Anticipating drought-related food security changes

Food insecurity early warning can provide time to mitigate unfolding crises; however, drought remains a large source of uncertainty. The challenge is to filter unclear or conflicting signals from various climatic and socio-economic variables and link them to food security outcomes. Integrating lag-1 autocorrelation diagnostics into remotely sensed observations from the Soil Moisture Active Passive (SMAP) mission and food prices, we found dramatic improvement in anticipating the timing and intensity of food crises, except in conflict settings. We analysed drought-induced food crises globally in the SMAP record (since 2015; approximately five per year). The change in soil moisture autocorrelation, which we term the Soil Moisture Auto-Regressive Threshold (SMART), signalled an accurate food security transition for all cases studied here (P < 0.05; n = 212), including lead time of up to three to six months for every case. The SMART trigger anticipates the timing of the transition and the magnitude of the food security change among small to large transitions, both into and out of crises (R2 = 0.80–0.83). While we do not evaluate out-of-sample forecast accuracy using our model, our findings suggest a significant advancement in the capabilities of food security early-warning diagnostics and could save lives and resources.