TROPICAL RAINFALL - A COMPARISON OF SATELLITE-DERIVED RAINFALL ESTIMATES WITH MODEL PRECIPITATION FORECASTS, CLIMATOLOGIES, AND OBSERVATIONS

Quantitative estimates of rainfall, as derived from satellite observations of cloud-top temperature, have been produced for each 2.5-degrees lat-long location from 30-degrees-N to 30-degrees-S for the period 1986-89. These remotely sensed estimates are a preliminary product of the Global Precipitation Climatology Project (GPCP), which is charged with the development of a global rainfall dataset during the period 1986-95, using all available in situ data supplemented with satellite-derived estimates of precipitation. Four-year means (1986-89) of monthly satellite-derived rainfall estimates were compared to the climatological rainfall datasets compiled by Jaeger and Legates and Willmott in the tropics, to assess the performance of the satellite estimates in the context of existing state-of-the-art global precipitation datasets. It is shown that the satellite estimates and the climatologies are in reasonable agreement both in temporal and spatial variation, although some large regional differences were found. A more quantitative assessment of these estimates is also presented by comparing them with ground-based observations from a high-density radar-raingage network over Japan. Short-range forecasts of precipitation from the National Meteorological Center (NMC) Medium-Range Forecast (MRF) and the European Centre for Medium Range Weather Forecasts (ECMWF) models are also compared with the satellite estimates in the tropics. The motivation for this comparison is to assess the utility of the model precipitation forecasts as a "first guess" of the global tropical precipitation field for use in an objective analysis scheme to integrate the satellite-derived estimates and in situ data for the GPCP. The results show that the distribution and magnitude of mean tropical rainfall agree reasonably well among the estimates and model forecasts during the 9-month study period (March-November 1989). However, temporal correlations between the satellite estimates and the model precipitation predictions reveal that the models currently do not represent the time variations of rainfall in much of the tropics, particularly 30-60-day oscillations, that are apparent in the satellite estimates.