Diagnosing the Prediction of Tropical Cyclones and their Environments in Global Models (GFS and MPAS)

Chris Davis
July 31st 1:15 in room 2890

Tropical cyclone prediction at lead times of 5-10 days is challenging because it requires accurate prediction of organized, deep convection across a range of scales. Related to this, the formation and decay of tropical cyclones contribute significantly to medium range TC prediction. This talk examines the prediction of northern hemisphere TCs and their large-scale environments during 2012, 2013 and 2014. Results from both the Global Forecast System (GFS) and the Model for Prediction Across Scales (MPAS) are considered. It is shown that the GFS exhibits episodic large-scale moisture errors over the Western Pacific Ocean that can inhibit TC formation. Regional errors in deep convection are also shown to affect the forecast track of TCs. In contrast, MPAS exhibits more false alarms than GFS, especially after forecast day 7. These false alarms appear sensitive to dissipation in the model. Despite differences between the representation of TCs in the two models, they exhibited similar overall track and intensity forecast skill during August and September of 2013. Plans for MPAS development are summarized, with emphasis on facilitating comparison with the GFS for tropical cyclone prediction.