Spatial patterns in smallholder crop yield uncertainty identify priority intervention areas

Abstract:

High variability in management and environmental conditions in smallholder agricultural regions often results in low, variable crop yields. Data limitations prevent farm-scale yield estimates that are both accurate and scalable. Regional yield mapping approaches that rely on spatial aggregation can bias estimates and neglect local inequities. Here we develop an approach to characterize uncertainty distributions of smallholder yields on a high-resolution grid to enable risk-based regional planning. The approach combines a process-based crop model, new applications of satellite data, and a novel cross-validation method. We apply it to a case study on maize in Nigeria, where variability in water availability and fertilization leads to large yield ranges (0-7000 kg/ha). Despite this large uncertainty, we demonstrate how analysis of the spatial and temporal variability of the bottom-tail, median, and upper-tail of the yield distributions can inform regional planning. We identify contiguous spatial areas that consistently underperform as candidates for permanent irrigation infrastructure. High year-to-year variability suggests a need for temporary, annual interventions driven by within-season forecasts, as well as the need for improved data collection. Finally, using the bottom-tail of yield distributions identifies different areas for intervention than using medians, highlighting tradeoffs between reducing inequity and improving overall production.