Downscaling reanalysis over continental Africa with a regional model: NCEP versus ERA Interim forcing

Abstract

Five annual climate cycles (1998)(1999)(2000)(2001)(2002) are simulated for continental Africa and adjacent oceans by a regional atmospheric model (RM3). RM3 horizontal grid spacing is 0.44 at 28 vertical levels. Each of 2 simulation ensembles is driven by lateral boundary conditions from each of 2 alternative reanalysis data sets. One simulation downscales National Center for Environmental Prediction reanalysis 2 (NCPR2) and the other the European Centre for Medium Range Weather Forecasts Interim reanalysis (ERA-I). NCPR2 data are archived at 2.5 grid spacing, while a recent version of ERA-I provides data at 0.75 spacing. ERA-I-forced simulations are recommended by the Coordinated Regional Downscaling Experiment (CORDEX). Comparisons of the 2 sets of simulations with each other and with observational evidence assess the relative performance of each downscaling system. A third simulation also uses ERA-I forcing, but degraded to the same horizontal resolution as NCPR2. RM3-simulated pentad and monthly mean precipitation data are compared to Tropical Rainfall Measuring Mission (TRMM) data, gridded at 0.5, and RM3-simulated circulation is compared to both reanalyses. Results suggest that each downscaling system provides advantages and disadvantages relative to the other. The RM3/NCPR2 achieves a more realistic northward advance of summer monsoon rains over West Africa, but RM3/ERA-I creates the more realistic monsoon circulation. Both systems recreate some features of July-September 1999 minus 2002 precipitation differences. Degrading the resolution of ERA-I driving data unrealistically slows the monsoon circulation and considerably diminishes summer rainfall rates over West Africa. The high resolution of ERA-I data, therefore, contributes to the quality of the downscaling, but NCPR2 lateral boundary conditions nevertheless produce better simulations of some features.