MODELS GENERATED BY MULTIPLE REGRESSION IN FILLING METEOROLOGICAL DATA FAILURES IN AN AUTOMATIC METEOROLOGICAL STATION IN ALAGOAS

• Autor(a) principal: Araújo Neto, Renato Américo
• Data de Publicação: 2020
• Outros Autores: Nascimento, Jonathan Willyan dos Santos, Oliveira, Francisco Freire de, Rebelo, Gil Rafael Pacífico, Silva, Francisco de Assis de Lima, Carvalho, André Luiz de
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Revista Geama
• Texto Completo: https://www.journals.ufrpe.br/index.php/geama/article/view/2830
• topic: Correction; Failures; Meteorology; Statistic Regression

The objective of this study was to evaluate the multiple regression method to fill in the faults of the following meteorological variables: Average Air Temperature (Tmean), Relative Humidity (RHmean), and Rain Precipitation (Prec). Multiple regression was considered using different models, through the different cofactors evaluated (varying Tmean, RHmean, Dew Point, Pressure and Prec), generating four different multiple regression models for each meteorological variable studied. The models were statistically compared by Mean Absolute Error (MAE), Pearson's coefficient (r), agreement index (d) and Camargo and Sentelhas index (c). The results presented showed that multiple regression can be reliably used in Tmean, RHmean in Models 2, 3 and 4 (R> 0.90). The Precipitation variable had a coefficient of determination below 50% (R2 <0.50) and Model 2 obtained a p value greater than 1% in the Intercept (p = 0.012) and in the Pressure cofactor (p = 0.015). It cannot be used to correct Rainfall faults. Model 2 (except for Prec) presented better statistical coefficients and can be used to correct faults in the automatic station of Maceió, Alagoas.