Mangrove forest identification using object-based approach classification

Abstract

As an essential indicator of coastal ecosystem, mangrove forest has unique characteristics making it different from terrestrial vegetation. Mangrove forest identification is essential to support the inventory and monitoring of mangrove diversity. This study was aimed to identify (1) the characteristics of mangrove and non-mangrove objects based on image classification, (2) mangrove forest mapping from remote sensing imagery, and (3) accuracy assessment. This study applied a Geographic Object-Based Image Analysis (GEOBIA) approach for high-resolution imagery of WorldView-2 (2 m) at Clungup Mangrove Conservation (CMC), Malang Regency, East Java, Indonesia. Rule-set for mangrove object detection was built from the segmentation of the algorithm and image classification. A multiresolution segmentation algorithm in WorldView-2 was applied to make a segmented object. All multispectral bands of WorldView-2 and some vegetation indexes were used as input variables for the segmentation and classification object. In the classification process, a threshold for a particular variable representing a significant object difference was used. This process resulted in two classes which are identified as mangrove and non-mangrove. The results showed that the use of the GEOBIA method for high-resolution imagery has the potential to identify and plot a mangrove forest with high accuracy of up to 90%. This study contributes to the development of an object-based approach using remote sensing imagery and is very potential to be applied in a more detailed mapping.