Simulations

about

Our preliminary solution

This provides a better monitoring of ocean microplastics and will support future research on microplastics greatly. We’ll provide a login platform, such that oceanographers can record their progress. We would need assistance in fluid modeling, microplastic classification, and a laboratory for microplastic testing

Microscopically

Microscopically, we propose a Hyperspectral Mask R-CNN to detect and classify microplastics within a water sample, since the process of detecting microplastics chemically is extremely tedious.

Macroscopically

Macroscopically, we would model and predict the future movement of microplastics in the seas with a neural network with data from NASA’s CYGNSS. Traditional methods like direct simulations have to choose between run-time performance or prediction accuracy, not to mention the lack of generalization since these models can only use past data. Using neural networks, the predictions can be run much computationally efficiently and extrapolated into the future.

By customizing the neural network with tensors, the network will treat movement after rotation and flipping in the same way. Simulation data generated with PARCELS will be used to train the neural network. To enhance the details of the flow field and turbulence, we can employ current image super-resolution techniques.