Why Geoid Models Exist
The geoid is a physical concept, but no receiver can observe the entire gravity-defined surface directly at every point on Earth. A geoid model is the practical tool that estimates the separation between the ellipsoid and that surface so the same GNSS coordinate can be expressed as an orthometric elevation.
A Surface We Cannot See
The geoid is often described as an idealized mean-sea-level surface extended beneath the continents. That description is useful, but it can also make the geoid sound like a physical shell that could be surveyed directly. It is not. There is no painted line on a mountain showing where the geoid passes, and there is no instrument that can occupy every point on that surface and record its height relative to the ellipsoid.
Instead, geodesists infer the shape of the geoid from measurements of Earth's gravity field. Those measurements come from several sources, including terrestrial gravity surveys, airborne observations, satellite missions, and information about the shape of the sea surface. Each source contributes part of the picture, but none by itself provides a perfect gravity surface everywhere. The observations must be combined mathematically into a model that can be evaluated at any location within its coverage area.
A Model Turns Gravity into a Usable Number
For a GNSS user, the output of a geoid model is usually simple: enter a latitude and longitude, and the model returns N, the estimated separation between the ellipsoid and the geoid at that location. The receiver or software then combines that value with the ellipsoid height using H = h - N. The complicated gravity information is therefore reduced to one number that can be used in the vertical equation.
That simplicity can hide what is happening. The geoid value is not measured by the rover during the RTK observation. It is looked up or interpolated from a stored model after the horizontal position is known. Two receivers using the same satellite observations can therefore compute the same ellipsoid height and still display different orthometric elevations if they evaluate different geoid models.
The GNSS solution supplies h. The geoid model supplies N. The displayed orthometric elevation H depends on both. Agreement in RTK status does not guarantee agreement in orthometric elevation if the model or vertical reference differs.
The Model Is a Grid, Not a New Observation
Many geoid models are distributed as a grid of values covering a region. The grid does not contain a value for every possible coordinate. Instead, it stores geoid heights at regularly spaced locations. When the receiver needs N for a point between grid nodes, the software interpolates from the surrounding values.
This is conceptually similar to estimating an elevation between contour lines, although the actual mathematics may be more sophisticated. The important point is that the model contributes an estimated value based on nearby stored information. Changing the interpolation method, grid resolution, or model version can change N slightly even when the underlying GNSS coordinate remains identical.
Why One Model Is Not Enough Forever
Gravity models improve as new measurements become available and older observations are reprocessed. Satellite gravity missions provide broader and more consistent information. Airborne surveys fill gaps. Terrestrial observations are corrected or densified. At the same time, national agencies update the reference systems and control data to which the model is fitted.
A newer model is therefore not simply a renamed copy of the older one. It may use additional gravity observations, improved processing methods, a different reference frame, or a revised relationship to the vertical datum. Those changes are intended to improve consistency, but they also mean the geoid height at a particular coordinate may not be exactly the same from one model to another.
Regional Models and Global Models
A global geoid model must represent the entire planet with one consistent dataset. That makes it valuable for worldwide applications, but it may not contain the dense local gravity information needed for high-accuracy engineering work in every region. A national or regional model can incorporate much more detailed observations and can be fitted specifically to the vertical datum used in that area.
This distinction matters because the word geoid is sometimes used as though every model represents exactly the same surface. In practice, a receiver may include a coarse worldwide model for general navigation while survey software uses a higher-resolution national model designed to transform GNSS heights into elevations compatible with local control. Both are geoid models, but they are not interchangeable for every purpose.
Selecting any available geoid model does not automatically produce the project elevation. The model must be appropriate for the location, the intended vertical datum, and the accuracy required by the work.
A Geoid Model Is Part of the Reference
When an orthometric elevation is reported, the model used to obtain N becomes part of the coordinate's history. Recording only the final value of H leaves out information needed to reproduce the result. A complete record should identify the ellipsoid or horizontal reference, the geoid model and version, the vertical datum or realization, the units, and any additional project calibration that was applied.
This is why two files can contain the same latitude and longitude yet produce different elevations when imported into different software. One application may apply a built-in global model, another may apply a national model, and a third may assume the height is already orthometric and apply no conversion at all. The disagreement begins in the processing path, not in the satellite observations.
The Model Does Not Repair a Bad GNSS Height
A geoid model cannot rescue an inaccurate ellipsoid height. If the antenna height was entered incorrectly, the rover solution was poor, multipath biased the observation, or the base coordinate was wrong, applying a geoid model simply carries that error into the orthometric elevation. The model changes the vertical reference; it does not improve the underlying position solution.
The reverse is also true. A high-quality ellipsoid height can be converted into an incompatible orthometric elevation if the wrong model is selected. Troubleshooting therefore requires separating the two contributors. First determine whether h is correct. Then determine whether the correct N was applied. Treating those as separate questions prevents a reference problem from being mistaken for a positioning problem.
If two receivers agree on ellipsoid height but disagree on orthometric elevation, where should the investigation begin? Not with the satellites. Begin with the geoid model, vertical datum, units, and software settings used after the GNSS solution was computed.
A geoid model is the practical bridge between ellipsoid height and orthometric elevation. It estimates N from gravity data at the coordinate location. Because models differ in coverage, resolution, observations, and datum compatibility, the model name and version are essential parts of any reproducible elevation.
Now we can explain how two correct receivers can occupy the same point, agree on the GNSS solution, and still report different elevations. In the next section, we will compare model choices directly and follow the discrepancy back to the exact step where the numbers diverge.