Sub-seasonal forecasting of cropland productivity anomalies using satellite soil moisture in water-limited environments

Soil moisture depletion precedes declines in vegetation productivity by days to weeks, creating a natural prediction window for agricultural drought impacts. We test whether satellite root-zone soil moisture (RZSM) from the Soil Moisture Active Passive (SMAP) Level-4 product can provide skillful sub-seasonal forecasts of crop productivity anomalies in Near-Infrared Reflectance of Vegetation (NIRv) across drought-prone croplands of Eastern and Southern Africa. We test three input configurations: NIRv-only, NIRv with concurrent RZSM, and NIRv with explicitly lagshifted RZSM, to forecast NIRv anomalies at lead times of 8 to 40 days using a Convolutional Long Short-Term Memory (ConvLSTM) model chosen for its ability to jointly capture local spatial dependence and temporal lag structure. All configurations outperform the climatological NIRv and reduce the RMSE by 22–24%. The concurrent RZSM configuration achieves the highest domain-averaged anomaly correlation coefficient (ACC = 0.560), with the soil moisture contribution growing monotonically from negligible at 8 days to +0.056 ACC at 40 days. When we stratify by aridity, the RZSM advantage concentrates in water-limited croplands (mean ΔACC =+0.022 and +0.024 across the 48-day and 80-day input window, respectively, versus ≈0 in energy-limited croplands), where soil moisture–vegetation coupling is stronger. In energy-limited regions, soil moisture adds minimal predictive value. Our case study of the 2024 southern African El Niño drought shows that the RZSM-informed model resolves the spatial pattern of crop-productivity anomalies approximately 32 days in advance, reducing spatial mean absolute error by 30% and improving spatial correlation by 29% over the autoregressive NIRv-only forecast. Our results indicate that SMAP root-zone soil moisture is a viable predictor of sub-seasonal crop-productivity anomalies in water-limited African croplands. Our forecast target is a satellite proxy for vegetation productivity (NIRv anomaly), not crop yield, production, or documented crop failure; translating these cropland-masked anomalies into agricultural-impact outcomes requires crop-specific calibration and integration with food-security indicators.