An evaluation of COSMO‐CLM regional climate model in simulating precipitation over Central Africa

In this study, an analysis of present day climate simulation (1998–2008) is presented for the Central African (CA) region with the COnsortium for Small‐scale MOdelling in CLimate Mode (CCLM) regional climate model, forced by the ERA‐Interim (ERAINT) reanalysis data. The ability of the CCLM to simulate the observed precipitation with particular focus on the mean spatial pattern, low‐level circulation, seasonal cycles, and daily characteristics is evaluated. Likewise, the added value of the regional model CCLM compared to the driving ERAINT reanalysis is also investigated. It is shown that ERAINT and CCLM exhibit quite different sign of bias, which is an indication of the importance of internal variability and fine scale processes representation for the simulation of surface climate. Despite the CCLM is constantly dry over southern CA, the model succeeds to reproduce reasonably the mean spatial patterns of precipitation and low‐level circulation features, along with the associated seasonal cycles over the whole CA and majority of the five selected analysis sub‐regions. Results also show that daily precipitation indices are well represented, although the better performance greatly depends on the considered seasons. Nevertheless, CCLM substantially outperforms the ERAINT daily precipitation characteristics, thus highlighting the added value of the downscaling exercise over the region. The analysis of daily precipitation indices also reveals that the dry character of the model could probably be connected to the underestimation of the simulated less intense events, which in turn result to an overestimation of the simulated dry spell duration. Key Results We found that CCLM substantially improves the simulation of both mean seasonal and daily precipitation compared to the driving ERAINT reanalysis, thus highlighting the added value of the downscaling exercise over the region. It is also shown that the dry character of the model is probably connected to the less intense events and too long dry spells simulated by the model. Figure Caption Mean (1998–2008) spatial distribution of seasonal precipitation (in mm/day), from observation (a–d) GPCP. Also shown are the spatial distribution of the difference TRMM minus GPCP (e–h), the biases with respect to GPCP from ERAINT (i–l) and from CCLM (m–p), and the difference between CCLM and ERAINT (q–t). Hatching (stippling) indicates grid points where there is an added value by the dynamical downscaling, with GPCP (TRMM) used