Leveraging USGS Hydrologic Imagery Visualization and Information System (HIVIS) for an operational monitoring of streamflow using computer vision

This project stems from the need to enhance streamflow forecasting and real-time monitoring of river conditions. Several USGS stations are equipped with cameras that are useful for visual inspection of flow conditions. The network of cameras operated by USGS can be useful beyond the mere visual inspection as their images and videos can be processed to derive information on water level, flow conditions, and even measure streamflow values. To this end, it is essential to develop algorithms capable of processing the images and videos generated at USGS stations and convert them into valuable information to support streamflow and flood inundation forecast. The technique that consists of processing and interpreting digital images and videos is commonly known as computer vision.

The goal of this project is to develop a computer vision-based system that processes the real-time feed from USGS cameras and generate information about flow conditions in various rivers across the US. The method consists of automating image analysis to detect changes, perform image segmentation, and quantify essential hydrologic and hydraulic variables like water level, water velocity, and discharge. Digital images and videos will be obtained from the USGS National Imagery Management System (NIMS). The Large-Scale Particle Image Velocimetry (LSPIV) is the primary method used to infer water velocity using cameras.

This research implements cutting-edge computer vision techniques to optically measure streamflow. This project will culminate in the development of three software packages, one for optical river hydraulic measurement, a second for flow condition characterization, and a third for NWM verification. A final system encompassing these packages will be delivered to USGS for testing and potential operational deployment. Several peer-reviewed publications will be produced.

This project will advance real-time river condition monitoring and eventually enhance streamflow forecasting accuracy. USGS observation records will be expanded with the introduction of new sources of quantitative data which should further support streamflow and flood inundation forecasts using the National Water Model (NWM) and contribute to achieve resilient and informed communities. This addresses USGS and NOAA’s missions and goals. Additionally, this project holds the potential to engage local communities in the monitoring of flood and streamflow conditions through the use of crowdsourced and citizen science-based data, especially in remote and ungauged regions.