Mapping of mangrove species using integration of field spectrometers and WorldView-2 imagery in Karimunjawa and Kemujan Islands

Abstract

Mapping of mangroves at the species level is very important for inventorying the biodiversity of mangrove forests and supporting the management of coastal ecosystems. Remote sensing data provides efficient in terms of time, cost, effort, and resources for mangrove mapping. This study aims to (1) map mangrove species using WorldView-2 satellite imagery and object spectral reflection based on the Spectral Information Divergence (SID) pixel-based classification algorithm, and (2) test the accuracy of mapping mangrove species on Karimunjawa and Kemujan islands, Central Java. The research location was chosen because of the natural condition of the mangrove ecosystem and high species diversity. Field data taken include (1) identification of mangrove species, (2) coordinates of the position of pure mangroves, and (3) spectral reflection of mangrove species from field spectrometer measurements. Dendrogram analysis using the Ward Linkage method was carried out to classify species based on the cluster distance of the closest spectral reflection pattern. The confusion matrix accuracy test method is used to get the overall accuracy value from the mapping. The results showed that 24 species could be grouped into 4 levels based on dendrogram analysis, 1st level (2 groups), 2nd level (4 groups), 3rd level (5 groups) and level 4th (single species). The visualization of the SID classification results shows the grouping of several species groups at each dendrogram level. The best accuracy was obtained by mapping 1st level, 2nd level and 3rd level with OA values of 49.7%, 22.6% and 15.3% respectively, while single species mapping had 0% OA.