An improved empirical quantile mapping approach for bias correction of extreme values in climate model simulations

Abstract Quantifying and correcting biases in modeling simulations is crucial for deriving meaningful findings across various scientific disciplines. Climate model simulations, in particular, often exhibit systemic biases when compared to observations. These biases may persist in future climate simulations, affecting the results of many climate change impact assessment studies. Empirical quantile mapping (QM) is a widely used method to correct these biases by mapping quantiles between observed and simulated cumulative distribution functions (CDFs). However, empirical QM faces a fundamental challenge when the CDF of future simulations differs from historical simulations, potentially leading to extreme values falling outside the historical CDF range. To address this issue, our study introduces a novel approach to extrapolate future extreme values for bias correction, preserving the rank order of simulated future extremes. By construction, our approach ensures that bias-corrected values are not exaggerated and retain the rank structure of the original simulated data while preserving climate change signals in the bias-corrected outputs. In addition, our approach includes a technique to adjust the wet-day frequency for precipitation by preserving the ratio of wet-day frequency between observations and historical model simulations.