Identification of mangrove forest via remote sensing combined with tidal level and DEM
To address the accurate extraction of sparse and low mangroves perplexed by the periodic change of tide level,we take the Beibu Gulf of Guangxi as the research area to construct a decision tree model for mangrove identi-fication using Landsat8 OLI images at low and high tidal levels and DEM(Digital Elevation Model)data,which is then evaluated by comparing with SVM(Support Vector Machine).The research results show that difference exists in the spectra of mangroves with different heights and canopy densities or under different tide levels,while the sparse and low mangroves share the same spectrum with shady slope forest and water-terrestrial vegetation mixed pixel.The SVM approach classifies the mangroves as high-dense type and low-sparse type,and improves the overall accuracy by 4.65,4.41 and 7.22 percentage points for low-tide,high-tide and multi-tide images,respectively.The proposed approach reaches 98.80%of overall accuracy and 0.973 of Kappa coefficient,which are 1.62 percentage points and 0.035 higher than the best values of SVM approach.It can be concluded that considering the mangrove height,density,tide level and DEM can significantly improve the identification accuracy of mangroves from remote sensing images.
mangrove forestLandsat 8 OLItidal leveldigital elevation model(DEM)decision tree
万方数据
维普
