Satellite Derived Bathymetry with Sentinel-2 Imagery: Comparing Traditional Techniques with Advanced Methods and Machine Learning Ensemble Models

Abstract

Abstract Accurate charting of nearshore bathymetry is critical to the safe and dependable use of coastal waterways frequented by the trading, fishing, tourism, and ocean energy industries. The accessibility of satellite imagery and the availability of various satellite-derived bathymetry (SDB) techniques have provided a cost-effective alternative to traditional in-situ bathymetric surveys. Furthermore, improved algorithms and the advancement of machine learning models have provided opportunity for higher quality bathymetric derivations. However, to date the relative accuracy and performance between traditional physics-based techniques, improved physics-based methods, and machine learning ensemble models have not been adequately quantified. In this study, nearshore bathymetry is derived from Sentinel-2 satellite imagery near La Parguera, Puerto Rico utilizing a traditional band-ratio algorithm, a band-ratio switching method, a random forest machine learning model, and the XGBoost machine learning model. The machine learning models returned comparable results and were markedly more accurate relative to other techniques; however, both machine learning models required an extensive training dataset. All models were constrained by environmental influences and image spatial resolution, which were assessed to be the limiting factors for routine use of satellite-derived bathymetry as a reliable method for hydrographic surveying.