Shadow-derived vs stereo-derived building height: which to trust for a desk survey

When a candidate rooftop site lands on your desk, you need a storey count before you commit to a scouting trip. Two measurement approaches turn up in vendor data sheets for this: shadow-derived height and stereo-derived height. They come from different math and they fail in different ways. Knowing which one to trust for a given building matters more than knowing the method name.

How shadow method building height actually works

Shadow-derived height uses a single image. The sensor knows the sun angle and azimuth at capture time, measures the shadow a building casts on the ground, and runs the trigonometry back to a height. It's cheap to compute and works with ordinary single-pass imagery, no stereo pair required.

The catch is that it only works when the shadow is clean. A building wedged between taller neighbors, a shadow falling across a parking structure or a sloped lot, or a capture taken near solar noon when shadows run short, and the number degrades fast. Dense urban blocks, the kind with adjacent rooftops and mechanical penthouses, are exactly where this method struggles most, which is unfortunate, because dense blocks are also where a lot of candidate rooftop sites sit.

How stereo method building height actually works

Stereo-derived height comes from two overlapping images taken from slightly different angles, the same principle your own depth perception runs on. Match the same rooftop edge in both frames, measure the offset, and you get an elevation model rather than an inferred one. No sun angle dependency, no shadow geometry to go wrong.

Stereo pairs cost more to acquire and process, and they need genuine overlap coverage rather than two images of the same area taken months apart. When that coverage exists, the height comes from geometry that holds regardless of time of day or season.

Which one to trust for a desk survey

For a telecom desk survey, where the deliverable is a storey count feeding into an antenna clearance estimate or a starting point for a lease negotiation, stereo-derived height is the one to lean on when it's available for the area. It doesn't degrade in dense blocks the way the shadow method does, and it doesn't care whether the capture happened at 10am or 2pm.

Shadow-derived height is a reasonable fallback for isolated structures in open terrain, a standalone warehouse, a suburban commercial building with no tall neighbors nearby, where the shadow falls clean on flat ground. Treat it with more caution in a downtown core or anywhere rooftops crowd together.

Either way, resolution sets a floor under both methods. At VHR building height accuracy below roughly half a meter per pixel, a rooftop edge is a handful of pixels wide, and the shadow line or the stereo match has something real to lock onto. Coarser imagery turns both methods into guesswork no matter which one you picked.

What this means for the storey count you write down

A height in meters isn't the number that goes into your site file. Floor-to-floor heights vary by building type, so converting a height estimate into a storey count takes a judgment call, usually a fixed assumption per building class (roughly 3 to 3.5 m for commercial, closer to 3 m for residential) applied consistently across a list rather than building by building. Whichever method produced the input height, keep a note of it. A stereo-derived 42 m and a shadow-derived 42 m pulled near solar noon in a dense block don't carry the same confidence, even if they look identical in a spreadsheet.

If you're triaging a list of forty candidate rooftops before anyone drives out to confirm, knowing which height method sits behind each number tells you where to spend your site-visit budget and where you can sign off from the desk.

Storey Estimation turns that height estimate, whichever method it came from, into a per-building storey count delivered as a CSV or layer attribute, refreshed annually from VHR imagery, so the desk survey starts with a number you can actually work from.

Worth a look if your next candidate list is sitting in a spreadsheet waiting on height data.