ESTIMATING SHALLOW-WATER BATHYMETRY FROM SENTINEL-2 IMAGERY USING MACHINE LEARNING: A CASE STUDY IN COASTAL AREA OF GIA LAI, VIETNAM

Abstract

This study investigates the potential of applying machine learning algorithms to estimate shallow-water bathymetry from Sentinel-2 satellite imagery in the coastal waters of Gia Lai province, Vietnam. A Sentinel-2 Level-2A image acquired on January 23, 2020, under clear and stable optical conditions, was used after atmospheric correction. In-situ bathymetric data were collected using a GPS-integrated single-beam echo sounder and employed as training and validation datasets. Five machine learning models, including Multiple Linear Regression (MLR), Random Forest (RF), Decision Tree (DT), K-Nearest Neighbors (KNN), and Long Short-Term Memory (LSTM), were developed and compared for bathymetric estimation accuracy. The results indicate that the coefficient of determination (R²) ranges from 0.80 to 0.91, with Root Mean Square Error (RMSE) values between 0.72 m and 1.05 m, and Mean Absolute Error (MAE) values ranging from 0.49 m to 0.72 m. These findings demonstrate the feasibility of integrating Sentinel-2 imagery with machine learning for accurate shallow-water bathymetric mapping. Furthermore, the proposed approach shows strong potential for application to other coastal regions of Vietnam with comparable environmental and optical characteristics.