Assessing Model Guidance for Hurricane Ian

Hurricane Ian made landfall in southwest Florida on 28 September 2023 as a Category 4 storm, causing at least 149 deaths and over $100 billion of damage in the state. EMC’s Model Evaluation Group (MEG) assessed model guidance for Ian, examining track and intensity forecasts for global and hurricane models, as well as Quantitative Precipitation Forecasts (QPF) from a range of models. An overview of the MEG findings will be presented, with both forecast maps and statistical plots.

Longer-range forecasts from deterministic and ensemble global systems correctly forecasted a significant tropical system in the Gulf of Mexico up to 9 days in advance. NOAA’s GFS/GEFS system showed a significant left-of-track bias with Ian forecasts up until very close to landfall, while the ECMWF system was notably better, and the UKMET system had the smallest track errors. The GFS/GEFS, however, had the smallest intensity bias of the global models.

The HWRF and HMON models both also exhibited a significant left-of-track bias for Ian, with both models, like the GFS, showing a landfall in the Florida Panhandle until two days prior to landfall. The left tracks likely contributed to all three models predicting the second U.S. landfall (along the Carolina coast) too far west. The HWRF and HMON did, however, correctly forecast that Ian would rapidly intensify into a major hurricane. The experimental HAFS system performed better than either the HWRF or HMON with Ian’s track and had comparable intensity forecasts.

Model QPF suffered from the left-of-track bias and showed a catastrophic rainfall event for Tampa, only correctly shifting the axis of heaviest rainfall south very close to the start of the period of the most intense rainfall. Overall, convection-allowing models (CAMs) performed the best with rainfall guidance, especially with the convection well in advance of the Ian landfall. That said, there was an overall high bias for the highest amounts and coverage of the heaviest rain; the localized probability-matched mean products from the High-Resolution Ensemble Forecast system (HREF) did an overall good job with the QPF and were an improvement of the “regular” probability-matched mean products.