Coral Reef Change Detection in Remote Pacific Islands Using Support Vector Machine Classifiers

A significant decline in coral-dominated benthic cover-type (CDBCTC) has been detected across six remote Pacific islands, with reductions ranging from 25% to 43%. This finding was achieved using a robust machine learning classifier based on support vector machines (SVM), trained and validated with ground truth observations and Landsat imagery. The study introduces a methodology for spatial generalization of coral cover classifiers, demonstrating high accuracy in cross-validation (75.3%) and when tested against new data from neighboring reefs. The research highlights the effectiveness of SVM-based change detection in assessing temporal changes in coral reef ecosystems, providing critical insights for global coral reef monitoring.

Gapper, J. J., El-Askary, H., Linstead, E., & Piechota, T. (2019). Coral Reef Change Detection in Remote Pacific Islands Using Support Vector Machine Classifiers. Remote Sensing, 11(13), 1525. https://doi.org/10.3390/rs11131525