Assessing the natural recovery of mangroves after human disturbance using neural network classification and Sentinel-2 imagery
llegal cultivation of shrimp ponds and rice paddies was carried out in the Wumbaik Reserved Mangrove Forest in Rakhine State, Myanmar; many of them have been abandoned under administrative guidance. Although mangrove reforestation has often been carried out around the world, it is costly and tends to be biased toward specific tree species, resulting in a state that differs from the natural state.
This study focuses on how abandoned sites can be regenerated in a natural state without human intervention.
AS abandoned areas are generally small in size and low-cost method is required especially in developing countries, we adopted Senitnel-2 imagery, which has relatively high resolution and can be used free of charge, and developed an automatic extraction method of mangrove forests using deep neural networks. In order to finely distinguish mangrove forests from other vegetation, we added tree canopy height information estimated from freely available global DEMs and DSMs, which improved the extraction accuracy of mangrove areas and proved to be sufficiently practical.
Maung, W. S. and Sasaki, J.: Assessing the natural recovery of mangroves after human disturbance using neural network classification and Sentinel-2 imagery. Remote Sensing, 13(1), 52, 2021. DOI