Variable-resolution Frameworks for the Simulation of Tropical Cyclones in Global Atmospheric General Circulation Models.

Abstract

The ability of atmospheric General Circulation Models (GCMs) to resolve tropical cyclones in the climate system has traditionally been difficult. The challenges include adequately capturing storms which are small in size relative to model grids and the fact that key thermodynamic processes require a significant level of parameterization. At traditional GCM grid spacings of 50-300 km tropical cyclones are severely under-resolved, if not completely unresolved. This thesis explores a variable-resolution global model approach that allows for high spatial resolutions in areas of interest, such as low-latitude ocean basins where tropical cyclogenesis occurs. Such GCM designs with multi-resolution meshes serve to bridge the gap between globally-uniform grids and limited area models and have the potential to become a future tool for regional climate assessments.

A statically-nested, variable-resolution option has recently been introduced into the Department of Energy/National Center for Atmospheric Research (DoE/NCAR) Community Atmosphere Model's (CAM) Spectral Element (SE) dynamical core. Using an idealized tropical cyclone test, variable-resolution meshes are shown to significantly lessen computational requirements in regional GCM studies. Furthermore, the tropical cyclone simulations are free of spurious numerical errors at the resolution interfaces. Utilizing aquaplanet simulations as an intermediate test between idealized simulations and fully-coupled climate model runs, climate statistics within refined patches are shown to be well-matched to globally-uniform simulations of the same grid spacing. Facets of the CAM version 4 (CAM4) subgrid physical parameterizations are likely too scale sensitive for variable-resolution applications, but the newer CAM5 package is vastly improved in performance at multiple grid spacings.

Multi-decadal simulations following 'Atmospheric Model Intercomparison Project' protocols have been conducted with variable-resolution grids. Climate statistics produced using ocean-centered, variable-resolution nests are almost identical to those from ``unrefined'' simulations using globally-uniform grids. Regional climatology is improved by refinement due to the better representation of topography. Lastly, tropical cyclone structure, intensity, and interannual variability in the Atlantic Ocean are all significantly improved with the use of variable-resolution grids. These attributes are well-matched to other modeling studies at similar horizontal resolutions while only requiring a fraction of the computational cost, highlighting the potential of using variable-resolution GCMs to simulate small-scale phenomena.