Comparison of bias‑correction methods for future projections of multi‑variable extremes through PMP estimation

Probable Maximum Precipitation (PMP) is a key input for the design of critical hydraulic infrastructure, yet its estimation under climate change is strongly affected by climate‑model biases and by how these biases are corrected. This study compares several quantile‑based bias‑correction methods for future projections of multi‑variable extremes by using PMP, derived from multiple interacting meteorological variables, as a diagnostic. daily minimum temperature, maximum relative humidity, and maximum wind speed from ERA5‑Land, together with daily precipitation from rain‑gauge stations and corresponding outputs from six CMIP6 models were used.Future projections are considered under SSP2‑4.5 and SSP5‑8.5 scenarios. Three statistical bias‑correction methods are applied to all relevant variables: Quantile Mapping, Quantile Delta Mapping, and CDF‑t, which are widely used in climate‑impact studies. Bias‑corrected historical and future time series are then combined through a PMP formulation of the storm maximization method. The three bias‑correction methods produce noticeably different historical PMP estimates and substantially different future PMP changes, even when they yield similar corrections at the level of individual variables. In particular, CDF‑t and Quantile Delta Mapping generally lead to pronounced increases in future PMP, whereas basic Quantile Mapping can produce decreasing PMP, especially under the high‑forcing SSP2‑4.5 scenario. This contrast arises from the way each method modifies the frequency and intensity of physically favourable combinations (temperature–humidity–wind–precipitation), leading to a wide range of future PMP outcomes.