@MastersThesis{Carrasco:2017:MéAvOr,
author = "Carrasco, Arletis Roque",
title = "M{\'e}todo de avalia{\c{c}}{\~a}o orientada a objeto aplicado
{\`a}s previs{\~o}es de precipita{\c{c}}{\~a}o sobre a
Am{\'e}rica do Sul",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2017",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2017-03-03",
keywords = "precipita{\c{c}}{\~a}o, t{\'e}cnicas de verifica{\c{c}}{\~a}o
espacial, avalia{\c{c}}{\~a}o orientada a objeto, atributos
geom{\'e}tricos, precipitation forecast, spatial verification
techniques, object-based evaluation, geometrics attributes.",
abstract = "O presente estudo tem como objetivo principal, baseado no
algoritmo do Method for Object-Based Diagnostic Evaluation (MODE),
propor e aplicar uma metodologia de avalia{\c{c}}{\~a}o
diagn{\'o}stica orientada a objeto nas previs{\~o}es de
precipita{\c{c}}{\~a}o sobre a Am{\'e}rica do Sul. A
metodologia proposta foi implementada dentro do Sistema
Comunit{\'a}rio de Avalia{\c{c}}{\~a}o de Modelos
Num{\'e}ricos de Previs{\~a}o de Tempo e Clima (SCANTEC) - este
um software estat{\'{\i}}stico que encontra-se em constante
desenvolvimento na Divis{\~a}o de Modelagem e Desenvolvimento
(DMD) do CPTEC/INPE. Um experimentos de casos idealizados permitiu
validar a implementa{\c{c}}{\~a}o do m{\'e}todo e mostrar
tamb{\'e}m as suas vantagens com rela{\c{c}}{\~a}o as
m{\'e}tricas convencionais de avalia{\c{c}}{\~a}o. Para a
aplica{\c{c}}{\~a}o dessa metodologia, como prova de conceito,
foram realizados experimentos usando as previs{\~o}es do modelo
BRAMS (Brazilian Regional Atmospheric Modelling System) e
comparadas com as previs{\~o}es do modelo GFS (Global Forecast
System) tendo como valores de refer{\^e}ncia os dados de
precipita{\c{c}}{\~a}o estimados no produto 3B42RT do TRMM
Multisatellite Precipitation Analysis (TMPA). Duas an{\'a}lises
foram realizadas, uma primeira usando o acumulado de
precipita{\c{c}}{\~a}o de 6 horas com previs{\~o}es de at{\'e}
24 horas de integra{\c{c}}{\~a}o e uma segunda usando o
acumulado de 24 horas de precipita{\c{c}}{\~a}o com
integra{\c{c}}{\~a}o dos modelos de 96 horas. Ambos modelos
apresentam uma ligeira tend{\^e}ncia de degrada{\c{c}}{\~a}o do
desempenho com a integra{\c{c}}{\~a}o no tempo, subestimando a
quantidade de objetos observados errando mais por falhas do que
por alarmes falsos. A aplica{\c{c}}{\~a}o da metodologia de
avalia{\c{c}}{\~a}o de precipita{\c{c}}{\~a}o orientada a
objeto aqui proposta, in{\'e}dita para as previs{\~o}es geradas
no CPTEC, evidenciou a seu potencial para caracterizar os
padr{\~o}es espaciais das previs{\~o}es de
precipita{\c{c}}{\~a}o fornecidas por esse centro permitindo
compara{\c{c}}{\~o}es mais justas com outros produtos
dispon{\'{\i}}veis. ABSTRACT: The present study has as the main
objective, based on the Method for Object-Based Diagnostic
Evaluation (MODE) algorithm, to propose and apply a methodology of
object-based diagnostic evaluation to precipitation forecasts over
South America. The proposed methodology was implemented within
code of the Community System for the Evaluation of Numerical
Models of Weather and Climate Prediction (SCANTEC), which is a
statistical software that is constantly in development at
CPTEC/INPE. An idealized case study allowed validating the
methodology implemented and also showed the advantages over
conventional evaluation metrics. For the application of this
methodology, as proof of concept, experiments were carried out
using BRAMS (Brazilian Regional Atmospheric Modelling System)
models forecasts and compared with the predictions of the GFS
(Global Forecast System) model using as reference values the
estimated precipitation data in the TRMM Multisatellite
Precipitation Analysis (TMPA) 3B42RT product. Two analyses were
performed: the first using 6-hour precipitation accumulated with
forecasts up to 24 hours of integration, and the second using
24-hour precipitation accumulated with models integration of the
96 hours. Both models present a slight tendency of skill
degradation in function of model integration time, underestimating
the amount of observed objects. They failed more by misses than by
false alarms. The application of the precipitation objective-based
evaluation methodology proposed here, unpublished for CPTEC
forecasts, demonstrated its potential to characterize the spatial
patterns of precipitation forecasts provided by this center
allowing fairer comparisons with other available products.",
committee = "Vila, Daniel Alejandro (presidente) and Sapucci, Luiz Fernando
(orientador) and Ferreira, Nelson Jesuz and Mattos, Jo{\~a}o Gerd
Zell de and Freitas, Edmilson Dias de",
englishtitle = "Method of object-based evaluation applied to precipitation
forecasts over South America",
language = "pt",
pages = "116",
ibi = "8JMKD3MGP3W34P/3NH9KMB",
url = "http://urlib.net/ibi/8JMKD3MGP3W34P/3NH9KMB",
targetfile = "publicacao.pdf",
urlaccessdate = "27 abr. 2024"
}