Weather Underground's latest Report on Irma
Hurricane hunter observations on Thursday evening showed lower winds in Irma than earlier in the day, and found a distinct double wind maximum, caused by concentric eyewalls. By my count, this may be Irma’s 7th eyewall replacement cycle, and appears to be the first one that has notably degraded the hurricane’s maximum winds. The Air Force hurricane hunter plane in the storm Thursday night found top surface winds near 165 mph on their first pass through the eye near 7:35 pm EDT, but only 140 mph on their second pass through near 9 pm EDT. The pressure stayed nearly constant in the two passes, at 920 and 921 mb, respectively. The reduction in the peak winds is not necessarily a good thing, since the hurricane hunter data showed that the hurricane-force winds of the storm had spread out over a larger area, which will increase the storm surge. In addtion, the fact that Irma maintained its very low pressure may mean the aircraft may have missed sampling the strongest winds of the hurricane. Total precipitable water loops showed that part of Irma’s apparent weakening may be due to the island of Hispaniola blocking the inflow of moisture from the south, which was reducing the amount of moisture to the storm. This effect will diminish on Friday morning, when Irma moves to the west of Hispaniola. The hurricane will then be drawing air from the south over high terrain in Cuba, but the blocking effect of Cuba is much less than that of Hispaniola. I don't see any reason for Irma to weaken further over through Saturday, and it may intensify again to 185 mph winds.
The ridge of high pressure steering Irma turned out to be stronger than expected on Thursday, and forced the hurricane to move farther south than the models had been predicting. This resulted in a westward shift in the predictions for Irma’s turn to the north by Florida this weekend. As of 10 pm EDT Thursday, the latest available runs (12Z or 18Z) of all of our top models for tracking hurricanes—the European, GFS, HWRF, HMON, and UKMET—had Irma making landfall in South Florida Sunday morning, then turning north. All of the members of the high-probability cluster of the European model ensemble (Figure 2) also showed this track. The only ray of hope for South Florida could be found in the ensemble members of the 12Z GFS ensemble (Figure 3), which had a few members missing the peninsula. Since the average error in a 3-day forecast is 120 miles, it is important to remember that the models may still have additional shifts, and one must pay attention to the NHC cone of uncertainty! If you are in the cone of uncertainty, you are in danger of a direct hit.
https://www.wunderground.com/cat6/triple-trouble-cat-5-irma-cat-3-jose-cat-1-katia
