Incorporating CART algorithm and i for mapping Mangrove using Landsat 8 imagery

The purpose of this study is to introduce a method for improving the accuracy of mapping mangrove covers by integrating CART algorithm and vegetation indices using Landsat 8 imagery. In this study, 7 vegetation indices were calculated including DVI, NDVI, NDII, IPVI, MNDWI, SAVI, and OSAVI. The important indicators and thresholds were identified by the CART algorithm in R-software and then the Mangrove cover besides surrounding areas was mapped using the decision tree technique. The results of accuracy assessment based on comparing control points which recorded by GPS with SPOT 6/7 images showed that the resulted map had a general accuracy of 80.97% and a kappa of 0.74. Using moderate resolution satellite imageries for mapping mangrove is difficult due to background reflectance effect (e.g. water and soil) and mixed pixels considering environmental conditions. Results demonstrated that this approach could produce information about mangroves for decision-makers in order to conservation and management planning.