The NA-CORDEX data archive provides climate change scenario data over North America for use in impacts, decision-making, and climate science. This data comes from RCM simulations dynamically downscaling CMIP5 GCMs run as part of the WCRP Coordinated Regional Climate Downscaling Experiment (CORDEX).

These climate model outputs often contain bias that hinder their use in impacts and adaptation analysis. To improve their usability, the NA-CORDEX data archive will provide both raw simulation outputs and outputs with statistical adjustments to reduce the bias. In applications that require multiple variables, it is desirable not just to correct the marginal distributions of the individual variables, but also to preserve and if possible correct the relationships between variables.

We perform both univariate and multivariate bias-correction of 7 different impacts-relevant variables from CORDEX simulations against the gridMET gridded daily observational dataset using using Kernel Density Distribtion Mapping (KDDM), a nonparametric quantile mapping technique). To perform the multivariate correction, we apply KDDM in conjunction with a sequence of random orthogonal rotations, following the approach of the N-pdft algorithm in Cannon (2017).

We compare the results of the two corrections in 10 study locations of interest across North America using a set of standard descriptive metrics developed for the DOE-funded Framework for Assesing Climate's Energy-water-land nexus by Targeted Simulations (FACETS) program. Making use of the "cascade of metrics" approach from the FACETS model evaluation framework, we use the results of simpler analyses to target the application of more complex analyses. We then identify cases where there is a comparatively large disagreement between the two approaches and investigate whether there is a systematic pattern in the bias that indicates likely targets for deeper analysis of the simulations or the bias-correction methodology.