Effects of Crop Model Complexity on Sorghum Yield Projections Under Climate Change in Data Scarce Environments (Koutiala, Mali).

In sub-Saharan Africa, there is significant concern about the potential impacts of climate change and future variability on agricultural production. Assessing the magnitude of such impacts on crop yields in the future remains challenging due to the confounding influence of other more potent drivers of change. In this study, we discuss yield projections generated under the Agricultural Model Intercomparison and Improvement Project (AgMIP) for a distribution of farms and agricultural practices representative of the Koutiala district in Mali’s cotton belt. Our analysis focuses on two local sorghum genotypes (CSM335 ‘Ceblen’ and CSM63E ‘Jakunben’) calibrated in four contrasted process-based models (APSIM, DSSAT, Samara, SarraH) themselves driven by the climate projections of five GCMs for the mid-century period (2040-2069) at 571ppm CO2 concentration. Climate change projections used indicate a consistent pattern of increase in the mean growing season temperatures and little or no change in precipitation with high agreement, which is expected for temperature but unusual for precipitation in that region. On the average, sorghum yields decline in the absence of any adaptation measures. Simple plant adaptation options such as lengthened crop duration, altered root distribution are usually sufficient to offset this decline. Simpler models such as SarraH perform better at mimicking the observed distribution of yields than complex ones such as APSIM in data scarce environments. Randomization procedures required allocating farms across a discrete and limited sample of soil types and fertility management practices can however have a very significant effect on simulated yield distributions and on the uncertainty associated with yield projections.