Competing use of organic resources, village-level ...iks.ukzn.ac.za/sites/default/files/95.pdf ·...

Agricultural Systems 104 (2011) 175–190

Contents lists available at ScienceDirect

Agricultural Systems

journal homepage: www.elsevier .com/locate /agsy

Competing use of organic resources, village-level interactions between farmtypes and climate variability in a communal area of NE Zimbabwe

M.C. Rufino a,*, J. Dury a, P. Tittonell a,b, M.T. van Wijk a, M. Herrero c, S. Zingore d, P. Mapfumo e,f, K.E. Giller a

a Plant Production Systems Group, Wageningen University, P.O. Box 430, 6700 AK, Wageningen, The Netherlandsb Systèmes de Culture Annuels, Centre de Coopération Internationale en Recherche Agronomique pour le Développement (CIRAD), Francec International Livestock Research Institute (ILRI), P.O. Box 30709, Nairobi, Kenyad Tropical Soil Biology and Fertility Institute (TSBF) of the Centro de Agricultura Tropical (CIAT), P.O. Box 158, Lilongwe, Malawie Department of Soil Science and Agricultural Engineering, University of Zimbabwe, P.O. Box MP167, Mount Pleasant, Harare, Zimbabwef Soil Fertility Consortium for Southern Africa (SOFECSA), CIMMYT, Southern Africa, P.O. Box MP163, Mount Pleasant, Harare, Zimbabwe

a r t i c l e i n f o

Article history:Received 8 December 2009Received in revised form 26 April 2010Accepted 1 June 2010Available online 1 July 2010

Keywords:Crop–livestock interactionsDynamic modellingCattleCrop residuesGrasslands

0308-521X/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.agsy.2010.06.001

* Corresponding author. Tel.: +31 317 483045; fax:E-mail address: [email protected] (M.C. Rufin

a b s t r a c t

In communal areas of NE Zimbabwe, feed resources are collectively managed, with herds grazing ongrasslands during the rainy season and mainly on crop residues during the dry season, which createsinteractions between farmers and competition for organic resources. Addition of crop residues or animalmanure is needed to sustain agricultural production on inherently poor soils. Objectives of this studywere to assess the effect of village-level interactions on carbon and nutrient flows, and to explore theirimpact on the long-term productivity of different farm types under climate variability. Crop and cattlemanagement data collected in Murewa Communal area, NE Zimbabwe was used together with a dynamicfarm-scale simulation model (NUANCES-FARMSIM) to simulate village-level interactions. Simulationsshowed that grasslands support most cattle feed intake (c. 75%), and that crop residues produced bynon-cattle farmers sustain about 30% of the dry season feed intake. Removal of crop residues (0.3–0.4 t C ha�1 yr�1) from fields of non-cattle farmers resulted in a long-term decrease in crop yields. No-access to crop residues of non-cattle farmers increased soil C modestly and improved yields in thelong-term, but not enough to meet household energy requirements. Harvest of grain and removal of mostcrop residues by grazing cattle caused a long-term decline in soil C stocks for all farm types. The smallestdecrease (�0.5 t C ha�1) was observed for most fertile fields of cattle farmers, who manure their fields.Cattle farmers needed to access 4–10 ha of grassland to apply 3 t of manure ha�1 yr�1. Rainfall variabilityintensifies crop–livestock interactions increasing competition for biomass to feed livestock (short-termeffect) or to rehabilitate soils (long-term effect). Prolonged dry seasons and low availability of crop res-idues may lead to cattle losses, with negative impact in turn on availability of draught power, affectingarea under cultivation in consecutive seasons until farmers re-stock. Increasing mineral fertiliser use con-currently with keeping crop residues in fertile fields and allocating manure to poor fields appears to be apromising strategy to boost crop and cattle productivity at village level. The likelihood of this scenariobeing implemented depends on availability of fertilisers and decision of farmers to invest in rehabilitatingsoils to obtain benefits in the long-term. Adaptation options cannot be blind to what occurs beyond fieldand farm level, because otherwise recommendations from research and development do not fit the localconditions and farmers tend to ignore them.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Crops and livestock are integrated in the farming system thatpredominates in communal areas of northeast (NE) Zimbabwe(Kunjeku et al., 1998). Main interactions between crops and live-stock are the use of draught power for ploughing, animal manureapplied to crops, and the use of crop residues as feed for livestock

ll rights reserved.

+31 317 482952.o).

(Steinfeld, 1988). Manure is needed to sustain crop production be-cause soils are inherently poor and mineral fertilisers alone areinsufficient to achieve crop yields required to secure householdfood requirements (Rodel and Hopley, 1973; Grant, 1976). Cattleare economically the most important livestock kept by farmers,although only 40% of the households own cattle (Zingore et al.,2007a). Rainfall variability represents one of the largest risks tofarming in NE Zimbabwe, with a high frequency of occurrence ofdroughts (one out of five years) and recurrent dry spells (Matariraet al., 2004).

http://dx.doi.org/10.1016/j.agsy.2010.06.001

mailto:[email protected]

http://dx.doi.org/10.1016/j.agsy.2010.06.001

http://www.sciencedirect.com/science/journal/0308521X

http://www.elsevier.com/locate/agsy

176 M.C. Rufino et al. / Agricultural Systems 104 (2011) 175–190

Within the communal area of Murewa in NE Zimbabwe, each vil-lage has access to well-delimited communal grasslands where cattleare herded during the growing season to avoid crop damage. Duringthe growing season the feed value of grasses is much better than dur-ing the dry season (Frost, 1996), when cattle graze preferentiallymaize and groundnut residues available on croplands. Most cattleowners collect part of their crop residues to feed cattle in the dryseason, when feed shortages are critical (Mtambanengwe andMapfumo, 2005). Collection of crop residues may have negative con-sequences for crop production because of the continuous removal oforganic materials from fields, especially for farmers who have noaccess to other sources of carbon (C). Differential management offields has resulted in heterogeneity in soil fertility within and acrossfarms: farmers who own cattle concentrate manure on fields closestto their homesteads, which led to gradients of soil fertility withfertile homefields, and poor outfields (Zingore et al., 2007a). Thefields from poor farmers without livestock receive small amountsof nutrients and show poor soil fertility irrespective of distance fromhomesteads.

Collective management of cattle at village level, and toleranceof non-cattle farmers to the grazing of their crop residues maycontribute to the concentration of C and nutrients in the fieldsof the cattle owners. While the intensity of such interactions reg-ulates the degree of inequity between farmers (Ramisch, 2005),rainfall variability may have a large effect on the extent of theinteractions: in years of poor rainfall or occurrence of droughtthe competition for plant biomass to meet alternative uses in-crease. There is little quantitative information for southern Africaon the effects of communal management of feed resources on thesize of nutrient and C flows and on the long-term consequencesfor the productivity of croplands. Objectives of this study were toassess the effect of village-level interactions on C and nutrientflows, and to explore their impact on the long-term productivityof different farm types under climate variability. Focus wasplaced on the interactions under current and alternative manage-ment practices, and the comparisons between cattle farmers andnon-cattle farmers. We combined information available for studyarea collected through interviews, field measurements, observa-tions, experiments and literature. We used the NUANCES-FARM-SIM modelling framework (Giller et al., 2006; Van Wijk et al.,2009), which consists of relatively simple crop, cattle, manuremanagement and grassland models, developed and tested forthe conditions of smallholder farming in NE Zimbabwe. The spe-cific research questions were: (i) What is the size and dynamicsof nutrient and C flows mediated by cattle at farm and village le-vel? (ii) How do nutrient and C flows change according to alter-native management practices? (iii) What is the effect of climatevariability on farm- and village-level interactions? (iv) Whendoes competition for organic resources become most critical forcattle and for crop production? (v) What are the options forintensification of communal farming under climate variability?

Table 1Characteristics of farm types and resource groups classified according to the typology for

Farm type

Wealthier Medium-wealthier

Resource group RG1 RG2Proportion in the village (%) 6 35Livestock owned c. 10 cattle <10 cattleResource exchanges Hire labour and share draught

powerDo not sell or hire lshare draught powe

Land holding (ha) >3 2–3Food self-sufficiency Self-sufficient, able to sell grain

and vegetablesSelf-sufficient, ablegrain and vegetable

2. Methodology

2.1. Study area

The study took place in the Murewa smallholder area located80 km E of Harare in Zimbabwe, between 17 and 18�S and 31and 32�E, which belongs to Natural Region II, an agro-ecologicalzone of relatively high potential for agriculture (Vincent andThomas, 1960). Maize is the main staple crop in Murewa, withgroundnuts, sweet potatoes, sunflower and vegetables also grown.Cattle usually graze during the day and are tethered in the kraalclose to homesteads overnight. Crop residues are fed to cattle dur-ing the dry season and manure is used to fertilise maize crops andvegetable gardens.

2.1.1. Climate, soils and natural vegetationMurewa has a sub-tropical climate receiving 750–1000 mm

rainfall annually, distributed in a unimodal pattern (November–April), with an annual coefficient of variation of 30% (Kunjekuet al., 1998). Soils in Murewa are predominantly granitic sandysoils (Lixisols) with low inherent fertility (Nyamapfene, 1991). Asmall proportion of the area has relatively fertile dolerite-derivedclay soils (Luvisols) considered the best agricultural soils inZimbabwe. The natural vegetation at Murewa is Miombo wood-land dominated by Brachystegia spp. and Julbernardia spp. trees.The grass cover in the woodland is dominated by species of thegenus Hyparrhenia, and is therefore termed Hyparrhenia-veld type(Rattray, 1957). Andropogon, Digitaria, and Heteropogon spp. arealso common species especially where tree density is high.Sporobolus pyramidalis dominates where grazing intensity is rela-tively high, and in the wet ‘vlei’ area.

2.1.2. Farmers and farm typologyA common approach when modelling agro-pastoral communi-

ties is to stratify farm households using typologies (Thornton et al.,2007). A simplified village that resembles the Majonjo village ofMurewa was constructed using a farm typology developed byZingore et al. (2007a). The typology distinguishes four farmerresource groups (RG) based on livestock ownership, farm size, pro-duction orientation, labour hired, and food self-sufficiency (Table 1).Feeding strategies, herding patterns, crop residues, and manuremanagement were studied during the dry season of 2006 (June–September) and the rainy season of 2006–2007 (February–May).Cattle owners (RG1 and RG2), non-cattle farmers (RG3 and RG4)and other key informants such as the kraal head and herders wereinterviewed. Biomass production and species composition of thecommunal grassland were measured. Grain yields, amount of cropresidues and their management were estimated through interviewsand field measurements.

the communal area of Murewa. Source: Zingore et al. (2007a).

Medium-poor Poor

RG3 RG426 33No cattle No cattle

abour,r

Sometimes sell labour or exchangeit for draught power

Sell labour and/or exchange it fordraught power

<2 <1to sells

Purchase grain and sell vegetables Purchase food or receive food aid

M.C. Rufino et al. / Agricultural Systems 104 (2011) 175–190 177

2.2. Modelling framework

2.2.1. Farm-scale modelNUANCES-FARMSIM is a farm-scale model, where household

objectives, constraints and resource allocation are considered, link-ing simulation results from four sub-models. The coupling of thesub-models allows to represent short- and long-term feedbacksthat characterise interactions between crops and livestock, andtheir consequences on natural resources of farming systems (Titto-nell et al., 2009). Crop and soil modules are combined at field levelin the model FIELD (Field-scale resource Interactions, use Efficien-cies and Long-term soil fertility Development – Tittonell et al.,2007, 2010). Combinations of crop and soils can be simulated fordifferent field types. LIVSIM (LIVestock SIMulator–Rufino et al.,2009) simulates animal production based on genetic potential, feedavailability and its quality. The dynamics of nutrients throughmanure collection, and storage are simulated by HEAPSIM (Rufinoet al., 2007), estimating mass and nutrient cycling efficiencies ofmanure management. Weather and nutrient inflows constitute in-puts to FARMSIM that are accounted for during the simulations,and modified for scenario analyses. Sub-models incorporate pro-cesses and interactions in a descriptive fashion, and operate withdifferent time steps: monthly for LIVSIM, HEAPSIM, and seasonalfor FIELD.

2.2.2. FIELD, the crop-soil modelFIELD simulates long-term changes in soil fertility (C, N, P and

K), interactions between nutrients that determine crop production,

LIVSIM GrassS

FIELD

HEAPSIM

Farm type

(B)

Fig. 1. (A) Simplified village and model used for this study. The village consists of croplalivestock heads, labour availability. (B) The model NUANCES-FARMSIM takes into accounFIELD simulates crop production and the dynamics of C and nutrients in the soils, LIVSorganic resources management in the kraal and in the compost heap, GrassSIM descrmanagement is described using rules.

and crop responses to mineral and organic fertilisers. Resource-limited total dry matter and grain production are calculated inFIELD on the basis of seasonal resource (light, water and nutrients)availabilities through application of crop specific resource use effi-ciencies for capture and conversion, derived from literature, exper-iments and/or process-based modelling work. Simulation of soilprocesses and calibration and testing for this study are describedin Tittonell et al. (2007).

2.2.3. LIVSIM, the livestock modelLIVSIM simulates cattle production in time according to genetic

potential of the breed and feeding. Potential production is definedby mature weight, growth rate and milk yield. LIVSIM is based onthe concepts of the model of Konandreas and Anderson (1982) butdiffers from that model in that: (i) energy and protein require-ments calculations are based on AFRC (1993), (ii) feed intake is cal-culated with the model of Conrad (1966), (iii) excreta production issimulated, and (iv) the decision rules for herd management (i.e.weaning age, management of reproduction, lactation, and feedinggroups). For this study, LIVSIM was complemented with a grazingroutine that includes diet selection and restrictions to feed intake.The approach used includes functional relationships between in-take and herbage mass, grazing behaviour and animal size. Theinfluence of the spatial distribution of feed on the diet selectionwas treated using the concepts of the hierarchical foraging modelof Senft et al. (1987). The foraging model takes management intoaccount including herders’ preference to graze cattle (Senft,1989). Herders’ preferences are captured by using time spent at

IM

(A)

nd surrounded by grazing land. Resource groups were defined on the basis of land,t the interactions between resource groups due to cattle and manure management.IM simulates animal production and reproduction of the herd, HEAPSIM describesibes grass growth for each grazing unit. The models are linked dynamically and


each different grazing unit, an input to LIVSIM collected throughobservation (see Section 2.3.2). Diet selection is accounted for byusing a preference index based on crude protein and speciesabundance. Potential intake was adjusted with a relative intakecoefficient to calculate actual dry matter intake considering con-straints imposed by herbage availability (Johnson and Parsons,1985; Richardson and Hahn, 1991; Herrero et al., 1998). Feed allo-cation within the herd was calculated using a relative energyrequirement coefficient for each individual. For more details oncalibration and testing of LIVSIM see Rufino (2008).

2.2.4. HEAPSIM, the organic resources management modelWe adapted HEAPSIM to represent manure management for the

study site. Manure excreted during kraaling is usually left to accu-mulate during the dry season, mixed with feed refusals. AroundAugust, manure is heaped in the open air and later applied duringplanting around November. Cattle owners (n = 35) were inter-viewed on their manure management practices. Most farmers(85%) removed all manure from the kraal once a year. Only 20%of the farmers did not compost the manure collected from the kraalbut applied it directly to the fields. Most cattle owners apply man-ure (between 1.8–7.2 t ha�1 yr�1) to their maize fields. About 30%of the farmers removed small amounts of manure from the kraalto apply it to their vegetables garden.

2 12

22 3

3

Fig. 2. Map of the study village: territory is delimited by two rivers and two village bounhigh Miombo woodland vegetation (grazing units 8, 9 and 10). Cropland is located betweto the open grassland and high Miombo woodland. The houses show the location of the

2.2.5. Village level modelDifferent instances of FARMSIM were used to simulate four

farm types or resource groups (RG) at village level (Fig. 1). Themodel GrassSIM (Grass SIMulator) was developed to simulate grassgrowth and describes availability of green and dead grass from dif-ferent grazing units, which were identified through participatorywork with herders and complemented with measurements andobservational work (Fig. 2 and Tables A1 and A2 Appendix). Theherd, simulated by LIVSIM, grazed on the grassland during theday on grazing routes identified through field observations, andwas kept overnight on-farm within a kraal, where manure accumu-lates. Dynamics of manure decomposition before collection andduring composting are followed by HEAPSIM for each RG. Differentinstances of FIELD were used to simulate aboveground biomassand grain production, and soil C in the different field types of eachRG. During the dry season cattle was allowed to graze crop resi-dues on croplands, and manure produced was left in the grazedfield, and incorporated (after C and nutrient losses) into the soilmodule of FIELD.

2.2.6. Grassland model – GrassSIMGrassSIM describes dynamically grass and dead biomass for

landscape units of different soil quality and grazing intensity, asa function of rainfall use efficiency (RUE). This approach has been

Dust roadRivers

Open grassland

CroplandGrazing units

Interviewed householdsHigh Miombo Woodland

Low Miombo Woodland

...

daries. Territory is divided by a hill that extends from N to S, which is covered withen two blocks of grazing land: to the left the low Miombo woodland, and to the right

different households.

Table 2Farm characteristics and input use of the resource groups in the study village for thebaseline scenario, and input rates used for the targeted fertilisation scenario. Fordetails see Section 2.4.

Farm type

RG1 RG2 RG3 RG4

Household size (#) 7 5 6 4Farm size (ha) 3.5 2.2 1.9 0.9Homefield area (ha) 0.8 0.6 0.4 0.4Midfield area (ha) 0.8 0.4 0.6 0Outfield area (ha) 1.7 1.0 0.7 0.4Vegetable garden (ha) 0.2 0.2 0.2 0.1Cattle heads (#) 10 5 0 0

Input use in baseline scenarioFertiliser N (kg N farm�1) 100 45 35 13

Homefields (kg N ha�1) 45 45 50 24Outfields (kg N ha�1) 30 13 5 15

Fertiliser P (kg P farm�1) 17 10 4 2Homefields (kg P ha�1) 10 10 5 6Outfields (kg P ha�1) 4 2 1 0

Manure applied (t farm�1) 3–4 1.5–2 0 0

Input use in targeted fertilisation scenarioFertiliser N (kg N farm�1) 174 102 102 48

Homefields (kg N ha�1) 30 30 60 60Outfields (kg N ha�1) 60 60 60 60

Fertiliser P (kg P farm�1) 87 51 51 24Homefields (kg P ha�1) 15 15 30 30Outfields (kg P ha�1) 30 30 30 30

Manure applied (t farm�1) 3.5–5 1.8–2 0 0


used with success in semi-arid rangelands (Le Houérou, 1984; LeHouérou et al., 1988; Illius and O’Connor, 1999). GrassSIM usesconcepts of the model of Gambiza et al. (2000) for simulating pro-duction of Miombo woodlands in southern Africa. RUE ranged from1.7 to 3 kg dry matter (DM) per mm rain for soils with low (0.8%) tohigh (2.1%) SOC, and grazing intensities from 0 to 1 livestock units(LU) per hectare. RUE were calculated using data for grass strata ofMiombo woodlands (Barnes, 1956; Baars, 1996; Frost, 1996), andwere in the range of 2.7 ± 0.60 kg dry matter per mm rain esti-mated by Illius and O’Connor (1999). Data collected in the grass-lands of Manjonjo were used to calibrate grass at peak biomass,senescence and decay for each grazing unit (Appendix, Tables A1and A2). Feed qualities of the grass species used in the simulationswere extracted from Boudet (1991) and Topps and Oliver (1993)(Appendix, Table A3).

2.3. Simplified farm types, model inputs, parameters and assumptions

2.3.1. The village and the resource groupsThe simplified village consisted of 66 households, from which

four farmers (6%) belonged to RG1, 23 (35%) to RG2, 17 (26%) toRG3 and 22 (33%) to RG4 (Table 2). The land consisted of 116 haof cropland and 426 ha of communal Miombo woodland (Fig. 2),of which a large unit of about 160 ha was hardly used for grazingbecause of the difficulties for herding. Herbage biomass duringthe rainy season appeared not to be limiting cattle feed intake. Cal-culated average stocking rates were 0.3 and 0.5 LU ha�1 for rainyand dry season. It was assumed that the area under cropping re-mains constant, and the proportion of RG in the village does notchange: non-cattle farmers did not evolve into cattle farmers with-in the simulation, although the opposite could happen if cattle die.In reality, farm households are not static and poor households maygain resources and vice versa, but as the study examines villageinteractions and what is feasible within the boundaries of the re-sources available, such an assumption is justifiable.

2.3.2. Herd dynamics and herding patternsCattle owners shared the responsibility of herding during the

growing season. At the beginning of the simulations the herd con-sisted of 155 heads, 58% from the local Mashona breed and 42%Africander, 26% of the cattle belonged to RG1 and the rest toRG2. Initial composition of the herd was similar to that observedin Manjonjo, with 30% cows (calved at least once), 17% heifers,14% steers (males no more than 3 yr old), 25% adult males (includ-ing oxen and few bulls), and 15% calves (younger than a year old).Herding patterns, grazing itineraries and grazing units describedby herders did not change during the 10-yr simulation. Mortalityrates were set to those observed in the same area by Frenchet al. (2001). Offtake rates of live animals were assumed to be 3%per year (Hargreaves et al., 2004), cattle were removed from allclasses, and it was assumed there was no recruitment by purchaseinto the herd. These estimates of herd population dynamics are inagreement with observations by Steinfeld (1988) and Chinembiri(1999) in communal areas of Zimbabwe.

2.4. Scenarios

Over a series of 62 yr (1940–2002) of rainfall measured atMurewa, 25 yr were below the average of 800 mm, and 7 yr werebelow 600 mm (Fig. 3). We selected three consecutive series of10 yr for the simulations, 1945–1955, 1955–1965 and 1965–1975, each of them characterised by a different variability: the ser-ies 1955–1965 was on average wetter and less variable (mean900 mm, CV = 25%) than the series 1945–1955 (mean 860 mm,CV = 30%) used for the baseline, whereas the series from 1965 to1974 was relatively drier (mean 780 mm, CV = 35%) than the

baseline. Because rainfall variability and the probability of seasonaldrought (ca. 20%) are relatively high for Murewa, we compared allscenarios using the three rainfall series.

2.4.1. BaselineThis scenario represented current management described in

Section 2.3. Farmers from different RG removed crop residues fromtheir homefields, and allowed cattle to graze standing crop resi-dues in outfields. RG1 and RG2 removed 20% of the crop residuesto use it as bedding for cattle, whereas RG3 and RG4 removed only10% to mulch vegetable gardens. Fertilisers were applied at higherrates on homefields than on outfields. RG1 used more fertiliserthan others farmers (Table 2). RG1 and RG2 applied manure onlyto homefields and vegetable gardens. Manure management was as-sumed to be the same for RG1 and RG2. Manure accumulated inthe kraal, mixed with feed refusals and was removed once a yearduring the dry season, composted for a period of 3 months, andthen applied to fields.

2.4.2. No-access for cattle to crop residue of non-cattle farmersIn this scenario, non-cattle farmers (RG3 and RG4) incorporated

crop residues into the soils. Winter ploughing is a tillage practiceby which crop residues are ploughed into the soil after harvestaround May when the soils are still moist. Cattle would have lessfeed available during the dry season, which may negatively impactcattle productivity and manure production. The effect of this prac-tice on crop yields was evaluated for non-cattle farmers (RG3 andRG4) and cattle owners (RG1 and RG2) and animal productivity(herd dynamics and bodyweight changes).

2.4.3. Targeted fertilisationThis scenario evaluated the effect of increasing fertiliser use for

all RG following the recommendation of the Abuja declaration ofAfrican heads of state under NEPAD (www.africafertilizersum-mit.org). Available manure was distributed only in mid- and out-fields of RG1 and RG2. Manure management followed thesuggestions of Mtambanengwe and Mapfumo (2005) that organic

http://www.africafertilizersummit.org

http://www.africafertilizersummit.org

Fig. 3. (A) Annual average rainfall for 1942 to 2002, average 885 mm, CV = 27%, (B) Monthly rainfall for three series of 10 yr: (i) series from 1945–1955 represents an ‘average’series, (ii) series from 1955–1965 represent a ‘wet series’, and (iii) series from 1965–1975 a ‘dry series’. Each series starts in November and ends in October.


nutrient resources can be used most efficiently by reducingamounts applied in homefields, so that they can be spread equita-bly throughout farms to start rehabilitating degraded soils. Model-based explorations by Tittonell et al. (2008) suggest that fertilehomefields can be managed with maintenance fertilisation andconserving crop residues, while mid- and outfields need relativelyhigh fertiliser rates and C to stimulate biomass production. Fertil-iser rates of 60 kg N ha�1 and 30 kg P ha�1 used in the simulationswere the most efficient rates derived from the experimental workof Zingore et al. (2007b) in the study area (Table 2). Homefields ofRG1 and RG2 received 30 kg N ha�1 and 10 kg P ha�1 to simulatemaintenance fertilisation. In relatively poor homefields of RG3and RG4, crop residues were incorporated into the soil and fertilis-ers were added to all fields. Crop residues from mid- and outfieldsof all RG were assumed to be grazed.

2.5. Model simulations

To deal with the stochastic elements included in LIVSIM, i.e.conception and mortality, 100 replicate runs were used for eachscenario. Outcomes of 100 vs. 200 replicates were compared andfound not to differ. Length of the simulations was set to 10 yr tocapture long-term effects of the management scenarios on soil Cstocks.

3. Results and discussion

3.1. Size and dynamics of the nutrient flows at farm and village level

Simulations using baseline scenario and average rainfall showeda seasonal pattern in feed intake and excreta production for the vil-lage herd (Fig. 4): Grassland contributed 75% to annual feed intake,crop residues filled a critical feed shortage during the dry season(Fig. 4A). The sharp switch from feeding grass to feeding crop resi-dues was due to the harvesting of crops in the simulations, as farmerscease herding once crops were harvested and the headmen has de-clared that cattle can be released. Once cattle are free they grazeexclusively in cropland until crop residues are largely used up asthese are of better quality than grass. Cattle of RG1 consumed, on

average, twice as much grass and crop residues per farm as cattleof RG2 (Fig. 4B and C). Cattle of RG1, who had about 10 heads each,consumed about 26% of all feed consumed annually by the villageherd. Although RG2 owned an average of five cattle each, they collec-tively owned the largest part of the village herd. Nitrogen intake bycattle followed the seasonality in feed crude protein, with a peakduring the rainy season around January when herbage intake andquality were highest, and a peak during the dry season, around Juneafter harvest of grain (Fig. 4D). Depressed intake due to low quality ofgrass has been reported for Hyparrhenia-veld grasslands, and sup-plementation with rich protein sources has been advised since the1960s (Smith, 1961, 1962; Clatworthy et al., 1986), but adoption offodder legumes or the use of other supplements has been largelyunsuccessful in communal areas of Zimbabwe. Reasons for lack ofadoption of fodder legumes are several: (i) there have been difficul-ties to identify multipurpose legume trees adapted to long and cooldry seasons, phosphorus deficiencies and aluminium toxicity (Mat-imati et al., 2009), (ii) forage legumes often contain anti-nutritionalcompounds which reduce their acceptability by cattle (Dzowelaet al., 1997), (iii) forage has to be conserved as hay at the end ofthe rainy season (March–May) when there may not be labour avail-able, and there might be competition for land (Chakeredza et al.,2007) and (iv) the need for fencing to protect the forages when live-stock graze freely. Legume fodders or green manures often growpoorly on the depleted sandy soils that characterise outfields of com-munal areas and P fertilisers or liming may be required for them togrow well (Chikowo et al., 2004, 2006), inputs that farmers oftencannot afford on their maize fields.

Manure production followed the pattern of feed intake (Fig. 4E).Deposition of manure in cropland, grassland and accumulation inkraals were determined by feeding strategies and manure manage-ment. A small proportion of the faecal dry matter was left in crop-land during grazing of crop residues. Because cattle spent morethan half of the time in the kraal (12–14 h per day according toour observations), manure available for recycling on-farm was lar-ger than manure left in grassland and cropland. The amount ofrecyclable manure depended on the number of cattle, and there-fore each RG1 farmer may recycle about twice as much manureon their farm as each of the RG2 farmers (Fig. 4F). Excreta N leftduring kraaling followed the seasonal pattern of N intake

05

10152025

0

Total excreted Kraal Cropland Grassland

05

10152025

0

012345

0

00.20.40.60.8

1

0

Resource group 1 Resource group 2

0

20

40

60

0 96

01234

0

01234

0

0

20

40

60

0

12 24 36 48 60 72 84 96 108 120

12 24 36 48 60 72 84 96 108 120

12 24 36 48 60 72 84 96 108 120

12 24 36 48 60 72 84 96 108 120

12 24 36 48 60 72 84 108 120

12 24 36 48 60 72 84 96 108 120

12 24 36 48 60 72 84 96 108 120

12 24 36 48 60 72 84 96 108 120

Resource group 1 Resource group 2

Faec

al d

ry m

atte

r(t

mon

th-1

villa

ge-1

)Fa

ecal

dry

mat

ter

(t m

onth

-1fa

rm-1

)Ex

cret

a N

(kg

mon

th-1

farm

-1)

Cum

ulat

ive

kraa

l man

ure

(t

mon

th-1

farm

-1)

Dry

mat

ter i

ntak

e(t

mon

th-1

villa

ge-1

)D

ry m

atte

r int

ake

(t m

onth

-1fa

rm-1

)D

ry m

atte

r int

ake

(kg

mon

th-1

farm

-1)

N in

take

(k

g m

onth

-1fa

rm-1

)

Village

Resource group 1

Resource group 2

(A)

(B)

(C)

(D)

(E)

(F)

(G)

(H)

Time (months) Time (months)

Grass Crop residues

Fig. 4. Baseline scenario using the average rainfall series (A) simulated feed intake of dry matter for the village herd, (B) dry matter intake for the herd of a farmer fromResource Group 1 (RG1), (C) dry matter intake and (D) N intake for the herd of a farmer from Resource Group 2 (RG2), (E) simulated faecal dry matter for the village herd, (F)faecal dry matter for the herd of a farmer from RG1 and for herd of a farmer from RG2, (G) excreta N for herd of a farmer from RG1 and RG2 and (D) accumulated faecal drymatter and crop residues in the kraal of RG1 and RG2.


(Fig. 4G). Due to poor manure management, from the 80–120 kg N yr�1 left in the kraal by cattle of RG1 and 40–60 kg N yr�1

for RG2, only 15–32 and 8–18 kg N yr�1 were available to be ap-plied to crops for RG1 and RG2, with an efficiency of 20–30% be-tween N excreted and N available to be applied to the fields. Onaverage, 7.3 and 3.8 t dry matter yr�1 of manure accumulated inthe kraal before losses, and after storage this resulted in 2.8 and1.5 t dry matter yr�1 for RG1 and RG2. Efficiency of N retentionthrough collection and storage was about 35–38%. Overall lossesof C and N during storage were about 20–25% because manureused for composting was already exposed to large losses duringthe accumulation period in the kraal (Fig. 4H). At village level, be-fore the cropping season about 24% of the cropland was actively

Table 3Main soil characteristics for the field types of the resource groups of the study villageP = extractable P. Source: Zingore et al., 2007b.

Farmtype

Field type Area(ha)

Clay + silt(%)

Sand(%)

Bulk density(kg dm�3)

RG1 Homefield 0.8 12 88 1.42Midfield 0.8 14 86 1.45Outfield 1.7 15 85 1.51Garden 0.2 59 41 1.28



RG4 Homefield 0.4 12 88 1.56Outfield 0.4 14 86 1.49Garden 0.1 53 47 1.26

manured. This area is much larger than the 4–10% estimated byPowell et al. (2004) and Schlecht et al. (2004) for West African sav-annas for cropland under corralling contracts.

Large differences in farm-level grain production were observedbetween cattle owners of RG1 and RG2 (3–5 t grain yr�1 farm�1)and non-cattle farmers of RG3 and RG4 (less than 1 t grain yr�1

farm�1). Smaller yields for RG3 and RG4 can be explained withsmall size of cropped land, poor soil quality and low input use(cf. Tables 2 and 3). Maize grain yields followed the pattern of rain-fall variability, with averages of 3.9, 1.2 and 0.6 t ha�1 in home-fields, midfields and outfields of RG1, and 4.0, 0.6, 0.3 t ha�1 inhomefields and mid- and outfields of RG2. For RG3 and RG4 farm-ers grain yields were poor (0.5–1 t ha�1 in homefields and 0.1–0.3

. SOC = soil organic carbon, TSN = total soil N, CEC = cation exchange capacity, Ext.

SOC(g kg�1)

TSN(g kg�1)

CEC(cmolc kg�1)

Ext. P(mg kg�1)

pH (1:2.5water)

5.6 0.60 4.5 8 5.24.8 0.43 3.0 6 4.84.1 0.41 1.5 4 4.714 1.2 27 23 5.8

6 0.62 3 9 5.43 0.53 4 5 4.52.2 0.22 2 4 4.216 1.8 33 17 5.9

4 0.45 3 4 5.03.7 0.34 2 5 3.83.3 0.31 2 3 4.113 0.9 24 32 5.5

3.8 0.36 2 5 4.73 0.29 3 3 3.915 1.7 32 31 6.2

Tabl

e4

Sim

ulat

edca

ttle

and

crop

prod

ucti

on,m

anur

eex

cret

ed,c

olle

cted

and

appl

ied,

and

crop

resi

dues

inco

rpor

ated

for

each

reso

urce

grou

p(R

G).

Ave

rage

san

dst

anda

rdde

viat

ions

ofth

ela

st5

yrof

the

sim

ulat

ions

for

the

base

line

scen

ario

usin

gth

eav

erag

era

infa

llse

ries

.

Her

dsi

zeLi

vew

eigh

tG

rass

inta

keC

rop

resi

dues

inta

keM

ilk

prod

uce

dM

anur

een

ters

the

kraa

lC

rop

resi

dues

tokr

aal

Tota

lm

anu

reap

plie

dM

aize

stov

erpr

odu

ced

Cro

pre

sidu

esto

soil

sC

stoc

kG

rain

prod

uce

d(#

farm�

1)

(kg

farm�

1)

(tfa

rm�

1yr�

1)

(tfa

rm�

1yr�

1)

(kg

farm�

1yr�

1)

(tfa

rm�

1yr�

1)

(tfa

rm�

1yr�

1)

(tfa

rm�

1yr�

1)

(tfa

rm�

1yr�

1)

(tfa

rm�

1yr�

1)

(tfa

rm�

1)

(tfa

rm�

1yr�

1)

Base

line

RG

111

.9±

0.3

3316

±22

521

.1±

2.7

7.4

±0.

515

26±

220

7.0

±0.

50.

8±

0.2

2.9

(3.9

)a6.

7(2

.0)b

<0.5

38.1

(11.

5)b

5.1

(1.5

)b

RG

26.

1±

0.2

1660

±13

910

.7±

1.6

3.7

±0.

383

0±

119

3.5

±0.

30.

6±

0.1

1.6

(2.0

)3.

3(1

.5)

<0.5

18.8

(9.4

)2.

9(1

.4)

RG

30

00

00

00

0(0

.2)

1.3

(0.7

)<0

.314

.6(8

.6)

0.6

(0.4

)R

G4

00

00

00

00

(0.1

)0.

5(0

.6)

<0.3

6.5

(8.2

)0.

2(0

.3)

Non

-cat

tle

farm

ers

cons

erve

thei

rcr

opre

sidu

esR

G1

10.6

±0.

328

59±

159

19.2

±2.

05.

5±

0.5

1336

±17

46.

0±

0.4

0.8

±0.

12.

6(3

.7)

6.7

(2.0

)<0

.338

.0(1

1.5)

5.1

(1.5

)R

G2

5.4

±0.

114

14±

879.

6±

1.1

2.7

±0.

372

2±

952.

9±

0.2

0.6

±0.

11.

4(1

.9)

3.3

(1.5

)<0

.318

.7(9

.4)

2.8

(1.3

)R

G3

00

00

00

00

(0)

1.8

(1.1

)1.

7(1

.0)

18.5

(10.

9)1.

0(0

.6)

RG

40

00

00

00

0(0

)0.

8(1

.0)

0.7

(0.9

)8.

3(1

0.4)

0.4

(0.5

)

Targ

eted

fert

ilisa

tion

RG

113

.6±

0.4

3939

±32

523

.3±

3.6

9.8

±0.

518

53±

258

8.1

±0.

70.

7±

0.1

3.4

(4.5

)13

.0(3

.9)

3.4

(1.0

)49

.0(1

4.9)

12.6

(3.6

)R

G2

7.0

±0.

419

90±

183

11.9

±2.

05.

0±

0.3

958

±12

44.

0±

0.6

0.4

±0.

11.

8(2

.0)

5.4

(2.5

)2.

6(1

.2)

25.3

(12.

7)4.

4(2

.2)

RG

30

00

00

00

0(0

.4)

4.4

(2.6

)1.

6(0

.9)

20.0

(11.

8)3.

9(2

.3)

RG

40

00

00

00

0(0

.2)

2.1

(2.6

)1.

0(1

.2)

9.5

(11.

9)1.

8(2

.2)

aTo

tal

man

ure

excr

eted

duri

ng

graz

ing

plu

sad

diti

onof

com

post

.b

Bet

wee

npa

ren

thes

esex

pres

sed

ona

per

hec

tare

basi

s.


in outfields t ha�1) and showed little variation from year to year.High yield on homefields of RG1 and RG2 were mainly due to an-nual additions of 2–4 t manure ha�1 and fertiliser applications of40–50 kg N ha�1, and 7–10 kg P ha�1, rates far below the blanketrecommendations of 120 kg N ha�1, and 30 kg P ha�1 for the region(Chuma et al., 2000). Grant (1976) recommended application of10 t manure ha�1 yr�1 to maintain productivity of sandy soils ofcommunal areas of Zimbabwe. Rodel and Hopley (1973) calculatedthat six cattle were needed to provide the 10 t of manure needed tofertilise an hectare of cropland in experiments over 10 yr. Accord-ing to our simulations, the whole village herd with an average sizeof 187 cattle transferred annually 100 t faecal dry matter fromgrasslands to cropland. With minimum losses, that amount willnot suffice for 10% of the 116 ha of cropland, if it were applied atrecommended rates. Due to harvest of grain and removal of mostcrop residues by grazing cattle, soil C stocks of all farm types de-clined during the simulations. The smallest decreases(�0.5 t C ha�1 in 10 yr) were observed in homefields of RG1 andRG2 which received 1–2 t C ha�1 yr�1 from manure. Homefieldsof RG3 and RG4 and outfields of all farm types, showed changesof �1.5 to �3 t C ha�1 over the 10-yr simulation. Manure deposi-tion during grazing of the cropland was small (less than0.2 t manure ha�1) and did not compensate for removal of C in cropresidues (0.3–0.6 t ha�1).

3.2. Effect of different management scenarios

3.2.1. Effects on cattle productivityThe baseline was compared with two alternative management

scenarios. In the no-access to crop residues scenario (cattle couldnot graze the residues of RG3 and RG4 farms), growth of the herdwas restricted and weight losses during late dry season were pro-nounced, resulting later in lower calving rates than in the baselinewith a reduction in 20% in herd size. Crop residues produced byRG3 and RG4 represented 30% of all crop residues in the village,and were mostly removed from fields to mulch vegetable gardensor consumed in the field by cattle. Overall, the intake of crop resi-dues was reduced by about 25% when cattle could not graze infields of RG3 and RG4.

The targeted fertilisation scenario increased availability of cropresidues during the dry season and increased herd productivity(Table 4). Although in this scenario cattle did not graze homefieldsof any farm in the village, larger production of biomass of themid- and outfields than in the baseline allowed the herd to grow.Mid- and outfields occupied 63% of the cropland, and producedon average 32% of the 135 t yr�1 of crop residues of the village inthe baseline, and 55% of the 283 t yr�1 in the targeted fertilisationscenario, which supported the increase in intake of crop residueduring the dry season.

3.2.2. Effects on crop productivityNo-access to crop residues of non-cattle farms had a relatively

small negative effect (14–17%) on the amount of manure accumu-lated in kraals and available for recycling. Grain yields of RG1 andRG2 were not much affected (<5%) when crop residues on RG3 andRG4 farms were not available for grazing (Table 4). Mtamba-nengwe and Mapfumo (2005) observed in a village in Murewa thatthe amounts of standing maize residues in the fields were less than0.5 t ha�1 at the end of the dry season (30–50% of that standing atthe beginning of the dry season), and did not different betweenfarms of different wealth class. The reduction in biomass wasattributed to free-ranging cattle. An annual addition of 0.7–1.7 t ha�1 of crop residues to the poor soils of RG3 and RG4, re-sulted in our scenario in an increase in grain yields, and in soil Cstocks in the long-term (Fig. 5A and B). However, an increase of40–50% in production of grain, would not suffice to cover the

Gra

in y

ield

s (t

ha-1

)

Soil C stock (t ha-1)

RG1 RG2 RG3 RG4

Baseline

No-access to crop residues

Targeted fertilisation

(A)

(B)

(C)

0

1

2

3

4

5

0

1

2

3

4

5

0

1

2

3

4

5

5 10 15 20 25

Fig. 5. Simulated grain yields plotted against simulated soil C stock at each fieldtype for the cattle farmers of Resource Groups 1 and 2 (RG1 and RG2), and for thenon-cattle farmers (RG3 and RG4), for three different scenarios: (A) baseline, (B)cattle have no-access to the crop residues of RG3 and RG4, and (C) targetedfertilisation scenario, where all crop residues of the homefields are incorporatedinto the soil, manure is applied to the mid- and outfields and fertiliser use in all ofthe fields is increased (see Section 2.4 for details on scenarios).


household grain needs. A family of six people needs about 1.5–1.7 tof maize grain yr�1 to meet their energy requirements. RG3 andRG4 produced only 1.0 and 0.4 t maize grain yr�1 under the sce-nario of no-access to crop residues.

In the targeted fertilisation scenario, continuous incorporationof crop residues to homefields (between 3.7–4.0 t ha�1 yr�1), appli-cation of small doses of manure (0.7–0.8 t ha�1 yr�1) to mid- andoutfields, and increased fertilisation rates of N and P (30 and 15to homefields and 60 and 30 kg N and P ha�1 yr�1 to mid- andoutfields) increased the grain production on RG1 and RG2 farmsthreefold, and their soil C stocks by about 7–11 t per farm at theend of the simulation (Fig. 5C). Under this scenario, 41% of thecropland was manured, receiving a cumulative application of man-ure of 11–16 t ha�1 in 10 yr. This is a simplification of how themanagement scenario may be implemented. Farmers may opt toapply larger amounts of manure in a scheme designed to rehabili-tate soils. The increased soil C stocks of homefields of RG3 and RG4farms, brought about by the incorporation of crop residues(3 t ha yr�1) and application of 60 and 30 kg N and P ha�1 yr�1 toall their fields, increased farm grain production 6- to 7-fold. Thetargeted fertilisation in combination with management of organicresources appears as a promising strategy for improving crop pro-duction at farm and village level. However, in practice, there maybe practical considerations which discourage the rehabilitation ofpoor fields: (i) farmers may be reluctant to invest in outfieldswhich are often poorly supervised due to the distance from thehomestead, (ii) land preparation and planting in fields with resi-dues is more laborious than in fields with bare soil, (iii) incorpora-tion of crop residues may be limited by availability of oxen topractice winter ploughing during the early dry season, and thewillingness of cattle owners to share oxen, and (iv) farmers whomay incorporate crop residues into the soil are also those who needthe residues to feed livestock during the dry season.

We are aware of the limitations of simulations of long-term re-sponses to continuous addition of organic residues on poor sandysoils. FIELD takes into account availability and uptake of N, P andK for crop responses to nutrient applications, and therefore annualinputs of C to the soil may be overestimated when other nutrients(or factors) are limiting crop production in the field (Tittonell et al.,2007). For example, strong deficiencies of Ca and Zn have been re-ported for maize growing in depleted sandy soils in long- termexperiments in Murewa (Zingore et al., 2008). The effectivenessof the targeted fertility management remains to be tested throughparticipatory experimentation with farmers, an approach that isbeing implemented by the Soil Fertility Consortium for SouthernAfrica (SOFECSA) (www.sofecsa.org).

3.3. Effect of rainfall variability on management scenarios

3.3.1. Effect on herd dynamics and cattle productivityRainfall variability had a large effect on feed availability and

thereafter on herd dynamics (Fig. 6). Negative effect of no-accessto graze crop residues was particularly large for the cattle popula-tion and bodyweight for prolonged dry seasons. With average rain-fall (Fig. 6A and D), cattle population stabilised in month 60 for theno-access scenario, while it kept on growing for the targeted fertil-isation scenario. With the dry rainfall series (Fig. 6C and F), cattlepopulation started to decline in month 48 for the no-access sce-nario, while it still grew for the targeted fertilisation scenario. Inthe dry season cattle grazed crop residues and grass on the vlei –the lowest landscape position of the grassland, where intake wasmainly restricted due to the poor grass quality. When availabilityof crop residues for cattle was reduced in the no-access to crop res-idues scenario, cattle lost weight rapidly, and died due to starva-tion when the onset of the rainy season was delayed. Starvationmight have been overestimated because FARMSIM keeps track of

feeds available only within the area of exploitation described byfarmers for normal circumstances. Feed resources utilisation de-scribed in the study and used for the model explorations agreeswith previous studies in Zimbabwe (Steinfeld, 1988). However,adaptive strategies to minimise death of cattle such as movingherds to different areas where forage may be available (Scoones,1996), or destocking when the season becomes critically dry(Sandford and Scoones, 2006), were not considered.

3.3.2. Effects on the intensity of interactionsRainfall variability effect on feed availability generated feed-

backs into the crop–livestock system by affecting herd dynamics –through the intake of grass and crop residues – and finally theamount of C and nutrients transferred from grassland to cropland.In the simulations, poor rainfall during the growing season re-sulted in poor recovery of bodyweight of the herd, and conse-quently cattle entered the dry season in a poor condition that,added to low availability of crop residues, risked their survival(Fig. 6A and C). Poor feed availability during the dry season notonly risks the survival of cattle but also affect the reproductioncapacity of females in the long-term (Mukasa-Mugerwa, 1989).Cattle population was smallest for the no-access to crop residues

http://www.sofecsa.org


scenario, specially in the dry rainfall series, i.e. 12% smaller thanthe population of the baseline, and 27% smaller than the popula-tion of the targeted fertilisation scenario (Table 6). Cattle popula-tion, bodyweight, milk and manure, where largest for the wetrainfall series, i.e. 9–15% larger than the average rainfall seriesfor all scenarios, and 18–24% larger than the dry rainfall seriesfor all scenarios (Table 5). In the baseline scenario, cattle consumedmore crop residues (3.7 and 7.4 t farm�1 yr�1 for RG2 and RG1)than that produced on the farms of their owners (3.3 and6.7 t farm�1 yr�1 for RG2 and RG1); this deficit was larger in wetrainfall series (cf. Tables 4–6). In the targeted fertilisation scenario,RG1 and RG2 produced enough crop residues to feed their cattle,but free roaming cattle also grazed on outfields of RG3 and RG4,thus leaving only small amounts to be incorporated into the soilscomparable to that of the no-access to crop residues scenario. Goodrainfall was not necessarily associated with larger amounts of cropresidues being incorporated to the soils, as good rainfall resulted ina larger and heavier herd, more demanding in feed than in the dryrainfall series (Tables 5 and 6).

For all scenarios, the transfer of C as manure from the villagegrasslands to the cropland was about 10% larger for the wet(105–115 t manure yr�1) than for the dry rainfall series (80–102 t manure yr�1). In the baseline, the transfer of mass and nutri-ents from fields of non-cattle farmers to the kraals of cattle farmerswas also larger for the wet rainfall series than for the other scenar-ios. However, there are no large differences in soil C between rain-fall series (Fig. 7A) because biomass production, and manureapplications partly compensate for removals of crop residues. Con-serving residues and applying manure had a slightly larger effectfor the wet rainfall series (Fig. 7B), than for baseline and dry rain-fall series because the differences in biomass added to soils wererelatively small (only 0.2 t ha�1 yr�1). The targeted fertilisationscenario led to larger differences in soil C for the wet rainfall seriesthan for the others, particularly for RG1 and RG2 (Fig. 7C). Rainfallvariability has a relatively small effect on soil processes which haverelatively slow rates, but the effect on crop yields and on humanfood security can be large. In the baseline, when the whole village

Her

d si

ze (1

03kg

)N

umbe

r of c

attle

Time0 12 24 36 48

100

150

200

250

300

0

25

50

75

100

0 12 24 36 48 60 72 84 96 108 120

average series wet s(A)

(D)

*sem

Baseline No Access

Fig. 6. Simulated cattle population and aggregated bodyweight of the whole herd of thresidues of non-cattle farmers, and the targeted fertilisation scenario), and with three raindry series.

applied 2.3 and 0.4 t yr�1 of N and P mineral fertilisers, grain pro-duction fluctuated around 100 t of grain (Fig. 8A), which wouldfeed 330 people with an average requirement of 300 kg maizecapita yr�1. Grain required for food self-sufficiency, with a popula-tion annual growth rate of 3%, would increase from 99 to 129 t yr�1

over 10 yr. In the simulations, a dry year caused grain productionto halve (Fig. 8C), which may leave most people food insecure;especially where markets function poorly and trade cannot com-pensate food deficits (Tschirley and Jayne, 2010). Non-cattle farm-ers of RG3 and RG4, who represented 60% of the village population,produced in the baseline only about 15% of total grain, and slightlyless for the dry rainfall series (Fig. 8D, E and F). For the no-access tocrop residues scenario, the overall grain production of the villageincreased by about 10%, and the fraction of the production of theRG3 and RG4 rose to about 25%, which would not be enough forfood self-sufficiency (cf. Tables 4–6). Under the targeted fertilisa-tion scenario where production of grain more than doubled, thefraction of total grain produced by RG3 and RG4 increased to abouthalf of the total, and would have reached food self-sufficiency mostof the years. The targeted fertilisation scenario would implyincreasing use of mineral fertiliser 2.5-fold for N (from 2.3 to5.8 t yr�1) and 7-fold for P (from 0.4 to 2.9 t yr�1). At village level,this would mean putting into practice the aspiration of the Africangreen revolution of using about 50 kg of fertiliser per ha of crop-land and manuring about 40% of the land.

Large rainfall variability may not only have a direct, short-termeffect on food self-sufficiency as shown in the village-level produc-tion of grain but also a long-term effect on livelihoods risking live-stock assets. Therefore, in environments where farmers largelydepend on their harvests for their survival, interventions to in-crease community resilience must include interactions betweenfarmers, which often include the management of individual farmsand of the communal resources. Similar patterns of communal andindividual resource management have been described for the WestAfrican savannas (e.g. Powell et al., 1996; Manlay et al., 2004;Schlecht et al., 2004), where farmers manage heterogeneous soilfertility and rainfall variability through crop–livestock integration

(months)0 12 24 36 48 60 72 84 96 108 12060 72 84 96 108 120

eries dry series(B) (C)

(F)(E)

crop residues Targeted fertilisation

e village under three different management scenarios (baseline, no-access to cropfall series (A) and (D) average rainfall series, (B) and (E) wet series, and (C) and (F) a

0

2

4

6

8

10

0 60 1200

2

4

6

8

10

0 60 120

-6

-4

-2

0

2

4

0 60 120

-6

-4

-2

0

2

4

0 60 120

-6

-4

-2

0

2

4

0 60 120

-6

-4

-2

0

2

4

0 60 120

-6

-4

-2

0

2

4

0 60 120

-6

-4

-2

0

2

4

0 60 120

0

2

4

6

8

10

0 60 120Time (months)

Sim

ulat

ed S

oil C

cha

nge

(t fa

rm-1

)

RG1 RG2 RG3 RG4

average series wet series dry series

(A)Baseline

(B)No access to crop residues

(C)Targeted fertilisation

Fig. 7. Simulated changes in soil organic C (with respect to the year 0) after 10 yr of cultivation under different management scenarios: (A) baseline, (B) no-access to cropresidues of the non-cattle farmers (RG3 and RG4), and (C) targeted fertilisation where all crop residues of the homefields are incorporated into the soils, manure is applied tothe mid- and outfields and the fertiliser use in all the plots is increased (see Section 2.4 for details on scenarios).


at the scale of territories. Rainfall variability thus results in fluctu-ation in cattle populations and therefore in availability of manure,

Time

Average rainfall(A) (B)

(D) (E)

Gra

in p

rodu

ctio

n (t

villa

ge-1

)Fr

actio

n of

gra

in p

rodu

ced

by

non-

cattl

e fa

rmer

s

0

0.1

0.2

0.3

0.4

0.5

1 2 3 4 5 6 7 8 9 10 1 2 3 4

Average rainfall0

100

200

300

TaNo access CRBaseline

Fig. 8. Simulated grain production for the whole village under three management scenaritargeted fertilisation), and using three different rainfall series: (A) average series, (B) afarmers (RG3 and RG4) for (D) average rainfall series, (E) a wet rainfall series and (F) a

which is often the sole means to manage soil fertility and secureyields (Powell et al., 2004).

(years)

Wet series Dry series(C)

(F)5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Wet series Dry series

rgeted fertilisation Village food self-sufficiency

os (baseline, no-access to crop residues of the non-cattle farmers (RG3 and RG4), andwet series and (C) a dry series, and fraction of total grain produced by non-cattledry rainfall series.

Table 5Simulated cattle and crop production, manure excreted, collected and applied, and crop residues incorporated for each resource group. Averages and standard deviations of the last five of the simulations for the baseline scenario usingthe wet rainfall series.

Herd size Liveweight

Grass intake Crop residuesintake

Milk produced Manure entersthe kraal

Crop residuesto kraal

Total manureapplied

Maize stoverproduced

Crop residuesto soils

C stock Grainproduced

(# farm�1) (kg farm�1) (t farm�1 yr�1) (t farm�1 yr�1) (kg farm�1 yr�1) (t farm�1 yr�1) (t farm�1 yr�1) (t farm�1 yr�1) (t farm�1 yr�1) (t farm�1 yr�1) (t farm�1) (t farm�1 yr�1)

BaselineRG1 13.6 ± 0.4 3715 ± 413 24.5 ± 1.7 7.5 ± 1.6 2049 ± 115 7.9 ± 0.3 0.9 ± 0.2 3.3 (4.3)a 6.6 (2.0)b <0.5 38.4 (11.6)b 5.2 (1.6)b

RG2 6.9 ± 0.3 1844 ± 198 12.2 ± 0.8 3.7 ± 0.8 1106 ± 60 3.9 ± 0.2 0.6 ± 0.1 1.8 (2.2) 3.4 (1.7) <0.5 19.0 (9.5) 3.0 (1.5)RG3 0 0 0 0 0 0 0 0 (0.2) 1.2 (0.7) <0.3 14.6 (8.6) 0.5 (0.3)RG4 0 0 0 0 0 0 0 0 (0.1) 0.5 (0.6) <0.3 6.5 (8.2) 0.2 (0.2)

Non-cattle farmers conserve their crop residuesRG1 12.2 ± 0.3 3285 ± 339 22.7 ± 1.1 5.7 ± 1.4 1801 ± 55 7.0 ± 0.2 0.9 ± 0.1 3.0 (4.0) 6.6 (2.0) <0.3 38.4 (11.6) 5.1 (1.6)RG2 6.1 ± 0.1 1595 ± 169 11.1 ± 0.6 2.8 ± 0.7 951 ± 43 3.3 ± 0.1 0.7 ± 0.1 1.6 (2.0) 3.4 (1.7) <0.3 19.0 (9.5) 3.0 (1.5)RG3 0 0 0 0 0 0 0 0 (0) 1.7 (1.0) 1.6 (1.0) 18.6 (11.0) 0.9 (0.5)RG4 0 0 0 0 0 0 0 0 (0) 0.8 (1.0) 0.7 (0.9) 8.4 (10.5) 0.4 (0.5)

Targeted fertilisationRG1 14.5 ± 0.7 4157 ± 412 25.8 ± 2.0 9.6 ± 1.4 2212 ± 111 8.7 ± 0.3 0.8 ± 0.1 3.5 (4.9) 12.8 (3.9) 3.2 (1.0) 49.8 (15.1) 12.4 (3.7)RG2 7.5 ± 0.4 2118 ± 202 13.3 ± 1.0 4.9 ± 0.7 1194 ± 82 4.4 ± 0.2 0.5 ± 0.1 1.8 (2.1) 5.4 (2.7) 2.6 (1.3) 26.1 (13.0) 4.7 (2.4)RG3 0 0 0 0 0 0 0 0 (0.4) 4.1 (2.4) 1.4 (0.8) 20.0 (11.8) 3.7 (2.2)RG4 0 0 0 0 0 0 0 0 (0.2) 2.0 (2.4) 1.0 (1.0) 9.5 (11.9) 1.7 (2.1)

a Total manure excreted during grazing plus addition of compost.b Between parentheses expressed in a per ha basis.

Table 6Simulated cattle and crop production, manure excreted, collected and applied, and crop residues incorporated for each resource group. Averages and standard deviations of the last five of the simulations for the baseline scenario usingthe dry rainfall series.

Herd size Liveweight

Grass intake Crop residuesintake

Milk produced Manure entersthe kraal

Crop residuesto kraal

Total manureapplied

Maize stoverproduced

Crop residuesto soils

C stock Grainproduced

(# farm�1) (kg farm�1) (t farm�1 yr�1) (t farm�1 yr�1) (kg farm�1 yr�1) (t farm�1 yr�1) (t farm�1 yr�1) (t farm�1 yr�1) (t farm�1 yr�1) (t farm�1 yr�1) (t farm�1) (t farm�1 yr�1)

BaselineRG1 10.9 ± 0.2 3147 ± 291 20.7 ± 1.7 7.0 ± 1.5 1515 ± 121 6.6 ± 0.5 0.8 ± 0.2 2.7 (3.8)a 6.3 (1.9)b <0.5 38.0 (11.5)b 4.8 (1.5)b

RG2 5.7 ± 0.1 1562 ± 146 10.3 ± 0.8 3.5 ± 0.8 808 ± 55 3.2 ± 0.2 0.6 ± 0.1 1.5 (1.9) 3.2 (1.6) <0.5 18.7 (8.5) 2.7 (1.4)RG3 0 0 0 0 0 0 0 0 (0.2) 1.2 (0.7) <0.3 14.5 (8.6) 0.5 (0.3)RG4 0 0 0 0 0 0 0 0 (0.1) 0.5 (0.6) <0.3 6.5 (8.2) 0.2 (0.2)

Non-cattle farmers conserve their crop residuesRG1 9.6 ± 0.2 2631 ± 198 18.1 ± 1.2 5.1 ± 1.2 1289 ± 67 5.5 ± 0.4 0.8 ± 0.2 2.7 (3.4) 6.2 (1.9) <0.3 37.6 (11.4) 4.7 (1.4)RG2 4.9 ± 0.1 1332 ± 106 9.2 ± 0.6 2.6 ± 0.2 705 ± 39 2.4 ± 0.2 0.6 ± 0.1 1.3 (1.8) 3.1 (1.6) <0.3 18.6 (8.4) 2.7 (1.3)RG3 0 0 0 0 0 0 0 0 (0) 1.7 (1.0) 1.5 (0.9) 18.3 (10.8) 0.9 (0.5)RG4 0 0 0 0 0 0 0 0 (0) 0.7 (0.9) 0.7 (0.8) 8.2 (10.3) 0.4 (0.5)

Targeted fertilisationRG1 12.2 ± 0.3 3604 ± 351 22.2 ± 2.0 9.2 ± 1.2 1723 ± 199 7.4 ± 0.5 0.7 ± 0.2 3.0 (4.3) 12.1 (3.7) 3.1 (1.0) 48.7 (14.8) 11.6 (3.5)RG2 6.4 ± 0.2 1810 ± 182 11.2 ± 1.0 4.7 ± 0.6 923 ± 94 3.7 ± 0.2 0.4 ± 0.1 1.5 (1.9) 5.0 (2.5) 2.5 (1.1) 25.2 (11.4) 4.4 (2.2)RG3 0 0 0 0 0 0 0 0 (0.4) 4.1 (2.4) 1.4 (0.8) 19.8 (11.6) 3.6 (2.2)RG4 0 0 0 0 0 0 0 0 (0.1) 1.9 (2.4) 0.9 (1.1) 9.4 (11.7) 1.7 (2.1)

a Total manure excreted during grazing plus addition of compost.b Between parentheses expressed in a per ha basis.

186M

.C.Rufino

etal./A

griculturalSystem

s104

(2011)175–

190


3.4. Cattle productivity vs. crop productivity

More than 85% of communal farm households use animaldraught power to cultivate their land, but only 5–8% of the farmershave sufficient animal traction which leads either to poor cropyields because of delays in planting, reduction in the planted area,or failure to carry out winter ploughing (Shumba, 1984). A span oftwo oxen requires about three and a half days to plough a hectareof land on a wet soil (Francis, 1993). According to Spear (1968),maize yields are reduced by 1–3% per day when planting is delayeduntil mid-November. In the simulations it was assumed that thereare no constraints for land preparation which may delay planting,assumption which may be valid for oxen owners. In practice, yieldsattained by farmers without cattle would probably be lower andmore variable across years than those obtained in this study. Inthe simulations, the village had between 40–50 oxen across sce-narios, which would be enough to plough the 116 ha of croplandwithin a month, if access to oxen were guaranteed and exchangesbetween farmers were facilitated. To sustain the herd size of thebaseline scenario, cattle of RG1 (with an average of 10 heads) con-sumed between 20–25 t of grass biomass yr�1. The Hyparrhenia-veld grassland produces about 1.5–4 t biomass ha�1 yr�1 (Frost,1996), of which about half may be actually consumed by cattledue either to poor quality or constraints imposed by herbage avail-ability (De Ridder and Breman, 1993; Herrero et al., 1998). Withouttaking into account the negative effect of overgrazing on the pas-tures, each farmer of RG1 would need to have access to 12–27 haof grassland to apply about 3–4 t of manure yr�1 (under relativelypoor management) in their outfields, i.e. 4–10 ha of grassland perhectare of cropland depending on rainfall. Under the baseline sce-nario with current management, it should be possible to maintainthe herd size to guarantee the availability of oxen for ploughing. Itappears that the cattle population of the village of Manjonjo with acropland area of 116 ha, and a grassland area of 426 ha, could onlybe expanded if the production of feed (crop residues or forages)would increase, as in the targeted fertilisation scenario. Investingin livestock productivity is something that needs to be evaluatedwith the farming community, as increases in populations of live-stock may increase inequities in the use of the common resources(Ramisch, 2005). Inequities may be compensated by other ex-change mechanisms at community level (Dekker, 2004).

3.5. Options for intensification through crop–livestock integrationunder climate variability

There is consensus on the need of organic resources to sustaincrop production in communal farming areas of Zimbabwe(Campbell et al., 1998; Mapfumo et al., 2007; Nyamangara andNyagumbo, 2010), and researchers agree on the long-term effec-tiveness of manure applications (Grant, 1967; Mugwira et al.,2002; Nyamangara et al., 2003). However, recommendations donot match manure availability on smallholders farms (Zingoreet al., 2007a). Our study indicates that farmers with 10 cattlemay recycle about 4 t manure yr�1 with current manure manage-ment. This amount may be increased by adding crop residues,but the quality of manure compost will be reduced, and the costof transport back to fields will increase. Besides, adding crop resi-dues to the compost reduces feed availability to cattle, which inturn affects manure production. Surveys conducted in communalareas of Zimbabwe (Mugwira and Mugwira, 1997) indicated thatfarmers apply much larger amounts (between 10–20 t manurefarm�1 yr�1) than that calculated in our simulations. However,manure applied by farmers is usually mixed with large amountsof sand that comes from the bottom of the kraal when manure isdug out, which reduces the quality of the compost. Research inZimbabwe has shown some opportunities to increase manure

availability through storing manure in pits instead of in heaps(Nzuma and Murwira, 2000), which may require extra labour. Thismay reduce the gap between manure that accumulates during kra-aling, which was in our study between 7–8 t manure farm�1 yr�1

for farmers with 10 cattle, and the 3–4 t manure farm�1 yr�1 ap-plied to crops. Losses of about 30% for C and 20% for N were mea-sured for optimal manure management in the highlands of EastAfrica where manures were stored for 6 months (Rufino et al.,2007). In southern Africa, the storage of manure in kraals for longerthan 6 months leads to large C and nutrient losses. The lack of ac-cess to livestock by most smallholder farmers plus the poor effi-ciency of nutrient retention through manure management meanthat animal manure alone will not suffice to support crop produc-tion and a combination of means to manage soil fertility areneeded.

Thornton et al. (2003) identified some crop–livestock manage-ment strategies that show promise in increasing income and pro-ductivity of smallholders in maize-based cropping systems ofsouthern Africa. These strategies included improved livestock feed-ing with dry and green maize stover, and intercropping grain cropswith dual-purpose or forage legumes. Increasing use of crop resi-dues for cattle feeding may bring negative consequences for non-cattle farmers unless this is compensated somehow through socialagreements. Sumberg (2002) suggested that smallholder farmersshow little interest in increasing productivity per head of their cat-tle, because of the finance and insurance roles attached to cattle.Combinations of technologies where all farmers access fertilisers,non-cattle farmers keep their crop residues and cattle farmers pro-duce forage legumes to allow cattle to consume poor quality grass,may help reducing competition for organic resources at village le-vel. These technologies need to be explored together with farmersto evaluate how they fit into their broader livelihoods strategies.Scenario evaluation using models as presented here could supportsuch participatory analysis. Adaptation options cannot be blind towhat occurs beyond field and farm level, because otherwise recom-mendations from research and development simply do not fit thelocal conditions and farmers tend to ignore them.

4. Conclusions

An increase in biomass production result of the widespread useof small rates of fertiliser, partial retention of crop residues in thefields, and small rates of manure appear the best combination foreffective crop–livestock integration. Farmers make use of commonresources according to their social agreements. The removal of or-ganic residues by cattle, which is tolerated by non-cattle farmersleads to small crop yields in poor fields of these farmers. Free-graz-ing of crop residues has relatively little impact on-farms of cattleowners who use manure and fertilisers. Yet, the grazing of cropresidues may become crucial in dry years and have a long lastingeffect on livestock populations. Rainfall variability intensifiescrop–livestock interactions increasing the competition for biomassto feed livestock (short-term effect) or to rehabilitate soils (long-term effect). Prolonged dry seasons and low availability of crop res-idues may lead to cattle losses. This may impact in turn on theavailability of animal traction, negatively affecting the area undercultivation in consecutive seasons until farmers re-stock. Cattlepopulations are needed not only as assets but also due to the keyrole of animal traction on timing of planting and productivityand as means to manage soil fertility. Crop–livestock integrationat village level resulted in concentration of nutrients in farms withlarge herds, increasing the dependency of poor smallholders onexternal inputs and other types of exchanges within villages suchas labour for food, cash or manure. We showed in the targeted fer-tilisation scenario that three times more fertiliser than currently


used was necessary to boost productivity and to compensate theremoval of C and nutrients through grazing. This scenario, how-ever, may be unrealistic for a smallholder village in Zimbabwe un-der the economic and political circumstances that prevailed at thetime of the study – though a highly plausible scenario for donorassistance. To allow such changes to occur and be sustainable, pro-found institutional transformations would be required.

Acknowledgements

Funding for this research was provided by the European Unionthrough the AfricaNUANCES project (Contract No. INCO-CT-2004-003729), the International Development Research Centre (IDRC)and the Department for International Development (DFID) through

Table A1Landscape and grazing units for the communal grazing land of Manjonjo village, NE Zimdiscussions with key informants. Advantages were qualified with + (more is better) and d

Landscape unit Grazing unit Soil type Tree cover Water logging Land

Low Miombowoodland

1 Sandy Young trees No Wood

2 Vlei No trees Yes Grazi3 Sandy Young trees Partial Wood

Open grassland 4 Sandy No trees No Grazi5 Vlei No trees Yes Grazi6 Sandy No trees No Fallow7 Sandy No trees No Grazi

High Miombowoodland

8 Red clay Mature trees No Wood

9 Red clay Mature trees No Grazi10 Red clay Mature trees No Wood

Table A2Grazing units of the communal grassland of the Manjonjo village, NE Zimbabwe. Areadistinguish grazing units.

Landscapeunits

Grazingunits

Area(ha)

March April May

Standingbiomass

Litter Standingbiomass

Litter Standbiom

(kg ha�1) (kg ha�1) (kg ha�1) (kg ha�1) (kg h

Low 1 142 700 7 250 13 150Miombo 2 14 2400 288 2300 276 2100Woodland 3 11 3300 3 2450 74 1900

Open 4 17 2500 25 2350 141 2300

Grassland 5 39 3300 132 3100 31 29006 20 2000 0 1100 0 8507 8 1300 0 900 9 700

High 8 22 1550 78 1300 182 1200Miombo 9 154 1050 179 800 248 600Woodland 10 12 1350 95 1000 190 900

Table A3Feed quality parameters for the feed stuffs used in the simulations.

Grass species Early rainy season Early dry se

ME DMD CP ME(MJ kg DM�1) (g kg DM�1) (g kg DM�1) (MJ kg DM�

Hyparrhenia dissoluta 10.2 650 135 8.0Sporobolus pyramidalis 9.8 620 125 7.9Heteropogon contortus 10.5 670 110 7.5Digitaria gazensis 11.5 700 163 9.2Andropogon gayanus 9.7 620 158 8.0Cynodon dactylon 10.4 640 137 8.3Aristida congesta 10.0 650 109 7.8

Other feedsZea mays leaves – – – 8.0Zea mays stems – – – 6.8

Sources: Topps and Oliver (1993) and Boudet (1991).

the Climate Change and Adaptation in Africa (CCAA) programme‘‘Lack of resilience in African smallholder farming: Exploring mea-sures to enhance the adaptive capacity of local communities topressures of climate change”. We thank Nico de Ridder, Peter Frostand Barbara Maasdorp for their valuable contributions to thisstudy, and two anonymous reviewers for their comments on ourmanuscript. Any error or omission remains full responsibility ofthe authors. We are grateful to the farmers and herders of the Man-jonjo village who contributed with their valuable time to the datacollection with which this study was built.

Appendix A

See Tables A1–A3.

babwe. Criteria for defining the grazing units were derived from observations andisadvantages with � (more is worse).

use Grazingintensity

Grassquality

Distanceto fields

Waterlogging

Openview

Presenceof rocks

land clearing High � � � + + + + + + � �

ng Low + + + � � + + + +land clearing High + + + � � � � + + + + +

ng Low + � � � � + + + + + +ng Low + + � � � � � + + + + + +

High + + � + + + + + + + + +ng (fallow) High + � + + + + + + + + +

land clearing Low + + + + + + + + + � � � � �

ng Very low � � + + + + + + � � � � � �land clearing Low � � + + + + + + � � � � � �

and standing biomass and dead biomass (litter) and criteria used by herdsmen to

Most abundant species

ingass

Litter

a�1) (kg ha�1)

12 Hyparrhenia dissoluta, Sporobolus pyramidalis, Aristida congesta441 Sporobolus pyramidalis, Kyllinga erecta, Bulbostylis burchellii

19 Sporobolus pyramidalis, Cynodon dactylon, Eleusine indica

184 Sporobolus pyramidalis, Hyparrhenia dissoluta, Heteropogoncontortus

435 Sporobolus pyramidalis, Scleria lagoensis, Kyllinga erecta0 Sporobolus pyramidalis, Eragrostis heteromera, Cynodon dactylon,

56 Cynodon dactylon, Sporobolus pyramidalis, Hyparrhenia dissoluta

288 Hyparrhenia dissoluta, Andropogon gayanus, Heteropogon contortus264 Hyparrhenia filipendula, Andropogon gayanus, Aristida leusina243 Hyparrhenia dissoluta, Andropogon gayanus, Themeda triandra

ason Late dry season

DMD CP ME DMD CP1) (g kg DM�1) (g kg DM�1) (MJ kg DM�1) (g kg DM�1) (g kg DM�1)

510 45 6.3 400 40400 43 5.2 330 30410 32 5.8 370 24630 74 8.2 530 45470 58 6.2 400 47517 60 8.0 500 50420 52 5.8 370 43

520 72 7.6 500 50500 54 6.0 450 45


References

AFRC, 1993. Energy and Protein Requirements of Ruminants. An Advisory ManualPrepared by the AFRC Technical Committee on Response to Nutrients. CABIInternational, Wallingford, UK.

Baars, R.M.T., 1996. Condition and Management of the Rangelands in the WesternProvince of Zambia. Wageningen University, Wageningen, The Netherlands.

Barnes, D.L., 1956. Veld Management Studies at the Gwebi College of Agriculture,Southern Rhodesia. Witwatersrand University, Johannesburg, South Africa.

Boudet, G., 1991. Manuel sur les Pâturages Tropicaux et les Cultures Fourragères.Institut d’Elevage et de Médecine Vétérinaire des Pays Tropicaux (IEMVT).Maisons-Alfort, France.

Campbell, B.M., Frost, P., Kirchmann, H., Swift, M., 1998. Nitrogen cycling andmanagement of soil fertility in small-scale farming systems in north easternZimbabwe. J. Sust. Agric. 11, 19–39.

Chakeredza, S., Hove, L., Akinnifesi, F.K., Franzel, S., Ajayi, O.C., Sileshi, G., 2007.Managing fodder trees as a solution to human-livestock food conflicts and theircontribution to income generation for smallholder farmers in southern Africa.Nat. Res. For. 31, 286–296.

Chikowo, R., Mapfumo, P., Nyamugafata, P., Giller, K.E., 2004. Woody legume fallowproductivity, biological N2-fixation and residual benefits to two successivemaize crops in Zimbabwe. Plant Soil 262, 303–315.

Chikowo, R., Mapfumo, P., Leffelaar, P.A., Giller, K.E., 2006. Integrating legumes toimprove N cycling on smallholder farms in sub-humid Zimbabwe: Resourcequality, biophysical and environmental limitations. Nutr. Cycl. Agroecosyst. 76,219–231.

Chinembiri, F.M., 1999. Communal area livestock management systems inZimbabwe. In: Holness, D.H. (Ed.), Strategies for Dry Season Feeding ofAnimals in Central and Southern Africa. Zimbabwe Society for AnimalProduction and FAO, Harare, Zimbabwe. <http://www.fao.org/docrep/004/AC152E/AC152E08.htm#ses4.1>.

Chuma, E., Mombeshora, B.G., Murwira, H.K., Chikuvire, J., 2000. The dynamics ofsoil fertility management in communal areas of Zimbabwe. In: Hilhorst, T.,Muchena, F. (Eds.), Nutrients on the Move-Soil Fertility Dynamics in AfricanFarming Systems. International Institute for Environment and Development,London, UK, pp. 115–132.

Clatworthy, J.N., Maclaurin, A.R., Avila, M., 1986. The possible role of forage legumesin communal area farming systems in Zimbabwe. In: Haque, I., Jutzi, S., Neate,P.J.H. (Eds.), Potentials of Forage Legumes in Farming Systems of Sub-SaharanAfrica, ILCA Workshop held at Addis Ababa, Ethiopia, 16–19 September 1985,pp. 265–288.

Conrad, H.R., 1966. Symposium on factors influencing voluntary intake of herbageby ruminants – physiological and physical factors limiting feed intake. J. Anim.Sci. 25, 227–243.

De Ridder, N., Breman, H., 1993. A new approach to evaluating rangelandproductivity in Sahelian countries. In: Behnke, R.H., Scoones, I., Kerven, C.(Eds.), Range Ecology at Disequilibrium. Overseas Development Institute,London, UK, pp. 104–117.

Dekker, M., 2004. Sustainability and resourcefulness: support networks duringperiods of stress. World Dev. 32, 1735–1751.

Dzowela, B.H., Hove, L., Maasdorp, B.V., Mafongoya, P.L., 1997. Recent work on theestablishment, production and utilization of multipurpose trees as a feedresource in Zimbabwe. Anim. Sci. Feed Technol. 69, 1–15.

Francis, J., 1993. Effects of Strategic Supplementation on Work Performance andPhysiological Parameters of Mashona Oxen. Department of Animal Science,University of Zimbabwe, Harare, Zimbabwe.

French, N.P., Tyrer, J., Hirst, W.M., 2001. Smallholder dairy farming in the Chikwakacommunal land, Zimbabwe: birth, death and demographic trends. Prev. Vet.Med. 48, 101–112.

Frost, P., 1996. The ecology of miombo woodlands. In: Campbell, B.M. (Ed.), TheMiombo Woodland in Transition: Woodland and Welfare in Africa, vol. 266.CIFOR, Bogor, Indonesia, pp. 11–57.

Gambiza, J., Bond, W., Frost, P.G.H., Higgins, S., 2000. A simulation model of miombowoodland dynamics under different management regimes. Ecol. Econ. 33, 353–368.

Giller, K.E., Rowe, E., de Ridder, N., van Keulen, H., 2006. Resource usedynamics and interactions in the tropics: scaling up in space and time.Agric. Syst. 88, 8–27.

Grant, P.M., 1967. The fertility of Sandveld soil under continuous cultivation. Part II.The effect of manure and nitrogen fertilizer on the base status of the soil.Rhodesia Zambia Malawi. J. Agric. Res. 5, 117–128.

Grant, P.M., 1976. Peasant farming on infertile sands. Rhodesia Sci. News 10, 252–254.

Hargreaves, S.K., Bruce, D., Beffa, M.L., 2004. Disaster Mitigation Options forLivestock Production in Communal Farming Systems in Zimbabwe. 1.Background Information and Literature Review ICRISAT and FAO. Bulawayo,Zimbabwe and Rome, Italy. p. 69.

Herrero, M., Dent, J.B., Fawcett, R.H., 1998. The plant/animal interface in models ofgrazing systems. In: Peart, R.M., Curry, R.B. (Eds.), Agricultural SystemsModeling and Simulation. Marcel Dekker Publishers, New York, USA, pp. 495–542.

Illius, A.W., O’Connor, T.G., 1999. On the relevance of nonequilibrium concepts toarid and semiarid grazing systems. Ecol. Appl. 9, 798–813.

Johnson, I.R., Parsons, A.J., 1985. A theoretical analysis of grass growth undergrazing. J. Theory Biol. 112, 345–367.

Konandreas, P.A., Anderson, F.M., 1982. Cattle Herd Dynamics: An Integer andStochastic Model for Evaluating Production Alternatives. ILCA Research Report2, ILCA Publications, Addis Ababa, Ethiopia.

Kunjeku, E., Frost, P.G.H., Nabane, N., Murwira, H.K., Mushove, P., Chidziya, E.,Gambiza, J., Mkwana, R., Semwayo, D.T., Bond, I., 1998. Research, inventory andmonitoring. In: Chenje, M., Sola, L., Paleczny, D. (Eds.), The State of Zimbabwe’sEnvironment 1998. Ministry of Mines, Environment and Tourism, Goverment ofthe Republic of Zimbabwe, Goverment of the Republic of Zimbabwe, Harare,Zimbabwe, pp. 313–403.

Le Houérou, H.N., 1984. Rain use efficiency: an unifying concept in arid landecology. J. Arid Environ. 7, 213–247.

Le Houérou, H.N., Bingham, R.L., Skerbek, W., 1988. Relationship between thevariability of primary production and the variability of annual precipitation inworld arid lands. J. Arid Environ. 15, 1–18.

Manlay, R.J., Ickowicz, A., Masse, D., Feller, C., Richard, D., 2004. Spatial carbon,nitrogen and phosphorus budget in a village of the West African savanna – II.Element flows and functioning of a mixed-farming system. Agric. Syst. 79, 83–107.

Mapfumo, P., Mtambanengwe, F., Vanlauwe, B., 2007. Organic matter quality andmanagement effects on enrichment of soil organic matter fractions incontrasting soils in Zimbabwe. Plant Soil 296, 137–150.

Matarira, C.H., Makadho, J.C., Mukahanana-Sangarwe, M., 2004. Vulnerability andadaptation of maize production to climate change in Zimbabwe. In: Zhakata, W.(Ed.), Goverment of the Republic of Zimbabwe. Ministry of Environment andTourism, Harare, Zimbabwe, pp. 1–66.

Matimati, I., Maasdorp, B.V., Hove, L., 2009. On-farm productivity of Acaciaangustissima, Calliandra calothyrsus and Leucaena leucocephala in a subhumidarea in Zimbabwe. Afr. J. Range For. Sci. 26, 75–80.

Mtambanengwe, F., Mapfumo, P., 2005. Organic matter management as anunderlying cause for soil fertility gradients on smallholder farms inZimbabwe. Nutr. Cycl. Agroecosyst. 73, 227–243.

Mugwira, L.M., Murwira, H.K., 1997. Use of Cattle Manure to Improve Soil Fertility inZimbabwe: Past and Current Research and Future Research Needs. SoilFertNetWorking Paper 2. Soil Fertility Network For Maize-based Cropping Systems inSouthern Africa, CIMMYT, Harare, Zimbabwe.

Mugwira, L.M., Nyamangara, J., Hikwa, D., 2002. Effects of manure and fertilizer onmaize at a research station and in a smallholder (peasant) area of Zimbabwe.Commun. Soil Sci. Plant Anal. 33, 379–402.

Mukasa-Mugerwa, E., 1989. A review of reproductive performance of female Bosindicus (Zebu) cattle. ILCA, Addis Abada, Ethiopia.

Nyamangara, J., Nyagumbo, I., 2010. Interactive effects of selected nutrientresources and tied-ridging on plant growth performance in a semi-aridsmallholder farming environment in central Zimbabwe. Nutr. Cycl.Agroecosyst. doi: 10.1007/s10705-009-9282-7.

Nyamangara, J., Piha, M.I., Giller, K.E., 2003. Effect of combined cattle manure andmineral nitrogen on maize N uptake and grain yield. Afr. J. Crop Sci. 11, 289–300.

Nyamapfene, K.W., 1991. Soils of Zimbabwe. Nehanda Publishers (Pvt) Ltd., Harare,Zimbabwe.

Nzuma, J.K., Murwira, H.K., 2000. Improving the management of manure inZimbabwe Managing Africa’s soils 15. IIED-Drylands Programme, London, UK.p. 20.

Powell, J.M., Fernandez-Rivera, S., Hiernaux, P., Turner, M.D., 1996. Nutrient cyclingin integrated rangeland/cropland systems of the Sahel. Agric. Syst. 52, 143–170.

Powell, J.M., Pearson, R.A., Hiernaux, P.H., 2004. Crop–livestock interactions in theWest African drylands. Agron. J. 96, 469–483.

Ramisch, J.J., 2005. Inequality, agro-pastoral exchanges, and soil fertility gradientsin southern Mali. Agric. Ecosyst. Environ. 105, 353–372.

Rattray, J.M., 1957. The grasses associations of Southern Rhodesia. Rhodesia Agric. J.54, 197–234.

Richardson, F.D., Hahn, B.D., 1991. A model for the evaluation of differentproduction strategies for an animal production from rangeland in developingareas: an overview. J. Grass. Soc. S. Afr. 8, 153–159.

Rodel, M.G.W., Hopley, J.D.H., 1973. Investigations into systems of farming suitablefor tribal trust land. Rhodesia Agric. J. 70, 1–18.

Rufino, M.C., 2008. Quantifying the Contribution of Crop–livestock Integration toAfrican Farming. Doctoral thesis, Wageningen University, Wageningen, TheNetherlands.

Rufino, M.C., Tittonell, P., van Wijk, M.T., Castellanos-Navarrete, A., Delve, R.J., deRidder, N., Giller, K.E., 2007. Manure as a key resource within smallholderfarming systems: analysing farm-scale nutrient cycling efficiencies with theNUANCES framework. Livest. Sci. 112, 273–287.

Rufino, M.C., Herrero, M., van Wijk, M.T., Hemerik, L., de Ridder, N., Giller, K.E., 2009.Lifetime productivity of dairy cows in smallholder farming systems of thehighlands of Central Kenya. Animal 3, 1044–1056.

Sandford, S., Scoones, I., 2006. Opportunistic and conservative pastoral strategies:some economic arguments. Ecol. Econ. 58, 1–16.

Schlecht, E., Hiernaux, P., Achard, F., Turner, M.D., 2004. Livestock related nutrientbudgets within village territories in western Niger. Nutr. Cycl. Agroecosyst. 68,199–211.

Scoones, I., 1996. Hazards and Opportunities; Farming Livelihoods in Dry landAfrica: Lessons from Zimbabwe. Zed Books Ltd., London.

Senft, R.L., 1989. Hierarchical foraging models: effects of stocking andlandscape composition on simulated resource use by cattle. Ecol. Model.46, 283–303.

http://www.fao.org/docrep/004/AC152E/AC152E08.htm#ses4.1

http://www.fao.org/docrep/004/AC152E/AC152E08.htm#ses4.1


Senft, R.L., Coughenour, M.B., Bailey, D.W., Rittenhouse, L.R., Sala, O.E., Swift, D.M.,1987. Large herbivore foraging and ecological hierarchies. Bioscience 37, 789–799.

Shumba, E., 1984. Animals and the cropping system in the Communal Areas ofZimbabwe. Zimbabwe Sci. News 18, 99–102.

Smith, C.A., 1961. Studies on the Northern Rhodesia Hyparrhenia veld. 1. Thegrazing behaviour of indigenous cattle grazed at light and heavy stocking rates.J. Agric. Sci. 55, 369–375.

Smith, C.A., 1962. Utilization of Hyparrhenia-veld for nutrition of cattle in dryseason. 3. Studies on digestibility of produce of mature veld and veld hay, andeffect of feeding supplementary protein and urea. J. Agric. Sci. 58, 173–179.

Spear, A.J., 1968. Tillage Problems of Peasant Agriculture in Rhodesia. Supplementto Modern Farming June 1968.

Steinfeld, H., 1988. Livestock Development in Mixed Farming Systems.Wissenschaftsverlag Vauk Kiel, Stuttgart, Germany.

Sumberg, J., 2002. Livestock nutrition and foodstuff research in Africa: when is anutritional constraint not a priority research problem? Anim. Sci. 75, 332–338.

Thornton, P.K., Kristjanson, P.M., Thorne, P.J., 2003. Measuring the potential impactsof improved food-feed crops: methods for ex ante assessment. Field Crop Res.84, 199–212.

Thornton, P.K., Boone, R.B., Galvin, K.A., BurnSilver, S.B., Waithaka, M.W., Kuyiah, J.,Karanja, S., Gonzalez-Estrada, E., Herrero, M., 2007. Coping strategies inlivestock-dependent households in East and Southern Africa: a synthesis offour case studies. Human Ecol. 35, 461–476.

Tittonell, P., Zingore, S., van Wijk, M.T., Corbeels, M., Giller, K.E., 2007. Nutrient useefficiencies and crop responses to N, P and manure applications in Zimbabweansoils: exploring management strategies across soil fertility gradients. Field CropRes. 100, 348–368.

Tittonell, P., Corbeels, M., Van Wijk, M.T., Vanlauwe, B., Giller, K.E., 2008. Combiningorganic and mineral fertilizers for integrated soil fertility management in

smallholder farming systems of Kenya - explorations using the crop-soil modelfield. Agron. J. 100, 1511–1526.

Tittonell, P., van Wijk, M.T., Herrero, M., Rufino, M.C., de Ridder, N., Giller, K.E., 2009.Beyond resource constraints – exploring the biophysical feasibility of optionsfor the intensification of smallholder crop–livestock systems in Vihiga district.Kenya Agric. Syst. 101, 1–19.

Tittonell, P., Corbeels, M., van Wijk, M.T., Giller, K.E., 2010. FIELD-A summarysimulation model of the soil-crop system to analyse long-term resourceinteractions and use efficiencies at farm scale. Eur. J. Agron. 32, 10–21.

Topps, J.H., Oliver, J., 1993. Animal Foods of Central Africa. Zimbabwe AgriculturalJournal Technical Handbook No. 2. Modern Farming Publications, Harare,Zimbabwe. pp. 76–105.

Tschirley, D.L., Jayne, T.S., 2010. Exploring the logic behind Southern Africa’s foodcrises. World Dev. 38, 76–87.

Van Wijk, M.T., Tittonell, P., Rufino, M.C., Herrero, M., Pacini, C., Ridder, N.d., Giller,K.E., 2009. Identifying key entry-points for strategic management ofsmallholder farming systems in sub-Saharan Africa using the dynamic farm-scale simulation model NUANCES-FARMSIM. Agric. Syst. 102, 89–101.

Vincent, V., Thomas, R.G., 1960. An agricultural survey of Southern Rhodesia. Part I:agro-ecological survey. Government Printer, Salisbury, Rhodesia.

Zingore, S., Murwira, H.K., Delve, R.J., Giller, K.E., 2007a. Influence of nutrientmanagement strategies on variability of soil fertility, crop yields and nutrientbalances on smallholder farms in Zimbabwe. Agric. Ecosyst. Environ. 119, 112–126.

Zingore, S., Murwira, H.K., Delve, R.J., Giller, K.E., 2007b. Soil type, managementhistory and current resource allocation: three dimensions regulating variabilityin crop productivity on African smallholder farms. Field Crop Res. 101, 296–305.

Zingore, S., Delve, R.J., Nyamangara, J., Giller, K.E., 2008. Multiple benefits ofmanure: the key to maintenance of soil fertility and restoration of depletedsandy soils on African smallholder farms. Nutr. Cycl. Agroecosyst. 80, 267–282.

Competing use of organic resources, village-level ...iks.ukzn.ac.za/sites/default/files/95.pdf ·...

Documents

Transcript of Competing use of organic resources, village-level ...iks.ukzn.ac.za/sites/default/files/95.pdf ·...