Mangrove Classification Using an Integration of Radar and Optical Images of Sentinel 1 and 2: A Case Study of Can Gio, Ho Chi Minh City

Mangrove forests are of great importance to coastal communities, providing not only a source of food and resources but also protecting coastlines, preventing erosion and regulating our climate. However, mangroves have been changed significantly due to deforestation and restoration during the last decades. At present, remote sensing has been widely proven to be essential in monitoring and mapping mangrove forest. In this study, we exploited an integration of radar and optical images of Sentinel 1 and 2 to classify mangroves in the Can Gio district, Ho Chi Minh City. Sentinel-1 images were collected in February 2021 while Sentinel-2 images were in January 2021. Based on an analysis of training samples, a decision tree diagram was designed to classify the Can Gio mangroves with four major plants, consisting of nypa palm, rhizophoraceae, avicennia and ceriops tagal. The result presented a spatial distribution of the mangrove types inside the Can Gio mangrove forest and their sparse appearance in combination with other vegetation in the Can Gio district. The classifier for the Can Gio mangroves obtain an overall accuracy of 80% and Cohen’s Kappa coefficient of 0.75. The decision tree model on the integration of radar and optical images of Sentinel 1 and 2 was expected to contribute a large inventor of classification algorithms and to be effectively applied for mangrove classification.