Performance of Regional Climate Model RegCM3 over Thailand

Abstract

This study assesses the performance of the regional climate model RegCM3 and its sen- sitivity to selected physical parameterizations and 1-way double nesting over Thailand. A total of 16 simulation experiments were conducted using different combinations of convective and ocean flux parameterization schemes on a 60 km resolution (D1) and a nested 20 km resolution (D2) domain. The simulated results were compared with the Thai Meteorological Department (TMD) data for (near-sur- face) temperature and precipitation, and ERA40 reanalysis data for upper-level synoptic winds. In both the D1 and D2 experiments, considerable systematic underestimation of temperature was found for the upper sub-regions (Central-East, Northeast, and North) of Thailand with the maximum cold bias of 6.0°C. Seasonal discrepancies (between dry and wet seasons) in precipitation, as observed in these sub-regions, were well captured. Over the lower sub-region (South), cold biases were smaller than in the upper sub-regions, and the relatively high rates of precipitation normally observed in the dry season were reproduced by the model. The main features of seasonal synoptic winds were fairly simulated with better performance seen in the dry season. For the convective parameterization schemes, MIT-Emanuel performed best on temperature, followed by Anthes-Kuo. However, in the wet season, the former scheme produced large wet biases, particularly for the upper sub-regions. Grell-Arakawa-Schubert yielded smaller cold biases than Grell-Fritsch-Chappell, but their relative performances on precipitation were not conclusive. For ocean flux parameterization, the BATS scheme, in comparison with the Zeng scheme, provided better predictions for temperature and yielded more precipitation. The nested modeling enhanced the spatial details of model outputs but did not necessarily improve the overall performance in a particular sub-region. A gridded observa- tion dataset from the University of Delaware (UDEL) was used to qualitatively examine the model performance on temperature and precipitation, and reasonable agreement was found for both TMD and UDEL datasets. An additional simulation test was conducted to examine the effect of a land cover modification (here, from 'irrigated crop' to 'crop/mixed farming') as a potential technique to alleviate the problem of temperature underestimation. It was found that using the modified land cover helped reduce the degree of bias by 2.2 to 3.3°C in the upper sub-regions during the dry season.