Aplicação da metodologia de Box-Jenkins para a previsão da precipitação na região Sul de Moçambique

Abstract

Precipitation variability is an important factor for agricultural planning, water resource management, and mitigating the impacts of extreme events. This study aimed to model and forecast monthly precipitation in southern Mozambique for a six-month time horizon, focusing on the cities of Maputo, Xai-Xai, and Inhambane. The study used monthly precipitation time series data from the National Institute of Meteorology, covering the period 1992 to 2022. Data were analyzed using seasonal autoregressive integrated and moving average models (SARIMA). From the analysis, the models that fitted the data well were the SARIMA(3, 0, 3)(2, 0, 2) 12 model for the city of Inhambane, the SARIMA(0, 0, 1)(1, 0, 1) 12 model for the city of Xai-Xai, and the SARIMA(1, 0, 0)(1, 0, 1) 12 model for the city of Maputo. The in-sample predictions were consistent with the historical behavior of the series, suggesting the application of the models for planning and decision-making purposes. The study concluded that the monthly rainfall series for the cities of Maputo, Xai-Xai, and Inhambane fit the SARIMA models, capturing their seasonal patterns. The diagnostics confirmed the quality of the adjustments and the usefulness of the models for forecasting, providing relevant support for actions to adapt to climate change and reduce the risks associated with extreme weather events.