At the 2pm synoptic analysis and interim update, NHC upgraded Melissa to a hurricane .. here’s the satellite loop for the last couple of hours, you can see the distinctive spiral arms forming and very cold (pink) cloud tops consolidating near the center of circulation on the IR (left).
Tracking continues to be dire for Jamaica. Here’s a new auto-generated map, this one has the current model statistics in the data block on the right:
These are average errors in nautical miles at the specified times (er_2day is the forecast error two days or 48 hours from forecast time). Score is an abstract measure that not only incorporates the average error, but also includes consistency. As an extreme example, let’s say model X has two forecasts, one was exactly right, the other off by 90 miles, and model Y has errors of 45 and 55 at the same times. Model X has an average error of 45 miles, model Y 50 – but clearly the average is misleading, because you can’t trust X to get it right every time, so Y is actually a better model. -1.0 means there wasn’t enough data for a reliable estimate.
A score of around 5.5 or better is pretty good. Note that despite the public trashing of GFS, the GFS ensemble (AEMN) is actually better than the European model ensemble (EEMN). Also note how good the new HFS models (HFSA and HFSB) are at five days – half the error of the top scoring model overall (Google’s deep mind ensemble). The lesson here is beware of doom bloggers who are cheerleaders for a single model – a good forecaster uses all the tools and understands how they work and evolve over time for a given storm.
Your best bet is always to focus on the official NHC forecasts. Just because someone has the raw track data (and everyone has access to it these days) and can make dramatic graphics doesn’t mean they know how to interpret it.
Speaking of which: NHC’s Key messages regarding Hurricane Melissa (en Español: Mensajes Claves)
Thank you!