Student stipends for this work were funded by the Pacific Intelligence Innovation Initiative (P3I)
This work took place during June 2024 in the SPICE Summer Data Science Institute. With inspiration from NASA-IMPACT we became very interested in mapping plastic debris in the ocean of the Windward side of Oahu, Hawaii. This interest lead us to the paper Automatic Detection and Identification of Floating Marine Debris Using Multispectral Satellite Imagery and it's github repo. This amazing work contains data to train a model to detect plastic using Sentinel2. We trained a decision tree model (opposed to xgboost utilized in the paper) and achieved a similiar accuracy score of __%.
From there, we downloaded Sentinel 2 data for a bounding box off the Windward Coast of Oahu from the Copernicus Open Access Hub. We downloaded data for Bands 1-8, 11 & 12. This data was processed utilizing the Texas Advanced Computing Center (TACC); the workflow can be found in Sentinel-Query1.ipynb.
Lastly we utilized our trained model to make predictions on the processed data. The predictions were mapped utilizing the geemap Python package