Estimating excavation volume from shadow length in satellite imagery

A new cut shows up on your weekly pass over the site. The footprint is easy enough to trace. The depth is the harder question, and depth is what separates a drainage trench from something that needs a hardened roof later.
Shadow-length analysis is the standard workaround when you don't have a stereo pair or a DEM handy. It's cheap, it works off a single image, and it's been part of the imagery analyst's toolkit since the film era. It's also easy to get wrong if you skip a step.
The shadow-length method
The geometry is simple trigonometry. Sun elevation angle, which you pull from the image metadata (most VHR providers embed it in the product header alongside azimuth and off-nadir angle), gives you one side of a right triangle. The shadow length, measured on the ground, gives you the other.
For a vertical feature like a smokestack or a berm, height equals shadow length times the tangent of the sun elevation angle. For a pit or trench, you're running the same formula in reverse: the shadow falls from the lip of the excavation down onto the floor or the opposite wall, and its length tells you how far below the original grade that floor sits.
In practice:
- Pull sun elevation and azimuth from the scene metadata, not an assumption. A 5-degree error in elevation angle swings your depth estimate by more than you'd think at low sun angles.
- Digitize the shadow from the excavation lip to its tip in your GIS, then convert the pixel length to ground distance using the image's GSD. At 0.5 m GSD a trench 30 pixels deep gives you a shadow you can actually measure; at coarser resolution the shadow blurs into noise.
- Cross-check against a known object in the same frame, a parked vehicle or a shipping container of known height, to confirm the elevation angle you're using actually matches what the shadows in that scene are doing.
Multiply an average depth across the excavation footprint and you get a rough cut volume. That's a volumetric read good enough to flag "this got bigger" or "this is now deep enough to pour a basemat," not good enough to put in an engineering report.
Where the estimate falls apart
Shadow geometry assumes a flat reference plane and a shadow landing on it cleanly. Real excavations rarely cooperate.
Sloped pit walls throw the shadow at an angle that isn't the depth you think it is, and the discrepancy gets worse the shallower the sun angle. Spoil piled on the near rim shortens the visible shadow and makes a deep cut look shallow. An uneven floor, which is most floors mid-excavation, means the shadow's tip isn't landing on a single reference elevation at all. Low sun angle images (early morning, high latitude, winter passes) give you long shadows that are easier to measure but more prone to all three of the above. High sun angle images give you short, barely-there shadows that vanish into measurement error.
None of this makes the method useless. It makes it a screening tool: a way to turn "something changed" into "something changed by roughly this much," fast, from a single pass, without waiting on a stereo tasking window. For a confirmed program-level assessment you'd still want a DEM differencing pass once stereo coverage is available.
Doing this on a named site, every week
The shadow-length math doesn't change week to week, but doing it by hand on a new cut, a new berm, and a new structure footprint every time a fresh pass comes in is where the manual archive comparison starts eating your week. Facility Expansion runs that comparison against your named site automatically: weekly VHR optical passes, flagged construction and excavation changes, a volumetric read on the dig and a footprint read on anything new, delivered as an annotated site map instead of a stack of scenes you have to diff yourself.
If you're tracking a site for exactly this kind of change, it's worth seeing what a week of flagged activity actually looks like before the next pass comes in.