Investigating Scene Segmentation for Improving Satellite-Derived Bathymetry Accuracy and Coverage.

Saved in:

Bibliographic Details
Title:	Investigating Scene Segmentation for Improving Satellite-Derived Bathymetry Accuracy and Coverage.
Authors:	Affonso, Juliane¹ (AUTHOR), Kastrisios, Christos¹ (AUTHOR) Christos.Kastrisios@unh.edu, Parrish, Christopher E.^1,2 (AUTHOR), Calder, Brian¹ (AUTHOR)
Source:	Journal of Coastal Research. 2026, Vol. 42 Issue 4, p706-719. 14p.
Subjects:	Image segmentation, Bathymetry, Oceanographic maps, Spatial analysis (Statistics), Remote-sensing images
Abstract:	Affonso, J.; Kastrisios, C.; Parrish, C.E., and Calder, B., 2026. Investigating scene segmentation for improving satellite-derived bathymetry accuracy and coverage. Journal of Coastal Research, 42(4), 706–719. Charlotte (North Carolina), ISSN 0749-0208. Seafloor bathymetry is essential for assisting safe maritime navigation through up-to-date navigational charts. Satellite-derived bathymetry (SDB) can serve as a low-cost, rapid alternative or supplement to shipborne and airborne surveys, particularly for mapping remote and shallow areas. However, the accuracy of SDB is relatively low compared to other methods, and the spatial coverage is often limited. One contributing factor is that conventional SDB approaches assume that bottom type and water clarity remain constant throughout the entire area of interest, such that a single linear model can be used to retrieve bathymetric information. To enhance the accuracy and coverage of depth estimation, this work explores the segmentation of the scene into smaller spatial units and the incorporation of the water-column contribution in the SDB equation. The segmentation methods consisted of six approaches: three horizontal, one vertical (by depth), and two combinations of horizontal and vertical methods. Utilizing Sentinel-2 images over the Dry Tortugas and the St. Thomas East End Reserve, it was found that, with one exception, the geographic segmentation methods improved the accuracy of depth estimation. Accuracy improvements of up to 60% were observed, although the vertical segmentation results showed some indications of potential overfitting to the reference data, particularly at the smaller vertical bin sizes. Overall, the results provide strong indication that scene segmentation can improve both the accuracy and coverage of depth estimates compared to traditional SDB methods that employ only a single model. [ABSTRACT FROM AUTHOR]
	Copyright of Journal of Coastal Research is the property of KnowledgeWorks Global, Ltd and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Affonso, J.; Kastrisios, C.; Parrish, C.E., and Calder, B., 2026. Investigating scene segmentation for improving satellite-derived bathymetry accuracy and coverage. Journal of Coastal Research, 42(4), 706–719. Charlotte (North Carolina), ISSN 0749-0208. Seafloor bathymetry is essential for assisting safe maritime navigation through up-to-date navigational charts. Satellite-derived bathymetry (SDB) can serve as a low-cost, rapid alternative or supplement to shipborne and airborne surveys, particularly for mapping remote and shallow areas. However, the accuracy of SDB is relatively low compared to other methods, and the spatial coverage is often limited. One contributing factor is that conventional SDB approaches assume that bottom type and water clarity remain constant throughout the entire area of interest, such that a single linear model can be used to retrieve bathymetric information. To enhance the accuracy and coverage of depth estimation, this work explores the segmentation of the scene into smaller spatial units and the incorporation of the water-column contribution in the SDB equation. The segmentation methods consisted of six approaches: three horizontal, one vertical (by depth), and two combinations of horizontal and vertical methods. Utilizing Sentinel-2 images over the Dry Tortugas and the St. Thomas East End Reserve, it was found that, with one exception, the geographic segmentation methods improved the accuracy of depth estimation. Accuracy improvements of up to 60% were observed, although the vertical segmentation results showed some indications of potential overfitting to the reference data, particularly at the smaller vertical bin sizes. Overall, the results provide strong indication that scene segmentation can improve both the accuracy and coverage of depth estimates compared to traditional SDB methods that employ only a single model. [ABSTRACT FROM AUTHOR]
ISSN:	07490208
DOI:	10.2112/JCOASTRES-D-26TM1-00001.1