Performance of DEMETER calibration for rainfall forecasting purposes: Application to the July–August Sahelian rainfall
Résumé
This work assesses and compares the skill of direct and model-output-statistics (MOS) calibrated hindcasts of the July–August rainfall amounts for the dry period 1980–2000 over the Sahel issued from the Development of a European Multimodel Ensemble System for Seasonal to Interannual Prediction (DEMETER) experiment, with the aim to highlight among the simulated parameters, i.e., those potentially relevant for rainfall forecasts purposes. Three approaches were used: the DEMETER (1) direct rainfall, (2) MOS-calibrated rainfall, and (3) MOS-calibrated atmospheric dynamics and energy. Canonical correlation analyses (CCA) were employed in the two latter approaches to calibrate the direct rainfall fields from DEMETER. The main results are the following. First, the observed relationship between indexes depicting Sahelian rainfall (SRI) and West African Monsoon dynamics (MOD850 for the modulus of the wind at 850 hPa) is not correctly reproduced in DEMETER: the correlation between them achieves 0.25 only (versus 0.6 in observations). The calibration with atmospheric dynamics and energy fields (moist static energy fluxes at 850 hPa and zonal wind at 500 hPa) leads to slightly higher results than with rainfall. Second, the leading coupled mode of the CCA between observed rainfall and simulated atmospheric dynamics (∼28% of the total variance) connects Sahelian rainfall variability to a weakened African Easterly Jet (AEJ, somewhat shifted to the north) as well as to the strength of the MSE fluxes from the Guinean Gulf and the Mediterranean Sea. Finally, the teleconnection pattern between the CCA mode and simulated surface temperatures concurs well with observations, giving confidence into the physical basis of the mode, although the cross-validated skill scores obtained through the three approaches remain low.