If you ask AI what the key characteristics of a good flow gauging site are, it will insist on “Straight channel reaches with smooth, regular cross-sections and steady velocities” and probably suggest a spot where “Bed and banks made of stable material that will not erode, scour, or shift during floods”.

As if we needed anymore proof that AI has never spent a day in the field…

And so, I find myself standing in a small wiry mountain creek that’s in a hurry to get down the hill, with no regard for an AI hydrologist’s rule book.. It meanders like a drunk on a bicycle. Not a straight section in sight. The fall is considerable; the bed and banks are classic Otago silts and broken schist rock.

A few weeks ago, I was up here scoping out a spot to install a flume. The flume structure becomes the “Straight channel reach with smooth, regular cross-section” and includes a ramp or “choke” that is a nifty way to calculate flow based on the level of water.

But it still needs installing sensitively. The structure itself wants to be level in both planes and ideally wants an approach channel that facilitates a calm, dare I say… laminar flow into the flume.

The spot I have found is just about long enough to accommodate the flume and allow for the approach channel. We start massaging the banks into shape and rake the bed as flat as possible. Thank goodness, there are no boulders in the way on this occasion. I wonder how many builders have used a level underwater?

It takes a few goes to get the flume sitting level. We use AstroTurf under the flume. I have found from experience that it helps to bind the silts and stop water flow eroding the bed from under it. And it stops the outfall from scouring out.

Backfilling the structure is best done with rootbound sediments. They hold together well and stop water from pulling sediments through.

All told, it’s a tidy install that didn’t fight us too much. And from a few meters away looks quite innocuous. The Stilling well (the white plumbing pipe) will house a pressure datalogger. This pressure data can be converted to flow using a bit of funky maths.

Job Done!

Dirk Van Walt