Factoring Climate Variability and Change into Crop Models for Enhancing Sorghum Performance in the West African Semi-Arid Tropics

Abstract

In semi-arid West Africa, there is growing concern about the potential impacts of climate change and variability on the productivity of agricultural crops and hence food security. This region is highly vulnerable to variable rainfall pattern (amount, distribution and onset of growing season) and is also characterized by smallholder farmers practicing rain-fed farming activities. Thus, assessing the magnitude of impacts of climate change and variability on crop yields remain challenging due to the confounding influence of other more potent drivers of climate change. Studies were conducted on the potential of factoring climate variability and change into crop models for enhancing sorghum performance in the West African semi-arid. The objectives were to (i) compare four (4) known methods for the determination of the onset dates of rainfall and in relation to farmer‘s sowing window, establish the most suitable onset dates (OGS) and length growing season (LGS) across agroecological zones of Mali, (ii) from on-station field experiments , evaluate the effects of sowing date on the physiology and yield of ten (10) sorghum genotypes, (iii) compare and validate three process-based crop models against independent field scale data sets of the tested sorghum genotypes under non-limiting water and nutrient conditions, and (iv) to simulate the sensitivity of the crop models to current climate system (1980-2009) and near future climate scenario (2040-2069) and to propose possible adaptation measures. Long-term daily rainfall records were applied to four (4) known methods for the determination of the onset dates of rainfall in order to determine the most suitable method to estimate the onset date of growing season (OGS) and length of growing season (LGS) that fitted into farmer‘s planting window for major cereal crops for selected weather stations for each agroecological zone (AEZ) of Mali. The LGS was estimated using probability distribution for each zone and evaluated with the duration to maturity for selected crop genotypes according to FAO statistics. The hypothesis used also considered a time lag minimum of 7 days between the mean onset date and traditional farmers‘ sowing dates for the major cereal crops. The most suitable method used to estimates OGS was based on early, normal and late dates across some stations in Mali in order to account for inter-annual variability. Also, the performance of three process-based crop models (APSIM, DSSAT, and Samara) were compared for the simulation of growth and yield of contrasting sorghum genotypes (CSM63E ‗Jakunben‘, CSM335 ‗Ceblen‘, Fadda and IS15401 ‗Soumalenba‘ respectively) relative to the crop growth parameters and yield observations obtained from field trial conducted in the year 2013 at ICRISAT-Mali. The genotypes were calibrated and validated using the process-based models against the impacts of climatic risks over six (6) selected sites across West African semi-arid region for current climate system (1980-2009) and near future climate scenarios (2040-2069) using five (5) Global Climate Models (GCMs) projections at 571ppm CO2 concentration.The results show that of the methods for the determination of onset of rain (OGS) evaluated across AEZ of Mali, Def_4 was the most suitable for predicting OGS at Sahelian and Sudano-sahelian zones while Def_3 was found suitable for Sudanian and Guinea savanna zones. These methods (Def_3 and Def_4) exhibited superior capacity relative to the farmer‘s average planting date. They were found applicable to cropping system, for prevention of false OGS and crop failure, and ensured appropriate choice of crop variety according to LGS under climate variability and change currently being experienced in West Africa (Mali inclusive). The results further showed that the crop models reproduced the performance of the tested diverse photoperiod sensitivite sorghum varieties. Thus, the crop model simulation of phenology during calibration and validation were shown to be within the close range of observed values with the lowest RMSE, normalized RMSE (%) and strong coefficients of determination (R2). The best performance regarding total leaf number (TLN) estimation was Samara, which had lowest RMSE value of 1.3 leaves, followed by DSSAT (2 leaves) and APSIM (2.2 leaves). The model overestimated the leaf area index (LAI) across the cultivars though Samara was found as the best simulated, this indicated the strength of organogenesis modelling approach in accurate simulation of LAI. Differences which were found in the simulated yields are attributable to the contrasting ways for grain yield simulation, despite similar parameterization for simulating potential total biomass. On average, DSSAT and Samara showed the best performance for grain yield with the lowest RMSE values (814 and 874 kg/ha) and normalized RMSE (38.8 and 41.7 %). In terms of total biomass, DSSAT shows the best performance with lowest RMSE value (2586 kg/ha), good normalized RMSE (28.7 %) and strong R2 (0.8), followed by estimates from APSIM and Samara. Climate change projections based on GCM‘s indicate disparities in the mean growing season as well as the magnitude of precipitation across the chosen sites, but high agreement on warming (elevated temperatures) was observed. On the sensitivity of the current system to climate change, the average sorghum yields show decline relative to the baseline yields in the absence of adaptation measures. However adoption of adaptation options such as lengthened crop duration and growing degree days(GDD) indicates significant increase in grain yields for CSM335 (local guinea landrace) and Fadda (improved hybrid) compared to CSM63E and IS15401 genotypes.