Satellite height estimates vs. Google Earth measurements for cell site scouting
When you're scoring forty candidate rooftops for a small-cell or DAS node before anyone drives the list, pulling up Google Earth feels like the fastest way to rule half of them out. Zoom in, tilt the view, eyeball the roofline against the street below, done. The trouble is the tool was built to tell you ground elevation, not roof height, and that gap is exactly what gets a candidate bounced in the zoning review three weeks later, after you've already quoted a mount height to the carrier.
What Google Earth's elevation tool measures
The ruler and elevation profile tools read terrain elevation at whatever point your cursor sits on, pulled from the underlying digital elevation model. In a lot of the world that model is coarse, stitched from older radar passes, and it has no idea a building exists at that coordinate. Drop one point at the base of a structure and another on what looks like the roof edge in the tilted 3D view, and both readings come from the terrain grid near the footprint. The building's height above grade never enters the calculation.
The 3D buildings layer helps a little in cities where Google has modeled structures, but those models are frequently extruded from rough massing data or crowdsourced edits, not a survey. A building can show up eight stories tall in the 3D view and turn out to be six when the structural engineer walks the roof. For a quick visual gut-check that's forgivable. For antenna clearance, parapet height, or a floor count you're about to put in a lease exhibit, it's a liability you won't see until the site visit catches it, or doesn't.
Where a stereo pair or shadow length earns its keep
A height estimate pulled from very high resolution satellite or aerial imagery works differently. Instead of interpolating terrain under a cursor, it either triangulates the rooftop edge against the ground using two overlapping looks from different angles (stereo), or measures the shadow a building casts and works backward using the sun's known position at capture time. Either way, the number comes from the structure itself, not from a terrain grid that happens to pass near it.
The question that actually drives a shortlist decision isn't "what's the elevation here." It's "how tall is this specific rooftop, and how many floors does that work out to." A per-building height value, converted to an estimated storey count and delivered as a CSV or a layer attribute you can filter and sort, lets you knock out candidates that are too short for the mount height a carrier wants, or flag the ones tall enough to need a structural conversation, before you've committed a field day to any of them. That's the gap Storey Estimation is built to close: building height and storey count from imagery, refreshed on an annual cycle, without sending anyone up on the roof first.
When the desktop number is good enough, and when it isn't
For first-pass ranking across a candidate list, an imagery-derived height estimate beats a terrain-tool guess, because the measurement comes from the building and not the ground next to it. It's enough to cut a list of sixty rooftops down to the dozen worth a closer look, or to catch the three-story building someone's RF team assumed was mid-rise based on an old permit record.
It isn't a substitute for a structural survey once a site is shortlisted. Lease negotiations, loading calculations, and anything that ends up stamped by an engineer still need someone on the roof with a tape measure and a load assessment. The desktop estimate's job is narrowing the field so that visit only happens for sites worth the trip, not every rooftop on the long list.
If you're staring down a list of candidate buildings and don't want to burn a field day ruling out the wrong half of it, see what a per-building height and storey layer looks like for your market.