Report on the added value of dynamical and statistically downscaled data

This report details an ensemble of seasonal hindcast generated by a global model EC-EARTH and then downscaled by five regional models and by two statistical methods over eastern Africa with focus on Ethiopia. The hindcast includes 15 members, initialised on May 1st and covers the period 1991-2012. There are two sub-regions where the global hindcast has some skill in predicting June-September precipitation (northern Ethiopia - North-East Sudan and southern Sudan - northern Uganda), while there is no skill elsewhere in the region. The RCMs are able to capture and reproduce the signal evident in the driving EC-EARTH seasonal 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 skill and shows a weaker performance in terms of predictability. At the same time there are no clear evidences that the RCMs provide the added value compared to the driving EC-EARTH if we define the added value as a higher forecast skill in the RCM hindcast. This conclusion is only for Ethiopia in the June-September season and cannot be generalised for other regions and seasons.