High-Resolution Satellite Bathymetry Mapping: Regression and Machine Learning-Based Approaches

Abstract

Remote spectral imaging of coastal areas can provide valuable information for their sustainable management and conservation of their biodiversity. Unfortunately, such areas are very sensitive to changes due to human activity, natural phenomenona, introduction of non-native species and climate change. Thus, the main objective of this research is the implementation of a robust image processing methodology to produce accurate bathymetry maps in shallow coastal waters using high resolution multispectral WorldView-2/3 satellite imagery for the monitoring at the maximum spatial and spectral resolutions. Two different island ecosystems have been selected for the assessment, since they stand out for their richness in endemic species and they are more vulnerable to climate change: Cabrera National Park and Maspalomas Natural Protected area, located in the Balearic and Canary Islands, Spain, respectively. In addition, a third example to show the applicability of the mapping methodology to monitor the construction of a new port in Granadilla (Canary Islands) is presented. Contributions of this work focus on improving the preprocessing methodology and, mainly, on the proposal and assessment of new satellite derived regression and machine learning bathymetric models, which have been validated and compared with respect to measured reference bathymetry. After a thorough analysis of 9 techniques, using visual and quantitative statistical parameters, ensemble learning approaches have demonstrated excellent performance, even in challenging scenarios up to 35 m depth, with mean RMSE values around 2 m.

Satellite-Derived Bathymetry of an Inland Reservoir in India

“…in [7] are the two widely-accepted classical approaches, where the relationship between the reflectance values and the depth of the water column are considered to be linear. The classical Lyzenga and Stumpf models have been extended in various directions, including towards accounting for non-linear relationship between the spectral reflectance values and the depth of the water column [10]- [14] and for incorporating variations in atmosphere and water surface transmittance [15]. Several machine learning algorithms, such as multivariate regression, support vector machine (SVM), K-nearest neighbor (KNN), random forest [14], [16]- [20] have also been considered.…”

“…The classical Lyzenga and Stumpf models have been extended in various directions, including towards accounting for non-linear relationship between the spectral reflectance values and the depth of the water column [10]- [14] and for incorporating variations in atmosphere and water surface transmittance [15]. Several machine learning algorithms, such as multivariate regression, support vector machine (SVM), K-nearest neighbor (KNN), random forest [14], [16]- [20] have also been considered. Following the recent adoption and success of deep learning in image processing and other areas, basic fully-connected feedforward neural networks (FCFFNN) [16], [21], [22], convolutional neural networks (CNN) [23]- [29], gated recurrent unit (GRU) networks [30] and hybrid models where particle swarm optimization (PSO) and optimally pruned extreme learning machine (OPELM) are combined with neural networks, have been used for achieving better performance over the classical and machine learning models.…”

“…Most of the above mentioned models for SDB are supervised learning based models, where the models are trained to estimate depth of water columns from satellite data/images using certain ground truth data. The works use depth information collected from acoustic systems [11]- [13], [15], [16], [20]- [23], [31]- [35], airborne LiDAR systems [10], [11], [14], [17], [19], [26], [28], [36]- [38] or LiDAR data obtained from Ice, Cloud, and Elevation Satellite-2 (ICESat-2) satellites [18], [29], [30], as ground truth data. The study areas of interest in most of the above works belong to coastal regions [10]- [12], [14], [15], [17]- [19], [21], [28], [32], [34], [37], [39] and only a small number of works study SDB for inland water bodies [13], [16], [38].…”

“…The works use depth information collected from acoustic systems [11]- [13], [15], [16], [20]- [23], [31]- [35], airborne LiDAR systems [10], [11], [14], [17], [19], [26], [28], [36]- [38] or LiDAR data obtained from Ice, Cloud, and Elevation Satellite-2 (ICESat-2) satellites [18], [29], [30], as ground truth data. The study areas of interest in most of the above works belong to coastal regions [10]- [12], [14], [15], [17]- [19], [21], [28], [32], [34], [37], [39] and only a small number of works study SDB for inland water bodies [13], [16], [38]. As mentioned, most of the above works consider the reflectance values collected by satellites as input for estimating depth of water columns.…”

“…In the existing works, the trained models are evaluated on the same region in [11], [13], [21], [23], [30], [32]- [34], [37], [38] and different nearby regions in [10], [12], [14]- [19], [26], [28], [29], [35], [36]. We evaluate the proposed models in the same region, as in multiple works above.…”