Enhanced mangrove vegetation index based on hyperspectral images for mapping mangrove

As a specific forest community in tropical and subtropical coastal zones, mangrove has unique ecological functions and great social and economic value. Accurate mangrove mapping is important to the protection and restoration of mangrove ecosystem. Traditional classification methods rely on a large number of samples and complex classifiers, which are unsuitable for the large-scale extraction of mangroves because of low computational efficiency and poor generalization ability. This study proposes an Enhanced Mangrove Vegetation Index (EMVI) based on hyperspectral images. This index enhances the difference in greenness and canopy moisture content between mangroves and other vegetation using a green band and two shortwave-infrared bands in the form (Green-SWIR2)/(SWIR1-Green). Six typical mangrove areas (i.e., Qinglan Harbor in Hainan, Zhenzhu Harbor–Fangcheng Harbor in Guangxi, Lianzhou Bay in Guangxi, Zhangjiang Estuary in Fujian, Quanzhou Bay in Fujian, and Oujiang Estuary in Zhejiang) were selected as the study areas, and sample datasets were produced by field surveys and Google Earth high-resolution images. Compared with other VIs, such as the Normalized Difference Vegetation Index, Enhanced Vegetation Index, Moisture Stress Index, Mangrove Vegetation Index, and Combined Mangrove Recognition Index, EMVI exhibited better ability to distinguish mangroves and other vegetation. EMVI was applied to mangrove extraction in the six study areas based on ZY1-02D images, and the extraction results were compared with existing mangrove maps (GMW_2016 and CAS_Mangrove 2015) and the results of SVM. Results showed that EMVI featured the better overall accuracy and the Kappa coefficient than existing mangrove maps and the performance was similar to SVM. Further tests showed that EMVI was also suitable to other hyperspectral remote sensing images (i.e., GF-5, Hyperion, and PRISMA), but not to Sentinel-2 images. These results indicate that EMVI can be applied to different hyperspectral remote sensing images and different types of mangrove extraction. This index also has excellent application potential in mangrove mapping.