Probability Matching between ECMWF Forecasts and Rain Gauge Observations Using a Fuzzy Approach for Summer in Eastern United States

A better understanding of the characteristics and the quality of the precipitation forecasts is needed as more advanced and widespread model estimates of precipitation are available for the scientific community. This task is especially important for cases that include convective scale events since they are the most difficult to forecast. To reach this goal, this study compares and verifies ECMWF precipitation forecasts with rainfall gauge observations for selected summer seasons in the Eastern US region. Daily rainfall gauge data across the U.S. is obtained from the Global Historical Climatology Network (GHCN) database. The comparison and verification uses the probability based verification technique that compares the forecast and observed rain fractional coverage within spatial neighborhoods. This spatial verification approach known also as fuzzy approach, assumes that a slightly displaced forecast can still produce useful information of the observed value, in special in situations where is common to find that model results at the mesoscale level give displaced estimations in space and time. In addition to this method, traditional verification metrics based on deterministic techniques (e.g., frequency bias, false alarm ratio and probability of detection) were also used to compare the results provided by the probabilistic approach. The results presented here help to quantify the performance of the ECMWF forecasting system by reproducing the observed precipitation in convective environments. In addition, the application of this probability-matching technique using different spatial scales give promising information on the scales and intensities at which the ECMWF forecasts should be viewed as reliable.