ECMWF by far best model for tropics days 3-5 in 2006

[LarryWx]

Please refer to p. 25's table 4 at the following link, which shows 2006 model tropical Atl. verifications on average:

http://www.nhc.noaa.gov/verification/pd ... n_2006.pdf

(LINK ABOVE NOW CORRECTED)

Although not the best within 48 hours and actually next to last at 12 hours, ECMWF did improve from 12 to 48 and was the best by far at 72-120 hours

1) ECMWF at 72 hours:
- Was < 50% of both the UKMET/CMC 72 hour errors
- Was < all other models' 72 hour error by at least 20%
- Was slightly better than CMC's 36 hour error!

2) ECMWF at 96 hours:
- Was only ~40% of UKMET/CMC 96 hour errors
- Was < ALL other models' 72 hour error!
- Was < CMC's 48 hour error!

3) ECMWF at 120 hours:
- CMC error 2.8 times ECMWF error at 120!!
- Was < all other models' 120 hour error by at least 29%
- Was < all other models' 96 hour error except GFDL
- Was < CMC/UKMET 72 hour error by 20-25%!!

So that last statement means that, on average, a 120 hour ECMWF run had averaged a 20-25% smaller error than the corresponding CMC/UKMET runs made a whopping 48 hours later!! Folks, that is incredible!

With numbers like this for 72-120 hours, the several hour delay in getting ECMWF guidance is trivial imho.

[windstorm99]

Link doesn't work.

[LarryWx]

Link doesn't work.

Sorry about that. Here's the correction:

http://www.nhc.noaa.gov/verification/pd ... n_2006.pdf

[WindRunner]

The only disappointing part of it is that they keep it so bottled up . . . much like the FSSE here in the US. Charging people for one of the best models out there, while not surprising, is kind of disappointing. I know the folks I know up at the HPC always try to use it whenever possible, so it's no surprise that it works just as well for hurricanes . . .