On searching for optimized set of physical parameterization schemes in a multi-physics land surface process model

Abstract

Optimization of land surface models has been very challenging due to the increasing complexity of such models. Typical parameter calibration techniques often limit the so- lution of the spatiotemporal discrepancy in the modeling performance levels especially for regional applications. Thus, in this study, an attempt was made to perform scheme- based model optimization by designing a framework for coupling a micro-genetic algo- rithm (micro-GA) with the Noah land surface model that has multiple physics options (Noah-MP). Micro-GA controls the scheme selections in 10 different land surface pa- rameterization fields in Noah-MP in order to extract the optimal scheme combination for a certain region. This coupling framework was successfully applied to the optimiza- tion of the surface water partitioning in the Korean Peninsula, promising not only the effectiveness of the scheme-based optimization but also model diagnosis capability by exploring the scheme sensitivity during the micro-GA evolution process. Then, the method was applied to four different regions in East Asia that have different climatic characteristics. The results indicate that (1) the optimal scheme combinations vary with the regions, (2) schemes related to the surface water partitioning are important for the modeling accuracy, and (3) specialized post-parameter optimization for each region may be required.