Klasifikasi Habitat Bentik Berdasarkan Citra Sentinel-2 di Kepulauan Kei, Maluku Tenggara

Abstract

Imagery classification has long been used in analyzing remote sensing data. The use of the classification algorithm model can affect the results in interpreting benthic habitats in shallow water. This study aimed to determine the best classification algorithm model for mapping benthic habitat cover through Sentinel-2 satellite imagery. Three algorithm models were employed: Maximum Likelihood Classification (MLC), Minimum Distance Classification (MDC), and Mahalanobis Distance Classification (MaDC). The benthic habitat types were extracted using Lyzenga correction, giving three categories: coral, seagrass, and sand. The results showed that the application algorithm models of the MLC, MDC, and MaDC on the benthic habitat mapping resulted in an accuracy value that was not significantly different at the 95% confidence interval. However, of the three algorithms used, the MaDC algorithm provides the best results in overall accuracy (78.35%) than the MDC (74.45%) and the MLC (74.33%). It shows that the MaDC algorithm can be referred to as the mapped benthic habitat cover in the Kei Islands. However, this algorithm model needs to be continuously studied and compared to other models in other locations.
 
 Keywords: benthic, habitat classification, Kei Islands, sentinel

Benthic Habitat Mapping Using Sentinel-2A Imagery in the Waters of Pengudang Village, Bintan Regency

“…Mukrimin et al, (2021), also conducted research on object-based benthic habitat mapping using Sentinel-2A imagery with the application of SVM algorithm which resulted in an accuracy of 88%. In addition, other studies were also conducted using Sentinel-2A images with the application of MLC, MDC and MaDC algorithms which resulted in accuracy rates of 74.33%, 73.45%, and 78.35% respectively (La Ode Alifatri, 2022). The difference in the accuracy of benthic habitat mapping is caused by several factors such as differences in classification methods, the number of field observation points, the number of benthic habitat classes and the imagery used (Prabowo, 2018).…”