@MastersThesis{Fonseca:1998:EsCoDe,
author = "Fonseca, Josiane Ferreira Bustamante",
title = "Modelos regionais de previs{\~a}o num{\'e}rica de tempo: um
estudo comparativo das destrezas dos modelos FSU e ETA, sobre a
Am{\'e}rica do Sul",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "1998",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "1998-09-22",
keywords = "x.",
abstract = "Durante o per{\'{\i}}odo de Fev de 1997 {\`a} Fev de 1998 dez
importantes situa{\c{c}}{\~o}es sin{\'o}ticas de atividade
frontal sobre o Sudeste do Brasil s{\~a}o identificadas, sete
delas s{\~a}o passagens de sistemas frontais e tr{\^e}s s{\~a}o
situa{\c{c}}{\~o}es associadas de frontog{\^e}nese e
ciclog{\^e}nese. A performance da previs{\~a}o de curto prazo
(at{\'e} 72h) dos modelos regionais Eta (resolu{\c{c}}{\~a}o de
80 km) e FSU (resolu{\c{c}}{\~a}o de 1 °) s{\~a}o avaliadas
para essas situa{\c{c}}{\~o}es com an{\'a}lises iniciais das 00
e 12 UTC. Ao todo vinte rodadas s{\~a}o feitas com cada modelo.
As estat{\'{\i}}sticas S 1 {"}score{"}, RMSE, coeficiente de
correla{\c{c}}{\~a}o absoluta e erro m{\'e}dio, obtidas com
respeito {\`a}s an{\'a}lises, s{\~a}o usadas para a
avalia{\c{c}}{\~a}o. Importantes campos meteorologicos s{\~a}o
avaliados em compara{\c{c}}{\~a}o {\`a}s an{\'a}lises, e a
chuva {\'e} avaliada atrav{\'e}s de dados observacionais. Os
resultados s{\~a}o apresentados atrav{\'e}s de gr{\'a}ficos das
estat{\'{\i}}sticas para (a) o melhor caso previsto, (b) o pior
caso previsto, (c) a performance m{\'e}dia de todas as 20
rodadas, (d) a m{\'e}dia das rodadas das 00 UTC e (e) a
m{\'e}dia das rodadas das 12 UTC. Previs{\~o}es de um
par{\^a}metro {\'e} considerada boa se o coeficiente de
correla{\c{c}}{\~a}o da previs{\~a}o {\'e} maior do que a
correla{\c{c}}{\~a}o da persist{\^e}ncia. Press{\~a}o ao
n{\'{\i}}vel m{\'e}dio do mar, altura geopotencial de 850 e 500
hPa, umidade espec{\'{\i}}fica em 850 hPa e os componentes do
vento nos n{\'{\i}}veis 850 e 200 hPa s{\~a}o usados para a
aplica{\c{c}}{\~a}o das estat{\'{\i}}sitcas. As melhores
previs{\~o}es s{\~a}o obtidas a partir da an{\'a}lise de 12 UTC
de 13/06/97, onde o aprofundamento de um sistema de baixa
press{\~a}o pr{\'o}ximo a costa Argentina favorece o
desenvolvimento de uma frente, a qual move-se sobre o Sudeste do
Brasil at{\'e} o Rio de Janeiro. Nesse caso o S1 da press{\~a}o
na superf{\'{\i}}cie permanece abaixo de 70 at{\'e} as 72 horas
em ambos os modelos. As piores previs{\~o}es s{\~a}o obtidas com
os dados de 30/04/97 00 UTC. Este {\'e} um caso de
dissipa{\c{c}}{\~a}o frontal (enfraquecimento) pr{\'o}ximo a
costa do Esp{\'{\i}}rito Santo e subsequente
forma{\c{c}}{\~a}o e passagem de um fraco sistema frontal
avan{\c{c}}ado da Argentina para o sul do Brasil. Neste caso o S1
aumenta de 50 para 70 em 60 horas. As estat{\'{\i}}sticas
mostram as seguintes caracter{\'{\i}}sticas gerais: 1)
Previs{\~o}es do Eta mostram melhor performance para
previs{\~o}es de situa{\c{c}}{\~o}es frontais. Quando uma dada
situa{\c{c}}{\~a}o sin{\'o}tica permite uma boa previs{\~a}o a
performance do modelo Eta {\'e} melhor do que a do modelo FSU,
enquanto que em casos opostos a performance do modelo Eta {\'e}
pior do que a do modelo FSU. As previs{\~o}es de
precipita{\c{c}}{\~a}o do modelo Eta concordam melhor com as
observa{\c{c}}{\~o}es. 2) O vento meridional tem a menor
previsibilidade entre os par{\^a}metros estudados e a altura
geopotencial tem a melhor. 3) Rodadas das 12 UTC tem uma
habilidade ligeiramente melhor do que as rodadas das 00 UTC. 4) Os
erros RMS crescem gradualmente com o prazo de previs{\~a}o para
todas as vari{\'a}veis nos dois modelos. ABSTRACT: During the
period Feb 1997 to Feb 1998 ten important synoptic situations of
frontal activity over southem Brazil are identified, seven of
which are passive frontal passages and three are cyclogenesis and
frontogenesis associated situations. The short range (up to 72 h)
forecast performance of the Eta (80 km resolution) and the FSU (1
° resolution) regional models are evaluated for these situations
with 00 and 12 UTC initial analyses. Overall there are twenty runs
made with each of the models. The statisties S I score, RMS error,
absolute correlation coefficient and mean error, obtained with
respect to the analysis, are used for the evaluation. Important
meteorological fields are synoptically evaluated against the
verifying analyses, and the rainfall is evaluated against the
observational data. The results are presented through the graphs
of the statistics for (a) the best forecast case, (b) the worst
forecast case, (c) the mean performance of ali the 20 runs, (d)
the mean of the 0OUTC runs and (e) the mean of the 12 UTC runs.
Forecast of a parameter is considered good if the correlation
coefticient of the forecast is larger than the correlation of the
persistence. Sea levei pressure, geopotential height of 850 and
500 hPa, specific humidity at 850 hPa and the wind components at
850 and 200 hPa leveis are used for applying the statistics. The
best forecasts are obtained from the 12 UTC analysis of the
13/06/97, where the deepening of a low pressure system near the
Argentinian coast helped a front to develop, which later moved
over southem Brazil up to Rio de Janeiro. In this case the surfacc
pressure Si score remained below 70 through 72 hours both in the
Eta and FSU models. The worst forecast is obtained with the 00 UTC
data on 30 Apr 1997. This is a case of frontal dissipation
(wealcening) near the Esp{\'{\i}}rito Santo coast and subsequent
formation and passage of a weak front from Argentina to southem
Brazil. In this case the Si score was 50 for 12h forecast
increasing to 70 by 60 h.The statistics show the following general
characteristics: 1) Eta forecasts shows better performance for
frontal situation forecasts. When a given synoptic situation
permits a good forecast the Eta model performs better than the FSU
model, whereas in the opposite case the Eta model performs worse
than the FSU model. The precipitation forecasts of Eta model
agreed better with the observations. 2) The meridional wind has
the least predictability among the parameters studied and the 500
hPa geopotential has the best predictability. 3) 12 UTC runs have
a little more skill than the 00 UTC runs. 4) The RMS errors
increase gradually with the forecast range in ali variables in
both the models.",
committee = "Bonatti, Jos{\'e} Paulo (presidente) and Satyamurty, Prakki
(orientador) and Chan, Chou Sin (orientadora) and Franchito,
S{\'e}rgio Henrique and Gandu, Adilson Wagner",
copyholder = "SID/SCD",
englishtitle = "x",
language = "pt",
pages = "129",
ibi = "8JMKD3MGP3W34T/48ATN6B",
url = "http://urlib.net/ibi/8JMKD3MGP3W34T/48ATN6B",
targetfile = "publicacao.pdf",
urlaccessdate = "16 jun. 2024"
}