Yields Increases through Irrigation

The dataset is the result of a crop simulation and machine learning exercise to evaluate the irrigation potentials in sub-Saharan Africa. Samples of points (representative for each Agro-ecological zone in Africa) have been selected from a cropland map (from Tubliello et al., 2023) and, by using the DSSAT crop model fed with soil information from the ISRIC project (https://isric.org/projects/soil-information-system-for-africa-soils4africa/) and 10 years of daily climate data (from NASAPOWER), simulations have been conducted in each location under rainfed conditions and optimal irrigation. This last has been achieved y letting the crop model add irrigation water each time a water stress was perceived. Other management practices have been set as realistic as possible, particularly the amount of provided fertilization has been kept low to represent current conditions. Yields have been finally averaged along the different years of simulations to generate expected yields. Finally, by using the XGboost algorithm fed with the same soil and climatic variables used for the crop model, yields under both rainfed and irrigation conditions have been imputed to all remaining cropland locations in Sub-Saharan Africa. What is presented here are the yield increases, as percentages over rainfed yields, obtainable in each location for the four simulated cereals: maize, millet, rice and sorghum.