Simulation of extreme precipitation in the Rhine basin by nearest-neighbour resampling

Abstract

The use of the nonparametric nearest-neighbour resampling technique is studied for generating time series of daily rainfall and temperature for seven stations in the German part of the Rhine basin. The emphasis is on the reproduction of extreme N-day precipitation amounts. The daily temperatures are used to determine snow accumulation and melt in winter. Two versions of the resampling method, conditional on the atmospheric circulation and unconditional, show comparable results. For precipitation, the autocorrelation properties are well reproduced, whereas for temperature the autocorrelation coefficients are systematically under-predicted. The distributions of the N-day annual maximum precipitation amounts are adequately preserved. Despite the systematic underprediction of the temperature autocorrelation, the distributions of N-day maximum snowmelt are well reproduced. A 1000-year simulation for the seven stations shows that unprecedented rainfall situations can be generated.