Streamflow Monitoring Using Computer Vision Machine Learning
Streamflow affects both the human uses and biological health of aquatic communities, and water managers need to understand the flow of local streams to make policy decisions. Flow regimes can change for many reasons, including water withdrawals for human use, extreme weather events, urbanization, changes in precipitation, and many other natural factors. Although flow is often measured in larger streams and rivers around the country, measuring hydrological data in small to midsize streams can be costly and require special expertise. The USGS has set up the Flow Pictures Explorer website to test the use of photographs as a way to monitor streamflow. This project will deploy game cameras to provide a much larger set of images and expand the website to include multiple smaller streams across many states and communities. This project will also explore whether images can be used to estimate streamflow using machine learning techniques. This approach will allow for a cost-effective and user-friendly method of continuous flow monitoring and increase water quality modeling capabilities.
Contacts: Emily Nering; Kelly Krock; Katharine Marko; Britta Bierwagen