On bias correction in drought frequency analysis based on climate models

Abstract

Assessment of future drought characteristics based on climate models is difficult as climate models usually have bias in simulating precipitation frequency and intensity. In this study, we examine the significance of bias correction in the context of drought frequency and scenario analysis using output from climate models. In particular, we use three bias correction techniques with different emphases and complexities to investigate how they affect the results of drought frequency and severity based on climate models. The characteristics of drought are investigated using regional climate model (RCM) output from the North American Regional Climate Change Assessment Program (NARCCAP). The Standardized Precipitation Index (SPI) is used to compare and forecast drought characteristics at different timescales. Systematic biases in the RCM precipitation output are corrected against the National Centers for Environmental Prediction (NCEP) North American Regional Reanalysis (NARR) data and the bias-corrected RCM historical simulations. Preserving mean and standard deviation of NARR precipitation is essential in drought frequency analysis. The results demonstrate that bias correction significantly decreases the RCM errors in reproducing drought frequency derived from the NARR data. Different timescales of input precipitation in the bias corrections show similar results. The relative changes in drought frequency in future scenario compared to historical scenario are similar whether both scenarios are bias corrected or both are not bias corrected.