Considerations for the use of radar-derived precipitation estimates in determining return intervals for extreme areal precipitation amounts

To explore the feasibility of radar-based extreme precipitation climatologies, prototype radar areal reduction factor (ARF) curves are developed and compared to those based on traditional rain gauge networks. For both the radar and gauge data, increasing the spatial density of observations has little influence on the ARF relationship. However, independently, considerable differences between radar ARF and gauge ARF exist. Radar ARF decays at a faster rate (with increasing area) than gauge ARF. For a basin size of 20,000 km, the percent difference between radar ARF and gauge ARF ranges from 11 to 32%. This implies that radar-derived estimates of extreme point precipitation are disproportionately larger than radar-derived estimates of extreme areal precipitation, as compared to the corresponding relationship based on rain gauges. Between-station variance of same-day extreme precipitation, as well as the coefficient of variation tends to be larger for the radar-derived areal extreme events, favoring a smaller radar areal precipitation. Smaller radar ARF is also favored because, on average, a higher percentage of gauges have coincident annual maxima than do the radar pixels that correspond to these gauges. Radar ARF curves computed based on gauge-calibrated radar data decay at an even faster rate than the unadjusted radar ARF. The accuracy of the calibrated radar data for these extreme events is suspect, however. Areal precipitation amounts for the 2-, 5and 10-year return period were computed by fitting an extreme value distribution to the areal radar, (and separately gauge), maxima from 5 years of available data. In one study area, the radar estimates tend to exceed those based on the gauge, whereas in a different region the gauge estimates tend to exceed those based on the radar. These results emphasize that a smaller radar ARF does not necessarily imply a lower radar mean areal precipitation. q 2005 Elsevier B.V. All rights reserved.