Precipitação pluviométrica mensal no estado do Rio de Janeiro: sazonalidade e tendência

Carregando...
Imagem de Miniatura

Data

2009-01-01

Autores

Carvalho Araujo, Mirian Fernandes
Guimaraes, Ednaldo Carvalho
de Carvalho, Daniel Fonseca
de Araujo, Lucio Borges [UNESP]

Título da Revista

ISSN da Revista

Título de Volume

Editor

Universidade Federal de Uberlândia (UFU)

Resumo

The objective of this work was to carry a descriptive analysis in the monthly precipitation of rainfall stations from Rio de Janeiro State, Brazil, using data of position and dispersion and graphical analyses, and to verify the presence of seasonality and trend in these data, with a study about the application of models of time series. The descriptive statistics was to characterize the general behavior of the series in three stations selected which present consistent historical series. The methodology of analysis of variance in randomized blocks and the determination of models of multiple linear regression, considering years and months as predictors variables, disclosed the presence of seasonality, what allowed to infer on the occurrence of repetitive natural phenomena throughout the time and absence of trend in the data. It was applied the methodology of multiple linear regression to removal the seasonality of these time series. The original data had been deducted from the estimates made by the adjusted model and the analysis of variance in randomized blocks for the residues of regression was preceded again. With the results obtained it was possible to conclude that the monthly rainfall present seasonality and they don't present trend, the analysis of multiple regression was efficient in the removal of the seasonality, and the rainfall can be studied by means of time series.
O presente trabalho foi dirigido no sentido de realizar uma análise descritiva da precipitação pluviométrica mensal de estações climatológicas do Estado do Rio de Janeiro, utilizando medidas de posição e de dispersão e análises gráficas, e, através de um estudo sobre a aplicação de modelos de séries temporais, verificar a presença das componentes de sazonalidade e de tendência nestes dados. Os dados experimentais utilizados fazem parte do Banco de Dados Climatológicos do Estado do Rio de Janeiro. Selecionou-se três estações por apresentarem séries históricas consistentes. A estatística descritiva foi utilizada para caracterizar o comportamento geral da série. A metodologia de análise de variância em blocos casualizados e determinação de modelos de regressão linear múltipla, considerando anos e meses como variáveis preditoras, revelaram a presença de sazonalidade, que permitiu inferir sobre a ocorrência de fenômenos naturais repetitivos ao longo do tempo, e ausência de tendência. Aplicou-se a metodologia de regressão linear múltipla para a remoção da sazonalidade dessas séries temporais. Os dados originais foram subtraídos das estimativas feitas pelo modelo ajustado e procedeu-se novamente a análise de variância em blocos casualizados para os resíduos da regressão. Verificou-se que a regressão múltipla foi eficiente na remoção da sazonalidade da série.

Descrição

Palavras-chave

Climate, Time series, Rainfall statistics, Multiple regression

Como citar

Bioscience Journal. Uberlandia: Universidade Federal de Uberlândia (UFU), v. 25, n. 4, p. 90-100, 2009.

Coleções