Machine Learning and the GR2M Model for Monthly Runoff Forecasting

Hydrologic Model Runoff Forecasting Machine Learning GR2M Thailand.

Authors

  • Natapon Kaewthong Department of Civil Engineering, Faculty of Engineering Rajamangala University of Technology Srivijaya, Songkhla 90000,, Thailand
  • Torlap Kanplumjit Department of Civil Engineering, Faculty of Engineering Rajamangala University of Technology Srivijaya, Songkhla 90000,, Thailand
  • Naras Kwanthong Faculty of Engineering and Technology Rajamangala University of Technology Srivijaya, Trang 92150,, Thailand
  • Kritsana Sureeya Research assistant, Rajamangala University of Technology Srivijaya, Songkhla 90000,, Thailand
  • Chayanat Buathongkhue
    chayanat.b@rmutsv.ac.th
    College of Industrial Technology and Management, Rajamangala University of Technology Srivijaya, Nakhon Si Thammarat 80210,, Thailand https://orcid.org/0009-0001-3441-2156

Downloads

This article presents the results of an analysis of monthly rainfall into monthly runoff using Machine Learning algorithms, including Multiple Linear Regression, Multilayer Perceptron, and Support Vector Machine, which were compared with the GR2M hydrologic model to identify the most suitable approach for rainfall-runoff analysis in watersheds in the lower southern region of Thailand. This region is characterized by its unique geographic location at the border between Thailand and Malaysia. It faces challenges due to uncertainty in rainfall data, measured only on the Thai side, leading to a lack of corresponding data from Malaysia. The analysis found that the Machine Learning Support Vector Machine algorithm consistently provided the most accurate results across all sub-basins. Sub-basin TU02 achieved an MAE of 2.63 mm/month, while sub-basin X.119A had an MAE of 68.10 mm/month, sub-basin X.184 had an MAE of 145.05 mm/month, and sub-basin X.274 had an MAE of 66.08 mm/month. This research demonstrated the utility of advanced algorithms in rainfall-runoff analysis for areas with partial or incomplete data coverage. The findings confirm that the Machine Learning Support Vector Machine algorithm outperformed the Hydrologic Model (GR2M) in terms of accuracy and reliability. Therefore, this study concludes that applying the Machine Learning Support Vector Machine algorithm is an optimal approach for runoff prediction in the southern region of Thailand and provides a framework for potential applications in other areas with similar data and geographic challenges.

 

Doi: 10.28991/CEJ-2025-011-01-022

Full Text: PDF