Stereo satellite imagery vs airborne lidar for terrain roughness mapping

A siting engineer staring at a candidate ridge has two real options for terrain data: order a lidar flight, or pull stereo imagery and derive a surface model from what's already in orbit. They're not interchangeable, and treating them that way wastes either budget or time.

What each one measures

Airborne lidar sends laser pulses down from a plane or drone and times the returns. Some of those pulses sneak through gaps in canopy and bounce off bare ground, which is why lidar is the method of record for a true digital elevation model (DEM), the bare-earth surface under trees, scrub, and ground cover. A lidar point cloud at reasonable density can separate ground returns from vegetation and structure returns, which is exactly what you need for a foundation-grade terrain model.

Stereo VHR imagery works differently. Two images of the same ground, taken from slightly different angles, let photogrammetry software triangulate height the way your own eyes judge depth. The output is a digital surface model (DSM): the top of whatever is there, whether that's bare rock, a tree canopy, or a barn roof. It's RGB, not a laser return, so it can't see through canopy to bare earth the way lidar can. What it does have going for it is reach: a stereo pair from a VHR satellite or aerial platform at 0.5 m GSD or better can cover a ridge, or a whole prospect's worth of ridges, without anyone chartering a flight.

Where the gap matters for siting

If you're past the screening stage and locking a turbine layout for foundation design, you want the DEM. Bare-earth accuracy under forest canopy isn't something a stereo DSM can give you, full stop, and no amount of post-processing closes that gap when the canopy is thick.

But most ridges in a development pipeline never get that far. A developer might carry two dozen candidate sites and only mobilize a met mast on four or five of them. Flying lidar on all two dozen to find out which ones are worth a mast is backward: you're spending survey-grade money to answer a screening-grade question. That's the gap a stereo-derived DSM fills. Slope, surface roughness, and obstacle height pulled from a stereo pair are enough to rank ridgelines against each other and rule out the ones with obvious roughness problems or obstruction issues before a crew ever drives out. The sites that survive that first pass are the ones worth spending lidar budget on.

Obstacle height without a flight plan

Obstacle height mapping is the piece people underrate in this comparison. A DSM captures the top surface as it stands today, tree lines included, which makes it better suited than a bare-earth DEM for spotting obstructions that affect wind flow and micrositing, like a stand of trees upwind of a candidate pad or a ridge where the canopy height changes abruptly along the crest. The question at this stage is narrow: does this stretch of ridge look clean enough to justify putting a mast here, or does the surface tell you to cross it off the list now.

This is the use case Wind Farm Siting is built around: deriving slope, roughness, and obstacle height from stereo VHR imagery so a desk screen, not a field crew, rules out the ridges that were never going to pencil. It's delivered per project with an annual refresh, which actually fits how most pipelines move. A prospect gets screened once, carried for a season or two, then checked again before a final decision.

Neither method replaces the other end to end. Lidar earns its cost once you've narrowed the field. Stereo imagery earns its keep before that, when the question isn't "how accurate is this terrain" but "which of these ridges deserve the question in the first place."

If that first pass is where you're stuck right now, that's the problem this site is built to shorten.