Abstract

Climate model output often contains significant biases that need to be removed before the data can be used for impacts analysis or as a forcing input for other models. Many methods of bias correction have been proposed, and analysis of their ability to reproduce various statistical character... Show moreClimate model output often contains significant biases that need to be removed before the data can be used for impacts analysis or as a forcing input for other models. Many methods of bias correction have been proposed, and analysis of their ability to reproduce various statistical characteristics of the observational dataset has shown that distribution mapping performs best (Teutschbein and Seibert, J. Hydrol 2012). Distribution mapping corrects bias via a transfer function that systematically adjusts individual data points such that the cumulative distribution function (CDF) of the model data matches the CDF of the observations. Using cross-validation on synthetic data against an oracle method, we evaluate the performance of distribution mapping using both empirical CDFs (a.k.a. quantile mapping) and CDFs from fitted distributions (a.k.a. probability mapping) to determine the best implementation of distribution mapping. We then bias-correct daily temperature minimum and maximum data from the GCM-driven simulations in the NARCCAP output archive against the Maurer et al. 1/8-degree gridded daily observational dataset using distribution mapping with a fifteen-day moving window over the CONUS region. We examine the spatial characteristics of bias as revealed by this procedure, and the effects of bias correction on the climate change signal, finding regions both of enhancement and of diminishment. Show less