Quantifying 3D coral reef structural complexity from 2D drone imagery using artificial intelligence
Coral reef ecosystems hold important biological, ecological and economic value, but are in decline due to a combination of global and local stressors resulting in loss of reef structural complexity and biodiversity. Given the foundational role reef structure plays in marine ecosystem health and dive...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-03-01
|
| Series: | Ecological Informatics |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1574954124005004 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832595436409454592 |
|---|---|
| author | Aviv Suan Simone Franceschini Joushua Madin Elizabeth Madin |
| author_facet | Aviv Suan Simone Franceschini Joushua Madin Elizabeth Madin |
| author_sort | Aviv Suan |
| collection | DOAJ |
| description | Coral reef ecosystems hold important biological, ecological and economic value, but are in decline due to a combination of global and local stressors resulting in loss of reef structural complexity and biodiversity. Given the foundational role reef structure plays in marine ecosystem health and diversity, there is a need for tools capable of quantifying this complexity over larger scales and higher resolutions. In this study, we present an approach that integrates drones, various color space information, and deep learning neural networks to predict three-dimensional complexity metrics—height range, rugosity and fractal dimension—across large (thousands of m2) spatial scales. The validation of our model resulted in R2 values of 0.71, 0.65, and 0.56 for each metric, respectively, indicating a robust predictive capability. Because these three complexity metrics can be depicted through a plane equation, we could calculate the metric with the lowest color space predictive accuracy from the other two, thereby enhancing overall predictive accuracy. Our deep learning approach offers a scalable solution to measure structural complexity of shallow coral reefs, providing a balance of spatial resolution and extent that is more ecologically relevant than other large-scale, airborne approaches. The application of this method to four distinct patch reef habitats revealed unique structural signatures among seemingly similar reefs, highlighting the importance of site-specific considerations in assessing reef complexity. A scalable, high-resolution method such as we present will provide a more cost-effective and nuanced understanding of reef habitats, ultimately aiding in their conservation and management amid ongoing environmental challenges. |
| format | Article |
| id | doaj-art-f030582df9f8413cb67fde6a54a5d966 |
| institution | Kabale University |
| issn | 1574-9541 |
| language | English |
| publishDate | 2025-03-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Ecological Informatics |
| spelling | doaj-art-f030582df9f8413cb67fde6a54a5d9662025-01-19T06:24:39ZengElsevierEcological Informatics1574-95412025-03-0185102958Quantifying 3D coral reef structural complexity from 2D drone imagery using artificial intelligenceAviv Suan0Simone Franceschini1Joushua Madin2Elizabeth Madin3Corresponding author.; Hawai‘i Institute of Marine Biology, University of Hawai‘i at Mānoa, Kāne‘ohe, HI 96744, USAHawai‘i Institute of Marine Biology, University of Hawai‘i at Mānoa, Kāne‘ohe, HI 96744, USAHawai‘i Institute of Marine Biology, University of Hawai‘i at Mānoa, Kāne‘ohe, HI 96744, USAHawai‘i Institute of Marine Biology, University of Hawai‘i at Mānoa, Kāne‘ohe, HI 96744, USACoral reef ecosystems hold important biological, ecological and economic value, but are in decline due to a combination of global and local stressors resulting in loss of reef structural complexity and biodiversity. Given the foundational role reef structure plays in marine ecosystem health and diversity, there is a need for tools capable of quantifying this complexity over larger scales and higher resolutions. In this study, we present an approach that integrates drones, various color space information, and deep learning neural networks to predict three-dimensional complexity metrics—height range, rugosity and fractal dimension—across large (thousands of m2) spatial scales. The validation of our model resulted in R2 values of 0.71, 0.65, and 0.56 for each metric, respectively, indicating a robust predictive capability. Because these three complexity metrics can be depicted through a plane equation, we could calculate the metric with the lowest color space predictive accuracy from the other two, thereby enhancing overall predictive accuracy. Our deep learning approach offers a scalable solution to measure structural complexity of shallow coral reefs, providing a balance of spatial resolution and extent that is more ecologically relevant than other large-scale, airborne approaches. The application of this method to four distinct patch reef habitats revealed unique structural signatures among seemingly similar reefs, highlighting the importance of site-specific considerations in assessing reef complexity. A scalable, high-resolution method such as we present will provide a more cost-effective and nuanced understanding of reef habitats, ultimately aiding in their conservation and management amid ongoing environmental challenges.http://www.sciencedirect.com/science/article/pii/S1574954124005004Coral reefsStructural complexityDronesDeep learningPredictive modelingConservation |
| spellingShingle | Aviv Suan Simone Franceschini Joushua Madin Elizabeth Madin Quantifying 3D coral reef structural complexity from 2D drone imagery using artificial intelligence Ecological Informatics Coral reefs Structural complexity Drones Deep learning Predictive modeling Conservation |
| title | Quantifying 3D coral reef structural complexity from 2D drone imagery using artificial intelligence |
| title_full | Quantifying 3D coral reef structural complexity from 2D drone imagery using artificial intelligence |
| title_fullStr | Quantifying 3D coral reef structural complexity from 2D drone imagery using artificial intelligence |
| title_full_unstemmed | Quantifying 3D coral reef structural complexity from 2D drone imagery using artificial intelligence |
| title_short | Quantifying 3D coral reef structural complexity from 2D drone imagery using artificial intelligence |
| title_sort | quantifying 3d coral reef structural complexity from 2d drone imagery using artificial intelligence |
| topic | Coral reefs Structural complexity Drones Deep learning Predictive modeling Conservation |
| url | http://www.sciencedirect.com/science/article/pii/S1574954124005004 |
| work_keys_str_mv | AT avivsuan quantifying3dcoralreefstructuralcomplexityfrom2ddroneimageryusingartificialintelligence AT simonefranceschini quantifying3dcoralreefstructuralcomplexityfrom2ddroneimageryusingartificialintelligence AT joushuamadin quantifying3dcoralreefstructuralcomplexityfrom2ddroneimageryusingartificialintelligence AT elizabethmadin quantifying3dcoralreefstructuralcomplexityfrom2ddroneimageryusingartificialintelligence |