Integrating Spatial and Temporal Dynamics for Rainfall Prediction Using the STARMA Model in Uttar Pradesh

Abstract

Accurate rainfall forecasting is vital for agriculture, water resource planning, and disaster risk reduction, especially in regions like Uttar Pradesh, India, where communities are highly sensitive to climatic fluctuations. Conventional time series models such as autoregressive integrated moving average (ARIMA) and seasonal ARIMA, though proficient in modeling temporal trends and seasonality, often fail to capture spatial dependencies that significantly influence rainfall variability. This study adopts the space-time autoregressive moving average (STARMA) model, which seamlessly integrates both spatial and temporal dimensions. The dataset used in this study is monthly rainfall data (1981–2022) for 12 districts of Western Uttar Pradesh, sourced from NASA POWER Data Access Viewer (https://power.larc.nasa.gov/data-access-viewer/). A major advancement lies in the formulation of the spatial weight matrix (SWM) using the Riemannian great-circle (RGC) distance, allowing for the inclusion of spatial correlations up to the third-order neighborhood across selected districts. Empirical results highlight the STARMA model’s marked improvement in predictive accuracy over ARIMA and SARIMA, demonstrating its effectiveness in capturing regional rainfall dynamics.