Thermal cameras with high sensitivity to medium and long wavelengths can resolve features at the surface of flowing water arising from turbulent mixing.
Images acquired by these cameras can be processed with particle image velocimetry (PIV) to compute surface velocities based on the displacement of thermal features as they advect with the flow.
We conducted a series of field measurements to test this methodology for remote sensing of surface velocities in rivers.
We positioned an infrared video camera at multiple stations across bridges that spanned five rivers in Alaska.
Simultaneous non-contact measurements of surface velocity were collected with a radar gun.
In situ velocity profiles were collected with Acoustic Doppler Current Profilers (ADCP).
Infrared image time series were collected at a frequency of 10Hz for a one-minute duration at a number of stations spaced across each bridge.
Commercial PIV software used a cross-correlation algorithm to calculate pixel displacements between successive frames, which were then scaled to produce surface velocities.
A blanking distance below the ADCP prevents a direct measurement of the surface velocity.
However, we estimated surface velocity from the ADCP measurements using a program that normalizes each ADCP transect and combines those normalized transects to compute a mean measurement profile.
The program can fit a power law to the profile and in so doing provides a velocity index, the ratio between the depth-averaged and surface velocity.
For the rivers in this study, the velocity index ranged from 0.82 – 0.92. Average radar and extrapolated ADCP surface velocities were in good agreement with average infrared PIV calculations.