Adjustment of a time series model to predict rainfall

• Autor(a) principal: Santos, Pedro Henrique Alves Bittencourt
• Data de Publicação: 2021
• Outros Autores: Delfino, Otávio Augusto dos Santos, Santos, Ricardo Vitor Ribeiro dos, Nascimento, Mateus do
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Research, Society and Development
• Texto Completo: https://rsdjournal.org/index.php/rsd/article/view/15643
• topic: Meteorológicas; Previsões; Sazonalidade; Modelo SARIMA.; Meteorological; Forecasts; Seasonality; Model SARIMA.; Meteorológico; Previsión; Estacionalidad; Modelo SARIMA.

Precipitation is one of the most relevant meteorological variables for climate studies. Knowing its spatial and temporal variability allows planning various human activities, both from an economic and social point of view. Such importance is due to the consequences that it can cause, in excess or in lack, causing floods, floods, droughts, decrease in energy supply, low food production, among others. This study aimed to study the historical series of average monthly rainfall in the city of Lavras/MG in order to obtain a statistical model that allows predictions to be made. For this purpose, 228 observations were used corresponding to the period from January 2000 to December 2018 for this analysis, the existence of the trend and seasonality components was verified. The Box and Jenkins methodology was used to model the data. Some models were adjusted using the SARIMA class, as the series under study showed stochastic seasonality. The comparison between the models considered suitable for the series was performed using the NDE and AIC. The SARIMA (0,0,0) x (0,1,1)12 model was used to make predictions of future observations. The series of monthly average rainfall in the city of Lavras/MG presented a seasonal component with a periodicity of 12 months. The adjusted model obtained a very good result, since the 95% confidence intervals contained the twelve real values ​​of average monthly rainfall in the city of Lavras/MG for the year 2019, even in the face of unforeseen and uncertainties associated with climatic factors. The model in question can be used in decision making to carry out future strategic plans that involve public issues associated with the city of Lavras. These forecasts can also be used to assist the managers of the Funil/MG hydroelectric plant to schedule future water flow operations and maintenance properly, as it is close to the municipality of Lavras.