GPS elevation vs terrain elevation data

Observation and model

A device estimates its position. A terrain dataset estimates the surface below a coordinate.

Recorded elevation begins with observations made during the journey. Terrain-derived elevation begins with a route line and an existing model of the land. The first describes the device's processed vertical position at a time. The second assigns a modelled surface height to a horizontal location.

This distinction remains even when both appear as the same smooth profile. A planning service can add terrain heights to a line that was drawn with latitude and longitude only. A recording service can replace noisy device heights with terrain values after the activity. The resulting file may no longer reveal every step in that processing.

Neither label guarantees a particular datum. Before comparing the numbers, establish whether they are ellipsoidal heights, orthometric elevations or values transformed through a geoid model. A constant-looking offset often points to a reference difference rather than a taller landscape.

The device side

GNSS solves a three-dimensional position, then the device or app may transform it.

Satellite positioning natively relates a receiver to a mathematical reference frame and ellipsoid. A geoid model can convert ellipsoid height into an orthometric height intended to correspond more closely with mapped elevation. That conversion requires the appropriate model and compatible reference frames. Simply subtracting a generic offset is not enough.

Phones and watches may also use a barometer, recent calibration, motion sensors and software filtering. GPS elevation on screen can therefore be a convenient product label rather than a description of a raw satellite field. GPS altitude and barometric altitude explains what each input contributes. Vertical accuracy also behaves differently from horizontal accuracy, and poor satellite geometry or reflected signals can disturb both.

A barometer can capture relative climbing smoothly when calibrated well, but pressure changes can move its baseline. Satellite height can anchor that baseline yet remain noisy. Sensor fusion is intended to balance those weaknesses, not abolish them.

The model side

Terrain elevation inherits the surface and cell structure of the chosen model.

A bare-earth digital elevation model aims to represent ground terrain without buildings and vegetation. Software samples cells beneath or around the route and may interpolate between their stored values. The result depends on cell spacing, source measurements, vertical datum and the model's treatment of water and built features.

• BridgesA bare-earth model may describe the ground or water beneath a bridge rather than the deck travelled.
• TunnelsThe model describes the surface above, not the underground route.
• Cliffs and cuttingsA small horizontal offset can sample a sharply different height.
• Vegetation and structuresA surface model and a bare-earth model may intentionally represent different tops.

Fit the evidence to the question

Use terrain for a plan and recorded data for the journey, with exceptions made explicit.

Terrain sampling is useful when a planned route has no heights or when consistent profiles are needed across a collection. Recorded values can preserve what the device experienced, including travel on raised structures or other features the terrain model misses. They can also preserve noise and calibration drift.

When the sources disagree, compare the horizontal line first. Then check the vertical datum, terrain product, model resolution, device sensors and any correction setting. Look for whether the difference is a steady offset, a local feature or repeated noise. Each pattern points to a different cause.

The beacen Guide explains the current product workflow.

Keep the original recording before replacing its elevations, since a corrected copy has different provenance. A recorded profile and a terrain-derived profile remain different evidence even when their shapes look alike.