Abstract:
This study aims to develop a hybrid method Normalized Difference Vegetation Index (NDVI) and Normalized
Difference Water Index (NDWI) for the automatic delineation of the mangrove forests from Landsat 8 DCM imagery. This
method was developed because the visual interpretation is a method that is not very efficient in extracting spatial
information of mangrove forests. This hybrid method was tested on Landsat 8 imagery path/row 117/062, where many
mangrove forests in the region this coverage. The method used to separate the object of forest and non-forest on the
NDVI, including for separating water and non-water at NDWI is Otsu thresholding. While the ability to see the accuracy
of this hybrid method, mangrove forest extraction results automatically tested with mangrove forests visual interpretation
of the image of Sentinel-2 on a specific area designated as a sample. Field verification was conducted to ensure the
object of mangrove forests. Furthermore, the accuracy test conducted by Confusion Matrix. The results of this study
indicate that hybrid method NDVI and NDWI for delineation of mangrove forests is quite accurate. This is evident from
the overall accuracy of 95.5%, 8.34% omission error, and 0.76% commission error. However, the hybrid method is also
experiencing an error detection on the object cloud shadow, the shadow of the hill, and non-mangrove swamp forests
that have similar spectral characteristics of mangrove forests
