Methane Plume Segmentation

Problem: While carbon dioxide is the most prevalent greenhouse gas, methane is much more potent. And methane leaks can be difficult to detect. This project used multispectral satellite imagery to detect and segment methane plumes. The primary challenge in model training was the scarcity of true methane plumes with which to train a model.

Solution: This project involved reviewing data and modeling choices on the project and subsequent steps for refining the training data and model approach in order to advance the project as quickly as possible.