Fechar
Metadados

@MastersThesis{Rocha:2017:ImAjMa,
               author = "Rocha, Andr{\'e} Muniz Marinho da",
                title = "Impacto do ajuste da matriz de covari{\^a}ncia dos erros do 
                         background na assimila{\c{c}}{\~a}o de dados de radar",
               school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                 year = "2017",
              address = "S{\~a}o Jos{\'e} dos Campos",
                month = "2016-10-03",
             keywords = "assimila{\c{c}}{\~a}o de dados de radar, 
                         precipita{\c{c}}{\~a}o, 3D-VAR, radar data assimilation, 
                         precipitation.",
             abstract = "A assimila{\c{c}}{\~a}o de dados combina as 
                         informa{\c{c}}{\~o}es de modelos num{\'e}ricos e as 
                         observa{\c{c}}{\~o}es meteorol{\'o}gicas, atrav{\'e}s de um 
                         processo f{\'{\i}}sico-estat{\'{\i}}stico, gerando a melhor 
                         representa{\c{c}}{\~a}o poss{\'{\i}}vel do estado da atmosfera 
                         em um dado instante de tempo. O objetivo deste trabalho {\'e} 
                         ajustar a matriz covari{\^a}ncia do erro do background dentro do 
                         ciclo de assimila{\c{c}}{\~a}o de dados de radar Doppler, a fim 
                         de melhorar a an{\'a}lise e, como consequ{\^e}ncia, as 
                         previs{\~o}es de precipita{\c{c}}{\~a}o de curto prazo. O 
                         modelo atmosf{\'e}rico e o sistema de assimila{\c{c}}{\~a}o 
                         utilizados s{\~a}o o Weather Research and Forcasting (WRF) e o 
                         WRF Data Assimilation (WRFDA) 3D-Var. O dom{\'{\i}}nio abrange o 
                         oeste do sul do Brasil, incluindo os estados do Paran{\'a}, Santa 
                         Catarina e Rio Grande do Sul e parte do Paraguai com 
                         resolu{\c{c}}{\~a}o horizontal de 2 km e 45 n{\'{\i}}veis. O 
                         per{\'{\i}}odo de estudo {\'e} de 15 de outubro a 15 de 
                         novembro de 2014, com a avalia{\c{c}}{\~a}o da 
                         precipita{\c{c}}{\~a}o feita comparando os resultados da 
                         modelagem com os dados do Tropical Rainfall Measuring Mission 
                         (TRMM) 3B42, usando os {\'{\i}}ndices estat{\'{\i}}stico Root 
                         Mean Square Error (RMSE). Os outros campos meteorol{\'o}gicos 
                         tamb{\'e}m foram avaliados usando o mesmo {\'{\i}}ndice 
                         estat{\'{\i}}sticos comparando-o com as observa{\c{c}}{\~o}es 
                         de superf{\'{\i}}cie. Observa{\c{c}}{\~o}es das 
                         Esta{\c{c}}{\~o}es meteorol{\'o}gicas de superf{\'{\i}}cie 
                         foram usadas para compara{\c{c}}{\~a}o com os resultados do 
                         modelo com e sem assimila{\c{c}}{\~a}o de dados do radar. As 
                         esta{\c{c}}{\~o}es selecionadas foram Curitiba, Bacacheri, 
                         Londrina e Foz do Igua{\c{c}}u. Durante o processo de 
                         assimila{\c{c}}{\~a}o, os dados convencionais do Global 
                         Telecommunication System tamb{\'e}m foram assimilados. A matriz 
                         de covari{\^a}ncia do erro de background foi gerada utilizando um 
                         utilit{\'a}rio do WRFDA aplicando o m{\'e}todo NMC com 03 meses 
                         de simula{\c{c}}{\~o}es de 24 h a partir de 00UTC e 12UTC. O 
                         processo de gera{\c{c}}{\~a}o da matriz B espalha 
                         horizontalmente as informa{\c{c}}{\~o}es de uma determinada 
                         observa{\c{c}}{\~a}o usando um filtro recursivo, em seguida, o 
                         ajuste da matriz de covari{\^a}ncia do erro de background foi 
                         aplicado, ajustando os par{\^a}metros variance scaling, 
                         relacionada com a intensidade com que cada observa{\c{c}}{\~a}o 
                         ir{\'a} influenciar as vari{\'a}veis de estado nos pontos da 
                         grade do modelo, e o length scaling, relacionada com a 
                         influ{\^e}ncia do erro em escala de dist{\^a}ncia nos valores 
                         dos pontos da grade das vari{\'a}veis de estado do modelo, de 
                         modo a ajust{\'a}-los para a regi{\~a}o de estudo, os dados 
                         assimilados e o sistema meteorol{\'o}gico estudado. Foram 
                         testados diversos valores dos dois par{\^a}metros e os resultados 
                         baseado no {\'{\i}}ndice estat{\'{\i}}stico mostrou melhorias 
                         na previs{\~a}o da localiza{\c{c}}{\~a}o e intensidade da 
                         precipita{\c{c}}{\~a}o quando aplicado os ajustes na matriz de 
                         covari{\^a}ncia do erro de background. ABSTRACT: Data 
                         assimilation combines the information from numerical models and 
                         meteorological observations through a physical-statistical process 
                         generating the best representation of atmospheric state in a 
                         moment of time. The goal of this work is to tune the background 
                         error covariance matrix while assimilating Doppler radar data in 
                         order to improve the analysis and then the short-term 
                         precipitation forecast. The atmospheric model and the assimilation 
                         system used are the Weather Research and Forecasting (WRF) and the 
                         WRF Data Assimilation (WRFDA) 3D-Var. The domain covers the west 
                         of Southern Brazil, including the state of Parana, Santa Catarina 
                         and Rio Grande do Sul and part of Paraguay with horizontal 
                         resolution of 2-km and 45 levels. The period of study is from 
                         October 15 to November 15, 2014, and the evaluation of the 
                         precipitation was made by comparing the results from modeling 
                         against the Tropical Rainfall Measuring Mission (TRMM) 3B42 data, 
                         using statistical index such the Root Mean Square Error (RMSE). 
                         The other meteorological fields were also evaluated using the same 
                         statistical indice comparing them to the surface observations. 
                         Observations of the surface weather stations were used for 
                         comparison with the model results with and without radar data 
                         assimilation. The selected stations were Curitiba, Bacacheri, 
                         Londrina and Foz do Igua{\c{c}}u. During the assimilation 
                         process, the conventional data from Global Telecommunication 
                         System was also assimilated. The background error covariance 
                         matrix was generated using utility WRFDA applying the NMC method 
                         with 03 months of simulations of 24-h starting at 00UTC and 12UTC. 
                         The process of generating the matrix B horizontally spreads the 
                         information from a specific observation using a recursive filter, 
                         and then setting the error covariance matrix background was 
                         applied by adjusting the parameters variance scaling related to 
                         the intensity at each observation will influence the state 
                         variables in the model grid points , and the length scaling, 
                         related to the influence of the error in distance scale the values 
                         of the grid points of the model state variables, in order to 
                         adjust them to the region study, the assimilated data and the 
                         weather system studied. Different values of the two parameters 
                         were tested and the results based on statistical indicator showed 
                         improvements in predicting the location and intensity of 
                         precipitation when applied adjustments to the covariance matrix of 
                         background error.",
            committee = "Sapucci, Luiz Fernando (presidente) and Herdies, Dirceu Luis 
                         (orientador) and Vendrasco, {\'E}der Paulo and Correa, Cleber 
                         Souza",
           copyholder = "SID/SCD",
         englishtitle = "The impact of tuning the background covariance error matrix on the 
                         radar data assimilation",
             language = "pt",
                pages = "83",
                  ibi = "8JMKD3MGP3W34P/3MSRH7H",
                  url = "http://urlib.net/rep/8JMKD3MGP3W34P/3MSRH7H",
           targetfile = "publicacao.pdf",
        urlaccessdate = "27 nov. 2020"
}


Fechar