Does increased model resolution over complex terrain improve precipitation for the right reasons? A North American Monsoon case study

Accurately simulating precipitation characteristics in complex terrain is a notorious problem in global climate models. Increasing model resolution, which also increases topographic resolution, seems to be one of the easiest ways to improve model hydrometeorology. Higher resolution is relatively cheap to develop, but computationally costly to run. The number of grid cells increases exponentially and the number of time steps increases roughly linearly with decreasing resolution, which requires more computer time and data storage, transfer, and processing.

Are the higher-resolution simulations in fact better, or do they look better for the wrong reasons? What variables or parameterizations related to precipitation processes lead to unrealistic simulations?

To endeavor to answer these questions, Coupled Model Intercomparison Project Phase 5 (CMIP5) historical runs and high-resolution Weather Research and Forecasting (WRF) model runs are compared to Tropical Rainfall Measuring Mission (TRMM) observations and the Modern-Era Retrospective Analysis for Research and Applications (MERRA). The North American Monsoon (NAM) is used as a case study. WRF runs are conducted with high-resolution and low-resolution terrain to separate the effect of model resolution from terrain resolution. First, the realism of simulated precipitation is assessed against TRMM data. Then MERRA is used to evaluate the roles of underlying variables, such as specific humidity, temperature, and moisture flux, in the accuracy of model precipitation.