How to measure freeboard from satellite imagery

Freeboard is the vertical gap between the crest of your embankment and the surface of the pond behind it. It's one number, but it's the number an inspector, an insurer, or a dam safety engineer looks at first, because it tells you how much room is left before an overtopping event stops being theoretical. Measuring it by hand means a survey crew on the crest with GPS or total station gear, shooting the crest line and the pond edge, then doing the subtraction. That's accurate, and it's also expensive enough that most facilities only do it on a quarterly or annual cycle.

Freeboard estimation from remote sensing is the same subtraction, done from above instead of from a tripod on the crest.

The elevation problem, not the image problem

Getting two elevations, the crest edge and the pond surface, tied to the same vertical datum on the same day, is the hard part of freeboard estimation. A single optical image gives you a flat picture with no elevation information built in. To get the vertical, you need either stereo pairs processed into a digital surface model, or a sensor that already outputs elevation (lidar, interferometric radar), or a tie-in to a ground survey that anchors the imagery to known elevations.

Once you have a surface model, the workflow is roughly: trace the crest edge as a line feature, pull the elevation along that line, then trace the waterline of the pond (which is really just the edge of standing water in the optical or near-infrared band) and pull the elevation there too. Subtract. What you get is a freeboard estimate at however many points along the crest you sampled, which is actually an improvement over a single spot shot in the field, because embankments don't settle evenly and a crest that's fine at the inspection point can be six inches lower fifty metres down the alignment.

Pond elevation is the easier half. Water is flat, so one good elevation reading anywhere on the pond surface tells you the elevation everywhere on it, assuming no wind setup or short-term surge, which on a tailings pond is usually a safe assumption. Crest elevation is the part that takes resolution: at coarse ground sample distance the crest edge blurs into the downstream slope and you lose the line you're supposed to be tracing. Freeboard work needs imagery closer to the very-high-resolution end. The 10-30 m pixel satellite data used for regional flood mapping simply won't resolve that line.

What a baseline buys you

None of this works as an absolute measurement without a baseline. The surface model gives you relative elevation, crest versus pond, good to the vertical precision of the sensor and the quality of the stereo match. To turn that into an actual freeboard number you trust, you need at least one baseline tie to your own as-built crest elevation or your most recent certified survey. After that, every new pass gets compared against the baseline rather than reconstructed from scratch, which is also how you start catching movement: a crest segment that reads a few centimetres lower than last month's baseline before it reads inches lower is the kind of thing a quarterly walkdown schedule just doesn't catch in time.

That's the actual case for doing this on a weekly cadence instead of a quarterly one. A slow-building problem, rising phreatic surface, a beach that's narrowed, a crest segment settling under its own saturation, has weeks to develop between walkdowns with nobody watching. A weekly log against your own baseline closes that gap without putting a drone or a survey crew on site every week.

Where the method runs out

Vegetation on the crest or downstream face will throw off a photogrammetric surface model, because the sensor is measuring the top of the canopy, not the ground under it. Cloud cover takes out an optical pass entirely, which is one reason a thermal channel earns its keep, it also reads through haze and picks up seepage staining and warm discharge points that visible imagery misses outright. And no remote method replaces the engineer of record's judgment on what a given freeboard reading means for your specific spillway design and storm case. What it does is make sure that judgment gets fed fresh numbers every week instead of once a quarter.

If you're running that comparison by hand right now, it's worth seeing what a standing weekly read looks like before the next walkdown gap opens up.