Coordinately

Accuracy

How accuracy is measured and reported across Coordinately tools and articles.

By . Published . Last updated .

Every numerical value on Coordinately has an error bound. Some bounds are extremely tight (the math behind coordinate-format conversion is accurate to floating-point precision); some are loose (elevation in remote terrain may be off by tens of metres). This page documents what those bounds are for each kind of value the site publishes, so that a reader can decide whether a number is fit for their purpose.

The procedural side — how we decide which sources to cite and when — lives at /methodology and /editorial-policy. This page is the per-domain accuracy reference.

Precision versus accuracy

The two terms are often used interchangeably, but they are different. Precision is how many digits a measurement reports. Accuracy is how close those digits are to the true value.

A GPS reading of 40.7484°N, 73.9857°W is precise to four decimal places — about 11 m at the equator, narrower at higher latitudes. A reading of 40.7484170°N, 73.9857138°W is precise to seven decimal places — about 1.1 cm at the equator. Whether either reading is accurate to within that precision depends on the receiver. A typical smartphone GPS is accurate to roughly 4.9 m under open sky per GPS.gov, so seven decimal places of precision on a smartphone reading is false precision — the trailing digits are noise.

Coordinately uses the NIST definition of measurement uncertainty as the operating model: every published value comes with an understood error bound, and we report values to a number of digits consistent with that bound.

GPS accuracy

Civilian GPS performance figures are published by GPS.gov, the official US National Coordination Office for Positioning, Navigation, and Timing:

  • Smartphone receivers under open sky: approximately 4.9 m horizontal, 95% of the time.
  • High-quality FAA-grade receivers: ≤ 1.82 m horizontal, 95% of the time.
  • GPS signal-in-space user range error (URE): ≤ 2.0 m globally, 95% probability. Measured global URE on 2021-04-20: 0.643 m, 95%.
  • Time-transfer accuracy: ≤ 30 ns versus UTC.

These are the headline numbers. In practice, accuracy degrades in urban canyons (multipath error from buildings), indoors (signal attenuation), and under heavy tree cover. Dilution of precision — the geometric effect of which satellites are above the horizon — adds a multiplicative factor that varies through the day.

Elevation accuracy

Coordinately uses two elevation sources, selected by region. The full rationale is in the architectural-decisions record; the practical bounds:

  • USGS 3DEP (US points). The 1/3 arc-second seamless DEM provides sub-2 m vertical accuracy on continental US land, better than 1 m in well-surveyed areas. Coverage: continental US, Alaska, Hawaii, Puerto Rico, US Virgin Islands.
  • OpenTopoData / SRTM30m (non-US points). Approximately 10 m vertical accuracy on rural terrain, ~30 m in urban or steep mountainous terrain. Coverage: land worldwide between 60°S and 60°N (the SRTM mission's latitude limit).
  • Ocean coordinates.Return “no data” rather than zero. The elevation tool surfaces the dataset provenance alongside every value so users can see which source produced the number.

Both 3DEP and SRTM30m are referenced to specific vertical datums. 3DEP uses NAVD88 (transitioning to NAPGD2022 with NSRS modernization); SRTM uses EGM96. The difference between vertical datums can be several metres at any given point — a separate kind of inaccuracy from the dataset's own grid uncertainty.

Geocoding accuracy

Forward geocoding (address → coordinates) uses the Mapbox Geocoding API v6. Mapbox publishes a confidence band on every result: high, medium, or low. Our tools surface the band — when confidence is medium or low, the tool presents alternative candidates rather than auto-selecting the top match.

Confidence and accuracy are related but not identical. A high- confidence match means Mapbox is confident it found the address; it does not mean the address is necessarily a precise point within its building. Address-level geocoding is typically accurate to ~5-15 m for residential addresses, ~30 m for commercial buildings with multiple entrances, and ~100 m or more for rural addresses without precise interpolation data.

Reverse geocoding (coordinates → address) is bounded by the spatial resolution of Mapbox's address dataset. The nearest-address match for an arbitrary point may be tens to hundreds of metres from the query point.

Distance accuracy

Distance calculations use one of two methods depending on the tool:

  • Vincenty's formula on WGS84 (ellipsoidal inverse) — sub-millimetre numerical accuracy. Used by the distance-calculator tool. The formula can fail to converge on near-antipodal points; in that case the tool falls back to the spherical haversine result with a visible caveat.
  • Haversine (spherical) — accurate to about 0.5% on Earth (worst case ~0.55%, at small distances along meridians near the equator). Used for routine display where a 0.5% error is acceptable, and as the fallback for Vincenty.

Both methods compute the great-circle (or geodesic) distance — the shortest path along the surface. Real-world transportation distances (driving, flying) are always longer.

Coordinate format precision

Different coordinate notations imply different precisions:

  • Decimal degrees at 6 decimal places ≈ 11 cm at the equator.
  • Decimal degrees at 5 decimal places ≈ 1.1 m.
  • Decimal minutes (DDM) at 3 decimal places ≈ 18 cm.
  • Degrees-minutes-seconds (DMS) with arcseconds to 2 decimals ≈ 30 cm.
  • UTM at 1 m precision (whole-metre easting/northing) is 1 m.
  • MGRS at 5-digit precision is 1 m; 3 digits is 100 m; 1 digit is 10 km.
  • Plus Codes at length 10 (the default) describe a ~14 m × 14 m cell; at length 11, ~3.5 m × 2.8 m; at length 14, sub-metre.

These figures are at the equator. Latitude-based shrinking applies to longitude-derived values — at 60° N or S, one degree of longitude is half the distance of one degree at the equator.

What “approximately” means here

When an article on Coordinately says “approximately”, we mean “within the stated error bound.” Where no error bound is stated, the qualifier should be read as “to the precision of the source we cite.”

When we do not know the error bound, we say so explicitly — not with a vague qualifier, but with a sentence acknowledging the uncertainty.

How accuracy appears in tools

Every tool that displays a numerical result also displays the provenance of that result and (where applicable) its confidence or precision band:

  • The elevation tool shows the dataset name (“USGS 3DEP” or “SRTM30m via OpenTopoData”) alongside the value.
  • The forward-geocoding tool shows the Mapbox confidence band and alternative candidates for medium/low confidence.
  • The distance-calculator tool shows which formula produced the result (Vincenty or haversine) and, if Vincenty failed to converge, a caveat.
  • Format-conversion tools (DMS, UTM, MGRS, Plus Codes) display all formats at their natural precision and a note about what that precision corresponds to on the ground.

If a tool you use does not show accuracy information you expect, write to info@coordinately.org — we want to know.

Related

Frequently asked questions

What's the difference between precision and accuracy?

Precision is how many digits a measurement reports; accuracy is how close those digits are to the true value. A GPS receiver that reports 40.748817°N, 73.985428°W is precise to six decimal places (about 11 cm at the equator). Whether the reported point is within 11 cm of the actual location is a separate question — that depends on the receiver's accuracy, which is usually several metres. NIST's definition of measurement uncertainty distinguishes these two concepts formally.

How accurate is my phone’s GPS?

Civilian GPS smartphone receivers achieve approximately 4.9 m horizontal accuracy under open sky 95% of the time, per GPS.gov. Indoors, in urban canyons, and under heavy tree cover, accuracy degrades — phones can report locations tens of metres from the true position. Multipath errors (signals bouncing off buildings) are the most common cause of urban inaccuracy. Cold-start latency (the time for a phone to acquire enough satellites for a fix) is a separate issue from accuracy.

Why is GPS more accurate at night?

GPS accuracy varies with satellite geometry (the dilution-of-precision factor, DOP), which depends on which satellites are above the horizon. DOP changes throughout the day in a roughly 12-hour pattern as satellites move through the sky. Night-versus-day differences are real but usually small — perhaps a metre at most for a typical receiver — and they don't apply uniformly. A more important factor for accuracy is open sky versus obstruction.

What does "no data" mean for elevation?

The elevation tool returns "no data" rather than zero when the queried coordinate is on or near the ocean and our elevation source has no value to report. The USGS 3DEP dataset covers US land only; OpenTopoData's SRTM30m covers land worldwide but excludes oceans. Returning zero for ocean coordinates would falsely imply sea level at every point and would round-trip badly through downstream calculations.

How accurate are coordinates I get from Google Maps?

Google Maps' coordinate display is precise to seven decimal places (about 1 cm), but the actual accuracy depends on what was geocoded. For a clicked map point, accuracy is essentially the screen-pixel resolution of the underlying tile at the current zoom — sub-metre at deep zoom, tens of metres at low zoom. For a geocoded address, accuracy depends on the address geocoder's confidence band, which Google doesn't always surface to the user. Sources like Mapbox publish explicit confidence levels (high, medium, low); Coordinately tools that use Mapbox surface those to the user.

Sources

  1. NIST — Measurement uncertainty — definitions of precision and accuracy · https://www.nist.gov/pml/owm/measurement-uncertainty · Accessed .
  2. GPS.gov — GPS accuracy — performance standards · https://www.gps.gov/systems/gps/performance/accuracy/ · Accessed .
  3. USGS — 3D Elevation Program (3DEP) accuracy standards · https://www.usgs.gov/3d-elevation-program · Accessed .
  4. OpenTopoData — SRTM 30 m dataset and stated accuracy · https://www.opentopodata.org/datasets/srtm/ · Accessed .
  5. Vincenty (1975) — Direct and Inverse Solutions of Geodesics on the Ellipsoid with application of nested equations · https://www.ngs.noaa.gov/PUBS_LIB/inverse.pdf · Accessed .
  6. NGA — WGS84 ellipsoid parameters · https://earth-info.nga.mil/index.php?dir=wgs84 · Accessed .
  7. Mapbox — Geocoding API v6 — confidence semantics · https://docs.mapbox.com/api/search/geocoding-v6/ · Accessed .

Cite this article

APA format:

Steve K. (2026). Accuracy. Coordinately. https://coordinately.org/accuracy

BibTeX:

@misc{coordinately_accuracy_2026,
  author = {K., Steve},
  title  = {Accuracy},
  year   = {2026},
  publisher = {Coordinately},
  url    = {https://coordinately.org/accuracy},
  note   = {Accessed: 2026-06-05}
}