This paper is a contribution to answer the question on how well do global p
recipitation fields, often used for climate model verifications, represent
reality. Here we focus on the experimental version of the GPCP-1DD product,
the daily precipitation estimates based on satellite measurements from the
Global Precipitation Climatology Projects (GPCP) and t+6 to t+30 hours mod
el predictions from the European Centre for Medium. Range Weather Forecast
(ECMWT). The spatial/temporal resolution of both data sets is 1 degree/dail
y. The ground truth is represented by 3100 daily rain gauge measurements op
erating during June/July 1997 in the region of The Mesoscale Alpine Program
me (MAP). These observations have been analyzed within the MAP section of t
he global grid by statistical interpolation. Verification results are given
in terms of difference fields (mean error GPCP = -0.59 mm/day; ECMWF = -1.
13mm/day), rank-order correlation coefficients (mean monthly value,GPCP = 0
.52; ECMWF = 0.67) as well as accuracy scores (probability of detection GPC
P = 0.62; ECMWF = 0.80 and false alarm ratio GPCP = 0.21; ECMWF = 0.18) and
skill scores (true skill statistics GPCP = 0.36; ECMWF = 0.54). These scor
es indicate that both global datasets have deficits in estimating realistic
ally precipitation amounts. However, the ECMWF Predictions have a high perf
ormance in forecasting the spatial distribution of the precipitating areas.
A major result of this study is also that accuracy and skill of the GPCP-1
DD estimates have been shown to be significantly lower than those of the EC
MWF forecasts. Only the mean error of the GPCP-1DD products is low, due to
the calibration with monthly synoptic data.