Field Crops Research 113 (2009) 105–115
Contents lists available at ScienceDirect
Field Crops Research journal homepage: www.elsevier.com/locate/fcr
Modeling the impacts of contrasting nutrient and residue management practices on grain yield of sorghum (Sorghum bicolor (L.) Moench) in a semi-arid region of Ghana using APSIM Dilys S. MacCarthy *, Rolf Sommer 1, Paul L.G. Vlek Center for Development Research (ZEF), University of Bonn, Walter Flex Strasse 3, 53113, Bonn, Germany
A R T I C L E I N F O
A B S T R A C T
Article history: Received 24 August 2007 Received in revised form 7 April 2009 Accepted 8 April 2009
The cropping systems model APSIM (Agricultural Production Systems sIMulator) was applied to assess the response of sorghum grain yield to inorganic fertilizers applications and residue retention in diverse farmers’ management systems (homestead fields and bush farms). The model was parameterized using data collected from experiments under optimum growth conditions (limited water or nutrient stress). Independent data from field experiments with three levels of P and four levels of N fertilizers conducted at two different locations and soils were used to evaluate the model. Soil water and fertility parameters measured were used for simulations while same starting conditions were assumed for unmeasured parameters for all trials. APSIM predicted the grain yield response of sorghum to both N and P applications with an overall modified internal coefficient of efficiency of 0.64. Following model parameterization, a long-term simulation study was conducted using a stochastic weather data derived from historical weather data to assess the effects of crop residue management on grain production. A gradual decline in sorghum grain yield was simulated over the 30-year simulation period in both the homestead fields and the bush farms, with yields being much lower in the latter under farmers’ management practices. Half the amount of mineral N fertilizer used in the bush farms was needed in the homestead fields to produce the average grain yields produced on the bush farm with full fertilization, if crop residues were returned to the fields in the homestead. Year-to-year variability in grain yield was consistently higher with the removal of crop residues, irrespective of management systems. APSIM was responsive to both organic and inorganic fertilizer applications in the study area and also highlighted the essential role of crop residues and inorganic fertilizer in influencing the temporal sorghum grain production and hence the impact of farmers’ management practices on food security. This was evident in the rapid decline in soil organic carbon (SOC) accompanied by a decline in grain yield over the 30 years of cropping. The use of inorganic fertilizer and retention of crop residues (SOC) are critical for attaining food security in the study area. ß 2009 Published by Elsevier B.V.
Keywords: Grain yield Crop residues Smallholder farming systems Sorghum
1. Introduction Soil degradation poses a serious threat to crop production and consequently to food security in sub-Saharan Africa (De Jager et al., 2003). Cereal crops constitute a crucial part of the staple food in Ghana and other West African countries. Production of cereals in the semi-arid areas of Ghana, as in other West African Savannah regions, is strongly affected by inadequate or poorly distributed rainfall as well as by low levels of nitrogen (N) and phosphorus (P)
* Corresponding author. Current address: University of Ghana, College of Agriculture and Consumer Sciences, Institute of Agricultural Research, Kpong Research Centre, P.O. Box LG 68, Legon, Accra, Ghana. Tel.: +233 244090502. E-mail address:
[email protected] (D.S. MacCarthy). 1 Current address: International Center for Agricultural Research in the Dry Areas (ICARDA), P.O. Box 5466, Aleppo, Syrian Arab Republic. 0378-4290/$ – see front matter ß 2009 Published by Elsevier B.V. doi:10.1016/j.fcr.2009.04.006
in the soil (Bationo et al., 2003). The maintenance of sustainable yields would require considerable investment in inorganic fertilizers (Vlek et al., 1997), as nutrient recycling does not compensate for the removal of P from the soil through crop harvesting. However, mineral fertilizer use is notoriously low in these regions (De Jager et al., 2003). On average, a mere 8 kg ha1 of mineral fertilizer is applied yearly (Henao and Baanante, 1999). Phosphorus deficiency is a widespread constraint to crop production in tropical soils. In the semi-arid region of Ghana, the soils are inherently low in plant available P. The mean available soil P (Bray 1) values measured in the top 0–15 cm of the soil types in the study area range from 4.4 to 28 mg kg1, a range that is far below the required level needed for optimum crop production. The low levels of available P in the soils have been attributed to their advanced weathering, low to moderate sorption and poor organic matter content and recycling (Abekoe and Tiessen, 1998).
106
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
Abekoe and Tiessen (1998) in their studies in Northern Ghana further established that the presence of lateritic nodules in the soils increase the sorption capacities for P. It also reduces available root space, hence, limiting root growth and the plant’s ability to explore available P within the soil profile. Phosphorus deficiency has been shown to reduce crop response to mineral N (Smalberger et al., 2006). P deficiency results in the reduction in photosynthesis and thus directly reduces crop growth. Generally, low P supplies of soils in the tropics have a high potential of limiting crop production hence external inputs of P are necessary for adequate crop production. It is also important to consider spatial variations in terms of changes in soil fertility, even in smallholder farming systems. Variation in soil fertility could be a result of natural factors such as underlying soil types (geology), location within topography (Franzen et al., 2002) or to dynamic processes such as land use histories and/or management activities. The latter is the case in smallholder systems as more fertile soils are typically located close to the homesteads and fertility reduces with increasing distances from the homestead. Farmers’ management practices have been shown to have generated gradients in soil carbon and nutrient stocks (Rowe et al., 2006) through the uneven allocation of organic inputs and the export of crop residues from the bush farms to farms closer to the homesteads. Moreover, most resources are allocated to the homestead fields (e.g. labour, manure) resulting in wide variations in crop yield. Cropping systems models such as Agricultural Production Systems sIMulator, APSIM (Keating et al., 2003) describe the dynamics of crop growth, soil water, soil nutrients, and plant residues as a function of climate, cropping history and soil/crop management in a daily time step. Through the linking of crop growth with soil processes, APSIM is particularly suited for the evaluation of likely impacts of alternative management practices on the soil resource and crop productivity. The model has been used successfully in the search for strategies for more efficient production, improved risk management, crop adaptation, and sustainable production (Keating et al., 2003; Van Ittersum et al., 2003). The capability to simulate crop growth in response to low soil P is one of its recent capabilities, providing an opportunity to simulate crop production in the tropics where poor P nutrition affects crop yield and efficient use of applied mineral N fertilizers. This work therefore seeks to assess the yield response of sorghum (grain) to inorganic fertilizer application in distinct farm types and to assess the sustainability of current crop residue management practices of farmers in the study area using APSIM. To achieve this, the following objectives were set: (i) parameterize APSIM for sorghum growth in the study area, (ii) evaluate the performance of the model for different management systems and soils, (iii) apply the model in analyzing selected farmers’ management practices.
2. Materials and methods 2.1. Description of study area and farming system This study was conducted in Navrongo in the Upper East Region of Ghana, bordered by latitude 1081500 and 1181000 N and 08 and 18000 W. It lies in the semi-arid part of the Volta Basin, falling in the transitional zone of Guinea and Sudan Savannah Agro-Ecological Zones. The area is characterized by a uni-modal rainfall pattern with an annual average rainfall of 950 mm. The rainy season begins in May and ends in September/October, with some variation from year to year. The soils used in the study are Endoeutric-stagnic
Plinthosol and Eutric Gleyic Regosol (FAO classification). For ease of reading, Endoeutric-stagnic Plinthosol and Eutric Gleyic Regosol will be referred to as Plinthosol and Regosol, respectively. They are very low in organic matter. The vegetation comprises of scattered trees and shrubs with grass under growths. Fallows are now almost non-existent. The farming system is characterized by low input subsistence farming. The main crops cultivated are millet, sorghum, cowpea and groundnut. Livestock populations are very low. The area also experiences annual bush burning, rendering the land surface bare during the dry season. There are two main farm types which can be distinguished from one another based on their distance from settlements and differences in soil fertility; the homestead fields and the bush farms. Farmers in their attempt to improve crop production apply organic manure and crop residues harvested from the bush farms (mainly groundnut cultivated in the bush farms) on the homestead fields. This has resulted in soil fertility gradient between the homestead fields located in the settlements and the bush farms. The stovers of sorghum and millet are used as fuel and building materials while crop residues of groundnuts are used for fodder. Both the homestead fields and bush farms are referred to as management systems in this study. The homestead fields also benefit from animal and domestic waste while the bush farms do not receive any input and are normally cultivated for cowpea and groundnut. On the relatively fertile soils in the homestead farm, sorghum and millet are cropped in annual rotation with groundnuts or cowpea. The principal method of restoring fertility to the soil in the homestead fields is through animal wastes. There are limited quantities of animal waste due to low livestock population. On the average, about 1 t ha1 of animal waste is used. Thus, due to the inadequacy of the animal waste and given the fact that it has to be transported over long distances to the bush farms, their application is mainly restricted to the homestead fields. 2.2. Model description and parameterization The key APSIM (version 4.0) modules deployed in this study were Sorghum, SoilN2 (soil nitrogen), SoilP (soil phosphorus), SoilWat (soil water balance), MANURE and RESIDUE2. Crop modules within APSIM simulate phenological development, biomass accumulation, grain yield, N and water uptake, among others. Genetic coefficients used by APSIM for sorghum are expressed in thermal degrees and photoperiod (Table 1). Crop development is controlled by temperature (thermal degree days) and photoperiod. Thermal time accumulations were derived using algorithm described in Jones and Kiniry (1986) using observed phenology and weather data, a base temperature of 8 8C and an optimal temperature of 30 8C. Weather data were collected from a near-by weather station, which is about 1 km from the homestead experiments and 2 km away from the bush farm experiments. Summary of weather data for the experimental period is presented in Table 2. Potential biomass growth is a function of the intercepted radiation and the radiation-use efficiency. Water-limited growth is a function of water supply and the transpiration efficiency of the crop, which varies daily as a function of vapour pressure deficit. Actual biomass increase is simulated from either potential or water-limited growth as modified by temperature and N stresses. Soil water dynamics between soil layers were defined by the cascading water balance method (Ritchie, 1998). Its characteristics in the model are specified by the drained upper limit (DUL), lower limit of plant extractable water (LL15) and saturated water content (SAT). Soil water content measurements before sowing defined the initial soil water content of the soil. All soil water characteristics were measured from the study site.
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
107
Table 1 Genetic coefficients used for modeling sorghum in APSIM. Coefficient
Definition
tt_emerg_to_endjuv tt_flower_to_maturity tt_flag_to_flower tt_flower_to_start_grain photo_crit 1 photo_crit 2 photo_slope
Thermal time accumulation from seedling emergence to end of juvenile phase (8C days) Thermal time accumulation from flowering to maturity (8C days) Thermal time accumulation from flag stage to flowering (8C days) Thermal time accumulation from flowering to start of grain filling (8C days) Critical photoperiod 1 Critical photoperiod 2 The extent to which growth is affected by photoperiod increases beyond photo_crit 1 and 2
Parameters influencing soil fertility are mainly represented in APSIM-SoilN2 module. Initial state variables (NO3, NH4, soil organic carbon, pH and C:N ratio for soil and roots) were measured for each soil layer from each experimental site and used for simulations. The conceptual soil organic carbon pool is represented in the module as HUM, BIOM and FOM. HUM is the more stable component. BIOM is the active component of soil carbon. FOM is the fresh soil organic matter pool including plant roots and aboveground matter incorporated into soil through tillage. Flows between these pools are regulated by the C:N ratio of the receiving pool. Estimates of these pools for the study area were obtained by measurements for the 0–15 and 15–30 soil layers. Soil P module parameterization was achieved using measured labile P content of each soil layer as described in Tiessen and Moir (1993). P sorption capacity of soils was determined through inverse modeling. Also, the C:P ratio of roots and sorghum residues were calculated using measured field data. 2.3. Experiments for model parameterization To parameterize the APSIM model, two different sets of data were collected on the sorghum cultivar ICSV III used in this study on two different planting dates (June 12 and 26, 2005) and experiments conducted in the homestead (Regosol). Each planting date experiment was replicated three times and grown under limited growth stress conditions. Individual plot size measured 12 m 7.5 m on two planting dates (June 12 and 26, 2005). Sorghum was sown by direct dibbling at about 5 cm depth, 0.75-m row spacing and plant density of 12 plants m2. About 12 mm of supplemental irrigation was applied between the end of juvenile and panicle initiation stages using watering cans. Weeds were controlled by manual weeding with hoes three times before final harvest. Organic manure (cow dung) was incorporated into the soil a day before planting at a rate of 3 t ha1. Compound fertilizer (NPK – 15 15 15 at the rate of 60 kg ha1 of each of the nutrients) was applied 2 weeks after emergence and 60 kg N ha1 in the form of sulphate of ammonium applied in the 7th week. Phenological data including planting date, date of flowering, date for start of grain filling, date of physiological maturity and
date of flag leaf appearance were collected. These were noted when 50% of plant population per plot attained each of these stages. Start of grain filling was determined by observing the presence of milky substance in grain at the base of the panicles. Physiological maturity is attained when dark layer forms at the point of attachment of the grain to the panicle. Biomass N and P concentrations were determined from dried plant biomass harvested bi-weekly till flowering and at soft dough stage and maturity from 1 m2 plots. At final harvest, total above-ground biomass, yield and N and P concentrations were determined. Grain yield was determined by harvesting panicles from an area 9 m2 and grains separated from it. Sub-samples with known weight were dried at 70 8C to a constant weight. Dried weight of sub-samples are used to determine dry weight from the harvested area and then expressed as t ha1. Above-ground biomass at maturity was harvested by cutting plants just above the surface of the ground and fresh weight noted. Sub-samples with known fresh weight were taken for each replicate and dried to a constant weight at 70 8C. Above-ground biomass per hectare was then determined as in the case of grain yield. Soil-related modules were parameterized mainly with measured data from experiments carried out under optimal growth conditions, and from related literature, e.g. Fening et al. (2005) and Abekoe and Tiessen (1998). Disturbed and undisturbed soil samples which were taken in soil profiles (0–15, 15–30, 30–50, 50–75 and 75–100 cm) prior to sowing were analysed for organic carbon (OC%), labile P (as described in Tiessen and Moir, 1993), nitrate (NO3-N), ammonium (NH4-N), pH, bulk density and particle size distribution as described in Hoogenboom et al. (1999). P sorption capacities of soils were determined through inverse modeling with values within the limits of known boundary for the study area (Abekoe and Tiessen, 1998). 2.4. Experiments for model evaluation In order to evaluate the APSIM model, experiments were conducted in homestead fields (June 12 on one soil type; Regosol) and bush farms (on two different soil types: Regosol and Plinthosol) with the latter being planted on two different dates
Table 2 Mean monthly solar radiation, maximum and minimum temperature, and monthly total rainfall in 2005 at Navrongo, Ghana. Month
Solar radiation (MJ m2 day1)
Maximum temperature (8C)
Minimum temperature (8C)
Rainfall (mm)
January February March April May June July August September October November December
17.6 18.9 20.2 19.9 21.0 18.8 17.2 17.4 20.7 21.0 21.7 21.9
34.5 39.7 41.1 40.2 37.0 32.1 31.1 30.8 31.5 35.0 38.1 38.9
20.4 25.7 27.5 28.2 26.1 24.1 23.0 22.8 23.1 22.4 20.9 20.5
0 4.5 0 20.7 13.7 176.6 179 205.5 98.3 28.7 0 0
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
108
(June 12 and 26, 2005). The experiments were laid out in a randomized complete block design on each of the soil types and management systems. Four levels of inorganic N (0, 40, 80 and 120 kg ha1) in the form of ammonium sulphate and three levels of P (0, 30 and 60 kg ha1) in the form of triple super phosphate (TSP) were applied. Planting was at a density of 0.65 cm 0.25 cm. In the homestead, farmer’s fields were used for the study. Treatments were replicated seven times in the homestead fields, three times on the Regosol and four times on the Plinthosol in the bush farms for each planting date. Weeds were controlled manually using hoes twice before flowering. In the homestead, organic manure (N = 1.2%, P = 0.86%) was applied a day before sowing at a rate of 1 t ha1 (the average amount used in the study area). Soil samples were collected from soil profiles (Table 3) dug on the experimental sites. They were air-dried, sieved and analysed for organic carbon, pH, bulk density, plant wilting point field capacity and labile P (as in Tiessen and Moir, 1993). Ammonium and nitrate were also determined. Plants and soil samples were analysed as described in Hoogenboom et al. (1999). Sorghum phenological development (date of flowering), total biomass and grain yield were noted as described earlier (experiments for
model parameterization) to evaluate the performance of model. Final harvests for both grain and biomass yield were done on an area of 5.5 m2. Plant tissue N and P contents of above-ground biomass were also determined for two treatments in the homestead: (i) rain-fed condition, 120 N, 60P kg ha1 fertilizer and 1 t ha1 manure applied, (ii) supplementary irrigation, 120 N, 60P kg ha1 fertilizer and 3 t ha1 manure applied (this treatment was only for plant tissue analysis). Some of the soil parameters (from the experimental sites) used in evaluating the performance of the APSIM model are presented in Table 3. Experiments were also conducted only with manure (2, 4 and 6 t ha1) to evaluate model performance in utilising manure as a source of nutrients. 2.5. Data analysis 2.5.1. Evaluation of model performance The performance of the APSIM model in predicting grain yield was evaluated using the root mean square error (RMSE), the median unbiased absolute percentage error (MdUAPE), modified coefficient of efficiency (E1), correlation coefficient (r) and coefficient of determination (r2). Simulated and observed values
Table 3 Soil properties used for modeling sorghum yield in the study area. Soil parameters
Layer 1 (150 mma)
Homestead (Regosol) BD (g cm3) SAT (cm cm1) LL (cm cm1) DUL (cm cm1) Soil-C parameters Organic C (g 100 g1) Finertb Fbiomc Soil P parameter Labile P (mg kg1) P sorption (mg kg1) Bush farms (Regosol) BD (g cm3) SAT (cm cm1) LL (cm cm1) DUL (cm cm1) Soil-C parameters Organic C (g 100 g1) Finert Fbiom Soil P parameter Labile P (mg kg1) P sorption (mg kg1) Bush farms (Plinthosol) BD (g cm3) SAT (cm cm1) LL (cm cm1) DUL (cm cm1) Soil-C parameters Organic C (g 100 g1) Finert Fbiom Soil P parameter Labile P (mg kg1) P sorption (mg kg1)
2 (150 mma)
3 (200 mma)
4 (250 mma)
5 (250 mma)
1.54 0.353 0.054 0.231
1.53 0.357 0.094 0.219
1.62 0.369 0.106 0.212
1.63 0.341 0.161 0.219
1.64 0.338 0.130 0.197
0.58 0.35 0.02
0.56 0.40 0.02
0.45 0.50 0.01
0.37 0.80 0.01
0.32 0.80 0.01
21 79
6.2 150
5.7 150
3.2 200
1 200
1.56 0.352 0.046 0.203
1.58 0.321 0.096 0.209
1.56 0.320 0.110 0.205
1.58 0.372 0.122 0.209
1.56 0.246 0.139 0.195
0.39 0.35 0.015
0.36 0.35 0.01
0.32 0.60 0.01
0.37 0.80 0.01
0.32 0.80 0.01
15.0 50
5.2 75
5.0 120
2.0 180
1.0 200
1.59 0.353 0.049 0.201
1.61 0.357 0.099 0.203
1.56 0.369 0.119 0.200
1.58 0.341 0.129 0.197
1.56 0.338 0.130 0.199
0.40 0.35 0.015
0.37 0.35 0.01
0.23 0.60 0.01
0.25 0.80 0.01
0.32 0.90 0.01
11.5 50
5.8 75
5.0 120
BD: bulk density; SAT: volumetric water content at saturation. LL is wilting point and DUL is upper limit. a Layer thickness (mm). b Proportion of soil carbon assumed not to decompose. c Proportion of decomposable soil carbon in the more labile soil organic matter pool.
2.0 180
1.0 200
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
were also assessed for significant differences by using the Tukey test and Mann–Whitney test: X 2 0:5 RMSE ¼ ½n1 ðyieldsimulated yieldobserved Þ where n is the number of replicates in each planting date experiment. The MdUAPE is: jsimulatedi observedi j MdUAPE ¼ 100 median 0:5ðobservedi þ simulatedi Þ MdUAPE avoids problems such as bias in favor of lower prediction that occurs when using the regular MUdAPE (which does not use absolute differences between simulated and observed data) in expressing goodness of fit between predictions and observations (Armstrong and Collopy, 1992). The modified coefficient of efficiency E1 is defined as: Pn jobservedi simulatedi j E1 ¼ 1 Pni¼1 i¼1 jobservedi meanobserved j It was originally defined by Nash and Sutcliffe (1970). E1 values range from 1 to 1.0, with higher values indicating better agreement between model simulations and observations. An E1 value of zero indicates model performance is as good as the mean observed value of treatments. E1 = 1 implies a perfect fit for simulated and observed values. If E1 < 0.0, then the observed mean value is a better predictor than the model. The modified version of the squared difference terms are replaced by their respective absolute values, hence reducing the sensitivity of the coefficient to outliers as in the original coefficient (Legates and McCabe, 1999). 2.5.2. Scenario analysis APSIM has the capability to simulate long-term dynamics of soil water, organic matter, nutrients, crop growth and yield (Nelson et al., 1998) in response to management practices and weather conditions. In this study, APSIM was used to simulate sorghum grain yield in response to inorganic fertilizer (N and P) and also different residue management practices. These are seasonal removal of crop residues from the fields and retaining them in the soil. Soil chemical and physical parameters, initial soil conditions and agronomic information observed from field experiments were used as baseline information. The sorghum cultivar ICSV III was used in the simulations. To project (forecast) long-term yield trends for sorghum, the LARS-WG (Semenov and Brooks, 1999) climate simulation model was used to generate weather data by random re-sampling of 15-year historical weather data from the study site. The climatic inputs generated were daily solar radiation, maximum and minimum air temperatures and rainfall. To evaluate the sustainability of food production in the region, the following crop residue management scenarios were formulated:
annual removal of crop residues mineral fertilizer inputs; annual removal of crop residues + mineral fertilizer inputs; annual incorporation of crop residues mineral fertilizer inputs; annual incorporation of crop residues + mineral fertilizer inputs.
The model was applied in both management systems to simulate the long-term effects of the removal of crop residue compared to its incorporation into the soil on sorghum grain production and soil organic carbon. One ton per hectare of manure (N = 1.2%, P = 0.86%) was applied annually in the homestead. The organic manure was incorporated into the soil annually, after each first rain in May. Subsequently, the long-term implications of not using inorganic fertilizer in sorghum production in the study region were evaluated. To contrast the zero-fertilizer input
109
scenarios, a fertilization scheme of 80 kg N ha1 was used on the bush farm, 40 kg N ha1 on the homestead were used. Additionally, 30 kg P ha1 was applied in both locations for the simulation. These were the most economical fertilization practices (Kpongor, 2007). 3. Results 3.1. Model parameterization An average of 10.9 t ha1 of total biomass with a standard deviation of 0.32 was obtained under limited growth stress conditions from both planting date experiments. Grain yield was measured at 4.4 t ha1 with a standard deviation of 0.13. The total number of growing degree days taken from emergence to physiological maturity averaged at 1274 with a standard deviation of 15. Genetic coefficients developed using data set from the first planting date experiment obtained lower RMSE values for anthesis date, total biomass and grain yield (Table 4). It was therefore more appropriate for the cultivar and used for further modeling. 3.2. Evaluation of model performance Sorghum grain yield ranged from 0.50 t ha1 in the control on the bush farms (Plinthosol) to a maximum value of 4.53 t grains ha1 with the application of 120 kg N ha1 and 30 kg P ha1 on the homestead fields (Table 5). Murty et al. (1998) reported yields of between 3.85 and 5.34 t ha1 under good fertility (91, 45, 45 kg ha1 N–P–K respectively) and management conditions in Bagauda, Nigeria. Growing degree days from emergence to flowering for the various treatments are presented in Table 6. The ability of APSIM to reproduce observed sorghum phenology (anthesis), grain yield and total biomass was tested under different levels of mineral N and P applications on the homestead and bush farm soils. Also, the ability of the model to simulate N and P content of total biomass were assessed. The general trend of the growth duration (emergence to flowering) of sorghum in response to the different treatment of N and P fertilizer was reasonably well predicted by the model (Table 6). The model exaggerated the impact of nutrient stress in delaying crop phenology. This is expressed by the deviations between observations and predictions of GDDs at lower levels of input. In general, the model predicted the trend of total biomass production under the various N and P fertilizer treatment combinations well (Fig. 1). There was a good correlation between the observed and predicted total dry biomass values with an r value of 0.86. The model’s performance in predicting total biomass was good with an internal model efficiency coefficient of 0.50 and RMSE of 1.17 t ha1. Grain yield predictions in response to the various levels of inorganic N and P fertilizer applications were well predicted and better simulated for the homestead farms than for the bush farms (Fig. 2). MdUAPE values of 28 and 44% were obtained for the homestead and bush farm respectively. The RMSE of grain yield prediction in the homestead was lower than that in the bush farm on the Regosol (Table 7). The trend of sorghum grain yield in Table 4 RMSE values of selected growth and yield parameters using data from two planting dates. Parameter set
Grain (t ha1)
Biomass (t ha1)
Anthesis (days)
First planting date Second planting date
0.213a 0.281
0.284 0.297
14 14
a
RMSE values.
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
110
Table 5 Comparison of observed and simulated sorghum grain yields (t ha1) in response to P treatments with different levels of N in Navrongo, Ghana. Grain (t ha1)
N applied (kg ha1)
P applied (kg ha1)
B-R1
B-R2
H-R1
B-P1
B-P2
0.54 (0.59)a 0.70 (0.61) 0.63 (0.61)
0.88 (0.61) 0.84 (0.61) 0.81 (0.61)
1.3 (1.17) 1.67 (1.22) 1.43 (1.22)
0.51 (0.46) 0.48 (0.46) 0.61 (0.46)
0.50 (0.53) 0.75 (0.53) 0.82 (0.53)
0 30 60
0 0 0
0 30 60
40 40 40
1.14 (1.29) 2.01 (2.44) 2.42 (2.37)
1.35 (1.78) 2.48 (2.56) 2.88 (3.11)
2.78 (2.72) 2.79 (2.81) 2.78 (2.81)
1.14 (1.47) 1.65 (2.22) 1.81 (2.20)
1.17 (1.64) 2.03 (2.59) 2.42 (2.20)
0 30 60
80 80 80
2.44 (1.63) 3.15 (2.95) 3.36 (3.89)
2.68 (1.78) 3.44 (3.30) 3.56 (3.99)
3.65 (3.24) 3.83 (3.90) 3.89 (3.99)
2.33 (1.48) 3.02 (3.27) 3.20 (3.34)
2.20 (1.64) 2.44 (3.30) 3.58 (3.52)
0 30 60
120 120 120
2.37 (1.82) 3.45 (3.08) 3.77 (4.16)
2.67 (1.78) 3.62 (3.30) 3.57 (4.00)
3.81 (3.24) 4.53 (4.63) 4.36 (4.58)
2.41 (1.48) 3.65 (3.33) 3.78 (4.02)
2.94 (1.69) 3.34 (3.24) 3.68 (3.69)
H-R: homestead farms on Regosol; B-P: bush farms on Plinthosol; and B-R: bush farms on Regosol. (1, 2): First and second planting dates respectively. a Figures in parenthesis are simulated values.
Table 6 Comparison of observed and simulated duration of sorghum (ICSV III) growth from emergence to flowering date in two management systems and two soils expressed in growing degree days (GDD) in Navrongo, Ghana. Amount of fertilizer applied (kg ha1)
GDD emergence to flowering (8C days)
P
H-R1
N
a
B-R1
B-R2
B-P1
B-P2
0 30 60
0 0 0
1453 (1622) 1453 (1382) 1453 (1382)
1567 (1722) 1567 (1722) 1567 (1682)
1589 (1763) 1570 (1803) 1570 (1803)
1604 (1804) 1604 (1804) 1604 (1804)
1649 (1803) 1649 (1803) 1649 (1763)
0 30 60
40 40 40
1342 (1342) 1381 (1421) 1381 (1421)
1362 (1421) 1362 (1381) 1362 (1305)
1326 (1363) 1326 (1326) 1326 (1326)
1400 (1324) 1362 (1453) 1362 (1453)
1326 (1344) 1326 (1344) 1344 (1344)
0 30 60
80 80 80
1246 (1210) 1284 (1266) 1284 (1266)
1324 (1305) 1305 (1284) 1305 (1266)
1307 (1363) 1270 (1307) 1270 (1307)
1305 (1266) 1284 (1305) 1284 (1305)
1326 (1344) 1270 (1307) 1270 (1289)
0 30 60
120 120 120
1246 (1210) 1227 (1210) 1227 (1210)
1324 (1305) 1305 (1266) 1305 (1246)
1307 (1363) 1230 (1307) 1230 (1307)
1305 (1266) 1266 (1241) 1266 (1241)
1307 (1344) 1249 (1289) 1249 (1289)
H-R: homestead farms on Regosol. B-P: bush farms on Plinthosol, and B-R: Bush farms on Regosol. (1, 2): First and second planting dates respectively. a Figures in parenthesis are simulated values.
response to inorganic N and P was reasonably predicted by the model in both management systems with MdUAPE of 39%. Organic manure was also predicted with MdUAPE of 46%. The model simulated inorganic P application with zero N application on the homestead better than on the bush farm. Large deviations in both grain and biomass prediction were observed in the bush farm simulation at 80 kg N ha1 when no P was applied, a situation which could not be explained. This contributed to the lower efficiency of the model 0.59 and 0.56 on the Regosol and Plinthosol respectively in the bush farm. The overall predictions, for the bush farms on the Regosol were better than those on the Plinthosol (Table 7). Inorganic P fertilizer applications on both soils without applying N fertilizer did not result in increased simulated grain yield on the bush farms. This is due to the low total N content of the soils which was well below the amount required for optimum crop production. The coefficient of model efficiency was in decreasing order, homestead > bush farm Regosols > bush farm Plinthosols with the least value being 0.56. This implies the model is credible in predicting the response of sorghum grain production to mineral N and P fertilizer applications. Considering the number of default sorghum model values used (Table 8), the internal predictive efficiency of the model could possibly be further improved by including other cultivar specific parameters such as, leaf area development measurements. Photoperiod slope of 0.01 was used
in order to eliminate any effect of photoperiod as the hybrid used in the study is insensitive to photoperiod (Murty et al., 1998). There was a good agreement between observed and simulated N and P content of total biomass under both rain-fed conditions with
Fig. 1. Comparison of measured and predicted total biomass yield of sorghum grown in both the homestead and bush farms under a range of N and P fertilizer applications in Navrongo, a semi-arid region of Ghana (data from first planting date was used).
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
111
Fig. 2. Comparison of measured and predicted sorghum grain yield grown in both the homestead and bush farms under a range of N and P fertilizer applications in Navrongo, a semi-arid region of Ghana (data from first planting date was used).
1 t ha1 manure applied and under supplementary irrigation with 3 t ha1 manure, each with 120 kg N and 30 kg P applications. APSIM however exaggerated N uptake particularly in early growth stages of plants, indicated by higher deviations of measured values from simulated ones. Though higher amount of manure was applied under supplementary irrigation experiment, amounts of N and P retained in plant tissues as simulated by the model were similar (Fig. 3). This could be due to the fact that the amount of inorganic N and P applied as well as the 1 t ha1 manure applied provided close to the optimum nutrient requirements in the study area. 3.3. Comparison of the impact of crop residue management on sorghum grain production Rainfall is an important and most variable weather parameter. The in-crop total rainfall amount over simulation period was
relatively stable with a CV of 19% (Fig. 4). In both management systems, simulated sorghum grain yield fluctuated over the simulation period (30 years) with a trend of yield decline in scenarios without application of mineral fertilizer (Fig. 5a and b). In contrast, sorghum yields remained relatively stable over the 30-year simulation period when mineral fertilizer was applied. But there was stronger fluctuation in yields from season to season. In both management systems, incorporating crop residues resulted in significant yield increases irrespective of applying mineral fertilizer. This was accompanied by an increase in the soil organic carbon content (0–15 cm) over the simulation period in response to the retention of crop residues (Fig. 6). The model also indicated that continuous removal of crop residue in the homestead, even with the application of 40 and 30 kg ha1 mineral N and P respectively over the 30 years period would result in declining soil organic carbon to levels close to what currently pertains in the bush farms (Fig. 6).
Table 7 Performance of APSIM to predict sorghum grain yield response to inorganic fertilizer. Location Homestead Bush Regosolsa Bush Plinthosolsa Overall a
RMSE (t ha1)
MdUAPE (%)
E1
R2
r
84 72 96
0.35 0.51 0.60
28 44 45
0.73 0.59 0.56
0.92 0.62 0.53
0.96 0.79 0.73
252
0.50
39
0.64
0.66
0.81
n
Combined data from the two planting dates.
112
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
Table 8 Model parameters of Sorghum used in simulations; values and sources. Parameter Thermal time accumulation Duration – end of juvenile to panicle initiation Duration – flag leaf to flowering stage Duration, flowering to start of grain filling Duration, flowering to maturity Duration – maturity to seed ripening Shoot lag (time lag before linear coleoptile’s growth starts) Day length photoperiod to inhibit flowering Day length photoperiod for insensitivity Photoperiod slope Plant height (max) Grain water content Base temperature Optimal temperature
Source
Value
Units
C C C C L D
280 231 59 650 1 15
8C day 8C day 8C day 8C day 8C day 8C day
D D L O O L D
12.3 14.6 0.01 2100 0.150 8 30
h h 8C/h mm g/g 8C day 8C day
L: literature; D: default value; C: calibrated; O: observed.
Fig. 4. Fluctuations in annual total in-crop rainfall amounts over simulation period.
Application of 40 and 30 kg ha1 mineral N and P respectively in the homestead fields with crop residue incorporation produced yields that were similar to those produced on the bush farm with 80 and 30 kg ha1 mineral N and P respectively over the simulation period (Fig. 7). The effect of incorporating crop residue on grain yield on the bush farm appeared only after approximately 6 years
of continuous residue retention whereas it showed up much earlier in the homestead (Fig. 5). 4. Discussion 4.1. Modeling sorghum growth and grain yield The P modules of other models such as CENTURY (Metherell et al., 1993) were developed for soils with mainly N-limiting
Fig. 3. Observed and simulated tissue nutrient (N and P) content of sorghum aboveground biomass in the homestead. (a) Rain-fed condition, 120 N, 60 kg P ha1 fertilizer and 1 t ha1 manure applied with no supplementary irrigation. (b) Rainfed with supplementary irrigation, 120 N, 60 kg P ha1 fertilizer and 3 t ha1 manure applied.
Fig. 5. Simulated effect of farmers’ management practice (crop residue removal) on the long-term dynamics of sorghum grain production in the bush farms (a) and homestead (b) Navrongo, a semi-arid region of Ghana.
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
Fig. 6. Simulated effects of farmers’ residue management practices on the long-term dynamics of soil organic carbon (0–15 cm) in the bush farm (a) and homestead (b) Navrongo, a semi-arid region of Ghana.
conditions where P dynamics have minor impact on crop growth (Gijsman et al., 1996). APSIM, however, appeared to be more suitable for this study area since it was developed with data mainly from the semi-arid tropics where resource poor smallholder farmers are faced with P limiting conditions in soils (Abekoe and Tiessen, 1998). Sahrawat et al. (2000) indicated a reduction of plant response to P as a result of the high P sorption capacity of soils and also observed that zinc deficiency in soils hindered responses of maize to P application. The study area is characterized with soils low in P sorption capacity and no study on zinc deficiency has been reported for the area. Abekoe and Tiessen (1998) also reported that the presence of lateritic soil nodules reduces P response by serving as a sink for P, thereby not making it available to crops. This was not an issue for this study as lateritic soil nodules were hardly present in the soils within 0–60 cm depths. Undoubtedly, N is the nutrient that limits crop production most in the study area as clearly visible in grain yield (Table 5). However, its efficiency is reduced by low P status of the soil. These two nutrients have been noted to be most limiting to crop production (Braimoh, 2003). When farmers are faced with the option of selecting one nutrient to focus on, based on their limited resources, it would be rational to go for N fertilizers as they give more returns. The predictive performance of APSIM for grain yield was high with an internal model efficiency of 0.64. An inaccurate phenology prediction predisposes essential physiological growth processes to be wrongly timed with the wrong dates possibly coinciding with adverse weather conditions. Sorghum phenological development was generally well predicted. Delayed time to flowering observed under low input (no inorganic N and P application) condition in
113
Fig. 7. Seasonal variability of sorghum grain yield under the different scenarios over the simulation period (30 years) in both homestead and bush farms. Each box in the graph shows the distribution of grain yield over the simulation period. The boundary of the box closest to zero indicates the 25th percentile, the broken line within the box marks the mean, the solid one the median, and the upper boundary of the box indicates the 75th percentile. Whiskers above and below the box indicate the 95th and 5th percentiles. CR: crop residue, B: bush farm, H: homestead farms, (a) no fertilizer application, (b) 80, 30 kg ha1 N, P for bush farm (B) and 40, 30 kg ha1 N, P for homestead (H) Navrongo, a semi-arid region of Ghana.
both the homestead and bush farms were none-the-less exaggerated by the model. This suggests the need for N and P stress factors influencing phenology to be improved with considerations for highly weathered low input soils. Low grain yield prediction by the model at low rate of fertilizer use was also reported for maize by Zingore (2006) in his study on smallholder farms in semi-arid Zimbabwe. This was attributed to poor internal nutrient use efficiency of the maize module of APSIM at low soil N contents. APSIM has also been reported to have performed creditably in predicting soil organic carbon dynamics in the semi-arid region of Africa (Micheni et al., 2004). The tissue N and P contents of total biomass were also well predicted in this study. This implies a good performance by the model in predicting soil processes such as soil water dynamics which are critical for nutrient transport from soil to plant rooting zone as well as N transformations. The high (0.64) overall modified internal coefficient of efficiency in predicting sorghum grain and total biomass production and the ability of APSIM to simulate tissue N and P contents, provides sufficient precision for a comparative evaluation of the long-term impact of farmers’ management practices on future grain yield production.
114
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
4.2. Implications of long-term sorghum grain yield predictions for future food security in the study region Although farmers are concerned about the deteriorating soil fertility of their fields, the driving force to adopting any soil fertility strategy is that of food security and growth in income. Increasing the fertility of their soils only comes as a by-product of these objectives. This was also reported by Snapp et al. (2002). This implies that the decreasing trend of soil organic matter as shown by the model (this needs to be proven with measured data) per se´ may not provide enough incentives for farmers to incorporate crop residues. Rather, the evidence of grain yield decline with the continuous reduction of soil carbon content which is, in part, a result of the continuous removal of crop residues. Incorporating crop residue would also require additional labour. Only farmers with sufficient labour stand to benefit from soil fertility improvements through incorporating crop residue (Snapp et al., 2002). Another draw back to retaining crop residue in the soil is their demand as fodder in the dry season and as fuel wood. It is worth noting that the high yields obtained from this study is due to the high yielding cultivar that was used as well as fertilizer management (two split applications) and weed control. This is however not yet widely adopted by farmers due to some of its characteristics such as the non-dropping nature of the panicles that makes it easy for birds to perch and feed on the seeds. Hence, additional labour is required to scare-off birds after the grainfilling stage. Given a population growth rate of 2.4% per annum (Ghana Statistical Services, 2002), and the fact that current levels of grain yield do not meet the current demand, any further reduction in yield, as indicated by the predicted negative yield trend in the absence of fertilizer application and with crop residue removal, poses a great threat to food sufficiency in the region. A negative trend in food production in the area between 1996 and 2000 was indeed reported by Braimoh (2003). Even under favorable climatic conditions, continuous adequate yields cannot be attained on poor soils (Ogunkunle, 1993) without investment in external inputs, particularly inorganic fertilizer as indicated by this study. The model highlighted the critical influence of both inorganic fertilizer and crop residue on reducing the declining trend and temporal variability in sorghum grain production. Simulating 30 years of farmers’ practice highlights the critical role of mineral fertilizer in maintaining and increasing soil organic carbon on these sandy soils (71% sand content). Scenario analysis using simulation modeling offers estimates of the long-term effects of farmers’ practices on the resource base (SOC) and its consequent effects on sorghum grain production and hence, on food security. Model outputs can also serve as inputs to longterm economic analysis of the management practices and thus, as a decision support tool for policy makers in the agricultural sector. SOC in the homestead decreased over the simulation period (30 years) to levels that currently pertain in the bush farm where the practice of crop residue removal (scenario) had been the norm for decades, thus increasing the credibility of model outputs. Similar trends of SOC decline were reported by Zingore (2006) using the FARMSIM model to simulate SOC content of virgin soils (sandy) that had been put under cultivation. More research is however needed to validate the long-term grain production and soil carbon dynamics in the study area to underpin current results. When farmers are faced with limited amounts of fertilizer, it will be more rational to invest it in the homestead fields rather than on fields in the bush farms. However, given the increasing demand for grains and the limited area with relatively fertile fields in the homesteads, mineral fertilizer will need to be used in the bush fields as well.
4.3. Contribution of crop residue to grain production APSIM simulations do reflect benefits of crop residue retention in grain yield over long-term period. Fig. 5a and b shows differences in grain yield between retaining and removing residues. Except for the bush farm where no fertilizer was applied, the differences were marginal. Over the simulation period, an average of about 18 and 7.2 kg ha1 N and P respectively were potentially available annually from crop residue for plant growth in the homestead fields, for instance. This was reflected in a consistently higher NO3 and NH4 content of soils where crop residues were incoporated. Without applying mineral fertilizer, a mean of 4.72 and 2.08 kg ha1 of nitrate and ammonia respectively were available daily over simulation period with the removal of crop residue. On the other hand, 6.37 and 3.72 kg ha1 nitrate and ammonia respectively were available when residue was incorporated. When 40 kg N ha1 was applied, a daily mean of 12.91 and 11.75 nitrate and ammonia were respectively available in the soil. Effects of crop residues were also higher with the application of mineral fertilizer in both management systems. This may be explained by the high C:N ratio of sorghum residue, hence requiring external N input to overcome microbial immobilization of N. It may also account for the higher effect of crop residues in the homestead as compared to the bush farms. Inputs to SOC from root biomass contributed little to the SOC as carbon derived from root biomass are described as highly labile (Balesdent and Balaane, 1992) and hence, have high turnover rate with most of it entering the active pool of SOC. Though crop residues help in reducing or preventing soil erosion through surface runoff, the contribution of crop residues to reducing surface runoff in this study was marginal, probably because crop residues were incorporated rather than used as mulch. There is, therefore, the need to encourage the partial retention of crop residues. This could be made possible by helping farmers with know-how to establish woodlots with fast growing tree species. This will help in reducing dependency on crop residues for fuel wood and fencing, reasons for which Singh et al. (2000) challenges the practicability of retaining them in the soil in these environments. No matter how effectively other problems hindering crop production are addressed, per capital food production will continue to decline unless issues such as crop residue management and others related to soil fertility are effectively addressed (Sahrawat et al., 2000). Seasonal variability in grain yield was consistently higher with the removal of crop residues both in the bush and homestead fields when no fertilizer was applied. This suggests a more stable grain production with the incorporation of crop residues (Fig. 7a). Similarly, applying fertilizer also reduced temporal variability in grain yield from a CV of 25–10% for bush farm when residue was incoporated. This implies inorganic fertilizer and organic carbon (manure and crop residues) serve to increase the resilience of the resource base (soil) and temper grain yield variability. Breman et al. (2007) in their synthesis on soil nutrient depletion in SubSaharan Africa also mentioned depletion of SOC due to the constant removal of crop residues. This is confirmed by Lithourgidis et al. (2006) in their study in Greece which reported stability in winter wheat yield over 25 years of continuous cultivation. This they attributed to the annual application of inorganic fertilizer as well as retaining crop residue in the soil after each harvest. 5. Conclusions APSIM provided a flexible modeling environment to configure a set of modules from a collection of crops, soil and management options to suit the semi-arid environment of northern Ghana. It successfully captured the effects of inorganic nitrogen and
D.S. MacCarthy et al. / Field Crops Research 113 (2009) 105–115
phosphorus fertilizer applications on grain and biomass yield of sorghum on bush and homestead fields in the study area. Even with the current level of manure application on the homestead fields, the SOC content in the topsoil (0–15 cm) declined over the simulation period to levels close to the current levels on the bush farms. Thus, land use activities that contribute to SOC are critical to the future of crop production, even with inorganic fertilizer applications. APSIM thus provided a scientifically sound prediction of sorghum grain yields in response to inorganic fertilizer use and crop residue management under contrasting farmers’ management systems. Incorporating crop residues from the preceding season improved grain yield in both the homestead fields and the bush farms. The current residue management practices of farmers in the study area cannot ensure a sustainable sorghum production and for that matter food production, hence threatening future food security unless measures are taken to address the improvement of organic matter and nutrient contents of these soils. Acknowledgements I am very grateful to the GLOWA-Volta project and the Challenge Program on Food and Water for providing me with funds without which this study would not have been possible. I am also grateful to the APSIM group for their support and for providing me with a free license to the APSIM version 4.0 which was used for the study. I also acknowledge the efforts of Dr. A.M. Manschadi for proof reading this manuscript. References Abekoe, M.K., Tiessen, H., 1998. Fertilizer P transformations and P availability in hillslope soils of Northern Ghana. Nutr. Cycl. Agroecosyst. 52, 45–54. Armstrong, J.S., Collopy, F., 1992. Error measures for generalizing about forecasting methods: empirical comparisons. Int. J Forecast. 8, 69–80. Balesdent, J., Balaane, M., 1992. Maize root-derived soil organic estimated by 13C natural abundance. Soil Biol. Biochem. 24, 97–101. Bationo, A., Mokwunye, U., Vlek, P.L.G., Koala, S., Shapiro, B.I., 2003. Soil fertility management for sustainable land use in the West African Sudano-Sahelian Zone. In: Gichuri, M.P., Bationo, Andre´, Bekunda, M.A., Goma, H.C., Mafongoya, P.L., Mugendi, D.N., Murwuira, H.K., Nandwa, S.M., Nyathi, P., Swift, M.J. (Eds.), Soil fertility management in Africa: A regional perspective, African Academy of Sciences Centro Internacional de Agricultura Tropical (CIAT); Tropical Soil Biology and Fertility (TSBF). Academic Science Publishers, Nairobi, Kenya, pp. 253–292. Braimoh A.K., 2003. Modeling land-use change in the Volta Basin of Ghana. PhD Dissertation. University of Bonn Germany. Ecology and Development Series No. 14. Breman, H., Fofana, B., Mando, A., 2007. The lessons of Drente’s ‘ESSEN’: soil nutrient depletion in Sub-Saharan Africa and the management strategies for soil replenishment. In: Braimoh, A.K., Vlek, P.L.G. (Eds.), Land Use and Soil Resources. Springer, The Netherlands, pp. 145–166. De Jager, A., Onduru, D., Walaga, C., 2003. Using NUTMON to evaluate conventional and low external input farming practices in Kenya and Uganda. In: Boutkes, S.T.E., Wopereis, M.C.S. (Eds.), Decision Support Tools for Smallholder Agriculture in Sub-Saharan Africa: A Practical Guide. IFDC/ACP-EU Technical Centre for Agricultural and Rural Cooperation (CTA), Muscle Shoals, USA/Wageningen, The Netherlands, pp. 44–53. Fening, J.O., Adjei-Gyapong, T., Yeboah, E., Ampontuah, E.O., Quansah, G., Danso, S.K.A., 2005. Soil fertility status and potential organic inputs for improving smallholder crop production in the interior Savannah zone of Ghana. J. Sustain. Agric. 25 (4), 69–92. Franzen, D.W., Hopkins, D.H., Sweeney, M.D., Ulmer, M.K., Halvorson, A.D., 2002. Evaluation of soil survey scale for zone development of site-specific nitrogen management. Agron. J. 94, 381–389. Ghana Statistical Services, 2002. 2000 Population and housing census. Special Report on 20 largest localities by local authorities. Ghana Statistical Service, Accra, Ghana. Gijsman, A.J., Oberson, A., Tiessen, H., Friesen, D.K., 1996. Limited applicability of the CENTURY Model to highly weathered tropical soils. Agron. J. 88, 894–903.
115
Henao, J., Baanante, C.A., 1999. Estimating rates of nutrient depletion in soils of agricultural lands in Africa. Technical Bulletin International Fertilizer Development Center (IFDC)-T-48. IFDC, Muscle Shoals, AL. Hoogenboom, G., Wilkens, P.W., Tsuji, G.Y. (Eds.), 1999. DSSAT v3, vol. 4. University of Hawaii, Honolulu, HI, pp. 234–235. Jones, C.A., Kiniry, J.R., 1986. CERES-Maize. A Simulation Model of Maize Growth and Development. Texas A&M Univ. Press. 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 of APSIM, a model designed for farming systems simulation. Eur. J. Agron. 18, 267– 288. Kpongor, D.S., 2007. Spatially explicit modeling of sorghum (Sorghum bicolor (L) Moench) production on complex terrain of a semi-arid region of Ghana using APSIM. PhD Dissertation. University of Bonn, Germany. Ecology and Development Series, No. 51. Legates, D.R., McCabe, G.J., 1999. Evaluating the use of ‘goodness-of-fit’ measures in hydrologic and hydro-climatic model validation. Water Resour. Res. 35, 233– 241. Lithourgidis, A.S., Damalas, C.A., Gagianas, A.A., 2006. Long-term yield patterns for continuous winter wheat cropping in northern Greece. Eur. J. Agron. 25, 208– 214. Metherell, A.K., Harding, L.A., Cole, C.V., Parton, W.J., 1993. CENTURY soil organic model environment. Technical documentation agroecosystem version 4.0. Great Plains System Res. Unit, Tech Rep. 4. USDA-ARS, Fort Collins, CO. Micheni, A.N., Kihanda, F.M., Warren, G.P., Probert, M.E., 2004. Testing the APSIM model with experimental data from the long-term manure experiment at Machang’a (Embu), Kenya. In: Delve, R.J., Probert, M.E. (Eds.), Modeling Nutrient Management in Tropical Cropping Systems. ACIAR Proceedings No. 114, pp. 110–117. Murty, D.S., Nwasike, C.C., Atokple, I.D.K., 1998. Registration of ‘ICSV III’ sorghum cultivar. Crop Sci. 38, 1716–1717. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I—discussion of principles. J. Hydrol. 10 (3), 282–290. Nelson, R.A., Dimes, J.P., Silburn, D.M., Paningbatan, E.P., Cramb, R.A., 1998. Erosion/ productivity modeling of maize farming in the Philippine uplands. Part II. Simulation of alternative farming methods. Agric. Syst. 58 (2), 147– 163. Ogunkunle, A.O., 1993. Soil in land suitability evaluation: an example with oil palm in Nigeria. Soil Use Manage 9, 37–42. Ritchie, J.T., 1998. Soil water balance and plant water stress. In: Tsuji, G.Y., Hoogenboom, G., Thornton, P.K. (Eds.), Understanding Options for Agricultural Production. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 41–54. Rowe, E.C., van Wijk, M.T., de Ridder, N., Giller, K.E., 2006. Nutrient allocation strategies across a simplified heterogeneous African smallholder farm. Agric. Ecosyst. Environ. 116, 60–71. Sahrawat, K.L., Abekoe, M.K., Diatta, S., 2000. Application of inorganic phosphorus fertilizer. In: Tian, G., Ishida, F., Keatinge, J.D.H. (Eds.), Sustaining Soil Fertility in West Africa Soil Science Society of America (SSSA) Special Publication No. 58. SSSA/ASA, Madison, WI, USA, pp. 225–245. Semenov, M.A., Brooks, R.J., 1999. Spatial interpolation of the LARS-WG stochastic weather generator in Great Britain. Climate Res. 11, 137–148. Singh, U., Diels, J., Henao, J., Breman, H., 2000. Decision support systems for improving the application of integrated nutrient management technologies. In: Tian, G., Ishida, F., Keatinge, J.D.H. (Eds.), Sustaining soil fertility in West Africa Soil Science Society of America (SSSA) Special Publication No. 58. SSSA/ ASA, Madison, WI, USA, pp. 305–321. Smalberger, S.A., Singh, U., Chien, S.H., Henao, J., Wilkens, P.W., 2006. Development and validation of a phosphate rock decision support system. Agron. J. 98, 471– 483. Snapp, S.S., Rohrbach, D.D., Simtowe, F., Freeman, H.A., 2002. Sustainable soil management options for Malawi: can smallholder farmers grow more legumes? Agric. Ecosyst. Environ. 91, 159–174. Tiessen, H., Moir, J.O., 1993. Characterization of available P by sequential extraction. In: Carter, M.R. (Ed.), Soil Sampling and Methods and Analysis. Canadian Society of Soil Science (Lewis Publishers), Boca Raton, FL, pp. 75–86. Van Ittersum, M.K., Howden, S.M., Asseng, S., 2003. Sensitivity of productions and deep drainage of wheat cropping systems in a Mediterranean environment to changes in CO2, temperature and precipitation. Agric. Ecosyst. Environ. 97 (1– 3), 255–273. Vlek, P.L.G., Ku¨hne, R.F., Denich, M., 1997. Nutrient resources for crop production in the tropics. Phil. Trans. R. Soc. Lond. B 352, 975–985. Zingore, S., 2006. Evaluation of resource management options for African smallholder farms using an integrated Modelling approach. In: Exploring diversity within smallholder farming systems in Zimbabwe: Nutrient use efficiencies and resource management strategies for crop production. PhD Dissertation. Wageningen University, Wageningen, The Netherlands, pp. 135–160.