Extreme rainfall analysis and estimation of depth-duration-frequency curves using weather radar

Rain gauge data are often employed to estimate the rainfall depth for a given return period. However, the number of rain gauge records of short-duration rainfall, such as 15 minutes, is sparse. The obvious advantage of radar data over most rain gauge networks is their higher temporal and spatial resolution. Further, the current quality of quantitative precipitation estimation with radar and the length of the available time series make it feasible to calculate radar-based extreme rainfall statistics. In this paper an 11-year radar data set of precipitation depths for durations of 15 min to 24 h is derived for the Netherlands (3.55 * 10^4 km^2). The radar data are adjusted using rain gauges by combining an hourly mean-field bias adjustment with a daily spatial adjustment. Assuming a Generalized Extreme Value (GEV) distribution, the index flood method is used to describe the distribution of the annual radar rainfall maxima. Regional variability in the GEV location parameter is studied. GEV parameters based on radar and rain gauge data are compared and turn out to be in reasonable agreement. Further, radar rainfall depth-duration-frequency (DDF) curves and their uncertainties are derived and compared with those based on rain gauge data. Although uncertainties become large for long durations, it is shown that radar data are suitable to construct DDF curves.