Mixed General Extreme Value Distribution for Estimation of Future Precipitation Quantiles Using a Weighted Ensemble - Case Study of the Lim River Basin (Serbia)

Abstract

Considering the recent extreme precipitation in southeast Europe, it has become necessity to investigate the impact of climate change on extreme precipitation. The aim of this study was to determine the change in precipitation quantiles with longer return periods under changing climate conditions. The study was conducted using the daily records gathered at 11 precipitation stations within the Lim River Basin, Serbia. The simulated precipitation datasets were collected from three regional climate models for the baseline period (1971–2000), as well as the future period (2006–2055) under the 2.6, 4.5 and 8.5 representative concentration pathways. The raw precipitation data from the climate models were transformed by employing four bias correction methods. Using the bias-corrected precipitation, an ensemble of annual maximum daily precipitation was developed. A weighted ensemble approach was applied to estimate the weights of each ensemble member favorizing the members whose quantiles were closer to observed measurements. The mixed general extreme value distribution was used to derive the projected quantiles with 100, 50, 25, 10, five and two year return periods based on the estimated quantiles and the normalized weights of all ensemble members. An overall increase of 69% and 56% for the 100 and 50 year return periods, respectively, can be expected within the northern part of the basin. Similarly, an overall increase of 50–57% and 39–42% for the 100 and 50 year return periods, respectively, may be expected for the central and southern parts of the Lim River Basin.