A Comparison of the NCEP–NCAR Reanalysis Precipitation and the GPCP Rain Gauge–Satellite Combined Dataset with Observational Error Considerations

Abstract

The Global Precipitation Climatology Project (GPCP) has released monthly mean estimates of precipitation that comprise gauge observations and satellite-derived precipitation estimates. Estimates of standard random error for each month at each grid location are also provided in this data release. One of the primary intended uses of this dataset is the validation of climatic-scale precipitation fields that are produced by numerical models. Nearly coincident with this dataset development, the National Centers for Environmental Prediction and the National Center for Atmospheric Research have joined in a cooperative effort to reanalyze meteorological fields from the present back to the 1940s using a fixed state-of-the-art data assimilation system and large input database. In this paper, monthly accumulations of reanalysis precipitation are compared with the GPCP combined rain gauge–satellite dataset over the period 1988–95. A unique feature of this comparison is the use of standard error estimates that are contained in the GPCP combined dataset. These errors are incorporated into the comparison to provide more realistic assessments of the reanalysis model performance than could be attained by using only the mean fields. Variability on timescales from intraseasonal to interannual are examined between the GPCP and reanalysis precipitation. While the representation of large-scale features compares well between the two datasets, substantial differences are observed on regional scales. This result is not unexpected since present-day data assimilation systems are not designed to incorporate observations of precipitation. Furthermore, inferences of deficiencies in the reanalysis precipitation should not be projected to other fields in which observations have been assimilated directly into the reanalysis model.