An Intercomparison of Observational, Reanalysis, Satellite, and Coupled Model Data on Mean Rainfall in the Caribbean

Abstract

This study examines the spatial variability of mean annual rainfall in the Caribbean in the satellite era 1979?2000. Intercomparisons of gridded rainfall fields from conventional stations, satellite estimators, reanalysis products, and coupled general circulation models (CGCMs) are made, with a focus on the Antilles island chain and their land?sea transitions. The rainfall products are rated for their ability to capture a number of key features, including (i) topographically enhanced precipitation over the larger western Antilles islands of Cuba, Jamaica, Hispanola, and Puerto Rico; (ii) the rain shadow west of Hispanola; (iii) the two dry zones where SSTs are low: north of Venezuela and north of the Lesser Antilles; and (iv) the wet axis extending north of Trinidad. The various monitoring and modeling systems produce gridded rainfall fields at resolutions from 50 to 280 km, from station reconstructions, satellite estimates, blended and reanalysis products, and CGCM climatologies with respect to surface forcing fields. Wet and dry biases were found in many of the reanalysis and satellite products, respectively?either over the whole Caribbean or in a certain sector. The intercomparison found some measure of consensus, but no single product is without discrepancy. High-resolution passive microwave satellite rainfall estimates [Climate Prediction Center?s multisaltellite passive microwave, IR morphed product (cMOR)] appear ?most representative?; however, the climatology is short (2003?07) and the field is generally drier than the consensus. Of the conventional products, decadal variability of climate interpolated rain gauges (DEKL), World Climate Research Programme?s (WCRP) blended rain gauges, the Comprehensive Ocean?Atmosphere Data Set (COADS), and an operational climate anomaly monitoring system of NCEP (CAMS) perform well. Among the satellite estimators, the Global Precipitation Climatology Project?s blended gauge and IR satellite (GPCP) and outgoing longwave radiation (OLR) capture the key features and ocean?island transitions. The Center for Ocean?Land?Atmosphere Studies [COLA; the coupled model, part of the Coupled Model Intercomparison Project (CMIP, phase 3)] and the climate forecast system of the NCEP (CFS) models perform reasonably, but NCAR?s Parallel Climate Model (PCM; the CGCM?s historical run of CMIP3) fares poorly. The version 2 hindcast of the operational Medium-Range Forecast (MRF) weather prediction model (REAN) captures the smaller wet zones and topographically enhanced features, but it does not handle the broad oceanic dry zones well, as the input from the operational climate data assimilation system of NCEP (CDAS) has a wet bias. Of the various key rainfall features, high rainfall over southern Cuba and the rain shadow west of Hispanola are poorly handled by most products. The wet axis north of Trinidad and the dry zone north of Venezuela are well represented in many climatologies.