Comparison of Precipitation Derived from the ECMWF Operational Forecast Model and Satellite Precipitation Datasets

Abstract

Abstract Precipitation is an important component of the climate system, and the accurate representation of the diurnal rainfall cycle is a key test of model performance. Although the modeling of precipitation in the cooler midlatitudes has improved, in the tropics substantial errors still occur. Precipitation from the operational ECMWF forecast model is compared with satellite-derived products from the Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TMPA) and TRMM Precipitation Radar (PR) to assess the mean annual and seasonal diurnal rainfall cycles. The analysis encompasses the global tropics and subtropics (40°N–40°S) over a 7-yr period from 2004 to 2011. The primary aim of the paper is to evaluate the ability of an operational numerical model and satellite products to retrieve subdaily rainfall. It was found that during the first half of the analysis period the ECMWF model overestimated precipitation by up to 15% in the tropics, although after the implementation of a new convective parameterization in November 2007 this bias fell to about 4%. The ECMWF model poorly represented the diurnal cycle, simulating rainfall too early compared to the TMPA and TRMM PR products; the model simulation of precipitation was particularly poor over Indonesia. In addition, the model did not appear to simulate mountain-slope breezes well or adequately capture many of the characteristics of mesoscale convective systems. The work highlights areas for further study to improve the representation of subgrid-scale processes in parameterization schemes and improvements in model resolution. In particular, the proper representation of subdaily precipitation in models is critical for hydrological modeling and flow forecasting.