Applying a New Integrated Classification Method to Monitor Shifting Mangrove Wetlands

Abstract

Remotely sensed technique is a potential method for mangrove management. Free multi-temporal Landsat imagery data provided a feasible approach to monitor shifting mangrove wetlands in China. In order to know the historic and present situation of the west shore of Donghai Island mangrove, a new integrated classification method with Support Vector Machines (SVM) and Decision Tree (DT) was applied, which is more effective than SVM algorithm. The integrated method combined the information of spectrum, spectral transformation and geographical position for land cover extraction. Post-classification change detection was used for analyzing the mangrove loss conditions and its possible reasons in the past three decades. The results informed that the mangroves in the study area has been most affected since the early 1970s, nearly 62% of the mangrove area has lost. The annual loss rate exceeded 3.18%, higher than world total loss rate as 2.07%.