Technical
Methodology, datums, and what each method cannot do.
This page exists because most surveying websites state an accuracy figure and nothing else. A number without a method behind it is decoration. Here is both, including the parts where each technique fails.
1 · Coordinates
Three different "heights", and why the difference is about fifty metres.
Most survey disputes that never happen would be avoided if everyone agreed which of these they were talking about. This is the single most common mistake we are asked to unpick.
WGS84 ellipsoidal
01
Raw GNSS output
Satellite observations are processed to the WGS84 ellipsoid. This is what the receiver computes first, and it is not yet a national grid coordinate.
OSGB36 / EPSG:27700
02
Horizontal coordinates you receive
Every easting and northing is transformed to the Ordnance Survey National Grid, with the Helmert transformation applied and residuals reported. This is the coordinate system UK design, setting-out and legal work is written in.
Orthometric height
03
Levels and finished levels
Heights are quoted orthometrically, referenced to mean sea level via the geoid model, not as raw ellipsoidal heights. This is the difference between a level that matches a drawing and one that is out by roughly fifty metres.
Undulation between the WGS84 ellipsoid and the geoid is not a constant. Across Norfolk and the South East it varies by tens of metres, and across our full coverage area by rather more, which is why quoting a raw GNSS height as a level above sea level produces a number that looks plausible and is wrong. Ask what the height is referenced to.
2 · Control
Nothing is measured until there is something proved to measure it from.
The network
Control is observed to a specification, not a vibe. Redundancy is built in so that a single blunder cannot hide inside a good-looking set of numbers. Occupied stations are used where they exist and can be trusted, rather than proliferating new marks that nobody else will ever find again.
The independent check
Survey-grade practice holds some points back. They are occupied before, then again after, and the difference is the error nobody optimised away. It is the only way an accuracy claim is worth more than the word of the person making it.
3 · Methods compared
Where each technique works, and where it quietly does not.
| Method | Good at | Fails at | Typical |
|---|---|---|---|
| Terrestrial laser scanning | Millimetre accuracy, geometry, interiors, vegetation | Clean reflective surfaces needed. Bare metal, water and glass scan poorly. Very long range degrades accuracy fast. | ±5mm on individual scans, ±10mm registered |
| UAV photogrammetry | Large open sites, quick coverage, design-grade surfaces | Vegetation, vertical or curved surfaces with no texture, and anywhere cameras cannot see each other from. Degrades badly over canopy. | ±30mm vertical with ground control |
| Terrestrial survey (total station and GNSS) | Hard detail, setting-out, levels, control, anything discrete | Slow across very large open areas, which is what the other two methods are for. | ±10mm horizontal, ±8mm vertical on control |
| GNSS precise point positioning | Control stations where a full network is overkill | Convergence time. Short sessions or recent orbits produce decimetre rather than centimetre results, and it will not see a session that ended. | Centimetre-level given a long session and final orbits |
| Drone-based ground-penetrating radar | Peat and fen depth profiling along a transect | Ground that will not return a signal, and peat so saturated it behaves as one continuous medium. No return, no interface, and we say so rather than drawing a plausible line between two points that did return. | ±100mm depth at the transect points themselves |
| Thermal imaging | Finding heat loss, failed insulation and moisture signatures | Absolute temperatures without a controlled reference surface. Results are meaningless without recorded weather and time of day. | 2°C or greater anomalies reliably detectable |
Note what is absent: no manufacturer or model appears anywhere on this page. The equipment is chosen to suit the requirement on each job, which is precisely why naming a specific unit on a website would misrepresent what you get.
4 · Quality control
What is actually checked, and when.
- Control is adjusted and residuals reported before any deliverable is drawn.
- Check points are left unmoved during the visit, so accuracy is demonstrated rather than claimed.
- If control does not meet the accuracy the job requires, it is re-observed. It does not get drawn around.
- Every deliverable is checked back against the raw observations, not just against itself.
- Point cloud registration is validated against check points, and the achieved figure is stated.
- Aerial work flies calibration targets and control, not just overlapping images.
- Limitations are written next to the claim they qualify, in the same document.
- Anything a method cannot see is stated as a limitation, never left implied.
Apply it to your site
Ask for the accuracy achieved, not the accuracy intended.
Any surveyor should be able to tell you what they measured against, what the residuals were, and what the method could not see. If they cannot, you now know the right question to ask.