How does bias correction impact simulated drought characteristics by Regional Climate Models?

While numerous studies have explored the impact of bias correction on climate projections of precipitation and temperature, few have investigated the influence of these corrected climate outputs on the estimation of drought characteristics. In response to this gap, this study aims to quantify and compare meteorological drought characteristics derived from climate simulation datasets with and without bias correction. We conducted our investigation using downscaled data of six Coordinated Regional Climate Downscaling Experiment-Southeast Asia (CORDEX-SEA) experiments for the period 1976 to 2005, with Quantile Mapping (QM) being employed as our bias correction (BC) technique. The drought characteristics examined in this study are frequency, duration, severity, intensity, and geographical extent, utilizing the 12-month Standardized Precipitation Evapotranspiration Index (SPEI). We performed univariate cross-validation QM BC for both precipitation and temperature at a grid-cell wide, demonstrating comparable performances for both the training (1976-1995) and testing (1996-2005) periods. The drought characteristics analysis was conducted for the entire temporal dataset (1976-2005) to utilize over 30 years of data for drought computation. Compared to the reference drought characteristics derived from APHRODITE, the drought characteristics from both the uncorrected and bias-corrected models show a similar pattern in the distribution of relative differences, with moderately lower differences for the models with bias correction, particularly in mainland Southeast Asia, except for the Myanmar region. Additionally, the application of QM showed moderate improvements in determining drought intensity and geographical extent. However, the implementation of QM only resulted in marginal improvements for drought duration and severity and did not perform well in reproducing the number of drought events. We further demonstrated that the effectiveness of bias correction on drought characteristics variation across topographies and land covers, with considerable improvement observed in low-elevation and grassland/arable land regions. To this end, this study provides a new perspective on the added value of bias correction techniques in climate change projections for drought indicator computation.