Datums & heights · 4 min read

h = H + N: ellipsoid height, orthometric height and geoid height

The single most useful equation in GNSS elevation work. What each term means, which sign convention applies, and where the error in the result comes from.

A GNSS receiver produces a height above the ellipsoid. A benchmark carries a height above the geoid. The equation that connects them is short enough to memorize and subtle enough to get wrong in the field on a regular basis.

h = H + N
Ellipsoid height = orthometric height + geoid height

Rearranged for the case you actually meet — you have a GNSS height and you want an elevation the client can use — it becomes H = h − N. Since N is negative throughout the conterminous United States, subtracting it makes the orthometric height larger than the ellipsoid height, typically by twenty to fifty meters.

The three terms

Ellipsoid height (h)
The distance from the reference ellipsoid to the point, measured along the ellipsoid normal. Purely geometric. It is what a GNSS solution natively produces and it has no physical meaning for drainage or flow.
Orthometric height (H)
The distance from the geoid to the point, measured along the curved plumb line. This is elevation in the ordinary engineering sense, the number on a benchmark and the number a level run carries.
Geoid height (N)
The separation between the ellipsoid and the geoid at a point, positive where the geoid is above the ellipsoid. Also called geoid undulation or geoid separation. It comes from a published geoid model, never from a measurement you make.

A worked example

Suppose a receiver on a control point returns an ellipsoid height of 384.126 m on NAD 83 (2011), and the geoid model interpolated at that latitude and longitude gives N = −29.512 m.

H = h − N = 384.126 − (−29.512) = 413.638 m
Subtracting a negative geoid height increases the orthometric height.

In feet that is about 1357.1 international feet. The sign is where crews lose points: N is negative, so the minus sign in H = h − N becomes an addition. Anyone who mechanically subtracts 29.512 lands 59 m low, which is an error large enough to be caught immediately — the dangerous version of this mistake is a small sign error in a geoid model applied to only part of a project.

Why the equation is an approximation

Strictly, h is measured along the straight ellipsoid normal and H along the curved plumb line, and those two lines are not collinear. The angle between them is the deflection of the vertical. Because the deflection is small — usually a few arcseconds — the resulting error in h = H + N is well under a millimeter in ordinary terrain, and the equation is treated as exact for all practical survey work.

There is a second and much larger source of inexactness, and it is not geometric. It is that H and N have to be consistent with each other. A hybrid geoid model is built specifically so that h from GNSS minus N from the model reproduces the published H on the existing benchmark network. Mix a geoid model from one generation with a datum realization from another and the equation stops balancing.

Where the error in H actually comes from

Contributors to the uncertainty in a GNSS-derived orthometric height
SourceTypical magnitudeCan you reduce it in the field?
GNSS vertical positioning error2 to 4 times the horizontal errorYes — longer occupations, better geometry, redundant sessions
Geoid model errorA few centimetres, worse in rough terrainNo — but you can constrain it by holding local benchmarks
Local distortion in the vertical datumCentimetres to a decimetre regionallyPartly — by tying to several benchmarks and checking agreement
Antenna height blundersWhole decimetres or metersYes — measure twice, record the antenna reference point

Notice that the geoid model is rarely the dominant term. Vertical GNSS error and antenna measurement blunders usually are. The standard defence is to occupy at least two published benchmarks bracketing the project, compare the modelled H against the published H, and apply the residual as a local shift or an inclined plane. That procedure absorbs geoid model bias and local datum distortion at the same time.

Practical rules

  1. Record and report h, H and N separately. A single elevation with no provenance cannot be checked later.
  2. Name the geoid model on the drawing, not just the vertical datum.
  3. Hold benchmarks, do not merely check them, when the project has any tolerance tighter than a tenth of a foot.
  4. Never mix an ellipsoid height from one realization with a geoid model built for another.
  5. Remember that h has no hydraulic meaning. Nothing drains relative to the ellipsoid.

The modernized vertical datum is designed to make this workflow the primary one rather than a workaround. Instead of a national levelling network defining heights and a geoid model bridging to GNSS, the gravimetric geoid becomes the definition, and an orthometric height is obtained from a GNSS position plus the model directly. The equation h = H + N does not change. What changes is which term is considered the authority.

Questions

Is it h = H + N or H = h + N?

It is h = H + N. Ellipsoid height is the sum of orthometric height and geoid height. To get an elevation from GNSS you rearrange to H = h − N, and because N is negative across the conterminous United States that subtraction increases the number.

How accurate is a GNSS-derived orthometric height?

With good practice — long occupations, several benchmarks held, a current geoid model — a few centimetres is realistic over a normal project. The dominant errors are usually the GNSS vertical component and antenna height measurement, not the geoid model.

Why does my GNSS elevation disagree with a nearby benchmark by 0.15 ft?

Most often because the geoid model and the local realization of the vertical datum disagree slightly at that spot, or because the benchmark itself has moved or been disturbed. Occupy two or three benchmarks around the site; if they all show a similar residual, apply it as a local shift rather than assuming one bad mark.

Sources

  • NOAA / National Geodetic Survey — Hybrid geoid models, the relationship between GNSS-derived and levelled heights, and guidelines for GNSS-derived orthometric heights.