As of the Wednesday morning cycle, the 1000-member Google Deepmind ensemble gives this a 27% chance of becoming a tropical storm late next week, and of those members, they show a 60% chance of it becoming the season's first hurricane and even a 22% chance of reaching major hurricane intensity (Cat3+). Other AI and conventional ensembles agree with a decent chance of something forming here in 7-10 days. There's a LOT of uncertainty of course, but the signal in the models is there.
 |
| Image from https://tcatlas.org/realtime_ir.html |
This area is not as favored for development in early October.as one might think, at least for storms that eventually reach hurricane intensity. The last one that did so was Milton in 2024, then Roxanne in 1995, then Fern in 1967. It's a very short list.
Just for curiosity's sake, I ran some preliminary homogeneous track and intensity verification statistics on the meager activity so far this season. This covers Arthur through Hanna, and you can see on the charts that there are actually ZERO 5-day (120-hour) verification times, but it's what we've got. All six of the 4-day (96-hour) verifying forecasts come from Fay.
For track, the Google Deepmind ensemble mean (GDMI) leads the pack overall, then it gets a bit more noisy. The National Hurricane Center's forecasts (OFCL) were very closely aligned with GDMI except for the six forecasts for Fay when the model was superior.
For intensity, it's definitely a mixed bag, but the thing I see is the two consensus aides, HCCA and IVCN, did the best overall. The Google Deepmind ensemble mean was quite good too, but lost a little bit of ground on the 4-day forecasts of Fay. Also worth noting is that the HWRF hurricane model (HWFI) beat the NHC's average forecast error at every lead time -- none of the other hurricane models did that (HMNI, HFAI, HFBI, CTCI).
No comments:
Post a Comment