Exploring a Multi-resolution Approach Using AMIP Simulations
Friday, 19 December 2014: 4:00 PM
This study presents a diagnosis of a multi-resolution approach to the Model for Prediction Across Scales - Atmosphere (MPAS-A) for modeling regional climate in comparison to global high and low resolution simulations. Four experiments are conducted in AMIP-style with prescribed sea surface temperature for 1999 – 2009. In the first two experiments, MPAS-A is configured using quasi-uniform grids at 120 km and 30 km resolution. In the other two experiments, MPAS-A is configured using variable resolution with local refinement at 30 km resolution over North America and South America, respectively, embedded inside a quasi-uniform resolution domain at 120 km. The variable resolution simulations are compared against the quasi-uniform resolution global simulations to determine how well the variable-resolution simulations reproduce the features simulated by the globally high-resolution model in the refined resolution domain, and evaluate the upscale effects from the local grid refinement in the coarse resolution region. In previous analysis of idealized aqua-planet simulations with a refined domain over the tropics, high-resolution features only developed near the outflow boundary of the refined region. In contrast, the AMIP simulations with variable-resolution grids are able to reproduce the features of the quasi-uniform high-resolution simulation within the refined domain, particularly in South America during warm season. This suggests that regional forcings such as topography and land surface heterogeneity have more dominant control over the regional climate in North and South America so simulations that resolve those forcings using global quasi-uniform high resolution or variable resolution with grid refinement produce similar results. Outside of the refined grids, some remote upscaled effects are detected that suggest potentially important linkages of processes over North America with East Asia and the western Pacific. Our results provide support for the multi-resolution approach as a computationally efficient method for modeling regional climate.