Mapping Mangrove Growth and Deforestation using Satellite Imagery

For my fifth and final Metis project, I trained a convolutional neural network classifier on satellite imagery in order to map the extent of mangroves in the south of Florida. I used satellite image data that I acquired through Google Earth Engine, including both images of the Florida coastline as well as a map of mangroves thoughout the world. I trained the model on data from the year 2000, corresponding to the map of mangroves, and then applied the model to year 2020 data in order to track changes in the last 20 years. Please see the blog post I wrote on the subject as well as the project GitHub page for more extensive details.