Coral Reef Mapping Using Remote Sensing Techniques and a Supervised Classification Algorithm

Abstract

The vitality of the Great Barrier Reef (GBR) is threatened by many human-made impacts. Monitoring this ecosystem makes it possible to study the general condition and the health of the GBR. However, due to the large extent of the GBR and limited accessibility in the ocean environment, mapping and monitoring this ecosystem has been always challenging task and connived. In this regard, Remote Sensing (RS) is an effective technique that provides valuable information for mapping and monitoring this ecosystem. In an attempt to monitor the GBR, this article applied a supervised machine learning algorithm to classify the Landsat 8 imagery collected over the GBR. To this end, the spectral responses of coral reefs, shallow water, deep ocean, rocks and sands, and green alga were initially determined from the satellite images. This information was then ingested to the Maximum Likelihood supervised classifier to map coral reefs in the GBR. Additionally, this study discusses how the GBR has been affected by anthropogenic disturbance. The results provide confirmatory evidence that RS techniques present great promise as a means of mapping coral reefs and monitoring their general conditions. We used the ambiguity matrix and validation data to estimate the accuracy of the proposed method. Overall, the proposed method was able to identify 5 different classes considered in this article with an average accuracy of 90%.