Weather-related shocks and climate variability contribute to hampering progress toward poverty reduction in Sub-Saharan Africa. Droughts have a direct impact on weather-dependent livelihood means and the potential to affect key dimensions of households’ welfare, including food consumption. Yet, the ability to forecast food insecurity for intervention planning remains limited and current approaches mainly rely on qualitative methods. This paper incorporates microeconomic estimates of the effect of the rainy season quality on food consumption into a catastrophe risk modeling approach to develop a novel framework for early forecasting of food insecurity at sub-national levels. The model relies on three usual components of catastrophe risk models that are adapted for estimation of the impact of rainy season quality on food insecurity: natural hazards, households’ vulnerability and exposure. The paper applies this framework in the context of rural Mauritania and optimizes the model calibration with a machine learning procedure. The model can produce fairly accurate lean season food insecurity predictions very early on in the agricultural season (October-November), that is six to eight months ahead of the lean season. Comparisons of model predictions with survey-based estimates yield a mean absolute error of 1.2 percentage points at the national level and a high degree of correlation at the regional level (0.84).