Uncertainties in Simulating Crop Performance in Degraded Soils and Low Input Production Systems

Abstract

Many factors interact to determine crop production. Cropping systems have evolved or been developed to achieve high yields, relying on practices that eliminate or minimize yield reducing factors. However, this is not entirely the case in many developing countries where subsistence farming is common. The soils in these countries are mainly coarse-textured, have low water holding capacity, and are low in fertility or fertility declines rapidly with time. Apart from poor soils, there is considerable annual variability in climate, and weeds, insects and diseases may damage the crop considerably. In such conditions, the gap between actual and potential yield is very large. These complexities make it difficult to use cropping system models, due not only to the many inputs needed for factors that may interact to reduce yield, but also to the uncertainty in measuring or estimating those inputs. To determine which input uncertainties (weather, crop or soil) dominate model output, we conducted a global sensitivity analysis using the DSSAT cropping system model in three contrasting production situations, varying in environments and management conditions from irrigated high nutrient inputs (Florida, USA) to rainfed crops with manure application (Damari, Niger) or with no nutrient inputs (Wa, Ghana). Sensitivities to uncertainties in cultivar parameters accounted for about 90% of yield variability under the intensive management system in Florida, whereas soil water and nutrient parameters dominated uncertainties in simulated yields in Niger and Ghana, respectively. Results showed that yield sensitivities to soil parameters dominated those for cultivar parameters in degraded soils and low input cropping systems. These results provide strong evidence that cropping system models can be used for studying crop performance under a wide range of conditions. But our results also show that the use of models under low-input, degraded soil conditions requires accurate determination of soil parameters for reliable yield predictions.