Modelagem preditiva da precipitação hídrica com algoritmos de séries temporais na Bacia do Rio Formoso (TO): subsídios para a gestão hídrica no contexto da agroenergia digital

Abstract

Efficient water management and the sustainable use of agricultural residues are central challenges for the advancement of bioenergy and second-generation (2G) biofuel production. This dissertation applied time-series statistical models, focusing on SARIMA (Seasonal Autoregressive Integrated Moving Average), to forecast rainfall in the Formoso River Basin (Tocantins State), using data from the GAN Project. Among the tested configurations, SARIMA(0,1,1)(0,1,1)₁₂ presented the best performance, with the lowest AIC and BIC values, residuals without autocorrelation, and error metrics (MAPE and RMSE) within accepted standards. The model demonstrated high predictive capacity, capturing annual rainfall trends and seasonality, providing reliable information for water and agroenergy planning. The integration of rainfall forecasts with agricultural waste management logistics demonstrated direct benefits to the production sector, optimizing biomass storage, transportation, and processing, resulting in energy efficiency gains, cost reductions, and increased profitability for rural producers. Scientifically, the study reinforces the relevance of Digital Agroenergy and predictive modeling in climate management. In practice, it offers a strategic tool for decision- making in the field, connecting science, technology, and sustainable production. The conclusion is that the use of advanced statistical models with local data strengthens sustainability policies and increases the competitiveness of regional bioenergy, paving the way for new research integrating climate forecasting and agroenergy waste management.