Limited‐area model error statistics over Western Africa: Comparisons with midlatitude results

Abstract In the context of the African Monsoon Multidisciplinary Analyses experiment, a new forecast error covariance matrix has been computed for data assimilation purposes, using the spectral mesoscale limited‐area model ALADIN. The so‐called ‘lagged NMC’ method has been applied using sets of forecast differences performed during the summer overWestern Africa. Multiple linear regressions on these differences have been computed in order to determine statistics and balance features between forecast errors of divergence, vorticity, specific humidity, temperature and surface pressure. The resulting auto‐ and cross‐covariances display very different features from those obtained over Western Europe, and they reflect the uncertainty of forecasting the main meteorological features of the West African Monsoon. While the geostrophic coupling with vorticity is predominant in midlatitudes, over western Africa the main mass/wind couplings instead involve divergence in the low troposphere and below the tropopause. The couplings between e.g. low‐level humidity, temperature and divergence are strengthened over Africa. The vertical correlations are broader for increasing horizontal scales but in a much less pronounced manner than in midlatitudes. The mean vertical correlations for humidity are much more localized in the low troposphere over Africa. Copyright © 2006 Royal Meteorological Society