Large-Extent Mangrove Species Mapping Using Landsat 9 OLI-2: A Subpixel Analysis

Mangroves have the capabilities to both mitigate and adapt to climate change impact. However, it varies between species. Therefore, it is substantial to upscale mangrove explorations and studies to the species level. This study aims to perform a spectral-library-based linear spectral unmixing (LSU) analysis technique on Landsat 9 OLI-2 imagery as an alternative to the conventional mangrove species mapping methods. We used the center wavelength of Landsat 9 OLI-2's B2, B3, B4, and B5 bands to define the spectra of Sonneratia alba, Rhizophora apiculata, and Avicennia marina. We performed the LSU analysis on the Muaragembong mangrove forest area, Bekasi, West Java, Indonesia as the area of interest. The result showed that the mangrove species has a unique spectral signature. The reflectance is slightly higher at around 500–600 nm and lower at 750–770 nm than the typical vegetation spectral signature. Most of Muaragembong is covered with R. apiculata and A. marina. However, there is a distinctive spatial distribution pattern for each species. Based on the RMSE result, the model can produce a ±0.3% error in each pixel. Empirical evidence from the ground truthing helped to validate the distribution pattern. It is associated with environmental factors, such as supporting substrate and water access. This paper concludes that it is possible to perform the LSU analysis using multispectral satellite data for a large-extent mangrove species mapping. However, it is mandatory to validate the result on a ground-truthing process.