beacen.

Elevation measurement field answer 053

GPS altitude vs barometric altitude

The short answer

GPS altitude is estimated from satellite measurements and a stated vertical reference. Barometric altitude is inferred from air pressure, so it is good at tracking short-term change but needs a pressure reference and can drift as weather changes. Phones and watches may combine these inputs. A displayed altitude should not be called purely GPS or barometric unless its source is known.

Range and pressure

GNSS solves a vertical coordinate. A barometer observes air pressure.

A GNSS receiver estimates three-dimensional position from satellite range measurements. Height is part of that solution. Reflections, blockage and atmospheric errors affect the result.

A barometric sensor measures ambient pressure. Since pressure generally falls with height, software can convert a change in pressure into a change in altitude. This makes a barometer useful for following climbs and descents over a short period.

Pressure is not height by itself. An absolute barometric altitude needs a sea-level pressure or another calibration reference. Android's sensor API warns that using standard atmosphere as an approximation does not give accurate absolute altitude. Apple also exposes relative altitude as the change since the first altimeter event.

What zero means

Two correct numbers can use different vertical references.

Raw satellite positioning naturally produces height relative to a mathematical Earth ellipsoid. Maps and everyday summit heights commonly use a mean-sea-level-related vertical datum. The separation between these surfaces varies by location, so relabelling ellipsoidal height as elevation above sea level creates a systematic difference.

Platform APIs make the distinction visible in different ways. Android Location defines its altitude field above the WGS 84 reference ellipsoid. Apple CLLocation provides an altitude above mean sea level and a separate ellipsoidalAltitude property. These are API definitions, not evidence that every underlying measurement came from satellites alone.

Barometric altitude has another reference: the pressure assumed at sea level or at a calibration point. Before comparing values, identify both the physical input and its zero surface. Vertical datums explains the map side of that comparison.

Strengths and drift

The two methods fail in different ways.

  • GNSS heightCan jump with satellite geometry, obstruction, reflection and receiver processing.
  • Pressure changeCan track a smooth climb well but also responds when weather changes the ambient pressure.
  • Absolute calibrationA known local height or current pressure reference can anchor a barometric estimate, but that anchor can become stale.

Neither source is universally superior. GNSS supplies a global coordinate without local pressure calibration. A barometer supplies stable relative change without needing a clear satellite view. Combining them can use those complementary qualities.

The number on screen

A phone may present a processed height rather than one sensor.

Operating systems and apps can combine GNSS, pressure, terrain models and prior calibration. They may correct drift, translate between vertical references or replace missing point heights after recording. Unless the product documents the calculation, the displayed altitude should be treated as a location estimate, not labelled as pure GPS or pure barometer.

Inspect any vertical-accuracy field, the data source and whether the value is absolute or relative. Compare a stable section before judging a single spike. When two services disagree, check their datum, terrain model, smoothing and ascent rules before deciding one is wrong.

Route elevation gain adds another processing layer because it sums selected changes across many points. A small amount of vertical noise can accumulate into a large total. The article on why elevation gain differs covers that calculation separately.

Sources & scope

What this answer is based on.

The platform documents define exposed altitude, uncertainty and pressure-derived change. Hardware and operating systems may fuse sensors internally without exposing the full calculation.