Verification of Multi-Model Ensemble Forecasts of North Atlantic Tropical Cyclones.

Verification of Multi-Model Ensemble Forecasts of North Atlantic Tropical Cyclones.

Nicholas Leonardo and Brian A. Colle

¹ School of Marine and Atmospheric Sciences, Stony Brook University, Stony Brook, NY

Medium-range (day 3-5) ensembles provide decision makers with probabilistic track and intensity forecasts for tropical cyclones (TC’s). However, verification of these forecasts is usually limited to assessing the mean error and spread of individual ensembles for one or two seasons. Very few studies have compared the ensemble track dispersion against the NHC’s climatology-based cone of uncertainty. There is also the question of whether the ensembles exhibit systematic position and intensity biases and if their errors have improved over the years. This study verifies the forecasts for the 2008 through 2016 North Atlantic seasons, focusing on three global ensembles (51 ECMWF members, 21 GEFS members, and 24 UKMET members) and their combination (“MMG”). The results are compared with several other models, such as the deterministic ECMWF (ECdet), HWRF, GFDL, and the NHC’s official forecast (OFCL). The probabilistic skill of the ensembles is quantified through Brier Skill Scores, using the ECdet and the NHC cone forecasts as references. The NHC’s best track data is used as the verifying analysis.

The total track errors of the MMG mean, ECdet, and OFCL are smallest on average at all lead times. All models have a significant negative (slow) along-track bias (80-230 km) by 120 h, while there is little bias in the cross-track direction. Much of the slow bias is associated with TC’s becoming extratropical at mid-latitudes. The 72 h mean track errors of the ensembles overall decrease by 70-140 km between 2010 and 2016, corresponding to 90-150 km decreases in the slow biases. All EPSs are underdispersed in the along-track direction, while the ECMWF is slightly overdispersed in the cross-track direction. The MMG and ECMWF track forecasts have more probabilistic skill than the ECdet and comparable skill to the OFCL cone forecast. TC intensity errors for the HWRF and GFDL are lower than the coarser models within the first 24 h, but are comparable to the ECdet at longer lead times. The ECMWF and MMG have comparable or better probabilistic intensity forecasts than the ECdet, while the GEFS’s weak bias limits its skill.