Coastal Bathymetry Estimation from Sentinel-2 Satellite Imagery: Comparing Deep Learning and Physics-Based Approaches

Abstract

The ability to monitor the evolution of the coastal zone over time is an important factor in coastal knowledge, development, planning, risk mitigation, and overall coastal zone management. While traditional bathymetry surveys using echo-sounding techniques are expensive and time consuming, remote sensing tools have recently emerged as reliable and inexpensive data sources that can be used to estimate bathymetry using depth inversion models. Deep learning is a growing field of artificial intelligence that allows for the automatic construction of models from data and has been successfully used for various Earth observation and model inversion applications. In this work, we make use of publicly available Sentinel-2 satellite imagery and multiple bathymetry surveys to train a deep learning-based bathymetry estimation model. We explore for the first time two complementary approaches, based on color information but also wave kinematics, as inputs to the deep learning model. This offers the possibility to derive bathymetry not only in clear waters as previously done with deep learning models but also at common turbid coastal zones. We show competitive results with a state-of-the-art physical inversion method for satellite-derived bathymetry, Satellite to Shores (S2Shores), demonstrating a promising direction for worldwide applicability of deep learning models to inverse bathymetry from satellite imagery and a novel use of deep learning models in Earth observation.

A Combined Color and Wave-Based Approach to Satellite Derived Bathymetry Using Deep Learning

“…In this work, we explore a hybrid approach to SDB, named H-DSPEB, that combines both types of information. After the image date selection based on the performance of S2Shores (we refer to (Al Najar et al, 2022) for further details on date selection), our preprocessing workflow is focused on detecting and separating the signals representing actual ocean waves from the remaining information found in our Sentinel-2 subtiles. For each subtile, a pass-band filter in the spectral domain is applied to retain signals corresponding the range of ocean-specific wavelengths (periods Tmin = 5 s to Tmax = 25 s).…”

“…These data products have been used to study a wide array of natural processes in the coastal zone (Brando and Dekker, 2003, Yuan et al, 2020, Wei et al, 2021, Schratz et al, 2021, da Silveira et al, 2021. In the domain of bathymetry inversion from satellite imagery, two main methodologies have been studied and can be identified in the literature (Al Najar et al, 2022). These can be categorized based on the natural phenomena or process they exploit to invert water depth.…”

“…In the context of satellite derived bathymetry, DL-based works in the literature mostly explore the color-based approach, example studies include (Dickens and Armstrong, 2019, Wilson et al, 2020, Lumban-Gaol et al, 2021. On the other hand, great expectations come from the combination of different methods, in particular, based on wave information (Danilo and Melgani, 2016, Benshila et al, 2020, Al Najar et al, 2021, Al Najar et al, 2022. In this work, we develop on our Deep Single-Point Estimation of Bathymetry method (DSPEB) as presented in our previous works (Al Najar et al, 2021, Al Najar et al, 2022.…”

“…On the other hand, great expectations come from the combination of different methods, in particular, based on wave information (Danilo and Melgani, 2016, Benshila et al, 2020, Al Najar et al, 2021, Al Najar et al, 2022. In this work, we develop on our Deep Single-Point Estimation of Bathymetry method (DSPEB) as presented in our previous works (Al Najar et al, 2021, Al Najar et al, 2022. Specifically, we train and apply the DSPEB method using a number of Sentinel-2 L2A images coupled with bathymetry surveys obtained from the French Naval Hydrographic and Oceanographic Service.…”

“…We present two different satellite image pre-processing techniques to augment wave kinematics and color information as inputs to two DSPEB models. In addition, as a continuation of our preliminary work on hybrid models in (Al Najar et al, 2022), we investigate a multi-headed convolutional neural network architecure (CNN) named Hybrid-DSPEB (H-DSPEB) with the aim of incorporating both color and wave kinematics information into a multi-input single-output model. We present the performances of the three DSPEB models and we compare them to a state-ofthe-art bathymetry inversion models based on wave kinematics, Satellite to Shores (S2Shores) (Bergsma et al, 2021).…”