Impacts of Land Surface Model Complexity on a Regional Simulation of a Tropical Synoptic Event

Abstract

A multimode Chameleon Surface Model (CHASM) with different levels of complexity in parameterizing surface energy balance is coupled to a limited-area model (DARLAM) to investigate the impacts of complexity in land surface representations on the model simulation of a tropical synoptic event. A low pressure system is examined in two sets of numerical experiments to discuss the following. (i) Does land surface parameterization influence regional numerical weather simulations? (ii) Can the complexity of land surface schemes in numerical models be represented by parameter tuning? The model-simulated tracks of the low pressure center do not, overall, show large sensitivity to the different CHASM modes coupled to the limited-area model. However, the landing position of the system, as one measurement of the track difference, can be influenced by several degrees in latitude and about one degree in longitude. Some of the track differences are larger than the intrinsic numerical noise in the model estimated from two sets of random perturbation runs. In addition, the landing time of the low pressure system can differ by about 14 h. The differences in the model-simulated central pressure exceed the model intrinsic numerical noise and such variations consistent with the differences seen in simulated surface fluxes. Furthermore, different complexity in the land surface scheme can significantly affect the model rainfall and temperature simulations associated with the low center, with differences in rainfall up to 20 mm day?1 and in surface temperature up to 2°C. Explicitly representing surface resistance and bare ground evaporation components in CHASM produces the most significant impacts on the surface processes. Results from the second set of experiments, in which the CHASM modes are calibrated by parameter tuning, demonstrate that the effects of the physical processes represented by extra complexity in some CHASM modes cannot be substituted for by parameter tuning in simplified land surface schemes.