Validation of ENSEMBLES/DEMETER seasonal forecasts over the Sahel: Intraseasonal features

EMS conference in Amsterdam (29/09 - 03/10/2008)

The Sahel is the region of the globe which has experienced the most severe drying, from 1970 to 2000, with dramatic social and economic impacts. As a consequence, forecasting rainfall over West Africa at seasonal time scales became essential regarding to impacts in terms of food, health and security managements since the 1970’s.
One of the aims of the recent initiated AMMA (African Monsoon Multidisciplinary Analysis) project is to improve the prediction of the West African Monsoon (WAM) system and its impacts on West African nations. In parallel, within the European DEMETER and ENSEMBLES projects framework, a well-validated European coupled multi-model ensemble forecast system for reliable seasonal to interannual prediction has been developed. Coordinating studies on seasonal prediction between ENSEMBLES and AMMA participants became obvious from the first AMMA/ENSEMBLES meeting that took place in Bamako, Mali.
Thus, in this paper we propose to both validate ENSEMBLES and DEMETER seasonal forecast simulations over the Sahel, with respect to different reference observation data sets during the monsoon period (from July to September) for a common period (1991-2001). An overview of the performance of the different models participating in the projects in capturing the mean features of the WAM, as well as of the multi-model hindcasts is achieved using typical indices, maps, determinist and probabilistic scores commonly used in seasonal forecast studies. Preliminary results highlight common rainfall bias as simulated by coarse resolution GCM, namely overestimation of precipitation over the high mountains (Ethiopia plateau) and underestimation over the low ones (Cameroon mounts, Senegal coast). Other common bias (similar to the ones highlighted in the CMIP3 IPCC coupled simulations) can be shown for relevant SST areas (which are strongly related to rainfall interannual variability over the Sahel), namely a warm bias over the Gulf of Guinea and the ENSO (eastern Pacific) region. Based on a perfect model approach, rainfall potential predictability (PP) is estimated. Strong rainfall PP values can be shown over the Guinea Coast whereas the predictability is relatively weak over the Sahel. Moreover, the model rainfall forecasts seem to be more skilful over the Guinea Coast and the tropical Atlantic Ocean than the Sahel.
The Impact Studies community (agriculture, health....) have expressed a significant need in terms of forecasting intra-seasonal features of the WAM (onset date of the monsoon, occurrence of dry spells during the rainy season...). The performance of DEMETER/ENSEMBLES hindcasts in reproducing these features will be investigated with respect to both ERA40 and NCEP reanalysis.
Similarities and divergence points between DEMETER and ENSEMBLES hindcasts system performance over sub-Saharan Africa will then be discussed as a conclusion. A discussion will also be done about the selected time period. Namely, is 10 year model climate meaningful to build significant statistical results (mean, variance, etc) over the Sahel? Development strategies to improve the model scores will be suggested as perspectives.

Associated presentation:
EMS2008_caminade.v2.pdf