Forecasting seasonal rainfall in Uruguay: A comprehensive analysis using ARIMA and SARIMA models

Climate significantly impacts societies and the environment, with precipitation crucial for ecosystem formation, agriculture development, and water resource management. This study analyzes monthly rainfall data from 1991 to 2021 across Uruguay using data from 17 Uruguayan Institute of Meteorology (Inumet) stations. The main objective is to use autoregressive models for rainfall modeling and forecasting, a task that has significant implications for our understanding and prediction of climate patterns. After temporal and spatial analyses, classical decomposition showed better results with the additive model. The augmented Dickey-Fuller test confirmed series stationarity, while the Mann-Kendall test found no significant trend in monthly rainfall across different regions. Additionally, ACF and PACF results showed no strong trend or stationarity. These findings highlight the autoregressive models relevance for understanding climate and its impacts on the environment. The different performance statistics in model development and validation phase of this study also confirmed higher prediction accuracy using SARIMA models. The SARIMA model gave the best results for time series predictions by region.