This study aims at a new supplementary approach to identify optimal configurations of physics parameterizations in regional climate models (RCMs) and earth system models (ESMs). Traditional approaches separately evaluate variable performance, which may lead to an inappropriate selection of physics parameterization combinations. Besides traditional approaches, we suggest an additional selection approach by considering the joint dependence structure (covariance structure) between key meteorological variables, i.e., precipitation P and temperature T. This is accomplished by empirical P and T copula functions and the χ 2
-test, and is demonstrated in two locations in Kenya with different major precipitation processes. It is shown that the selection based on traditional approaches alone may lead to nonoptimal decisions in terms of joint dependence structure between P and T. It was found that the copula-based approach may reduce the need for complex multivariate bias correction, as demonstrated using local intensity scaling for P and linear scaling for T. The new approach may contribute to improving RCM and ESM simulations and climate-impact studies worldwide.