@MastersThesis{Bacelar:2017:PrInBr,
author = "Bacelar, Luiz Carlos Salgueiro Donato",
title = "Progn{\'o}sticos de inunda{\c{c}}{\~o}es bruscas utilizando
conjuntos de previs{\~o}es em curto prazo de radar
meteorol{\'o}gico",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2017",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2017-08-30",
keywords = "modelagem hidrol{\'o}gica distribu{\'{\i}}da,
cross-correlation, ensemble, hidrometeorologia,
geoestat{\'{\i}}stica, distributed hydrological modeling,
cross-correlation, ensemble, hydrometeorology, geoestatistics.",
abstract = "As inunda{\c{c}}{\~o}es bruscas s{\~a}o, at{\'e} hoje, um
desafio para {\'o}rg{\~a}os de monitoramento e alerta de
desastres naturais no mundo inteiro. Suas ocorr{\^e}ncias
est{\~a}o associadas a fatores meteorol{\'o}gicos e
hidrol{\'o}gicos espec{\'{\i}}ficos. Um deles, {\'e} a chuva
deflagradora na bacia, que em muitos casos n{\~a}o pode ser
prevista em tempo h{\'a}bil para as analises de suscetibilidade.
Sob esta {\'o}ptica, nesta disserta{\c{c}}{\~a}o foram
aplicados dois m{\'e}todos de previs{\~a}o por conjunto para
expor as incertezas das previs{\~o}es por
extrapola{\c{c}}{\~a}o do Radar Pico do Couto: via time-lagged e
a partir do modelo geoestat{\'{\i}}stico SAMPOTBM. O objetivo
principal foi avaliar se os membros poderiam prever as
inunda{\c{c}}{\~o}es em um horizonte de previs{\~a}o de
at{\'e} 120 minutos. As avalia{\c{c}}{\~o}es hidrol{\'o}gicas
foram realizadas na bacia urbana (168.5 km\$^{2}\$) de Nova
Friburgo, no Estado do Rio de Janeiro, a partir de quatro casos de
inunda{\c{c}}{\~o}es bruscas. Os resultados demonstraram que o
m{\'e}todo de previs{\~a}o por cross-correlation apresenta na
m{\'e}dia {\'{\i}}ndices de probabilidade de
detec{\c{c}}{\~a}o superiores a 0.5 em pelo menos 1 mm de
precipita{\c{c}}{\~a}o. Ao aplicar os campos previstos no Modelo
Hidrol{\'o}gico Distribu{\'{\i}}do (MHD), na
resolu{\c{c}}{\~a}o espacial de 250 metros e passos de tempo de
30 minutos, algumas ondas de cheias puderam ser representadas ao
longo dos hidrogramas, at{\'e} mesmo com 2 horas de
anteced{\^e}ncia. O m{\'e}todo de composi{\c{c}}{\~a}o de
cen{\'a}rios de chuva pelo time-lagged aumentou em m{\'e}dia a
confiabilidade das previs{\~o}es de vaz{\~o}es em tr{\^e}s dos
quatro casos analisados. O mesmo ocorreu com os campos de chuva do
SAMPO-TBM. A diferen{\c{c}}a entre os m{\'e}todos foi mostrado
ao aplicar o diagrama de ROC, duas horas antes das
ocorr{\^e}ncias das inunda{\c{c}}{\~o}es bruscas. Os 120
membros do SAMPO-TBM conseguiram simular melhor os cen{\'a}rios
de chuva para o caso em que n{\~a}o havia uma grande quantidade
de precipita{\c{c}}{\~a}o prevista pelo radar. Conclui-se a
partir disso que a combina{\c{c}}{\~a}o dos dois m{\'e}todos
pode ser promissor para antecipa{\c{c}}{\~a}o dos alertas, o que
contribuiria para a gest{\~a}o de risco mais efetiva em uma das
regi{\~o}es mais afetadas pelas inunda{\c{c}}{\~o}es bruscas do
pa{\'{\i}}s. ABSTRACT: Flash floods are still a challenge for
natural disaster monitoring and early warning centers worldwide.
Their occurrences are associated with specific meteorological and
hydrological conditions. One of them is due to the rain over
basin, which in many cases can not be predicted in lead time to
susceptibility analysis. Therefore, in this dissertation two
nowcasting methods were applied to expose the uncertainties of the
predictions by extrapolation of the Pico do Couto Radar images:
using time-lagged and from the geostatistical model SAMPO-TBM. The
main objective was to assess whether members could predict floods
within a forecast horizon prediction up to 120 minutes. For this,
the hydrological evaluations were realized in the urban watershed
(168.5 km\$^{2}\$) of Nova Friburgo, in the State of Rio de
Janeiro, based on 4 flash flood cases. The results showed that the
cross-correlation nowcasting method presents on average, a
probability of detection higher than 0.5 in at least 1mm of
rainfall. When the precipitation fields were applied in the
Distributed Hydrological Model (MHD), within 250 meters of spatial
resolution and 30 minute time steps, some flood waves could be
represented along the hydrographs, even 2 hours in advance. The
time-lagged method by composing rainfall scenarios increased the
reliability of flow forecasts in 3 of the 4 cases analyzed. The
same occurred with the SAMPO-TBM rain fields. The difference
between the two methods was shown with the ROC diagram, 2 hours
before the occurrence of flash floods. The 120 members of
SAMPO-TBM were able to simulate better rainfall scenarios in a
event that there was no amount of precipitation predicted by the
radar in its nowcasting. Therefore, it is concluded that a
combination of the two methods can be promising to early warnings,
which would contribute to more effective risk management in one of
the regions most affected by flash floods in Brazil.",
committee = "Vila, Daniel Alejandro (presidente) and Angelis, Carlos Frederico
de (orientador) and Cuartas Pineda, Luz Adriana and Beneti, Cesar
Augustus Assis",
englishtitle = "Flash flood nowcasting using weather radar rainfall ensemble",
language = "pt",
pages = "179",
ibi = "8JMKD3MGP3W34P/3PHEPE2",
url = "http://urlib.net/ibi/8JMKD3MGP3W34P/3PHEPE2",
targetfile = "publicacao.pdf",
urlaccessdate = "27 abr. 2024"
}