Multi-Level Factorial Sensitivity Analysis of REGCM4 Physical Scheme Combinations Over Canada

Abstract

Climate change has been considered as one of the most important problems around the world since it always causes severe, extensive and irreversible consequences. It is thus desired to generate reliable high-resolution climate projections for supporting impact assessments under changing climate conditions. In this study, a multi-level factorial design has been proposed to conduct the sensitivity analysis of RegCM4 physical scheme combinations. In detail, the major contributing factors for four climatic variables are selected based on the predetermined contribution level; then the F-test has been applied to evaluate the statistical significance of each individual physical scheme or the interaction among multiple physical schemes. The proposed design has been applied to Canada to illustrate its effectiveness. The results indicate that there exist considerable spatial and temporal simulated differences when choosing varied physical scheme combinations. Although the single physical schemes have dominant influences on simulated differences, the effects explained by their interactions can not be neglected. Moreover, the PBL scheme, moisture scheme and land surface model are found to be the dominant factors of temperature, precipitation and wind speed simulations, respectively. Furthermore, the vegetation cover conditions may affect the cumulus convective activities to some extent, accounting for the spatial and temporal statistical significance variations of cumulus convective scheme. The obtained results could help screen out the most sensitive physical schemes or interactions considering specific region and season. On this basis, the computational costs of follow-up long-term simulations are considered affordable, which further assist in selecting the optimized ii physical scheme combination and then generating reliable high-resolution climate projections over Canada.