Analysis of Atlantic hurricane forecasts for the 2017 season shows that CFAN’s accuracy in predicting tracks was superior to other government and publicly available forecasts beyond 2 days. CFAN uses a proprietary tracking algorithm to improve upon the forecasts provided by the ECMWF model, including a calibration that corrects for track forecast errors.
At five days lead-time, the average track error for the NHC forecasts was 170 miles. CFAN’s five-day track forecast error was smaller than the official NHC forecast by an average of 39 miles. For the 2018 season, CFAN plans to offer real-time track verification as part of our TropiCast forecast product.
While the NHC did not provide forecasts beyond 5 days, the average track error for CFAN’s forecast was 287 miles at 10 days lead-time. Longer lead times for hurricane forecasts are becoming increasingly important for planning for emergency management, restoring electric power, anticipating financial losses, and accelerating insurance payments.