An intercomparison of global oceanic precipitation climatologies

Abstract

Large-scale patterns of precipitation are important for the changes they may effect upon the circulation of the ocean. However, marine precipitation is very hard to quantify accurately. Four independent climatologies are examined to compare their estimates of the annual mean precipitation, and the seasonal and interannual variations. One data set, Global Precipitation Climatology Project (GPCP), is based upon satellite data, the other three on output of weather forecast reanalyses from the National Centers for Environmental Prediction (NCEP) and the European Centre for Medium-range Weather Forecasts (ECMWF). Although all datasets have their errors, there is general agreement on the geographical patterns of precipitation. All the models had higher rain rates in the tropics than shown by the satellite data, and also greater seasonal ranges. However, GPCP has 10-25% more precipitation than NCEP and ECMWF in most of the southern regions, because of their weak representation of convergence zones; NCEP2, a more recent version of the NCEP reanalysis, shows a marked improvement in this area. However, in most regions NCEP2 exhibits a larger seasonal range than shown by other datasets, particularly for the tropical Pacific. Both NCEP and NCEP2 often show a seasonal cycle lagging two months or more behind GPCP. Of the three reanalysis climatologies, ECMWF appears best at realising the position and migration of rain features. The interannual variations are correlated between all four datasets, however the correlation coefficient is only large for regions that have a strong response to El Niño and La Niña event, or for comparisons of the two NCEP reanalyses. Of the datasets evaluated, GPCP has the most internal consistency, with no long-term trend in its regional averages, and it alone shows the deficit in Mediterranean precipitation coincident with the Eastern Mediterranean Transient.

More information

Identifiers

Catalogue record

Export record

Altmetrics