Project goals

The central focus of project fluvius aims to understand the response of river basins after forest disturbance and restoration. Other goals include the advancement of remote sensing, artificial intelligence, and ecosystem monitoring.

Progress / successes:

• Developed an automated pipeline for core datasets, including preprocessing and storage on a cloud based platform

• Trained and evaluated various deep learning models to predict total suspended sediment (TSS) using water station data and satellite imagery

• Identified the top performing architectures using a hyperparameter grid search

• Used observations from sites withheld from training to evaluate the out-of-sample predictive performance of the top model, and compared our performance to existing benchmarks

Future work:

• Deploy the deep learning model on the entire Itacaiúnas River watershed over space and time