Five days is a long time in Hurricane forecasting!
Just for fun, this was a post on August 20th with Harvey 5 days out from landfall:
cieldumort (Sun Aug 20 2017 12:12 PM ):
WSI Calibrated ECMWF EPS now showing a greater than 90% chance that Harvey regenerates.
While the GFS runs should not yet be fully discounted, it is my personal opinion that the new GFS (since the July 19 upgrade) has not performed as admirably with Atlantic tropical cyclones as might be desired, and even in the face of a dramatic convective resurgence this morning, the latest runs (20/0z, 20/06z and now the 20/12z) still do mostly nothing with now x-Harvey... maybe a TD crossing Belize, or forming in the extreme southern Bay of Campeche, at best. These runs now run entirely counter to the Euro, and defy the eyes.
Now awaiting the restarting of the TC-specific models HWRF & HMON. Run, runs, run.
ftlaudbob (Sun Aug 20 2017 01:43 PM ):
Harvey looks alive and well.Should be an upgrade soon.
ftlaudbob (Sun Aug 20 2017 10:56 PM ):
And just like that, it died out again.
Harvey did not die, and by the morning of August 22nd it was a totally different story:
cieldumort (Tue Aug 22 2017 09:30 AM ):
Model consensus for south Texas this Friday is growing much more concerning. HWRF, HMON, GFS, ECMWF, GEM & GFS Ensemble Members with almost 100% unanimity on a landfalling Texas hurricane Cat 1 to Major. Some runs hang it around just inland, others around the coast - with potential for very severe flooding if either of those scenarios verify (inland and/or coastal)...