It's not so crazy when you remove the "garbage" from the latest model plot. Most of the strange tracks that have it looping out near the East Pac then back to the Caribbean or Gulf, or NW toward the NW Gulf or BoC are from rather bad models (LBAR, CLIPPER, BAMs). I removed the various climo runs, LBAR and the BAMS/BAMM/BAMD and we see quite a good consensus developing.
The GFS seems to be at the heart of the disagreement. Note that there are now 3 divergent tracks that take Matthew more rapidly off to the NE - the GFS ensembles, HWRF and the consensus models (TVCN/TVCC). The HWRF at least in part incorporates GFS data, so it's not surprising it sort of follows the lead of the GFS. The consensus models incorporate the GFS and HWRF, so the consensus track is well to the NE, too. The consensus is good to follow most of the time, but not when 1 or 2 of its members are extreme outliers. All the other models, however, keep Matthew inland for 5 days. That yellow line emerging offshore near northern Belize is the GFS operational run.
![Image](http://myweb.cableone.net/nolasue/mattmodels.gif)