Bathymetry Estimation Using Machine Learning in the Ulleung Basin in the East Sea

Aug 3, 2026

Accurate bathymetry estimation is achieved by integrating sea-surface free-air gravity anomalies with depth-related observations derived from satellite radar altimetry. Short-wavelength residual gravity anomalies are crucial for reliable bathymetry when combining satellite altimetry-derived free-air gravity anomalies with shipborne depth and gravity anomaly measurements. An optimized ensemble machine-learning framework is applied to residual gravity anomalies to estimate bathymetry using the gravity–geologic method (GGM) in the Ulleung Basin, improving predictions by 32.3 m over baseline GGM.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Bathymetry mapping and gravity field relevance

Prior work using satellite altimetry and machine learning

Gap in deep-water bathymetry studies

Study purpose and approach

Sensors and Materials, Vol. 35, No. 9 (2023) 3351–3362 3351 MYU Tokyo S & M 3396 Bathymetry Estimation Using Machine Learning in the Ulleung Basin in the East Sea Kwang Bae Kim, 1 Ji Sung Kim,2* and Hong Sik Yun3 1Department of Civil, Architectural, and Environmental System Engineering, Sungkyunkwan University, 2066 Seobu-ro, Jangan-gu, Suwon, Gyeonggi-do 16419, Republic of Korea 2School of Geography, University of Leeds, Woodhouse Lane, Leeds LS2 9JT, United Kingdom 3Interdisciplinary Program in Crisis, Disaster and Risk Management, Sungkyunkwan University, 2066 Seobu-ro, Jangan-gu, Suwon, Gyeonggi-do 16419, Republic of Korea (Received April 3, 2023; accepted August 15, 2023) Keywords: machine learning, Ulleung Basin, gravity–geologic method, satellite altimetry-derived freeair gravity anomalies, residual gravity anomalies Accurate bathymetry estimation is made possible by combining depth data with free-air gravity anomalies on the sea surface recovered from the geoidal heights that are equivalent to the mean sea surface derived from satellite radar altimetry. The residual gravity anomalies that represent the short-wavelength effect are required to accurately estimate bathymetry by combining satellite altimetry-derived free-air gravity anomalies and shipborne data including depth and gravity anomalies. In this study, the optimized ensemble model of machine learning techniques was applied to the residual gravity anomalies to estimate bathymetry by the gravity– geologic method (GGM) from various geospatial information including shipborne depth, shipborne gravity anomalies, and satellite altimetry-derived free-air gravity anomalies, in the Ulleung Basin in the East Sea. From the results, the GGM bathymetry predicted using the optimized ensemble model of machine learning was improved by 32.3 m over the GGMbathymetry estimated using the original depth and gravity anomalies. The method presented in this study is for estimating deep-water bathymetry using machine learning, and it has been proven to have superior performance compared with conventional methods. 1. Introduction Bathymetry mapping is important in understanding Earth’s gravity field. Recently, various research studies have been performed to predict accurate bathymetry using shipborne data, including depth and gravity anomalies, by utilizing high-resolution satellite altimetry-derived free-air gravity anomalies. Various machine learning technologies have also been developed and applied to each field of surveying. In bathymetry estimation in particular, research using machine learning technology has been actively conducted. In fact, many studies have revealed that machine learning is accurate and fast in comparison with conventional bathymetry measurement methods. Collin * Corresponding author: e-mail: [gyjki@leeds.ac.uk](gyjki@leeds.ac.uk)[ ](gyjki@leeds.ac.uk)[https://doi.org/10.18494/SAM4415](https://doi.org/10.18494/SAM4415) ISSN 0914-4935 © MYU K.K. [https://myukk.org/](https://myukk.org/) 3352 Sensors and Materials, Vol. 35, No. 9 (2023) and Hench(1) extracted spectra from very high resolution satellite images, formed neural bands through an artificial neural network, and estimated shallow-water bathymetry. In their study, the authors demonstrated that bathymetry estimation using machine learning is superior to conventional methods. Misra et al.(2) extracted shallow-water bathymetry using multispectral images of satellites, similarly to Collin and Hench.(1) However, their study used a support vector machine (SVM) rather than an artificial neural network. A significant number of studies onbathymetry estimation in shallow areas using satellite imagery and artificial intelligence have been conducted since 2015.(3–10) Most of these studies used multispectral images obtained from satellites but different machine learning methods. Owing to the characteristics of multispectral images, it is difficult to measure a large water depth, so t

FAQs

Why are residual gravity anomalies required for bathymetry estimation?

Residual gravity anomalies capture short-wavelength effects. They enable accurate bathymetry when fusing satellite altimetry-derived free-air gravity anomalies with shipborne depth and gravity anomaly data.

What machine-learning approach is used in the study?

The study applies an optimized ensemble model of machine learning techniques to residual gravity anomalies, producing bathymetry estimates through the gravity–geologic method (GGM).

How much improvement does the optimized ensemble model achieve?

The GGM bathymetry predicted with the optimized ensemble model improves by 32.3 m compared with GGM estimates based on the original depth and gravity anomalies.

Aug 3, 2026

Upload by Emma Wilson