Deterministic Forecasts of Tropical Cyclones Using a Variable-Resolution Global Model

Tropical cyclones have traditionally been difficult features for global numerical models to accurately simulate given their small size relative to model grid spacing. Even with recent improvements in operational forecast model resolution, cyclones remain underresolved and subsequent prediction difficulties persist. In an effort to alleviate these issues, the use of limited area models (LAMs) at higher resolutions has become popular. However, by definition, these models require boundary conditions and may lack two-way communication with the exterior domain. Variable-resolution global dynamical models can serve as the bridge between traditional global forecast models and high-resolution LAMs. These models can utilize existing computing platforms to approach 10 km or finer resolution in low-latitude ocean basins of interest to hurricane forecasters. They do so while maintaining global continuity, therefore eliminating the need for the externally-forced and possibly numerically and physically inconsistent boundary conditions required by LAMs.

A statically-nested, variable-resolution option has recently been introduced into the Community Atmosphere Model's (CAM) Spectral Element (SE) dynamical core. We present deterministic CAM-SE model simulations of observed tropical cyclones and compare the model's prediction of storm track and intensity to other global and regional models used operationally by hurricane forecast centers. The simulations are run on a 55 km global cubed-sphere grid with additional refinement to 13 km over the Atlantic and Eastern Pacific Oceans. Each forecast was integrated for 10 days and was initialized twice daily for three months during 2012 and 2013. We also investigate cyclone genesis, and whether locally high resolution in a global model leads to improved forecast skill at longer lead times. In addition to general performance statistics, we also consider the behavior of existing parameterizations in CAM with respect to tropical cyclone forecasting. We also highlight the improvement in model throughput by the use of variable resolution and discuss the potential computational benefits of such a setup as a numerical weather prediction tool.