How accurate is aerial measurement, really?

Every person who uses a tool like this asks the same question within about ninety seconds, and they are right to: how do I know this number is real?

The honest answer has two halves. For the thing most people measure — a flat area on the ground, traced on current imagery, at a house-sized scale — aerial measurement is very good, and better than most people's tape work. But it fails in specific, knowable ways, and a tool that does not tell you what they are is not being straight with you.

What is under the picture

You are not tracing a photograph. You are tracing an orthophoto — an aerial image that has been geometrically corrected so it behaves like a map, with a consistent scale you can measure on.

Four sources matter for most of the United States:

Google satellite imagery is what this tool uses. In most towns and suburbs it is aerial photography flown from a plane rather than satellite. At our own test address in Amherst, NY it resolves to about 2 inches per pixel at full zoom, where Esri stops at about 8.6. Google does not publish a capture date for a given spot; the imagery credit at the bottom of the map names the provider.

Esri World Imagery is the default in most web mapping, and this tool's backup map. Esri publishes 30 cm resolution across the United States, with a mixed 30 cm and 15 cm product covering more than a thousand metro areas. Metro areas at 30 cm are refreshed roughly annually, with selected cities updated quarterly at 15 cm. Esri does not publish a guaranteed age for any given tile, so "how old is this picture" is genuinely unanswerable in general — you check the capture date for your specific location.

USDA NAIP is the nationwide fallback. It runs at 60 cm for about half the states and 30 cm for the rest as of 2025, on a cycle of no more than three years and generally every other year. Critically, NAIP is flown leaf-on, during the growing season — which is exactly the wrong condition for seeing the ground under trees. Its published horizontal accuracy is that well-defined points fall within 6 meters at 95% confidence, which is fine for agriculture and coarse for a flower bed.

New York State orthoimagery is the best product available in Erie County, and it is leaf-off. The statewide program has run since 2000 on a four-to-five-year cycle; Erie County has actually been flown in 2024, 2021, 2017, 2014, 2011, 2008, 2005 and 2002 — closer to every three years. The published specification for the western zone is 1-foot ground sample distance, flown in April with the leaves down, at ±4 feet at 95% confidence under the NSSDA standard. The next Western New York acquisition is scheduled for spring 2027.

What "accuracy" means on a map

Two standards get quoted, and they say different things.

NSSDA — the National Standard for Spatial Data Accuracy, FGDC-STD-007.3-1998 — reports accuracy as a distance at 95% confidence. When New York says ±4 feet, it means 95 of 100 well-defined test points fall within 4 feet of their true position.

ASPRS Positional Accuracy Standards, Edition 2 (February 2023) replaced that approach for new work. It reports a horizontal accuracy class equal to RMSE, and it explicitly stopped endorsing 95% confidence reporting — and, notably, stopped endorsing the use of ground sample distance or map scale as a proxy for accuracy at all. Which is a formal way of saying what field people already know: a sharp-looking picture is not the same thing as an accurate one.

Here is the part that matters for you. Positional accuracy and relative accuracy are different problems. If the entire image is shifted three feet north, every point you trace is shifted three feet north — and the distance between your points, which is what you are measuring, is unaffected. Absolute position matters for surveying and for stakeout. For "how many yards of mulch," relative accuracy is what counts, and it is much better than the positional figure suggests.

The five things that actually go wrong

1. Roof pitch

An overhead image sees the footprint, never the slope. A 2,000 square-foot roof footprint at an 8/12 pitch is 2,404 square feet of actual surface — 20% more shingles than the picture shows. The correction is √(1 + (rise/12)²), and no amount of image resolution substitutes for knowing your pitch.

2. Building lean

In a standard orthophoto, the correction is computed for the ground surface. Anything tall is not shifted to its true ground position, so the roof of a building appears displaced away from the image center — the taller the structure and the further from nadir, the worse it gets. Pima County's GIS office puts it plainly: an orthophoto "can be off by several feet" for elevated features.

This is why tracing a two-story roofline can put your outline a few feet off the building's actual footprint, even on perfect imagery. True orthophotos fix this by correcting individual buildings so roof detail sits in its correct position, but they are more expensive to produce and are not what most free basemaps use.

For a homeowner tracing a driveway, this is a non-issue — the driveway is on the ground. For a roof, trace the roof edges you can see rather than trying to match the foundation.

3. Tree canopy

You cannot measure what you cannot see. The timing of the flight is what decides this: NAIP is flown leaf-on, New York's orthos are flown in April with the leaves down. On a Buffalo lot with mature maples, that difference is the difference between seeing your beds and seeing a green blanket. The USGS orthoimagery specification actively encourages leaf-off acquisition for this reason.

I have not found a credible published figure for how much of a typical suburban lot is obscured by canopy, so I will not invent one — but if your trace looks like it is going around a tree rather than a bed, switch imagery sources or measure that section on the ground.

4. Stale imagery

The picture is a snapshot of a day. If the driveway was widened, the deck built, or the bed reshaped since the flight, the image is accurate and your measurement is still wrong. Check the capture date before trusting a trace of anything recent.

5. Projection error — the big one

This is the failure mode that produces catastrophically wrong numbers, and it is entirely a software problem.

Web maps use Web Mercator. Mercator stretches the world increasingly as you move away from the equator, and area is stretched by the square of that factor. At Buffalo's latitude of roughly 43° north, linear distances are inflated by about 37%, and areas by about 87%.

If a tool computes area directly in Web Mercator coordinates without applying the cosine-of-latitude correction, your 5,000 square-foot lawn comes back as roughly 9,350 square feet. Esri's own guidance is blunt about it: avoid measuring polygons in Web Mercator; reproject first. AerialQuote applies the correction, and we test against geodesic ground truth on every build — a traced 20 × 10 meter rectangle returns 2,153 square feet against a true 2,152.8. But it is worth knowing why one free measuring tool disagrees wildly with another.

How this compares

EagleView, the commercial aerial roof report used across the insurance and roofing industries, guarantees its reports are "more accurate than any other measurement method" with a free-report remedy — but its actual guarantee page states no percentage. Its marketing claims 98%+ accuracy. Those two things are not the same and it is worth reading which one you are being shown.

Nearmap publishes real specifications: 5.5 cm vertical imagery with absolute horizontal accuracy of 19.8 cm RMSE. That is dramatically better than free imagery, and it is a paid subscription.

A measuring wheel, the thing this replaces for most people, is described even by its defenders as good on smooth pavement and degraded by slippage and bounce on rough terrain — used, in surveying terms, for lower-accuracy work. I could find no authoritative error specification for one, only vendor claims, so treat "the wheel is more accurate" as an assumption rather than a fact.

So what should you trust it for

Trust it for: ordering bulk material, sizing a job, checking whether a quote's square footage is plausible, comparing two options, and knowing whether something fits.

Verify on the ground for: anything you are cutting to fit, any boundary or setback question, anything where you are about to buy a large quantity of an expensive material, and anything under trees.

Never use it for: property lines. An aerial image is not a survey, and a traced line is not a legal boundary.

Run your own numbers

When you trace a property on aerial imagery yourself, the map corner shows the imagery source and the resolution in inches per pixel. If it says the picture is enlarged past native detail, you are zoomed in further than the photo was taken, and edges will look soft.

Sources: Esri — World Imagery resolution and coverage · USGS EROS — NAIP program and accuracy · NYS GIS Clearinghouse — orthoimagery program and the western zone metadata · FGDC — National Standard for Spatial Data Accuracy · ASPRS Positional Accuracy Standards, Edition 2, 2023 · Pima County GIS on orthophoto building displacement · Esri on measuring in Web Mercator · Nearmap accuracy specification · EagleView guarantee

Educational use. Aerial measurements are estimates, not a survey, and must never be used to determine property boundaries.