A control survey establishes a network of monuments whose positions and elevations are known to a stated accuracy, so that every later measurement on the project can be referenced to the same framework. Control is not a deliverable anyone gets excited about, and it is the reason projects either fit together or do not.
The underlying idea is simple. Measurements contain error, and error accumulates. Rather than letting it accumulate freely from one setup to the next across an entire site, you build a small, strong, redundantly measured skeleton first, adjust it as a whole, and then hang detail work off it. Errors in the detail work stay local instead of propagating.
Horizontal and vertical are different problems
It is tempting to treat control as one job because a modern receiver reports a position and a height together. In practice the two components behave differently enough that they are usually designed, observed and assessed separately.
Horizontal control is established today mostly by satellite positioning, either in a static network processed as a whole or by real-time methods referenced to a network of reference stations. It can also be established conventionally, by traverse or by triangulation and trilateration, and conventional methods remain the practical choice under canopy, in urban canyons, and inside structures.
Vertical control is harder to get right by satellite. Satellite positioning yields height above an ellipsoid, a smooth mathematical surface that does not follow gravity. Elevation in the sense that governs drainage is orthometric height, referenced to a surface approximating mean sea level, and converting between them requires a geoid model. The geoid model is itself an approximation, so where centimetre-level or better vertical is required, differential levelling is still the method of choice.
H = h − NDesigning the network before observing it
Network design is the part most often skipped, and the part that determines whether the adjustment will succeed. Three questions decide it: what accuracy is needed, where must control physically be, and what redundancy will let errors be detected rather than absorbed.
- Monument placement: on stable ground, outside the construction footprint, with sky visibility for satellite work and intervisibility where conventional observations are needed. Concrete monuments or driven rods outlast rebar with a cap.
- Density: enough points that any detail work is within a comfortable observing distance of two of them, since a single nearby point offers no check.
- Connection to published control: tie to existing published monuments so the project sits in a national framework and can be related to adjoining work and to record data.
- Redundancy: every point observed from more than one place, and enough independent observations that the adjustment has degrees of freedom. A network with no redundancy cannot reveal a blunder; it simply distributes it.
- Geometry: avoid long thin chains and very acute intersections. Strong geometry makes the adjustment insensitive to small observational errors.
- Documentation: a description and a photograph for every monument, good enough that someone unfamiliar with the site can find it in five years.
Observing and adjusting
Observations are taken according to a procedure matched to the accuracy sought: multiple sets of directions and distances at conventional stations, repeated independent occupations with different satellite geometry for static work, and level runs observed in both directions with balanced sight lengths.
The observations are then adjusted, normally by least squares, which distributes residual error over the whole network in a way that is statistically defensible and, importantly, produces statistics that let you judge the result. A least squares adjustment reports the residual on each observation and an uncertainty for each coordinate. Those numbers are the point of doing it that way.
- Reduce the raw observations for instrument corrections, atmospheric conditions and, for conventional distances, the reduction to the mapping surface.
- Run a minimally constrained adjustment first, holding one point fixed. This tests the internal consistency of the observations without letting errors in published control contaminate the result.
- Inspect the residuals for outliers. A single large residual usually indicates a blunder such as a mis-entered instrument height or a misidentified point, not random error.
- Only when the network is internally clean, run the constrained adjustment holding the published control, and inspect how much the published points had to move.
- Report the accuracy achieved as uncertainty at a stated confidence level, together with what was held fixed and in which reference frame and epoch.
Expressing accuracy honestly
Two different things are commonly called accuracy and they answer different questions. Local, or relative, accuracy describes how well two nearby points agree with each other. Network accuracy describes how well a point agrees with the national reference frame. A project can have excellent relative accuracy and be shifted a foot as a whole, which is fine for internal consistency and fatal for connecting to a neighbouring project.
| Expression | Meaning | Example |
|---|---|---|
| Ratio of misclosure | Loop closure divided by loop length; a traditional traverse measure | 0.166 ft of misclosure on a 2847.65 ft perimeter is about 1:17,100 |
| Parts per million | Distance-dependent error component | 2 cm over 10 km is 2 ppm |
| Positional uncertainty at a confidence level | Statistical result from a least squares adjustment | Reported as a radius at 95 percent confidence |
| Levelling misclosure tolerance | Allowable loop misclosure scaled by the square root of loop length | With C = 8 mm, a 4 km loop allows 16 mm |
A ratio alone is a weak statement, because a loop can close beautifully while containing two compensating blunders. Redundant observations and a proper adjustment are what justify an accuracy claim.
Reference frames drift, so record the epoch
Positions in a modern geodetic reference frame are not permanent, because the ground moves. Tectonic motion, subsidence and post-glacial adjustment all change coordinates over time, at rates that are trivial for a fence line and significant for a long infrastructure corridor or a monitoring network.
For that reason, control results should always state the reference frame, the geoid model, and the epoch of the coordinates. Two surveys of the same monument in different frames or epochs will legitimately give different numbers, and without those labels a later user cannot tell whether the difference is motion, a change of frame, or an error.
Consult the National Geodetic Survey's published guidance for current frames, geoid models and recommended field procedures. That material is freely available and is the reference practitioners actually use.