The roles of bias-correction and resolution in regional climate simulations of summer extremes

Abstract

The suitability of dynamical downscaling in producing high-resolution climate scenarios for impact assessments is limited by the quality of the driving data and regional climate model (RCM) error. Multiple RCMs driven by a single global climate model simulation of current climate show a reduction in bias compared to the driving data, and the remaining bias motivates exploration of bias correction and higher RCM resolution. The merits of bias correcting the mean climate of the driving data (boundary bias correction) versus bias correcting the mean of the RCM output data are explored and compared to model resolution sensitivity. This analysis focuses on the simulation of summer temperature and precipitation extremes using a single RCM, the Nested Regional Climate Model (NRCM). The NRCM has a general cool bias for hot and cold extremes, a wet bias for wet extremes and a dry bias for dry extremes. Both bias corrections generally reduced the bias and overall error with some indication that boundary bias correction provided greater benefits than bias correcting the mean of the RCM output data, particularly for precipitation. High resolution tended not to lead to further improvements, though further work is needed using multiple resolution evaluation datasets and convection permitting resolution simulations to comprehensively assess the value of high resolution.