Impact Analysis of Univariate and Multivariate Bias Correction on Rice Irrigation Water Needs in Jiangxi Province, China

Abstract

Regional climate models (RCMs) provide an improved representation of climate information as compared to global climate models (GCMs). However, in climate-agricultural impact studies, accurate and interdependent local-scale climate variables are preferable, but both RCMs and GCMs are still subjected to bias. This study compares univariate bias correction (UBC) and multivariate bias correction (MBC) method to simulate rice irrigation water needs (IWNs) in Jiangxi Province, China. This research uses the daily output of Hadley Centre Global Environmental Model version 3 regional climate model (HadGEM3-RA) forced with ERAINT (ECMWF ERA Interim) data and 13 Jiangxi ground-based observations, and the observation data are reference data with 1989–2005 defined as a calibration period and 2006–2007 as a validation period. The result shows that UBC and MBC methods favorably bias-corrected all climate variables during the calibration period, but still no significant difference is noted between the two methods. However, the UBC ignores the relationship between climate variables, while MBC preserves the climate variables’ interdependence which affect subsequent analyses. In rice IWNs simulation analysis, MBC has better skill at correcting bias compare to UBC in ETo (evapotranspiration) and Peff (effective rainfall) components. Nonetheless, both methods have a low ability to correct extreme values bias. Overall, both techniques successfully reduce bias, even though they are still less effective for precipitation compared to maximum and minimum temperature, relative humidity and windspeed.