6.7 Datum Transformation in Practice
Key Takeaways
- A datum transformation is a three-dimensional operation between reference frames; a projection change is a two-dimensional operation within one frame.
- EPSG records several different Minna to WGS 84 transformations, each with its own area of validity, derivation and stated accuracy, so the correct set must be chosen for the project area.
- A three-parameter geocentric translation models only origin offset; a seven-parameter Helmert adds three rotations and a scale difference.
- Molodensky-Badekas applies rotations about the centroid of the local network rather than the geocentre, which greatly reduces correlation between the estimated parameters.
- Every plan must state the datum, the projection belt and the transformation used, because coordinates are meaningless without the frame they are expressed in.
Last updated: August 2026
6.7 Datum Transformation in Practice
1. Two operations that are constantly confused
| Datum transformation | Projection change | |
|---|---|---|
| What changes | The reference frame - the ellipsoid and its position and orientation in space | The mapping of a fixed frame onto a plane |
| Dimensions | Three-dimensional | Two-dimensional |
| Example | Minna (Clarke 1880 (RGS)) to WGS 84 | NTM Mid Belt to NTM West Belt; or NTM to UTM on the same datum |
| Needs parameters? | Yes - and they are empirical, area-dependent and uncertain | No - the formulae are exact given the projection constants |
| Typical magnitude in Nigeria | Of the order of hundreds of metres | Exact; no positional error introduced |
The practical rule: a projection change is arithmetic; a datum transformation is a measurement. The first is exact. The second carries an accuracy figure, and that figure belongs in the record.
2. Geographic to geocentric Cartesian
All three-dimensional transformations operate on Cartesian coordinates, so the first step is always to convert $(\phi, \lambda, h)$ on the source ellipsoid to $(X, Y, Z)$:
X &= (\nu + h)\cos\phi\cos\lambda \\ Y &= (\nu + h)\cos\phi\sin\lambda \\ Z &= \left[\nu(1-e^2) + h\right]\sin\phi \end{aligned}$$ where $\nu = \dfrac{a}{\sqrt{1 - e^2\sin^2\phi}}$ is the radius of curvature in the prime vertical, and $e^2 = 2f - f^2$. For **Clarke 1880 (RGS)**: $a = 6{,}378{,}249.145$ m, $1/f = 293.465$, giving $b = 6{,}356{,}514.87$ m and $e^2 \approx 0.00680348$. For **WGS 84**: $a = 6{,}378{,}137.0$ m, $1/f = 298.257223563$, giving $e^2 \approx 0.00669438$. Note the requirement for $h$ - **ellipsoidal** height, not orthometric. Feeding a levelled height into this equation introduces an error equal to the geoid undulation, which over Nigeria is of the order of tens of metres. --- ## 3. The parameter models ### Three-parameter geocentric translation $$\begin{bmatrix} X \\ Y \\ Z \end{bmatrix}_{target} = \begin{bmatrix} X \\ Y \\ Z \end{bmatrix}_{source} + \begin{bmatrix} \Delta X \\ \Delta Y \\ \Delta Z \end{bmatrix}$$ This models **only** the offset between the two ellipsoid centres. It assumes the axes are parallel and the scales identical - neither of which is true of a classical triangulation datum. It is simple, robust and coarse. ### Seven-parameter Helmert (Bursa-Wolf) $$\begin{bmatrix} X \\ Y \\ Z \end{bmatrix}_{target} = \begin{bmatrix} \Delta X \\ \Delta Y \\ \Delta Z \end{bmatrix} + (1 + s)\,\mathbf{R}\begin{bmatrix} X \\ Y \\ Z \end{bmatrix}_{source}$$ with three translations, three small rotations $(r_X, r_Y, r_Z)$ in the rotation matrix $\mathbf{R}$, and a scale difference $s$ expressed in parts per million. For small rotations, $$\mathbf{R} \approx \begin{bmatrix} 1 & r_Z & -r_Y \\ -r_Z & 1 & r_X \\ r_Y & -r_X & 1 \end{bmatrix}$$ **Watch the rotation sign convention.** The *position vector* convention and the *coordinate frame* convention differ by the sign of the rotations, and software packages disagree about which they implement. A seven-parameter set applied with the wrong convention produces errors of several metres that look like poor data rather than a blunder. Always record which convention a parameter set assumes. ### Molodensky-Badekas (ten-parameter) Identical in form to Bursa-Wolf except that the rotations are applied about the **centroid of the local network** rather than about the geocentre: $$\begin{bmatrix} X \\ Y \\ Z \end{bmatrix}_{target} = \begin{bmatrix} X_m \\ Y_m \\ Z_m \end{bmatrix} + \begin{bmatrix} \Delta X \\ \Delta Y \\ \Delta Z \end{bmatrix} + (1+s)\,\mathbf{R}\left(\begin{bmatrix} X \\ Y \\ Z \end{bmatrix}_{source} - \begin{bmatrix} X_m \\ Y_m \\ Z_m \end{bmatrix}\right)$$ Why it matters: when rotations are taken about a geocentre thousands of kilometres from the data, a small rotation and a large translation produce almost the same effect on the observations. The normal matrix becomes badly conditioned and the estimated parameters are strongly correlated - individually meaningless even though the fit is good. Rotating about the network centroid decorrelates them. The price is that the three centroid coordinates must be carried with the parameter set; quoting Molodensky-Badekas translations without the centroid is useless. ### Abridged Molodensky A closed-form approximation that produces $\Delta\phi$, $\Delta\lambda$, $\Delta h$ directly from the three translations and the differences in $a$ and $f$, without going through Cartesian coordinates. Convenient for hand computation; equivalent in accuracy to the three-parameter model it is built from. --- ## 4. Why there is no single "Nigeria" parameter set Candidates often ask for "the" Minna to WGS 84 parameters. There is no single authoritative set, and understanding why is the professional point. The EPSG registry contains **several distinct Minna to WGS 84 transformations**, each with its own **area of validity**, **derivation** and **stated accuracy**. Examples of the kind of entry recorded include a national-scope transformation derived at six stations from US Defense Mapping Agency TR8350.2 (September 1987) with stated accuracies of the order of a few metres per axis; **Minna to WGS 84 (3)** (EPSG:1754), derived at eight stations across the Niger delta and used by Shell SPDC throughout southern Nigeria onshore; and **Minna to WGS 84 (15)** (EPSG:15493), scoped to Nigeria between 4 deg N and 5 deg N and 6 deg E and 8 deg E, adopted by a joint-venture operator from 1 January 1996. The reason for the proliferation is the datum itself. Minna is a classical datum oriented at a single station, **L40**, where the deflection of the vertical and the geoid undulation were assumed to be zero, and its primary triangulation carries scale and orientation defects. A datum with internal distortion cannot be tied to a geocentric frame by one rigid transformation. Each parameter set is a **best fit over its own area**, and applying a delta-region set in the far north, or a national set to a delta project, degrades the result. **The professional conclusions:** 1. **Select the transformation whose area of validity covers the project.** Do not accept a software default. 2. **Record which transformation was used**, by EPSG code where one exists, in the job file and in the plan's description of the origin of coordinates (regulation 31(d)). 3. **Carry the accuracy figure.** A transformation with a stated 3 m to 6 m accuracy caps the accuracy of any coordinate that passes through it, no matter how well the field observations closed. 4. **Prefer a local fit where control exists.** Where several points have coordinates in both frames across the project area, a locally estimated transformation will outperform any published set. 5. **State the epoch when working in ITRF.** NGD2012 is realised in **ITRF2008 at epoch 2012.0**; ITRF coordinates carry velocities, and a modern frame without an epoch is incompletely specified. --- ## 5. A defensible workflow ``` 1. Observe in the modern frame (GNSS, NIGNET/CORS). Record: frame, realisation, epoch. 2. Convert geographic -> geocentric Cartesian on WGS 84. Use ELLIPSOIDAL height h, never orthometric H. 3. Apply the SELECTED transformation to Minna. Record: EPSG code, parameter values, convention, stated accuracy. 4. Convert geocentric Cartesian -> geographic on Clarke 1880 (RGS). 5. Project into the correct NTM belt. West / Mid / East by LONGITUDE; k0 = 0.99975; lat of origin 4 deg N. 6. CHECK against existing Government control. Any residual beyond the transformation's stated accuracy is a blunder to be found, not noise to be averaged away. 7. State on the plan (reg. 31(b), (d)): grid, north used, and the description of the origin of coordinates. ``` Step 6 is the one that separates a professional job from a plausible one. A transformation always produces an answer; only a check tells you whether it produced the right one.Loading diagram...
Test Your Knowledge
Why does the EPSG registry contain several different Minna to WGS 84 transformations rather than one national set?
A
B
C
D
Test Your Knowledge
What distinguishes the Molodensky-Badekas model from the Bursa-Wolf seven-parameter model?
A
B
C
D
Test Your Knowledge
A surveyor converts GNSS positions to Minna using published parameters with a stated accuracy of 3 to 6 metres, then reports cadastral coordinates to the nearest centimetre. What is wrong?
A
B
C
D
Test Your Knowledge
In the geographic-to-geocentric conversion used at the start of every three-dimensional transformation, which height must be used?
A
B
C
D