Jon Moskaitis, Will Komaromi, and James Doyle
Here we make a preliminary assessment of the COAMPS-TC track and intensity forecast performance for Hurricane Ian, focusing in particular on the model predictions valid at the time of Ian’s Florida landfall. COAMPS-TC is the real-time deterministic tropical cyclone prediction system supporting the NHCI project, run by NRL-Monterey. For every Atlantic tropical cyclone, forecasts are generated every 6 h out to a lead time of 120 h, using the “CTCX” version of COAMPS-TC (i.e. the global model used for initial and lateral boundary conditions is NOAA’s GFS). The observed track and intensity values used in the validation are from the National Hurricane Center’s (NHC) working “best track” analysis. These values are produced at six hour intervals, so for the purposes of this model validation exercise, landfall is considered to be at 1800 UTC 28 Sept 2022, which is the closest best track time to the actual landfall time of 1905 UTC 28 Sept 2022.
Fig. 1 shows all CTCX track forecasts issued for Ian from the storm’s formation time up to the time of its landfall on the Florida peninsula.
Early CTCX track forecasts (initial times between 0600 UTC 23 Sept 2022 and 0600 UTC 25 Sept 2022) all indicated the storm would cross Cuba and then recurve toward the NNE, but the model solutions ranged from Florida landfall in the Keys to a storm approaching the Big Bend area of the Florida Gulf coast. However, the 1200 UTC 25 Sept 2022 CTCX forecast and all subsequent CTCX forecasts leading up to the Florida landfall indicated a landfall location along the west coast of the Florida peninsula between Ft. Myers and Tampa Bay. These CTCX forecasts are shown in more detail in the upper-left panel of Fig. 2. Here it can be seen that the timing of landfall was excellent in CTCX, though storm center generally crossed the coast somewhat too far north.
For context, Fig. 2 also shows results for three other operational TC prediction models run by NOAA and routinely relied on for guidance at NHC, the regional tropical cyclone prediction models HMON and HWRF, and the GFS global model. These other models tended to take the forecast track too far west, with a position at landfall time often located over the Gulf of Mexico. Ian was a very challenging track prediction case, and CTCX did exceptionally well compared to the other U.S. operational TC prediction models.
Fig. 3 shows the CTCX intensity predictions corresponding to the track predictions shown in Fig. 1. All CTCX forecasts indicated significant intensification of Ian, with nearly all the forecasts peaking at major hurricane intensity (>= 96 kt) with some as high as Category 5 (>=137 kt). The observed peak intensity was 135 kt, a strong Category 4 storm.
The observed landfall intensity was 130 kt as analyzed by NHC, but seven hours prior to landfall (1200 UTC 28 Sept 2022) when the entire core of the storm was still over water, NHC analyzed the intensity at 135 kt. This could perhaps be described as the “pre-landfall intensity”, representing the intensity of the storm before the inner core begins to experience the much higher frictional energy dissipation over land. Table 1 shows the pre-landfall intensity for the CTCX forecasts represented in Fig. 2, which all had reasonably good forecasts of landfall position and time. The average pre-landfall intensity for these CTCX forecasts is 120 kt, 15 kt below the observed value. Most CTCX forecasts predicted a Category 4 pre-landfall intensity, but only one forecast erred on the high side of the observed value. The intensity forecast leading up to landfall was extremely challenging, as in reality Ian rapidly intensified coming out of an eyewall replacement cycle. Ian will be a great case for testing higher-resolution versions of COAMPS-TC that we know have a better chance of accurately simulating complicated inner-core dynamics such as eyewall replacement cycles.
Much validation work for the COAMPS-TC forecasts of Hurricane Ian remains to be done in the coming days and months, but these preliminary results regarding our forecasts of the Florida landfall indicate the model performed quite well for a very challenging storm.
by Martha Schönau, Luca Centurioni, Steve Jayne, and Elizabeth Sanabia