benefits of DLR systems. We contrasted cropping systems basedon an assumption of zero pigeonpea yields, and any grain yields obtain-ed from pigeonpea therefore constituted an added contribution tohousehold food needs. Likewise, pigeonpea can yield valuable fuelwoodand animal fodder, which were also excluded from our analysis (Orret al., 2015).

5. Conclusions

Our results indicate that both DLR systems and traditional maize-grain legume rotations can allow farmers in central Malawi to achievehigher maize yields when compared with continuous maize at lowfertilizer rates or a maize pigeonpea intercrop. Rotations also allowfarming households to maintain a greater or equal chance of meetinghousehold calorie and protein needs with their own production in allgrowing seasons. At the same time, rotation systems require less fertil-izer than systems where maize is grown continuously. We show that

Page 10: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

148 A. Smith et al. / Agricultural Systems 145 (2016) 139–149

intercropping pigeonpea with a grain legume in DLR systems has theadditional advantage of increasing total soil C and N over time whencompared with traditional rotations. However, this improvement insoil quality does not lead directly to increased maize yields in DLRsystems due to reduced soil N supply associated with treatments thatinclude pigeonpea.

We also found that DLR systems are only beneficial in appropriateagroecological zones. The benefits of DLR systems were reduced in theagroecological zone with greatest water limitation, due to the poorperformance of pigeonpea in intercrop with a grain legume underthose conditions.

We conclude that DLR systems have the potential to improve soils,maintain food production in the face of a highly variable climate, andreduce dependence on costly fertilizers in maize-based systems. DLRsystems are not a one-size-fits-all solution, however, and may not bebeneficial under inappropriate agroecological conditions. Based on theresults of this modeling study, we advocate farmer-participatory exper-imentation with DLR systems in the field across a wide range of agro-ecological zones. While modeling exercises such as this one canprovide insight into the potential long-term performance and dynamicsof cropping systems, they are no substitute for direct observation ofworking systems. Through practical experimentation, resource-limitedfarmersmay be able to determinewhether DLR systems are appropriatefor their area, and establish these systems locally as part of a strategy tobuild soil fertility while providing for immediate household needs.

Acknowledgments

This work originates from an USAID (Grant AID-OAA-A-13-00006)funded project entitled Africa RISING (Research In Sustainable Intensifi-cation for theNextGeneration),with additional funding from theGlobalCenter for Food Systems Innovation at Michigan State University. Theauthors would like to thank Jimmy Dinesi, Mark Freeman, AmosGanizani, Michele Hockett, Emmanuel Jambo, Isaac Jambo, BlessingsKadzembuka, Dziwani Kambauwa, Kondwani Khonje, EmmanuelMbewe, Miriam Mhango, Elian Mjamanda, Benart Msukwa, EdwardMzumara, Wilson Ndovie, Mary Ollenburger, Jakena Phiri, and all ofthe farmers involved in the Africa RISING project.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.agsy.2016.03.008.

References

Adiku, S., Carberry, P., Rose, C., McCown, R., Braddock, R., 1995. A maize (Zea mays)–cow-pea (Vigna unguiculata) intercrop model. In: Sinoquet, H., Cruz, P. (Eds.), Ecophysiol-ogy of Tropical Intercropping. INRA Editions, pp. 397–406.

APSIM Initiative, 2013. APSoil. http://www.apsim.info/Products/APSoil.aspx.Bray, R., Kurtz, L., 1945. Determination of total, organic, and available forms of phosphorus

in soils. Soil Sci. 59, 39–46.Burk, L., Dalgliesh, N., 2008. Estimating Plant Available Water Capacity—A Methodology.

CSIRO Sustainable Ecosystems, Canberra, Australia.Carberry, P.S., Adiku, S.G.K., McCown, R.L., Keating, B.A., 1996. Application of the APSIM

cropping systems model to intercropping systems. In: Ito, O., Johansen, C., Adu-Gyamfi, J.J., Katayama, K., Kumar Rao, J.V.D.K., Rego, T.J. (Eds.), Roots and Nitrogenin Cropping Systems of the Semi-Arid Tropics. International Crops Research Institutefor the Semi-Arid Tropics, pp. 637–648.

Carberry, P.S., Muchow, R.C., McCown, R.L., 1989. Testing the CERES-maize simulation-model in a semi-arid tropical environment. Field Crop Res. 20, 297–315.

Chikowo, R., Mapfumo, P., Nyamugafata, P., Giller, K.E., 2004. Woody legume fallow pro-ductivity, biological N-2-fixation and residual benefits to two successive maize cropsin Zimbabwe. Plant Soil 262, 303–315.

Chirwa, P.W., Black, C.R., Ong, C.K., Maghembe, J.A., 2003. Tree and crop productivity ingliricidia/maize/pigeonpea cropping systems in southern Malawi. Agrofor. Syst. 59,265–277.

Drost, N., Unpublished results.Food and Agriculture Organization, 2004. Human energy requirements: report of a joint

FAO/WHO/UNU Expert consultation. FAO Food and Nutrition Technical Report Series,Rome, Italy.

Gwenambira, C., 2015. Below and Aboveground Pigeonpea Productivity in On-farm Soleand Intercrop Systems in Central Malawi. Michigan State University, Department ofPlant, Soil and Microbial Sciences (MS Thesis).

Hendershot, W.H., Lalande, H., Duquette, M., 1993. Ion exchange and exchangeable cat-ions. Soil Sampling and Methods of Analysis Vol. 19, pp. 167–176.

Hockett, M.T., 2014. “They Say Wealth is in the Soil”: Local Knowledge and AgriculturalExperimentation among Smallholder Farmers in Central Malawi M.S. thesis MichiganState University.

Kassie, B.T., Rotter, R., Hengsdijk, H., Asseng, S., Van Ittersum, M.K., Kahiluoto, H., VanKeulen, H., 2013. Climate variability and change in the Central Rift Valley ofEthiopia: challenges for rainfed crop production. J. Agric. Sci. 152, 58–74.

Keating, B.A., Carberry, P.S., Hammer, G.L., Probert, M.E., Robertson, M.J., Holzworth, D.,Huth, N.I., Hargreaves, J.N.G., Meinke, H., Hochman, Z., McLean, G., Verburg, K.,Snow, V., Dimes, J.P., Silburn, M., Wang, E., Brown, S., Bristow, K.L., Asseng, S.,Chapman, S., McCown, R.L., Freebairn, D.M., Smith, C.J., 2003. An overview ofAPSIM, a model designed for farming systems simulation. Eur. J. Agron. 18, 267–288.

Kellogg Biological Station LTER, 2003. Carlo Erba NA1500 Series II Combustion Analyzer.Kellogg Biological Station LTER, 2008. KBS058: Particle Size Analysis for Soil Texture De-

termination (Hydrometer Method).Kerven, G.L., Menzies, N.W., Geyer, M.D., 2000. Soil carbon determination by high temper-

ature combustion: a comparison with dichromate oxidation procedures and the in-fluence of charcoal and carbonate carbon on the measured value. Commun. Soil Sci.Plant Anal. 31, 1935–1939.

Lightfoot, C.F., Dear, K.G., Mead, R., 1987. Intercropping sorghum with cowpea in drylandfarming systems in Botswana. 2. Comparative stability of alternative cropping sys-tems. Exp. Agric. 23, 435–442.

Mabapa, P.M., Ogola, J.B.O., Odhiambo, J.J.O., Whitbread, A., Hargreaves, J., 2010. Effect ofphosphorus fertilizer rates on growth and yield of three soybean (Glycine max) culti-vars in Limpopo Province. Afr. J. Agric. Res. 5, 2653–2660.

Malezieux, E., Crozat, Y., Dupraz, C., Laurans, M., Makowski, D., Ozier-Lafontaine, H.,Rapidel, B., de Tourdonnet, S., Valantin-Morison, M., 2009. Mixing plant species incropping systems: concepts, tools and models. A review. Agron. Sustain. Dev. 29,43–62.

Morris, R.A., Garrity, D.P., 1993. Resource capture and utilization in intercropping—water.Field Crop Res. 34, 303–317.

Myaka, F., Sakala, W., Adu-Gyamfi, J., Kamalongo, D., Ngwira, A., Odgaard, R., Nielsen, N.,Hogh-Jensen, H., 2006. Yields and accumulations of N and P in farmer-managed inter-crops of maize-pigeonpea in semi-arid Africa. Plant Soil 285, 207–220.

Ngwira, A., (Personal communication).Ollenburger, M., Snapp, S., 2014. Model applications for sustainable intensification of

maize-based smallholder cropping in a changing world. In: Ahuja, L., Ma, L.,Lascano, R. (Eds.), Practical Applications of Agricultural System Models to Optomizethe Use of Limited Water. American Society of Agronomy, pp. 375–398.

Orr, A., Kambombo, B., Roth, C., Harris, D., Doyle, V., 2015. Adoption of integrated food-energy systems: improved cookstoves and pigeonpea in southern Malawi. Exp.Agric. 51, 191–209.

Robertson, M.J., Carberry, P.S., Chauhan, Y.S., Ranganathan, R., O'Leary, G.J., 2001.Predicting growth and development of pigeonpea: a simulation model. Field CropRes. 71, 195–210.

Robertson, M.J., Sakala, W., Benson, T., Shamudzarira, Z., 2005. Simulating response ofmaize to previous velvet bean (Mucuna pruriens) crop and nitrogen fertilizer inMalawi. Field Crop Res. 91, 91–105.

Rurinda, J., vanWijk, M.T., Mapfumo, P., Descheemaeker, K., Supit, I., Giller, K.E., 2015. Cli-mate change and maize yield in southern Africa: what can farm management do?Glob. Chang. Biol.

Rusinamhodzi, L., Corbeels, M., Nyamangara, J., Giller, K., 2012. Maize-grain legumeintercropping is an attractive option for ecological intensification that reducesclimate risk for smallholder farmers in central Mozambique. Field Crop Res. 136,12–22.

Saha, S., Moorthi, S., Pan, H.L., Wu, X.R., Wang, J.D., Nadiga, S., Tripp, P., Kistler, R., Woollen,J., Behringer, D., Liu, H.X., Stokes, D., Grumbine, R., Gayno, G., Wang, J., Hou, Y.T.,Chuang, H.Y., Juang, H.M.H., Sela, J., Iredell, M., Treadon, R., Kleist, D., Van Delst, P.,Keyser, D., Derber, J., Ek, M., Meng, J., Wei, H.L., Yang, R.Q., Lord, S., Van den Dool,H., Kumar, A., Wang, W.Q., Long, C., Chelliah, M., Xue, Y., Huang, B.Y., Schemm, J.K.,Ebisuzaki, W., Lin, R., Xie, P.P., Chen, M.Y., Zhou, S.T., Higgins, W., Zou, C.Z., Liu, Q.H.,Chen, Y., Han, Y., Cucurull, L., Reynolds, R.W., Rutledge, G., Goldberg, M., 2010. TheNCEP climate forecast system reanalysis. Bull. Am. Meteorol. Soc. 91, 1015–1057.

Saha, S., Moorthi, S., Wu, X., Wang, J., Nadiga, S., Tripp, P., Behringer, D., Hou, Y., Chuang,H., Iredell, M., Ek, M., Meng, J., Yang, R., Mendez, M., van den Dool, H., Zhang, Q.,Wang, W., Chen, M., Becker, E., 2014. The NCEP climate forecast system version 2.J. Clim. 27, 2185–2208.

Saxton, K.E., Willey, P.H., 2006. The SPAWmodel for agricultural field and pond hydrolog-ic simulation. In: Singh, V., Frevert, D. (Eds.), Mathematical Modeling of WatershedHydrology. CRC Press, pp. 401–408.

Snapp, S.S., Bezner, Kerr R., Smith, A., Ollenburger, M., Mhango, W., Shumba, L., Gondwe,T., Kanyama-Phiri, G.Y., 2014. Modeling and participatory, farmer-led approaches tofood security in a changing world: a case study from Malawi. Scheresse 24, 350–358.

Snapp, S.S., Blackie, M.J., Gilbert, R.A., Bezner-Kerr, R., Kanyama-Phiri, G.Y., 2010. Biodiver-sity can support a greener revolution in Africa. Proc. Natl. Acad. Sci. U. S. A. 107,20840–20845.

Snapp, S., Kanyama-Phiri, G., Kamanga, B., Gilbert, R., Wellard, K., 2002. Farmer andresearcher partnerships in Malawi: developing soil fertility technologies for thenear-term and far-term. Exp. Agric. 38, 411–431.

Snapp, S.S., Mafongoya, P.L., Waddington, S., 1998. Organic matter technologies for inte-grated nutrient management in smallholder cropping systems of Southern Africa.Agric. Ecosyst. Environ. 71, 185–200.

Page 11: Doubled-up legume rotations improve soil fertility and maintain … · Doubled-up legume rotations improve soil fertility and maintain productivity under variable conditions in maize-based

149A. Smith et al. / Agricultural Systems 145 (2016) 139–149

Snapp, S.S., Silim, S.N., 2002. Farmer preferences and legume intensification for low nutri-ent environments. Plant Soil. 245, 181–192.

Thierfelder, C., Cheesman, S., Rusinamhodzi, L., 2012. A comparative analysis of conserva-tion agriculture systems: benefits and challenges of rotations and intercropping inZimbabwe. Field Crop Res. 137, 237–250.

Thornton, P.K., Jones, P.G., Ericksen, P.J., Challinor, A.J., 2011. Agriculture and food systemsin sub-Saharan Africa in a 4 degrees C+ world. Philos. Trans. R. Soc. A Math. Phys.Eng. Sci. 369, 117–136.

Trenbath, B.R., 1999. Multispecies cropping systems in India—predictions of their produc-tivity, stability, resilience and ecological sustainability. Agrofor. Syst. 45, 81–107.

Waddington, S.R., Mekuria, M., Siziba, S., Karigwindi, J., 2007. Long-term yield sustainabil-ity and financial returns from grain legume-maize intercrops on a sandy soil in sub-humid north central Zimbabwe. Exp. Agric. 43, 489–503.

World Health Organization, 2007. Protein and amino acid requirements in humannutrition: report of a joint WHO/FAO/UNU expert consultation. WHO Technical Re-port Series. World Health Organization, Geneva, Switzerland.