I was hoping for a little tropical storm this year, but a Cat 3, yikes!
According to this, we should have a more solid idea of the track tomorrow morning:
Track forecast for Irene
The computer models show good agreement that Irene will pass along the north coast of Hispaniola today, but just a slight wobble in Irene's track to take it farther offshore--or push it onshore, over the mountains--will have major impacts on the ultimate path and strength of the hurricane. A trough of low pressure is expected to move across the Eastern U.S. on Wednesday and Thursday, turning Irene more to the northwest by Wednesday. The timing and strength of this trough varies considerably from model to model, and will be critical in determining where and when Irene will turn to the north. Irene's strength will also matter--a stronger Irene is more likely to turn northward earlier. The most popular solution among the models is to take Irene to the northwest through the Bahamas on Wednesday and Thursday, then into the Southeast U.S. coast in South Carolina or North Carolina on Saturday. Irene would then travel up the mid-Atlantic coast, arriving near Long Island, New York on Monday morning as a strong tropical storm or Category 1 hurricane. One of the models proposing this solution is our best model, the ECMWF. However, we have two other of our very good models suggesting a landfall near Miami on Thursday night is likely (the GFDL and UKMET models.) NHC forecaster Stacy Stewart gave some good reasons in this morning's discussion to favor a track close to the east coast of Florida, but just offshore. Last years' worst performing major the model, the NOGAPS, predicts that Irene will pass out to sea, missing the Southeast U.S. coast. Keep in mind that the average error of a 4-day forecast from NHC is 200 miles, and just a small deviation in the path of a storm moving roughly parallel to the coast will make a huge difference in where it ultimately makes landfall.
The NOAA jet will be flying its first dropsonde mission into Irene today, which should result in a more reliable set of model runs first thing Tuesday morning.