The storm was relatively well predicted early on by the European Center, National Centers for Environmental Predictions, and the United Kingdom Meteorological Office global forecast systems. All of the systems forecast the storm to develop and move steadily to the west into the Leeward Islands and the Caribbean then miss Puerto Rico, before moving along the north coast of Cuba.
Longer range forecasts initially forecast the storm to make a sharp north turn. Forecasts from 3 September for example, had the storm move northward just east of the Florida with some potential tracks along the coast. Successive forecasts progressively shifted the storm track to the west. It will be shown that the EC ensemble forecast system had about 1 day of additional lead-time correctly tracking Irma over the Florida Keys and into the West Coast of Florida. The GEFS and UKMO systems showed a similar track with slight shorter lead times.
The potential threat to the Caribbean and eventually Florida with such long lead-times was impressive. Despite these successful forecasts, decisions related to what areas to evacuate were difficult due to the uncertainty of the storm track. Earlier forecasts suggested landfall on the United States as far north as South Carolina while shorter range forecasts placed the landfall farther south and west with time. This raises interesting questions related to when and where to consider evacuation.