Comparison of quantile mapping methods for daily precipitation correction and future projections under the SSP5-8.5

This study applied four Quantile Mapping (QM) methods (Parametric Linear, PL; Parametric Scale, PS; Empirical Quantile Mapping, EQM; and Multivariate Bias Correction using N-dimensional Probability Density Function Transform, MBCn) to correct daily precipitation data from 61 stations in South Korea. The performances of these methods were evaluated using six precipitation indices (r10 mm, r20 mm, SDII, R95ptot, R99ptot, and Rx1day) and five evaluation metrics (Percent bias, Pbias; Mean absolute error, MAE; Kling-Gupta efficiency, KGE; Euclidean Distance, ED and Jensen-Shannon divergence, JSD) for 14 General Circulation Models (GCMs) from the Coupled Model Intercomparison Project Phase 6(CMIP6). Additionally, the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), a multi-criteria decision-making method, was used to rank the GCMs based on performance. The entropy theory was applied to calculate the weights for each evaluation criterion. For future projections, the Shared Socioeconomic Pathway (SSP) 5-8.5 scenario was used to predict future daily precipitation for each GCM and QM method, and a Multi-Model Ensemble was constructed based on TOPSIS weights. Furthermore, the uncertainty of future precipitation projections for each QM method was quantified using Reliability Ensemble Averaging (REA). The results showed that MBCn and EQM performed the highest in historical reproducibility, particularly in correcting extreme precipitation events (MBCn-EVS:0.99, KGE:0.99, Pbias:0.42; EQM-EVS:0.94, KGE:0.97, Pbias:-1.2). In contrast, PL showed the lowest performance, with differences from observed values exceeding 500 mm, indicating significantly lower reproducibility and extreme precipitation correction compared to other methods. MBCn's future precipitation projections effectively reflected the greenhouse gas emission trends in the SSP5-8.5 scenario, while PL failed to account for extreme precipitation events. Additionally, MBCn exhibited the highest reliability in uncertainty analysis at 0.51, while PL showed the lowest at 0.016. This study enhances the reliability of precipitation projections under climate change through the comparison of bias correction methods and the quantification of uncertainties, providing valuable insights for policy-making and water resource management strategies.