Cost-Effective and high-resolution Bathymetric Mapping in Rivers: Leveraging Sentinel-2 and the Band-Ratio Algorithm

Accurate bathymetric mapping is essential for a wide range of applications, including coastal management, navigation, hydrodynamical modeling, and environmental monitoring. Traditional methods such as sonar and LIDAR surveys, while precise, are often cost-prohibitive, time-consuming, and limited in spatial coverage, particularly for remote or inaccessible areas. This study explores the application of Sentinel-2 satellite imagery (about 10m resolution) combined with the band-ratio algorithm as a high-resolution and cost-effective approach to estimating bathymetry in riverine environments. The Rhone river, a critical waterway in the Mediterranean region, has been selected as case study due to its environmental and economic significance.

The band-ratio algorithm utilizes the differential attenuation of light in the blue and green spectral bands to estimate water depth. Sentinel-2’s high spatial resolution and multispectral capabilities make it an ideal source for this method. A key aspect of this study was the evaluation of several atmospheric correction techniques to preprocess the satellite images by mitigating atmospheric interference  and ensuring accurate reflectance values. The tested correction methods included QGIS Dark Object Subtraction (DOS), ACOLITE Dark Spectrum Fitting (DSF), ACOLITE Exponential Rayleigh (EXP), and the C2RCC processor in ESA’s SNAP software. These methods were compared to identify the optimal approach for handling the optically complex waters of the study area.

EMODnet bathymetry data in the shelf off the Rhone river mouth was used to train the band-ratio algorithm through regression models that related the computed band-ratio index to observed water depths. The accuracy of the derived bathymetry was assessed using statistical metrics, including root mean square error (RMSE), correlation coefficient (R²), mean bias, and mean absolute error (MAE).

A subset of the Sentinel-2 images has been selected based on cloud cover, water clarity, and temporal relevance to the study period and among them the data acquired on September 11, 2022, provided the most accurate results. This image achieved an R² value of 0.8, an RMSE of 0.79 meters, and an index of agreement of 0.88 for depths ranging from 0 to 10 meters. These results demonstrate that the combined use of Sentinel-2 imagery (after proper atmospheric correction) and the band-ratio algorithm can yield reliable bathymetric estimates in shallow, moderately turbid river environments.