How to identify a turbine model from hub height and blade length in satellite imagery

If you manage a wind portfolio that's crept past its first decade, you've probably done this exercise at least once: pull up an aerial pass of a site, eyeball the turbines, and try to work out which model is out there when the original procurement records are thin or the asset changed hands twice since commissioning. It's slower than it should be, and it's the kind of task that falls on whoever drew the short straw that quarter.

Here's the method that actually works, and where it stops working once you're doing it across more than a handful of sites.

Measuring the two numbers that matter

Turbine model identification from imagery comes down to two measurements: hub height and blade length (rotor radius, really, but blade length is the proxy everyone uses because it's easier to eyeball).

Hub height you get from the shadow. On a clear day, with known sun elevation angle and azimuth for the capture time, the shadow cast by the nacelle gives you a reasonably solid hub height estimate. Most people doing this by hand use a simple trig calculation: height = shadow length × tan(sun elevation angle). It works, but it's only as good as your capture metadata, and a lot of archival imagery doesn't come with precise acquisition time stamps.

Blade length is more straightforward to pull straight off the image, tip to tip across the rotor disc, assuming the blades aren't foreshortened by the capture angle. At 0.5 m GSD or better you can usually get a measurement you'd trust to within a couple of meters. Coarser than that and you're guessing.

Once you've got both numbers, you cross-reference them against an OEM footprint matching table, hub height and rotor diameter banded by model family and rating. A 2 MW machine from one manufacturer and a 2.3 MW machine from another can sit within a meter or two of each other on both dimensions, which is where this gets annoying.

Where the manual match breaks down

The hub height and blade length lookup works fine for one turbine. It gets slow for a site of twelve, and it falls over for a portfolio of forty sites built across three procurement cycles with inconsistent records.

A few things trip people up consistently. Blade pitch at capture time changes the apparent rotor diameter if the image was taken at an angle rather than straight down, so you need near-nadir imagery or you're correcting for foreshortening on every single turbine. Shadow-based hub height estimates drift if the capture happened near solar noon, when shadows are short and the margin of error balloons. And the OEM spec tables themselves aren't always clean. Model variants with the same nameplate rating sometimes ship with different hub heights depending on the site wind class ordered, so two turbines that look identical in the data sheet can look different in the photo and still be the same model.

None of that is a reason to give up on the method. It's a reason to stop doing it turbine by turbine, site by site, in a spreadsheet, every time someone asks which sites are nearing the end of their design life.

What this is for

The reason anyone bothers with turbine model identification from imagery in the first place is usually end-of-life planning. You want to know, across the whole portfolio, which sites are running older model families with shorter remaining design life, and you want that list ranked by how straightforward the repowering consent is likely to be, before you commit a planning team to a site visit.

That's the exact cross-reference Repowering Candidates runs from one annual VHR pass: it ages the installed fleet from model footprint and imagery history and hands back a ranked list, so your team spends its time on the consent push instead of measuring shadows off a satellite image.

If you're the one who currently does this match by hand against a spreadsheet of OEM spec sheets, it's worth seeing what a portfolio-wide pass looks like instead of a site-by-site one.