GFDL bias?

[curtadams]

Many have noted the GFDL's tendency to overestimate cyclones in their early development. It's very eager to crank a TD right on up to a Cat3+ in 2 days, which is pretty exceptional albeit not impossible. I've noticed that it has a marked bias with eye size. A day or too after the run starts it likes to predict a very tight eye, which is probably partly responsible for the tendency to predict popup majors. But, as you head out to day 5 the eye usually seems to get HUGE - and the peak intensity drops. Is there some quirk with its gridding or fudge factors that produces these results? I think it would be more accurate if this could be fixed. It seems particularly out of place with TD 24, which is so huge and sprawling now but likely to tighten up as it develops.

[Recurve]

Sure did overestimate Katrina development, predicted rapid intensification over the gulfstream, looked sickening at the time and didn't verify, but it did get the SW track right.

Have noticed the eye tightening on many runs this season, hadn't noticed the later expansion but you are right that something in its calcs might be doing that for no real reason.

[curtadams]

Extreme example in the current Wilma. http://moe.met.fsu.edu/cgi-bin/gfdltc2. ... =Animation

At 24 hours - almost no eye. A 1/10 of a degree of sub- TS winds, or about 5 miles across. At 126 hours - huge eye. Sub-TS winds more than 1/2 a degree, about 30 miles. If the eyewall ends with hurricane force winds it's 50 miles across! This is the reverse of normal eye dynamics, which is to *shrink* with time and periodically dissipate to be replaced by an outer wall.