It is interesting to me that the GFS model went from multi-day, multi-run consistency for development of 94L...for a weeks worth of runs for that matter...and then all at once last night's 00z run abandoned the idea almost entirely, leaving only an open wave to pass harmlessly into the Caribbean.

That strikes me suspiciously as some kind of data ingest and/or assimilation errors being fed into the model. The 06z run continued that no development idea.

The GFDL went south and weaker, but that is understandable because the GFDL is initialized/parameterized off the GFS data; perhaps another clue the that GFS has been dealt some odd data for processing.

It should be noted, however, that the ECM has not really been too happy about this system, for about the last 3 day's worth of runs. It did have it 4 and 5 days back, more robustly, but since has been paltry at best.

Currently, there is a fairly obvious cyclonic curl to the convective debris/showers of 94L. I don't see any overwhelmingly negative factors for development. The deep layer shear analysis has a general easterly component at all levels; and given that 94L is in fact moving west means that here storm relative shear is actually quite low. That said, with amply warm water and decent capacity for instantiating anticyclonic outflow above any convective concentrations that get going near the core, this should have an opportunity to develop. Interestingly, the last 1 hour of loop shows a rather obvious banded feature on the northern semi-circle, with cooling cloud tops.

It is important to note that even though the models did exceptionally well sniffing out Bertha prior to her birth, that does not necessarily mean they will all handle every situation with the same result. Sometimes the models will be vague about development and we wind up with a major hurricane, such as Felix (2007).

An ending idea to this... If a deeper system gets going, it may in fact end up more around the NE Windwards as opposed to cutting straight west in the Caribbean.

John