Rice yield response forecasting tool (YIELDCAST) for supporting climate change adaptation decision in Sahel

•We used gene-expression programming (GEP) with weather data to develop YIELDCAST tool.•YIELDCAST was forced with GCMs downscaled outputs to forecast upland rice yields.•GEP is capable to downscale climate variables and forecast rice yield in Sahel region.•Increase in temperatures and decrease in rains will hinder yields to increase.•Yield forecasts suggested no major yield increase, unless measures are taken. Rice yield responses forecast (YIELDCAST) is a very useful decision support tool in climate adaptation in Sahel, where crops are purely rainfed climate-stressors sensitive. This study aims to construct upland rice yield responses forecasting algebraic formulation code referred as YIELDCAST by using gene-expression programming (GEP) based on observed rainfall and temperatures data (1979–2011), and forcing with global climate model (GCM) downscaled outputs under CO2 emission scenarios SR-A1B, A2 and B1 (2012–2100) over Bobo-Dioulasso, a Sahelian region. Statistically, GEP is a capable tool to downscale climate variables in the region (R = 0.746−0.949), and construct reliable rice YIELDCAST tool (R = 0.930; MSE = 0.037 ton/ha; MAE = 0.155 ton/ha, RSE = 0.137 ton/ha). Yields forecasted (2012–2100) showed a noticeable statistically significant difference between scenarios; however, fluctuating with no substantial increase (average below 1.60 ton/ha); suggesting that the increase observed in temperatures and decrease in rains will either reduced or hindered yield to largely increase in Sahel. With no such YIELDCAST tool to support adaptation decision, Sahel will still be under the trap of the broad array of adaptation strategy, which is a trial and error, less specific and costly. The model can help anticipate adaptation decision support on-farm water management, shift to suitable planting periods, and use of improved drought resistant and short duration varieties adapted to a local weather pattern.