Image-based measurements made by multiple operators to test and refine a remote sensing approach for estimating river discharge via critical flow theory

This data release makes available the data used to test and refine a remote sensing approach for estimating river discharge via critical flow theory. In rivers where flow conditions are near critical and well-defined standing waves are present, discharge can be inferred from measurements of wavelength and channel width made on an image. The goals of the study were to: 1) evaluate the sensitivity of discharge estimates to inter-operator measurement variability; 2) investigate scale-dependent biases in manual wavelength and width measurements; 3) compare manual and semi-automated approaches to wavelength measurement; and 4) to provide guidance for effective application of the critical flow approach.
 

This study used the same set of sites near established USGS gaging stations with standing waves (also known as undular hydraulic jumps, or UHJs) as the initial paper describing the critical flow-based approach to remote sensing of river discharge (Legleiter et al., 2025). The Inferring Q (discharge) from Undular Hydraulic Jumps (UHJs) Identified in River Images (InQUIRI) software was used to obtain imagery available in Google Earth from each of these locations and dates; these images are compiled in RiverImages.zip.
 

Each of six independent operators used the images to make the following measurements within InQUIRI: 1) one wavelength measurement between the first and second crests in a standing wave train; 2) a second wavelength measurement between the second and third crests in the same wave train; and 3) channel width, measured perpendicular to the two waves at the midpoint between them. In most cases, the operators also digitized a profile along the wave train to test various semi-automated methods for inferring a dominant wavelength. InQUIRI was used to calculate discharge estimates based on the mean of the two measured wavelengths (or inferred from the digitized profile) and the channel width. The image acquisition times inferred using the sundial method (Bae et al., 2025) and gage discharges recorded at those times were used to assess the accuracy of the estimated discharges.
 

A custom MATLAB script was used to ingest the *.mat InQUIRI session files from each operator for each site and date and extract the wavelength and width measurements and digitized profiles along the wave train when available. These data were compiled to create the Measurements.csv file.
 

A single individual visually inspected each of the 82 images and assigned a rapid type and wave type. The rapid categories include: 1) upstream tongue; 2) central core; and 3) downstream tail. The wave categories include: A) standing wave trains without lateral shock waves; B) more complex wave structure with intersecting lateral shock waves; and C) breaking waves exhibiting air entrainment and rollers. This information was compiled in the file ImageMetadata.csv.
 

For each combination of operator, site, and date, the measurements and profiles were overlain on the image and used to produce a screenshot. The MeasurementScreenshots.zip folder contains a separate subfolder with screenshots from each of the six operators and another subfolder that contains screenshots with the measurements from all operators overlain on the image, with the measurements color-coded by the level of agreement between the discharge estimated based on those measurements and the discharge recorded at a nearby gaging station on the date the image was acquired.
 

Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.