Evaluation of Multispectral Image for Mangrove Health Assessment Using Sentinel 2A and Field Spectrometer Data

Abstract

Mangrove health is one of the parameters to show the quality of mangrove forest ecosystems that can be compared with other locations. To preserve mangroves as an ecosystem with high services, it is necessary to map the health of mangroves. Ketapang Subdistrict, South Lampung Regency is one of the areas where the conversion of land into a pond area has a significant impact on mangrove damage. This study aims to identify mangrove health based on the vegetation index from Sentinel 2A imagery and spectral measurements in the field. The research methods used are supervised classification with support vector machine algorithm, GNDVI, SAVI and TSAVI, and direct measurement of object reflection. the result of the GNDVI, SAVI, and TSAVI correlation coefficient $(\mathrm{R}^{2})$ values are 0.71, 0.6, and 0.66 respectively. If the value of the correlation coefficient is in the range of 0.50-0.70, it can be said that the relationship is very strong between parameters. Mangrove health classification is classified into 3 classes: poor, moderate and healthy. The mangrove health area obtained ranged from poor class 17.0 ha with an area percentage of 7.13%, medium class 23.78 ha with a percentage of 9.98% and healthy class 197.63 ha with a percentage of 82.89%. The total area of mangroves is about 238.42 ha. Based on 30 sample points of field observation, the results of the accuracy test show an overall accuracy of 86.67%.