Variable-Resolution Tropical Cyclone Prediction using the Model for Prediction Across Scales (MPAS)

The Model for Prediction Across Scales (MPAS) was used to produce 92 global forecasts of tropical cyclones (TCs) across the Northern Hemisphere from August-October 2014. The results from fixed-resolution (15 km) and variable-resolution (ranging from 60 km to 15 km) 10-day forecasts will be compared, with a focus on the East Pacific domain because of its high-level of TC activity in 2014. In the variable-resolution MPAS, the 15-km mesh covers the East Pacific region and Central America. MPAS forecasts will also be compared with those from the NCEP Global Forecast System model for the same period. It will be shown that the variable-resolution MPAS performed nearly as well as the uniform resolution over the Eastern Pacific basin. Storm structure over the East Pacific (and other) basins will be compared where it will be shown that MPAS produced more realistic variability in storm size compared with GFS. It will be shown that biases in MPAS were reduced substantially in 2014, especially concerning surface moisture fluxes. Nonetheless, we will show that MPAS produces too many TCs at long lead times, whereas GFS produces too few. Physical reasons for these biases will be discussed.