Analysis of climate signals in the crop yield record of sub-Saharan Africa.

Food security and agriculture productivity assessments in sub-Saharan Africa (SSA) require a better understanding of how climate and other drivers influence regional crop yields. In this paper, our objective was to identify the climate signal in the realized yields of maize, sorghum, and groundnut in SSA. We explored the relation between crop yields and scale-compatible climate data for the 1962-2014 period using Random Forest, a diagnostic machine learning technique. We found that improved agricultural technology and country fixed effects are three times more important than climate variables for explaining changes in crop yields in SSA. We also found that increasing temperatures reduced yields for all three crops in the temperature range observed in SSA, while precipitation increased yields up to a level roughly matching crop evapotranspiration. Crop yields exhibited both linear and nonlinear responses to temperature and precipitation, respectively. For maize, technology steadily increased yields by about 1% (13 kg/ha) per year while increasing temperatures decreased yields by 0.8% (10 kg/ha) per °C. This study demonstrates that although we should expect increases in future crop yields due to improving technology, the potential yields could be progressively reduced due to warmer and drier climates.

Statistical-dynamical modeling of the maize yield response to future climate change in West, East and Central Africa using the regional climate model REMO.

Africa is vulnerable to the impacts of climate change, particularly in terms of its agriculture and crop production. The majority of climate models project a negative impact of future climate change on crop production, with maize being particularly vulnerable. However, the magnitude of this change remains uncertain. Therefore, it is important to reduce the uncertainties related to the anticipated changes to guide adaptation options. This study uses a combination of local and large-scale empirical orthogonal function (EOF) predictors as a novel approach to model the impacts of future climate change on crop yields in West, East and Central Africa. Here a cross-validated Bayesian model was developed using predictors derived from the regional climate model REMO for the period 1982-2100. On average, the combined local and large-scale EOF predictors explained around 28 % of maize yield variability from 1982 to 2016 of the entire study regions. Notably, climate predictors played a significant role in West Africa, explaining up to 51 % of the maize yield variability. Large-scale climate EOF predictors contributed most to the explained variance, reflecting the role of regional climate in future maize yield variability. Under a high-emissions scenario (RCP8.5), maize yield is projected to decrease over the entire study region by 20 % by the end of the century. However, a minor increase is projected in eastern Africa. This study highlights the importance of incorporating climate predictors at various scales into crop yield modeling. Furthermore, the findings will offer valuable guidance to decision-makers in shaping adaptation options.