Mapping Mangroves: New Integrated Spectral Protocol for Rapid Mangrove Identification Using Machine Learnin

Mangrove forests are a critical component of tropical and sub-tropical coastal habitats that provide a variety of benefits including coastal protection, water purification, species habitats for nursing and carbon sequestration. Despite their unique ecological role, staggering degradation of mangroves has occurred worldwide since the 1950's driven by rapid urbanization and popular land use practices such as aquaculture, charcoal/timber harvest, and agricultural production. Deforestation has been most pronounced in Southeast Asian countries, a hotspot for the world's mangrove stocks. In response to this biological concern, researchers have utilized remote sensing and GIS technologies to monitor changes in mangrove forests worldwide. Although many projects have been successful in detecting mangrove forests for specific case studies, there are growing opportunities to use GIS for rapid classification of mangroves at a global scale. To contribute to our understanding of global mangrove classification, this study proposes a new protocol for spectral mangrove identification. Employing machine learning classifiers (Support Vector Machine, Random Forest, Multi-Layer Perceptron) with Landsat 8 OLI spectral and ancillary data, this study will compare the effectiveness of the new Integrated Spectral Mangrove Protocol (Tasseled Cap Bands, DEM, Band 7, Distance to Coast, Filtered Bands) to other traditional approaches for mangrove identification using the Irrawaddy Delta (Myanmar) as a case study. This research helps to contribute foundational methods for regional -- scale automated mangrove monitoring efforts.