Extreme value modeling of areal rainfall from weather radar

Abstract

An 11 year high-quality radar rainfall data set is used to abstract annual maximum rainfall depths for durations of 15 min to 24 h and area sizes of 6 to 1.7 × 10³ km² for the Netherlands. Generalized extreme value (GEV) distributions are fitted to the annual maxima. A new method is presented to describe the distribution of extreme areal rainfall depths by modeling GEV parameters as a function of both duration and area size. This leads to a semiempirical expression from which quantiles of extreme rainfall depths can be obtained for a chosen duration, area size, and return period. The uncertainties in these quantiles are calculated using the bootstrap method. Radar-based areal reduction factors (ARFs) are derived. These ARFs are comparable to those based on high-density rain gauge networks derived from the literature. It is concluded that radar data, after careful quality control, are suitable to estimate extreme areal rainfall depths.