Disagreements in Low-Level Moisture between (Re)Analyses over Summertime West Africa

Abstract

eanalysis and operational analysis products are routinely used as the best estimates of the atmospheric state for operational and research purposes. However, different models, assimilation techniques, and assimilated datasets lead to differences between products. Here, such differences in the distribution of low-level water vapor over summertime West Africa are analyzed, as reflected in the zonal mean position of the leading edge of the West African monsoon [the intertropical discontinuity (ITD)] using five reanalyses [NCEP?NCAR, NCEP?Department of Energy (DOE), the Modern-Era Retrospective Analysis for Research and Applications (MERRA), the Climate Forecast System Reanalysis (CFSR), and the Interim ECMWF Re-Analysis (ERA-Interim)] and two operational analyses [Global Forecast System (GFS) and ECMWF] during the 11 monsoon seasons (April?September) from 2000 to 2010. Specific humidity differences regularly reach 50% of the mean value over areas spanning hundreds of kilometers and often coincide with northward excursions of the ITD that last several days and bring unusual rainfall to the Sahel and Sahara. The largest disagreements occur during the southward retreat of the ITD and are connected with anomalously high values of aerosol optical depth, consistent with the production of haboob dust storms. The results suggest that known errors in the representation of moist convection and cold pools may contribute to the identified disagreements. A large reduction in disagreement occurs in 2006, when upper-air observations were enhanced during the African Monsoon Multidisciplinary Analysis (AMMA) campaign, pointing to an insufficient observational constraint of the (re)analyses in other years. It is hoped that this work will raise awareness of the limited reliability of (re)analysis products over West Africa during the summer, particularly during northward surges of the ITD, and will instigate further work to improve their quality.