X hits on this document

PDF document

web-61st-IHC-Booklet.pdf - page 19 / 122





19 / 122

2006 Atlantic and Eastern North Pacific Forecast Verification

James L. Franklin NOAA/NWS/National Hurricane Center

NHC official track forecasts in the Atlantic basin set records for accuracy from 12-72 h in 2006. They consistently beat the individual dynamical guidance models, but trailed the consensus models slightly. Examination of trends suggests that there has been some increase in skill in recent years for the 24-48 h forecasts. Among the operational consensus models, GUNA performed the best overall. The GFDI, GFSI, and NGPI provided the best dynamical track guidance at various times, while the performance of the UKMI trailed considerably. No routinely-available early dynamical model had skill at 5 days. The ECMWF (a late model) performed extremely well, especially at longer times, but it was rarely available in time to be considered by the forecasters. A small improvement in the timeliness of this model would be of great value.

Atlantic official intensity errors were very near the previous 5-year means, but skill levels in 2006 were down sharply. Official errors trailed the GHMI and ICON (a consensus of the GHMI1 and DSHP) guidance, and had a significant high forecast bias. For the first time, dynamical intensity guidance (GHMI) in 2006 was superior to the statistical DSHP guidance on average.

Official track errors in 2006 for the eastern North Pacific were slightly lower than the 5-year mean errors, but were slightly higher than in 2005. The official forecast beat the individual dynamical models but not the consensus models. The consensus track models GUNA and CONU in the eastern Pacific were substantially better than their components, indicating a very strong independence of the consensus members. On the other hand, the GFS ensemble mean (AEMI) was inferior to its control run (GFSI). Among the dynamical models, the GFDI and UKMI were the best performers overall.

Eastern North Pacific official intensity errors were near the 5-year averages. There has been no detectible trend in intensity error since 1990, although skill appears to have increased slightly during this time. GHMI beat DSHP after 36 h, but ICON generally was superior to either one. The FSU super-ensemble also performed well.

1 GHMI is a new interpolated or early version of the GFDL, in which an offset correction is applied only to the first 24 h of the late model. Over a large sample, this has been shown to be superior to GFDI, in which the offset correction is applied uniformly at all time periods.

Document info
Document views181
Page views181
Page last viewedMon Oct 24 06:05:35 UTC 2016