There is an increasing availability and application of precipitation fields estimated by remote sensing or calculated by atmospheric circulation models, which are frequently used as input for distributed hydrological models. The spatial distribution of the estimated precipitation fields is extremely important and must be verified against observed precipitation fields. This paper proposes a method for spatiotemporal comparison between observed and estimated precipitation fields based on a pixel by pixel comparison and on contingency tables. Two distinct approaches are carried out: (i) the spatial integrated analysis produces skill scores denoting the ability of the estimated precipitation field in reproducing the occurrence of observed precipitation along the time; (ii) the time integrated analysis generates maps of skill scores showing the reliability of the precipitation estimates in each pixel. As an example of application, the estimated precipitation climatology of the CPTEC/COLA global circulation model over the Rio Grande basin is assessed. Using five different skill scores, the proposed method identified seasonal variations and spatial patterns in the performance of the estimated precipitation fields in relation to precipitation fields derived from pluviometers measurement.
skill score; rainfall estimate; contingency table; hydrologic model
