Idealized Tropical Cyclone Test Cases for Global Models

Increasing computational capabilities and the evolving computer architectures delivering those capabilities have spurred the development of new formulations for the atmospheric dynamical cores (fluid-flow solvers) that are the basis of operational global NWP and climate models.  In addition, research global atmospheric models are already being configured and used at convection permitting mesh spacings which necessitate the solution of the nonhydrostatic equations, providing further impetus for dynamical core development.  There are a limited number of idealized tests used to evaluate global dynamical cores, including dry and moist baroclinic waves, mountain waves and convection on reduced-radius spheres, and idealized tropical cyclones in an aquaplanet configuration, and these tests have formed the basis for the Dynamical Core Model Intercomparison Project (DCMIP) in 2012 and 2016 and other dynamical core evaluation projects.  

Given the obvious importance of tropical cyclone prediction, in this work we revisit the idealized tropical cyclone test case introduced by Reed and Jablonowski (2012, JAMES) that uses an aquaplanet configuration perturbed with an initial weak vortex and that employs simplified physics.  The test results in their paper, and subsequent papers presenting results from other dynamical cores, are for configurations using resolutions only as high as 0.25 degrees. Using the global Model for Prediction Across Scales (MPAS), we have computed solutions for this test case on meshes with cell spacing as small as 3 km and high vertical resolution, and we find that important features of the idealized cyclone are not resolvable at coarser-than convection permitting resolutions, including the converged eyewall diameter.  We also find that the converged solution is markedly different from observed tropical cyclones, for example the radius of maximum winds is much too small and the inflow depth is much too shallow.  These features stem from the configuration of the environment and the simplified physics used in the test.  We present modifications to both the environment and the simplified physics resulting in an idealized tropical cyclone that more-closely resembles observed cyclones.  Finally, we stress the importance of using test cases with known solutions in the evaluation of dynamical cores, and with simulated atmospheric phenomena that resemble observed phenomena.  The former is absolutely necessary to scientifically evaluate the performance of dynamical cores, and the latter is necessary if the test is to be relevant to global atmospheric model applications.