Verificação das Previsões NHC e Modelos em 2007


23 Jan 2007
O NHC publicou o relatório de verificação e análise estatística dos erros de previsão oficiais do NHC e de todos os modelos referentes à época ciclónica no Atlântico e Pacífico Leste em 2007.

-> 2007 National Hurricane Center Forecast Verification Report

Uma leitura do relatório muito na diagonal permite concluir que a época foi desafiante para o NHC e para os modelos pois foi bastante invulgar. O relatório é extenso e muito detalhado, por exemplo pormenoriza mesmo individualmente tempestades, verifica a nível de previsão de intensidades e trajectos, etc,etc.

Quanto às previsões do NHC normalmente eram iguais ou melhores que os modelos de consenso, quanto aos modelos individualmente, o GFS foi o grande vencedor, seguido do UKMET e do BAMM. Mas como se percebe facilmente olhando para as estatísticas, os consensos entre modelos e as previsões oficiais são em termos globais geralmente melhores do que um modelo individualmente.

NHC official track forecasts in the Atlantic basin set records for accuracy from
36-96 h in 2007. They beat or matched the consensus models at most time periods, but
generally trailed the best of the dynamical models. Examination of trends suggests that
there has been little net change in forecast skill over the past several years. Among the
consensus models, CGUN (the corrected version of GUNA) performed the best overall.
The GFSI and UKMI/EGRI provided the best dynamical track guidance, while the GFDI
and NGPI performed relatively poorly. The performance of the EMXI in 2007 was
The 2007 Atlantic season, which featured two category 5 hurricanes and several
episodes of rapid deepening, presented some unusual challenges. Intensity forecast
difficulty, as measured by Decay-SHIFOR, was highly elevated compared to the previous
5-year mean, and official intensity errors in 2007 were also larger than normal. Skill
levels, however, were higher in 2007 than in 2006. The statistical DSHP and LGEM
models provided the best objective guidance.
In the eastern North Pacific, official track errors set records at 12-36 h. Forecast
errors were below the previous 5-year mean even though the CLIPER error in 2007 was
higher than its 5-year mean. The official forecast beat the individual dynamical models
on average but trailed the consensus guidance. Among the dynamical models, EMXI
provided the best guidance by a wide margin.
Eastern North Pacific official intensity errors were well below the 5-year averages
at many time periods, setting accuracy records at 12-48 and 120 h. Despite the low errors
in 2007, there has been little or no overall trend in intensity error since 1990; skill,
however, appears to have increased slightly during this time. Either DSHP or LGEM,
both statistical models, provided the best intensity guidance at each time period.
The 2007 season marked the first year of operational availability of the HWRF
regional hurricane model. The model generally lagged its GFDL benchmark for
intensity, although it significantly outperformed the GFDL for track forecasts in the
Atlantic. A combination of the two models, however, generally was superior to either
one alone.
Also initiated in 2007 were in-house probabilistic forecasts of tropical
cyclogenesis. The verification was sufficiently favorable to begin experimental public
genesis forecasts in 2008.

Recent trends in NHC official track forecast error (top) and skill (bottom)
for the Atlantic basin.


Homogenous comparison of the primary Atlantic basin track consensus
models for 2007.