Improving vegetation condition forecasting for drought early warning in East Africa

Droughts are a recurring global climate hazard that incur human, economic and environmental costs. In Eastern Africa, pastoralist communities whose livelihoods depend on the availability of pasturelands are particularly vulnerable to the impacts of drought. In response to this vulnerability,  the University of Sussex developed vegetation condition forecasts for pastoralist communities using remote sensing data and machine learning techniques. These forecasts are designed to be used by the Kenyan National Drought Management Authority in monthly drought early warning bulletins. Building on stakeholder feedback and given the impacts of drought vary within a county/sub-county we identify a need for higher-resolution forecasts of the onset of drought. Here we present the initial findings from a comparative study exploring a range of machine learning techniques to generate higher resolution vegetation condition forecasts for transboundary pastoralist regions in eastern Africa.  We aim to evaluate how the forecast skill varies depending on:  machine learning technique, resolution of input data and satellite indicators included. This work is part of PASSAGE, a CLARE (clareprogramme.org) funded project working towards strengthening pastoralist livelihoods through effective anticipatory action.