Local Mean Bias Correction in a Regional Model Downscaling: A Case Study of the South China Sea Summer Monsoon of 1998

Abstract

Regional climate models used to resolve high-resolution local circulation by dynamically downscaling from a coarse-resolution analysis may have difficulty in keeping large-scale information the same as in the analysis. Such difficulty produces large-scale error or bias, mainly due to mathematically ill-posed lateral boundary conditions. To reduce this type of error, in this paper the authors propose a scheme called local mean bias correction (LMBC). LMBC has means in x and in y directions locally over the model domain, but not a single mean over the entire domain. In spectral space, local means are represented by the zero wavenumber in the x direction with all wavenumbers in the y direction, and the zero wavenumber in the y direction with all the wavenumbers in the x direction. The local mean perturbation can be removed to correct the local mean bias where the perturbation is defined as the difference between a regional field and a base field from the analysis. The LMBC constitutes a simple methodology to ameliorate the influence of the lateral boundary conditions (LBCs) on the integration of the regional climate model (RCM); using it in this study improved the regional climate simulation of the 1998 South China Sea (SCS) summer monsoon. Generally speaking, this scheme proved to have a well-simulated circulation during all periods of monsoon activities examined. Circulation errors during the SCS summer monsoon onset can be corrected using this LMBC scheme, though the rainfall amount was always underestimated as compared to the observation. This research demonstrated that shifting the domain north, east, west, and south as well as enlarging domain size produced results that are consistent in different cases over different years.