Estimativa Da Precipitação No Espírito Santo Por Intermédio De Regressão Polinomial

Abstract

A precipitação é um dos principais elementos da hidrologia, sendo uma variável de grande importância para a compreensão da dinâmica do ciclo hidrológico. Apesar da sua importância, a disponibilidade de dados hidroclimáticos é baixa. Dentre as alternativas para suprir a necessidade de informações da precipitação, a modelagem matemática é uma importante ferramenta que visa e sua estimativa. Assim, este trabalho avaliou as precipitações mensais e anuais de 110 estações pluviométricas do estado do Espírito Santo e avaliou o ajuste de modelos polinomiais de ordem 1 a 4 utilizando a longitude, latitude e altitude como variáveis explicativas para a previsão dessas precipitações. A precipitação no Espírito Santo mostrou variabilidade considerável, indicando grande influência do relevo, sendo observado também que localidades com maiores altitudes apresentaram maiores totais precipitados. A regressão polinomial de quarto grau se mostrou a mais adequada em representar as precipitações médias mensais e anual. Os ajustes foram considerados suficientes para representar as precipitações do Espírito Santo, com coeficientes de determinação superiores a 0,7 e erros percentuais absolutos médios entre 5,9% e 16,6%. Foi observada uma leve tendência dos modelos em subestimar os valores observados. De maneira geral, os meses do período seco, especialmente de maio a julho, obtiveram melhor desempenho dos modelos.Palavras-chave: modelagem matemática, climatologia, hidrologia, chuva. ESTIMATION OF PRECIPITATION IN THE ESPÍRITO SANTO STATE BY POLYNOMIAL REGRESSION ABSTRACT: Precipitation is one of the main elements of hydrology, being a variable of great importance for understanding the dynamics of the hydrological cycle. Despite their importance, the availability of hydroclimatic data is low. Among the alternatives to meet the need for precipitation information, mathematical modeling is an important tool that aims its estimate. This work evaluated the monthly and annual rainfall of 110 rainfall gauges in the state of Espírito Santo and evaluated the adjustment of polynomial models of order 1 to 4 using longitude, latitude and altitude as explanatory variables to predict these precipitations. Precipitation in Espírito Santo showed considerable variability, indicating great influence of the relief, being observed that location with higher altitudes presented higher precipitated totals. The fourth-degree polynomial regression proved to be the most adequate to represent the mean monthly and annual precipitations. The adjustments were considered sufficient to represent the Espírito Santo precipitation, with coefficients of determination higher than 0.7 and mean absolute percentage errors between 5.9% and 16.6%. A slight trend of the models was observed in underestimating the observed values. In general, the months of the dry period, especially from May to July, obtained better performance of the models.Keywords: mathematical modeling, climatology, hydrology, rainfall.

Analysis of monthly and annual rainfall variability using linear models in the state of Mato Grosso do Sul, Midwest of Brazil

“…In the first degree multiple regression, the c‐index scores were: in November—median (0.61 ≤ c < 0.65); in April and December—good (0.76 ≤ c < 0.85); in October and for annual value—very good (0.76 ≤ c < 0.85); in the rest of the months—excellent (c > 0.85). These results are considered satisfactory in the provision of monthly rainfall and, especially in the case of third degree regression, better than those obtained in studies using similar techniques (Mello and Silva, 2009; Bagirov et al ., 2017; Abreu et al ., 2019).…”

“…Understanding the spatiotemporal distribution of the rainfall is extremely important for climatological and water resources management. The applications of this knowledge vary from more accurate climatological classifications (Mello and Silva, 2009) to better water availability modelling (Silva et al ., 2011; Asarian and Walker, 2016; Abreu et al ., 2019).…”

“…Despite its importance, the availability of hydroclimatic information is insufficient in spatiotemporal terms in developing countries, a fact that is aggravated due Brazil's large territorial extension (Abreu et al ., 2019). Some alternatives to meet hydrological information needs had been satisfactorily applied, such as rainfall mapping by spatial interpolation (Viola et al ., 2010; Silva et al ., 2011; Correa et al ., 2014; Lyra et al ., 2018), usage of satellite images (Almeida et al ., 2015; Prakash et al ., 2018; Pessi et al ., 2019) or even through multiple polynomial regressions (Mello and Silva, 2009; Bagirov et al ., 2017; Abreu et al ., 2019).…”

“…On the other hand, the usage of multiple polynomial regression is a simple statistical method that can be applied without the need for more sophisticated computational resources. For comparison, the spatial interpolation of rainfall data (Viola et al ., 2010; Silva et al ., 2011; Lyra et al ., 2018) as well as the satellite images (Almeida et al ., 2015; Bagirov et al ., 2017; Silva and Oliveira, 2017; Pessi et al ., 2019) performed worse than multiple polynomial regression (Mello and Silva, 2009; Abreu et al ., 2019) when predicting rainfall spatial distribution.…”

“…Hydrological variables, such as rainfall, are often correlated with geographical coordinates—latitude, longitude, and altitude. Therefore, these geographic features can be used in multiple polynomial regression as predictors of average monthly and annual rainfall in regions where this information is scarce (Mello and Silva, 2009; Bagirov et al ., 2017; Abreu et al ., 2019).…”