Gravity-conditioned diffusion for seafloor bathymetric resolution enhancement

Global bathymetric products such as SRTM15+ provide extensive ocean coverage but cannot resolve many fine-scale seafloor features captured by shipborne multibeam surveys. Generative reconstruction methods can reconstruct more detailed terrain, but they may also introduce plausible-looking structures that are not supported by observations. We propose GravBathy-Diff, a conditional diffusion framework that uses observed low-resolution bathymetry as its primary input and incorporates satellite-altimetry-derived marine gravity as an auxiliary physical constraint. The gravity-differential enhancement module jointly encodes bathymetry, terrain derivatives, and the co-registered gravity field at the input stage. Gravity-conditioned time–frequency modulation regulates the contribution of gravity across diffusion stages and feature bands, while gravity-guided attention focuses this information on the spatial locations and feature channels where it is most useful. Deflections of the vertical, gravity anomalies, and vertical gravity gradients derived from Surface Water and Ocean Topography (SWOT) wide-swath altimetry are compared as alternative conditioning fields. Experiments are conducted on real SRTM15+ –multibeam bathymetry pairs from four ocean regions and assessed using depth error, perceptual similarity, terrain derivatives, and spectral fidelity. GravBathy-Diff achieves the strongest overall performance among the evaluated regression, adversarial, and diffusion methods, with the exception of one regional gradient-RMSE result. It ranks first on all six metrics averaged over four additional ocean areas and remains effective on an independent public digital elevation model (DEM) benchmark at 4×. Among the three gravity representations, the vertical gravity gradient provides the most effective guidance for reconstructing local relief. These results show that marine gravity can support the recovery of fine-scale seafloor structure while reducing high-frequency artifacts that are inconsistent with the available observations.

Authors

Institutions

Publication Details

Journal
GIScience & Remote Sensing
Published
2026-11-14
DOI
https://doi.org/10.1080/15481603.2026.2757465
Primary Topic
Underwater Acoustics Research
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Gravity-conditioned diffusion for seafloor bathymetric resolution enhancement

Dechao An, Xingyuan Yan, Zeyuan Zhang, Min Zhong et al.
5 links
GIScience & Remote Sensing
Underwater Acoustics Research
article

Gravity-conditioned diffusion for seafloor bathymetric resolution enhancement

Dechao An, Xingyuan Yan, Zeyuan Zhang, Min Zhong, Meng Yang, Zhongshan Jiang, Mingzhi Sun, Wei Feng
article en

Abstract

Global bathymetric products such as SRTM15+ provide extensive ocean coverage but cannot resolve many fine-scale seafloor features captured by shipborne multibeam surveys. Generative reconstruction methods can reconstruct more detailed terrain, but they may also introduce plausible-looking structures that are not supported by observations. We propose GravBathy-Diff, a conditional diffusion framework that uses observed low-resolution bathymetry as its primary input and incorporates satellite-altimetry-derived marine gravity as an auxiliary physical constraint. The gravity-differential enhancement module jointly encodes bathymetry, terrain derivatives, and the co-registered gravity field at the input stage. Gravity-conditioned time–frequency modulation regulates the contribution of gravity across diffusion stages and feature bands, while gravity-guided attention focuses this information on the spatial locations and feature channels where it is most useful. Deflections of the vertical, gravity anomalies, and vertical gravity gradients derived from Surface Water and Ocean Topography (SWOT) wide-swath altimetry are compared as alternative conditioning fields. Experiments are conducted on real SRTM15+ –multibeam bathymetry pairs from four ocean regions and assessed using depth error, perceptual similarity, terrain derivatives, and spectral fidelity. GravBathy-Diff achieves the strongest overall performance among the evaluated regression, adversarial, and diffusion methods, with the exception of one regional gradient-RMSE result. It ranks first on all six metrics averaged over four additional ocean areas and remains effective on an independent public digital elevation model (DEM) benchmark at 4×. Among the three gravity representations, the vertical gravity gradient provides the most effective guidance for reconstructing local relief. These results show that marine gravity can support the recovery of fine-scale seafloor structure while reducing high-frequency artifacts that are inconsistent with the available observations.

GIScience & Remote SensingVol. 63(1)
Sun Yat-sen University (CN)
National Key Research and Development Program of China
Underwater Acoustics Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon