Remote sensing for coral reef and seagrass cover mapping to support coastal management of small islands

Abstract

Coral reefs and seagrass are critical coastal resources due to their role in the ecosystem benefits for the coastal environment in terms of biodiversity, coastal protection, fisheries, and tourism. It is therefore important to preserve and protect these species. Coral and seagrass percent cover mapping is a simple approach to assess coral and seagrass condition. The application of remote sensing of coral and seagrass percent cover mapping is very challenging with respect to performance and accuracy. This research aims to utilize remote sensing data for coral and seagrass percent cover mapping. Linear and machine learning regressions (RF and SVM) were used to develop a coral and seagrass percent cover model from a Sentinel-2 MSI images. The Sentinel-2 MSI images were transformed into deglint, water column (DII), principal component, and mean texture analysis as input bands for the model. The results showed that coral percent cover mapping accuracy is relatively low (RMSE = ±17%) due to various problems, limitations, and an inaccurate model, whereas the results of the seagrass percent cover map had higher accuracy, with RMSE ±11%. The results obtained indicate that the seagrass percent cover map is suitable for use as basic information to support coastal management. However, the coral percent cover map is not an optimal information source due to its low accuracy.