Detecting Coral Reef Presence Using ICESat-2 Data and Machine Learning Methods - Conference Proceeding

Aug 5, 2026

Upload by Miles

Ocean warming and more frequent severe storms increase the vulnerability of benthic habitats, including coral reefs, to mass bleaching and infectious diseases. Effective monitoring is often limited by labor-intensive in-situ survey methods such as fixed-site photography and visual counts. This study uses ICESat-2 green-laser elevation data as a primary source to detect coral reefs near Heron Island, Australia. Classic ICESat-2 variables are combined with window-based seafloor pseudo-rugosity features, evaluated with logistic regression and convolutional neural networks, with accuracy gains from Sentinel-2 derived bathymetry.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages2

Reading Time5 min

AccessFree to download

Overview and Motivation

Monitoring Need and Limitations

ICESat-2 Data Role

Data and Methods

Study Area: Heron Island, Australia

Feature Construction From Seafloor Characteristics

Machine Learning Models

Results and Comparative Performance

Logistic Regression Outcomes

Convolutional Neural Network Comparison

Impact of Sentinel-2 Bathymetry

Implications and Future Work

Automated Reef Identification

Extension to Other Reefs and Benthic Habitats

University of New Hampshire University of New Hampshire Scholars Repository | Center for Coastal and Ocean Mapping | Center for Coastal and Ocean Mapping | | --- | --- | | 5-30-2024
Detecting Coral Reef Presence Using ICESat-2 Data and Machine Learning Methods
Gabrielle Trudeau
University of New Hampshire, Durham
Lowell Kim
University of New Hampshire, Durham, [Kim.Lowell@unh.edu](Kim.Lowell@unh.edu)
Follow this and additional works at: [https://scholars.unh.edu/ccom](https://scholars.unh.edu/ccom) | | Recommended Citation Trudeau, Gabrielle and Kim, Lowell, "Detecting Coral Reef Presence Using ICESat-2 Data and Machine Learning Methods" (2024) . Canadian Hydrographic Conference, Saint John's, Newfoundland, Canada, May 27-30. 1421. [https://scholars.unh.edu/ccom/1421](https://scholars.unh.edu/ccom/1421) This Conference Proceeding is brought to you for free and open access by the Center for Coastal and Ocean Mapping at University of New Hampshire Scholars Repository. It has been accepted for inclusion in Center for Coastal and Ocean Mapping by an authorized administrator of University of New Hampshire Scholars Repository. For more information, [please contact](please contact Scholarly.Communication@unh.edu)[ Scholarly.Communication@unh.edu](please contact Scholarly.Communication@unh.edu). Detecting coral reef presence using ICESat-2 data and machine learning methods Gabrielle Trudeau 1, Lowell Kim 1 1 Centre for Coastal and Ocean Mapping, University of New Hampshire, Durham, New Hampshire, USA [Gabrielle.trudeau@unh.edu](Gabrielle.trudeau@unh.edu) As ocean temperatures and sea levels continue to increase and weather storms become more severe and frequent, benthic habitats such as coral reefs become more vulnerable to deadly conditions such as mass bleaching events and infectious diseases. With these ever-changing conditions, it becomes imperative that monitoring efforts are made to ensure longevity of the world’s coral reefs. Current coral reef monitoring techniques require significant manpower to collect in-situ data, such as fixed site surveys using photography and visual counts. The Ice, Cloud, and land Elevation Satellite-2 (ICESat-2), while initially intended for collecting data regarding changes in the cryosphere, utilizes a green laser thus opening the door for an abundance of oceanic and bathymetric applications. ICESat-2 is currently an underutilized data source for ocean-related purposes, despite its high resolution and frequency, as well as the economical alternatives it potentially offers the remote sensing community. Using ICESat-2 as the primary data source, in this study machine learning methods are used in the detection of coral reefs located around Heron Island, Australia. Classic ICESat-2 variables such as date, depth and geographical location are used in conjunction with algorithmically extracted features of the seafloor such as a window-based pseudo-rugosity measurement. Binary logistic regression results are promising, motivating a comparison with convolutional neural network results. Both machine learning models show that the addition of Sentinel-2 satellite derived bathymetry values increase accuracies of coral detection. This research suggests ICESat-2 to be a useful data source in future coral reef monitoring methodologies. Ongoing work examines the value of automated reef identification in developing monitoring methodologies, as well as the value of other information that can be extracted from ICESat-2 data alone. Future steps will explore the applicability of these results to other types of reefs or benthic habitats.

FAQs

Why is coral reef monitoring important in the context of climate change?

Rising ocean temperatures and more severe weather increase the risk of mass bleaching events and infectious diseases, making long-term monitoring essential for reef longevity.

How does this study use ICESat-2 data for coral reef detection?

ICESat-2 elevation measurements are treated as a primary remote-sensing input, combined with algorithmically extracted seafloor features such as window-based pseudo-rugosity.

Which models were tested and what improved detection accuracy?

Binary logistic regression and convolutional neural networks were compared, and both improved when Sentinel-2 satellite-derived bathymetry values were added.

Aug 5, 2026

Upload by Miles