Combinatorial optimization of dynamics and physics in RegCM5 using a micro-genetic algorithm for precipitation and temperature simulations in Southeastern China

The fifth version of the regional climate model (RegCM5) has recently been released, incorporating updates in several model components such as the dynamic core and physical parameterizations. Traditionally, sensitivity tests based on random selection have been employed to identify the optimal sets from various combinations of model dynamics and physics. However, this approach is largely limited by computing power, often failing to explore the complete range of possible combinations necessary for an accurate representation of the regional climate. To overcome these limitations, advanced optimization techniques have emerged to efficiently explore the complete range of possible combinations, without relying solely on random-based sensitivity tests.  In this study, we employ a micro-genetic algorithm (micro-GA) for combinatorial optimization of the dynamic cores, cumulus parameterizations, and microphysics parameterizations in RegCM5. The model domains consist of one 20km mother domain covering the majority of East Asia, and two 2.5km nested domains covering the Yangtze River Delta (YRD) and Pearl River Delta (PRD), two densely populated regions in China. The focus is on conducting comparative assessments of simulated precipitation and temperature patterns in Southeastern China based on a series of experiments using the coupled RegCM5-micro-GA interface. The findings from this study will provide valuable insights to facilitate the wider use of RegCM5 by customizing its performance over the target regions.  [Acknowledgements] This research was supported by project GRF16308722, which was funded by the Research Grants Council (RGC) of Hong Kong.