Leveraging Hydroclimate and Earth Observation to Predict Grain Production in Sub-Saharan Africa

The importance of forecasting agricultural production in Sub-Saharan Africa (SSA) is increasing for the management of agricultural supply chains, market forecasting, and targeting of food aid. In particular, agricultural forecasts enable governments and humanitarian organizations to respond more effectively to shocks in food production and price spikes resulting from extreme droughts. In this study, we use hydroclimate, earth observations (EO) and machine learning to develop an operational, sub-national grain production forecast system for a number of SSA countries, including food-insecure regions where rapid response is critical. Before creating the forecast, we collect and organize crop production data from the Famine Early Warning Systems Network in order to identify trends and variability in agricultural technology, climate, and vegetation. In addition, we demonstrate the capability of hydroclimate and EO data to capture favorable or unfavorable crop development conditions during the growing season. In addition, we demonstrate a unique capability that explains how EO characteristics influence current grain production forecasts, thereby enhancing the forecasts' reliability and efficacy. This research lays the groundwork for the development of a large-scale, operational crop yield forecasting system that will provide actionable predictions of food shocks for famine early warning and guide advanced preparedness and response strategies.