Applying a time-varying GEV distribution to correct bias in rainfall quantiles derived from regional climate models

Abstract

Climate warming is causing an increase in extreme hydrometeorological events in most parts of the world. This phenomenon is expected to continue and will affect the frequency and intensity of extreme precipitation events. Although bias correction in regional climate model simulations has also been used to assess changes in precipitation extremes at daily and longer time steps, trends in the series predicted have seldom been considered. We present a novel bias correction technique that allows for the correcting of biases in the upper tails of the Generalized Extreme Value (GEV) distribution, while preserving the trend in projected precipitation extremes. The concept of non-stationary bias correction is demonstrated in a case study in which we used four EURO-CORDEX RCM models to estimate future rainfall quantiles. Historical observations have been used to correct biases in historical runs of the RCMs. The mean relative change in rainfall quantiles between the 1991–2021 historical period and the time horizon of 2080 was found to be 13.5% (st. dev.: 2.9%) for the return period of 2 years, which tends to decline with increasing return periods. Upon the return periods of 50 and 100 years, the mean relative change was predicted to be 5.5% (st. dev.: 1.1%) and 4.8% (st. dev.: 1%), respectively.

Keywords

• non-stationarity
• climate change
• trends
• bias correction
• gev distribution