| "-Uncertainty in the NHC Cone of Uncertainty-" A research project funded by NOAA known as the Joint Hurricane Testbed has produced a remarkable number of tools now in operational use at the National Hurricane Center to improve hurricane forecasts and warnings. One of these projects, called "Prediction of Consensus TC Track Forecast Error and Correctors to Improve Consensus TC Track Forecasts", was an effort by Dr. Jim Goerss at the Navy Research Lab to improve the accuracy of the NHC "cone of uncertainty" (AKA the "Cone of Death") showing where a storm is expected to track 2/3 of the time. The radius of the circles that make up the cone are based on error statistics of the official NHC forecast over the past five years. We can expect in certain situations, such as when the models are in substantial disagreement, a consensus forecast made using these models will have much greater than average errors. Since the NHC typically bases their forecast on a consensus forecast made using a combination of reliable hurricane forecasting models, it is instructive to view the "GPCE" (Goerss Prediction Consensus Error) circles to see if the uncertainty cone should be smaller or larger than usual. The consensus forecast I'll look at is called "TVCN", and is constructed by averaging the track forecasts made by most of (or all) of these models: GFS, ECMWF, NOGAPS, GFDL, HWRF, GFDN, and UKMET. In the case of this morning's 12 UTC (7am CDT) June 28 run of these models, here is what the radius of the "cone of uncertainty" should be, in nautical miles: 12 hours: 42 nm 24 hours: 73 nm 36 hours: 96 nm 48 hours: 112 nm 72 hours: 173 nm 96 hours: 327 nm 120 hours: 376 nm And here is the radius of NHC's "cone of uncertainty" for their official forecast, based on the average errors for the past five years: 12 hours: 36 nm 24 hours: 62 nm 36 hours: 85 nm 48 hours: 108 nm 72 hours: 161 nm 96 hours: 220 nm 120 hours: 285 nm So, the GPCE error estimates are showing that the latest forecasts for Alex out to 72 hours are 4% - 17% higher in uncertainty than average. The 4 - 5 day forecasts are 32% - 49% more uncertain than average--but of course, we expect Alex to be inland at those times." |