Integrating bathymetry data and high-resolution satellite imagery for benthic habitat mapping in a marine-protected area in Brazil

• Published in: International journal of remote sensing Vol. 46; no. 3; pp. 1086 - 1116
• Main Authors: Andrade Galvão de Medeiros, Thais, Botha, Elizabeth J., Cardoso Duarte, Fernando, Macedo De Mello Baptista, Gustavo, Sabino de Brito Pereira, Beatriz, Kampel, Milton, Seoane, José Carlos Sicolí
• Format: Journal Article
• Language: English
• Published: London Taylor & Francis 01.02.2025; Taylor & Francis Ltd
• Subjects: Aquatic habitats; Bathymeters; Bathymetry; benthic ecosystems; Benthic environment; Benthos; Brazil; Climate change; Coral reef ecosystems; Coral reef habitats; Coral reefs; corals; Disasters; Environmental management; Glint; Global warming; Habitats; High resolution; humans; hydroacoustics; Image resolution; Machine learning; Mapping; Marine invertebrates; Marine parks; Marine protected areas; Marine technology; Natural disasters; Protected areas; Reflectance; Remote sensing; river plume; River plumes; Satellite imagery; satellites; Spatial resolution; Support vector machines; Brazil
• ISSN: 0143-1161; 1366-5901; 1366-5901
• DOI: 10.1080/01431161.2024.2424510

Global warming, natural disasters, and increasing human pressure are threatening coral reefs worldwide. It is vital to document their distribution and spatial extent in order to quantify the threats they are subject to. Technological advances in satellite remote sensing have been instrumental in mapping the benthic habitats of coral reefs. In this study, we derived a benthic coral reef habitat map of the Coroa Alta Municipal Marine Park (Brazil) from a very high spatial resolution WorldView-2 (WV2) image, hydroacoustic bathymetry data and in situ substrate data using two machine learning classifiers: Random Forest (RF) and Support Vector Machine (SVM). An attempt was made to identify the most effective input data for benthic habitat mapping. The best overall classification accuracy for both RF and SVM was achieved using the sunglint-corrected reflectance bands and in situ measured bathymetry as input (97.29% and 95.19% respectively). However, spectral similarity between some benthic classes led to misclassification in areas affected by the São João de Tiba river plume and in highly glint-affected areas within the image. The results highlight the need to test different data sources and correction methods to obtain more accurate representations of the benthic habitats of an entire reef system. We expect this study to be useful for the conservation, protection, and environmental management of coral reef ecosystems.