kowoma.de

Map datum and reference systems

The receiver shows one coordinate, the map shows another for the same point, and both are correct. In Germany the difference can reach 200 metres. The culprit is a line of small print in the map legend that almost everybody overlooks.

The Earth has three shapes, and all three are needed

A coordinate is not a property of a place. It is the result of a computation on an assumed figure of the Earth, and there are three of those.

The terrain is the actual surface. You stand on it, but you cannot compute on it. The geoid is the surface of equal gravity potential corresponding to mean sea level, continued in thought beneath the continents. Because the Earth's interior varies in density it is not smooth, undulating by up to about a hundred metres worldwide. It is the physically correct reference for heights but mathematically awkward. The ellipsoid is the smoothed approximation: a sphere flattened at the poles, described by two numbers, on which latitude and longitude can be computed exactly.

EllipsoidSemi-major axis aFlattening 1/fUsed in
Bessel 18416,377,397.155 m299.1528128Potsdam datum, CH1903, MGI
International 1924 (Hayford)6,378,388.000 m297.0ED50
Clarke 18666,378,206.400 m294.9786982NAD27, North America
GRS806,378,137.000 m298.257222101ETRS89, ITRS
WGS846,378,137.000 m298.257223563GPS

The last two differ only in the tenth digit, some 0.1 millimetres on the semi-minor axis. For every practical purpose GRS80 and WGS84 are the same ellipsoid.

Why the ellipsoid alone is not enough

An ellipsoid describes shape only. A geodetic datum additionally fixes how that shape sits relative to the Earth: origin, orientation of the axes and scale. Historically every country surveyed itself long before anyone could measure the whole planet, choosing an ellipsoid that fitted its own territory well and anchoring it at a fundamental point on the ground. Such ellipsoids are therefore offset from the Earth's centre of mass by hundreds of metres. They fit beautifully at home and become useless abroad.

WGS84 does the opposite. Its origin is the centre of mass of the Earth, determined from satellite orbits. It fits nowhere particularly well and everywhere equally well, and only such a system works for a global navigation system.

The offset between two datums is therefore not an error and not an inaccuracy. Both describe the same point correctly, in different reference systems. It only becomes wrong when numbers from one system are entered into a device computing in the other.

Typical offsets

Reference systemEllipsoidAreaOffset from WGS84
ETRS89GRS80Europebelow 1 m
DHDN, Potsdam datumBessel 1841Germany, former West120–200 m
System 42/83KrassovskyGermany, former East≈ 130 m
ED50International 1924Western and Central Europe110–170 m
CH1903Bessel 1841Switzerland160–180 m
MGIBessel 1841Austria70–105 m
NAD27Clarke 1866North America40–95 m
Tokyo datumBessel 1841Japan, Korea≈ 450 m

Computed with the official transformations registered in the EPSG dataset. Values vary within a country, so ranges are given.

ETRS89 and WGS84 drift apart

The official reference system in Germany and much of Europe is ETRS89. It was defined in 1989 to agree with WGS84 at that time, but it is fixed to the Eurasian plate, while WGS84 follows the global mean and is updated regularly. The plate moves about 2.5 centimetres a year to the north-east, so nearly a metre has accumulated since 1989. Irrelevant for hiking and navigation, decisive for surveying and RTK, which is why correction services always state both the reference frame and the epoch.

Transforming between datums

The conversion does not run directly between latitude and longitude but through cartesian coordinates: the point is converted from latitude, longitude and height into an earth-fixed XYZ frame, shifted, rotated and scaled there, and converted back on the target ellipsoid. That is the Helmert transformation with seven parameters.

X′ = ΔX + (1 + s) · ( X − rz·Y + ry·Z ) Three translations in metres, three rotations in arc seconds, one scale factor in parts per million. The equations for Y′ and Z′ are analogous.

A warning worth the space. For the Potsdam datum the translations 582, 105 and 414 metres are widely quoted, and they are frequently used on their own. They are not a valid three parameter set: they belong to a seven parameter set whose rotations are 1.04″, 0.35″ and −3.08″ with a scale of 8.3 ppm. Drop the rotations and the result moves about 90 metres west in southern Germany, which is large enough to matter and small enough to look plausible.

Even a complete parameter set cannot represent the regional distortions of an old survey network. National mapping agencies therefore publish grid files that interpolate the residual differences point by point; the German dataset is called BeTA2007 and reaches a few centimetres.

In practice

  • The datum is in the map legend, usually next to the grid information. If it is missing and the map predates about 1995, a national datum is more likely than WGS84.
  • When in doubt, WGS84. Everything digital uses it: online maps, GPX files, geocaching listings, smartphone apps.
  • Never mix. A waypoint list assembled from two sources is the classic failure. Note the datum along with the coordinate, as you would note a unit with a measurement.
  • The offset is systematic. If all points of a route are displaced by the same amount in the same direction, it is the datum. If they scatter irregularly, it is reception.