Rainfall Forecasting using Spatio-Temporal and Neural Network Study Case: Meteorological Data of Madura Island
Year: 2025
Type: article
Abstract: Rainfall forecasting is crucial in meteorological studies due to its significant impact on sectors such as agriculture, which is the main livelihood on Madura Island. This study aims to forecast rainfall on Madura Island using a hybrid approach that combines the Generalized Space-Time Autoregressive-X (GSTARX) model and Neural Network (NN). The data used consist of daily rainfall records from Bangkalan, Sampang, Pamekasan, and Sumenep, covering the period from January 2013 to December 2023. Data from January 2013 to September 2023 were used for training, while data from October to December 2023 were used for testing. The GSTARX model was employed to capture spatio-temporal patterns, while the NN was applied to learn the non-linear relationships in the residuals. The results show that the GSTARX model effectively captures rainfall patterns, though some differences remain compared to the actual data, with RMSE values of Bangkalan (1.514), Sampang (0.256), Pamekasan (0.477), and Sumenep (0.127). Meanwhile, the hybrid GSTARX-FFNN model achieved improved forecasting performance in Sampang (0.392), Pamekasan (0.679), and Sumenep (0.412), although Bangkalan recorded a higher RMSE (1.359). Overall, the GSTARX model proved more effective in forecasting rainfall on Madura Island, delivering smaller and more consistent prediction errors.
Authors: Ryanta Meylinda Savira, Regita Putri Permata, Amalia Nur Alifah, Yohanes Setiawan, Adzanil Rachmadhi Putra
Language: English
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: gold
Estimated APC paid: $110
Funder: Universitas Telkom
OpenAlex ID: W4413179491