Hydroacoustic monitoring · Grand Rapids, Michigan
thalweg
n. — the deepest continuous line along a river channel
Channel cross-section
Deepest point at 30% of width
It is also a plain description of what our software goes looking for. Thalweg listens to a river with sound and reports what swam through it — in water too dark and too turbid for any camera, across a whole migration rather than a handful of survey nights.
Sonar solves the seeing problem. It listens instead of looking, so darkness and sediment stop mattering. What it does not solve is the reading: in a raw echo, a fish can look like a log, a bubble, or a patch of weed — and a log, a bubble and a patch of weed can all look like a fish. A season produces thousands of recordings, and almost none of what they contain is a fish.
Thalweg takes that first pass automatically and returns a short, ranked list a biologist can actually work through.
Sound, not sight
A fixed split-beam array watches one channel continuously — through night, flood and turbidity, for an entire spawning run.
Every target, in three dimensions
Each crossing is reconstructed as a path through the beam, with a distance, a direction and a duration — not a smudge on a screen.
Signal, separated from river
The overwhelming majority of what a river returns is not a fish. Expert time goes to the fraction that is.
Where this goes
Lake sturgeon on the Grand River is the problem we know best, and the one the instrument was built against. It is not the only place where somebody owes a regulator a number and has no continuous way to produce one — dam relicensing and fish passage, restoration projects that need a before and an after, agency and tribal programmes setting seasons against run timing.