COMPARISON OF DIFFERENT VEGETATION INDICES FOR ASSESSING MANGROVE DENSITY USING SENTINEL-2 IMAGERY

Abstract

Vegetation mapping provides important information for understanding ecological condition
through calculation of vegetation density. It based on vegetation indices developed through algorithms of a
mathematical model within the visible and near-infrared reflectance bands. The index is an estimate of either
leaf density per species or vegetation types, respectively. This study aimed to evaluate those indices and find
the best algorithm using Sentinel-2 satellite image. Twenty four algorithms of vegetation indices were analyzed
for mangrove density mapping, i.e., BR, GNDVI BR, GR, SAVI, MSAVI, NDRE, NDVI, NDVI2, NDWI,
NNIP, PSRI, RR, RVI, VIRE, SVI, VIRRE, MTV1, MTVI2, RDVI, VARI, VI green, MSR, and TVI. During
pre-processing stage, a digital number of a Sentinel-2 image was converted into radiance and reflectance value.
The analysis resulted in three algorithms that provide the highest accuracy, i.e., NDVI (normalized difference
vegetation indices) with exponential regression approach, RVI (Ratio Vegetation indices) with the exponential
approach and NDVI (normalized difference vegetation indices) with a polynomial approach. The mangrove
biomass spatial modeling NDVI with exponential regression approach (RMSE = 89 kg) showed the average of
each pixel (10x10m) was 0.97 ton / 100 m2. Total mangrove biomass for above ground and underground
vegetation in the study area was 22,365.6 tons. Sentinel-2 satellite image may best use one of those three
algorithms, especially applied for mangrove vegetation.