Factorial uncertainty source quantification for RegCM projections

Abstract:

Generating high-resolution climate projections is highly important for regional-scale impact assessments under changing climate conditions. Regional Climate Model (RegCM4) is considered as an effective dynamical downscaling tool, which has been widely utilized in future climate projections. It is well known that the choices of physical schemes, input General Circulation Model (GCM) datasets, and emission scenarios may cause significant uncertainties in future climate projections. Ensemble approaches are always used to deal with these uncertainties. However, it is challenging to quantify the sources of these uncertainties due to their multi-level interactive characteristics.

In this study, 24 RegCM runs (3 GCMs × 2 emission scenarios × 4 physical scheme combinations) have been conducted covering the periods of 2050 – 2059 and 2090 – 2099. Then a factorial uncertainty source quantification approach has been proposed to analyze the contributions of each uncertainty source (i.e., GCMs, physical schemes, and emission scenarios) and their interactions for several climatic variables (i.e., temperature, precipitation, soil moisture and evapotranspiration). The advantages of the proposed approach can be summarized as: (1) it can deal with not only the quantitative uncertainty sources, but also the qualitive ones; and (2) it can analyze the interactions between different uncertainty sources (like the interaction between GCM and emission scenarios). The obtained results could be used to evaluate the sensitivity of each uncertainty source and their interactions, as well as its developing trend in the future.