Precipitation simulation in global climate models: Impact of horizontal resolution and improved land surface scheme.

Abstract

This dissertation examines the improvement in the simulation of precipitation by using a high resolution model and a highly physically-based land surface scheme. To investigate the impact of the horizontal resolution on the simulated fields, six versions of NCAR's latest Community Climate Model (CCM2) are analyzed. The simulations were conducted for periods of 4-20 years. To study the effects of the improved land surface scheme in the model, the coupling of BATS to the standard model and a revised version of CCM2 with modified optical properties for clouds are compared. It is evident that the refinement of the model resolution adds further accuracy to the simulation. The increased resolution of the model improves the definition of the major mountain ranges and the average surface topography of lands and oceans. The taller mountains in a high resolution model play an important role in modifying the vapor flux over the land surface. They obstruct more water flow and reduce the specific humidity and the total precipitable water over land. The vapor transport to the land surface is also partly influenced by the changes in the large scale circulation. The adjusted surface pressures over the sea surface levels simulate more closely the position and magnitude of the observed pressure systems. A major problem in CCM2 is the excessive surface radiation, which is up to 50-100 Wm⁻² higher than the Surface Radiation Budget (SRB) dataset over vast areas in the summer hemisphere. At a high resolution, the lower land clouds reduce the planetary albedo and allow more solar radiation to reach the surface. But in spite of this, the high resolution model improves the amount and distribution of the land precipitation. The RCCM2-BATS model, which increases the optical thickness of the summer clouds, substantially reduces the overestimate of the surface radiation. This helps reduce the large discrepancies obtained in summer precipitation. The land surface scheme predicts the regional climate more accurately than the standard model, but its performance is restricted by the inputs and does not show any major improvement in precipitation predictions which are mainly controlled by various processes in the atmospheric model.