Dynamical and statistical downscaling of a global seasonal hindcast in eastern Africa

Within the FP7 EUPORIAS project we have assessed the utility of dynamical and statistical downscaling to provide seasonal forecast for impact modelling in eastern Africa. An ensemble of seasonal hindcasts was generated by the global climate model (GCM) EC-EARTH and then downscaled by four regional climate models and by two statistical methods over eastern Africa with focus on Ethiopia. The five-month hindcast includes 15 members, initialised on May 1 st covering 1991–2012. There are two sub-regions where the global hindcast has some skill in predicting June–September rainfall (northern Ethiopia – northeast Sudan and southern Sudan - northern Uganda). The regional models are able to reproduce the predictive signal evident in the driving EC-EARTH hindcast over Ethiopia in June–September showing about the same performance as their driving GCM. Statistical downscaling, in general, loses a part of the EC-EARTH signal at grid box scale but shows some improvement after spatial aggregation. At the same time there are no clear evidences that the dynamical and statistical downscaling provide added value compared to the driving EC-EARTH if we define the added value as a higher forecast skill in the downscaled hindcast, although there is a tendency of improved reliability through the downscaling. The use of the global and downscaled hindcasts as input for the Livelihoods, Early Assessment and Protection (LEAP) platform of the World Food Programme in Ethiopia shows that the performance of the LEAP platform in predicting humanitarian needs at the national and sub-national levels is not improved by using downscaled seasonal forecasts.