fvGFS Forecasts of the Rapid Intensification of Hurricane Matthew (2016)

Hurricane Matthew was the strongest Atlantic tropical cyclone of the 2016 season, and rapidly intensified into a Category 5 storm in the Caribbean before affecting Haiti, Cuba, and the Southeast US coast. In this study, the fvGFS (FV3 dynamical core with GFS physics) model at 2- km horizontal resolution is used to analyze the evolution of Matthew in the Caribbean sea and western Atlantic. 20 different forecasts were performed, with five different lead times (centered around genesis) and 4 different physics configurations, including turning the convective scheme on and off and making alterations to the microphysics scheme.

The fvGFS track forecasts all had a right and fast bias, similar to the operational GFS forecasts for Matthew from the same time period. Analysis of the track forecasts shows that the runs with the convective scheme turned off have a stronger (more realistic) deep-layer ridge, indicating that the convective scheme may have been partially responsible for some of the track bias.

The intensity forecasts from the fvGFS runs showed a wide range of variability. The rapid intensification of Matthew was somewhat unexpected due to close proximity to a region of strong shear. In the fvGFS forecasts, vortex tilt is found to be critical to the intensity evolution, with the cases that showed the vortex becoming aligned at the correct time having an intensity evolution that was closer to the observed storm. Interestingly, almost all of the fvGFS forecasts, including those that initiated intensification correctly, showed a 6-12 hour pause in the intensification process that led to some intensity forecast error and lag in the peak intensity. This was found to be at least partially due to the model representation of a convective feature on the north and east side of the TC, with outflow from this feature briefly hampering the outflow over the TC core. Although this convective feature was present in the observed TC, it did not hamper intensification at the same time as shown in the model forecast. These results indicate that model configuration is key to track and intensity evolution of forecasted TCs, and also show how critical forecasted TC structure is to the evolution of TCs in numerical models.