Streamflow Monitoring Using Computer Vision Machine Learning (ORD w R5 and other regions, states and tribes)
This project will develop a lower-cost method to quantify stream flow that can be used by states, tribes and other organizations. This approach can supplement the current methods, i.e., deployment of hydrological measurement equipment (stream gauges), that is costly and requires specialized expertise. This user-friendly alternative relies on continuous photo imagery and machine learning to estimate streamflow. Water volume is a critical variable in maintaining both the health of aquatic and riparian biological communities and for human uses of water. However, in many parts of the country, flow regimes are changing due to anthropogenic changes and natural impacts. Today, the USGS maintains over 7,000 continuous water monitoring gages, but most are in large rivers and reduced funding has led to removal of gages. Therefore, many states, tribes, and other groups want to collect more hydrologic data, especially in small to mid-size streams. In this proof of concept project, the website 'Flow Pictures Explorer' is being expanded as a publicly available repository for photos and hydrographs. State and tribal partners are supplying thousands of continuous photos from game cameras together with hydrological data for multiple stream sites. By using continuous imagery of streams with novel machine learning, states, tribes, nonprofits/watershed associations, and other groups will gain a better understanding of altered hydrology.
Project URL: http://fpe.ecosheds.org/
Geographic Scope: Round 1 data submitted from the following locations: Browns Brook, Raymond, NH; Parkers Brook, Oakham, MA; Peavine Creek, Pokagon Township, MI; Additional locations to be added from EPA Regions 1, 2, 3, and 5 before completion of project.
Project Status: Active - not recruiting volunteers
Participation Tasks: Data analysis, Data entry, Geolocation, Observation, Photography,
Start Date: 12/01/2019
Project Contact: Bierwagen.Britta@epa.gov
Federal Government Sponsor:
Other Federal Government Sponsor: U.S. Department of the Interior (DOI)
Fields of Science: Computers and technology, Ecology and environment, Geology and earth science, Ocean/water and marine
Intended Outcomes: The goal of this project is to determine whether these streamflow images and hydrographs can be used together to understand how much the stream, including flow regime, changes over time. In the subsequent phase, computer vision machine learning will be tested as an approach to estimate flow from images, using the large number of images generated from partner organizations. The resulting tools may be applicable to protection of aquatic life from the impacts of hydrological alterations.