Evaluating Tropical Cyclones Simulated by a Global Convection-Permitting Model.

Global models traditionally have low resolutions (~10-15 km) that impede accurate tropical cyclone intensity forecasts. Recent breakthroughs in global convection-permitting numerical weather prediction models have led to cost-efficient, high-resolution forecasts at the global scale. While numerous studies have examined the predictive skill of tropical cyclones in coarse resolution global models, few have investigated their predictive skill at convection-permitting resolutions. For this study, a 20-day simulation was produced with the Model for Prediction Across Scales (MPAS) using a globally uniform 4-km mesh. Three additional perturbed runs with lengths of 20, 15, and 10 days were produced using the same mesh. In order to determine MPAS’s predictive skill with regard to tropical cyclone tracks and intensities, the GFDL Vortex Tracker was adapted to the high-resolution model output, and the simulated tracks and intensities compared with “best track” data. The simulations were furthermore compared against surface wind analyses and satellite observations of rainfall. Overall, the evaluation demonstrated that MPAS tended to spin up more cyclones than observed (4 hits, 2 misses, and 18 false alarms). While reproducing the intensity of Typhoon Son-Tinh (2012), the model tended to over-intensify cyclones in the Indian Ocean. These biases indicate that the global 4-km MPAS needs to be improved before it can be a “one stop shop” for accurate tropical cyclone track and intensity forecasts across the globe.