How Survey Control Networks Maintain Accuracy

TL;DR: Survey control networks maintain accuracy through three mechanisms working together: built-in measurement redundancy that lets errors be detected and removed, least-squares adjustment that distributes residual error mathematically, and scheduled re-observation that catches the ground movement, marker damage, and datum drift that degrade any network over time. A control network is not a one-off deliverable — it is a living reference frame that holds millimetre accuracy only when it is checked, re-adjusted, and tied back to the national datum (GDA2020 and AHD) on a defined cycle.

Key takeaways

• Accuracy is held by redundancy, not single measurements — a braced network observed with multiple ties to every mark lets a least-squares adjustment find and reject a blunder; a simple open traverse cannot, so it silently propagates error into every set-out point downstream.
• Re-observation against a stable reference is what keeps a network "true." ISS recommends re-observing primary control quarterly on active mine sites and annually on stable infrastructure, comparing each mark against its adjusted history to detect movement above the network's noise floor (typically 1-3 mm horizontal for a Second Order network).
• The datum matters. GDA2020 is a plate-fixed datum, so marks tied to it stay coordinated as the Australian plate moves ~7 cm/year north-east; a network left on the old GDA94 frame drifts ~1.8 m relative to global positioning systems and must be transformed, not ignored.
• Marker stability sets the ceiling on accuracy. A Leica TS60 total station resolving 0.5" of angle is wasted if the monument it sits on heaves with the clay; deep-founded concrete pillars or driven steel star pickets to rock are what let the instrument's precision survive into the coordinates.
• Documented uncertainty is the proof of accuracy. Every maintained network should carry per-mark uncertainty estimates from its latest adjustment so users know whether a mark is fit for 1 mm alignment work or only ±50 mm earthworks set-out.

Why accuracy degrades — and why maintenance is the answer

A control network is often treated as permanent the moment the adjustment report is signed off. In practice, the instant the field crew demobilises, the network begins to drift. Understanding how survey control networks maintain accuracy starts with understanding what attacks it.

The dominant cause across Australian sites is ground movement. Reactive clay soils across much of NSW, SA and inland WA heave and shrink seasonally by 20-40 mm; mine blasting shifts marks within hundreds of metres of a bench; tailings dams and reclaimed ground consolidate for years; and tunnelling induces settlement troughs at the surface. None of this announces itself — a mark can move 15 mm and still look perfectly intact.

The second cause is physical damage. Construction traffic, excavation, earthmoving plant and simple vandalism remove or disturb survey marks routinely. On a busy Bowen Basin coal project or a Pilbara iron-ore expansion, losing 10-20% of working control over a campaign is normal, not exceptional.

The third, less visible cause is datum and reference drift. The Australian continent moves north-east at roughly 7 centimetres per year. GDA2020 accounts for this by fixing coordinates to the plate at epoch 2020.0, but networks established on the legacy GDA94 datum are now offset by about 1.8 metres from real-time GNSS — a discrepancy that quietly corrupts any work that mixes old and new control.

Key point: Maintenance is not optional housekeeping. A network that is never re-checked does not stay at its commissioned accuracy — it decays at a rate set by the site's soil, activity and the datum it sits on. The whole discipline of maintaining a control network is the disciplined detection and correction of that decay.

Mechanism 1: Redundancy and network geometry

The single most important reason a properly designed control network holds accuracy is that it measures everything more times than strictly necessary. This is redundancy, and it is what separates a survey-grade control framework from a string of points joined by single measurements.

Consider a simple open traverse: a chain of marks where each is measured only from the one before it. If one angle is read 10 seconds in error, that error flows into every subsequent point and there is no way to detect it — the geometry offers no contradiction. Now consider a braced network, where each control mark is observed from three or four others. The redundant observations create geometric constraints that must agree. When they don't, the disagreement (the misclosure) reveals that an error exists, and its pattern usually reveals where.

In practice ISS builds redundancy in through:

• Closed loops, not open chains. Every traverse returns to a known mark so its misclosure can be computed and checked against the order's tolerance (for example, ICSM SP1 Second Order angular misclosure of roughly 15"√n).
• Multiple GNSS baselines per mark. Static GNSS sessions are planned so each pillar is connected by independent baselines to at least three others, observed in separate sessions and even on separate days to break correlated atmospheric error.
• Inter-visibility for total-station ties. A Leica TS60 or Trimble S9 reads each connection in multiple faces and rounds, so a single misread cannot survive.

The geometric strength this creates is measured by the network's redundancy number (degrees of freedom). A network with 30 observations solving for 18 unknowns has 12 redundant observations — 12 independent chances to catch a problem. The more redundancy, the more reliably a single blunder can be isolated and the smaller the undetectable error that can hide in the network.

Mechanism 2: Least-squares adjustment

Redundancy creates contradictions; least-squares adjustment resolves them mathematically. This is the engine that turns a cloud of slightly inconsistent measurements into a single, defensible set of coordinates — and it is central to how survey control networks maintain accuracy.

When a network is over-determined (more observations than unknowns), no single set of coordinates satisfies every measurement exactly. Least-squares finds the coordinates that minimise the sum of the squared residuals — the most statistically probable position for every mark given all the evidence. Each observation is weighted by its expected precision: a 0.5 mm + 1 ppm EDM distance carries more weight than a 5 mm RTK position, so the adjustment trusts the better measurement more.

The adjustment does three things that directly protect accuracy:

1. Distributes random error fairly. Instead of dumping all misclosure onto the last mark (as a crude traverse adjustment does), it spreads residual error across the whole network in proportion to measurement quality.
2. Flags blunders statistically. Standardised residuals and data-snooping tests highlight observations that disagree with the network far more than their stated precision allows — the classic signature of a misidentified mark or a transposed reading.
3. Produces honest uncertainty. The output is not just coordinates but an error ellipse and a standard uncertainty for every mark. This is the number that tells a user whether a mark is good enough for kiln alignment (sub-millimetre) or only for bulk earthworks.

Software such as Leica Infinity, Trimble Business Centre or dedicated packages like Star*Net run these adjustments and report whether the network has passed its specified order. Crucially, the same adjustment is re-run every time the network is re-observed during maintenance — comparing the new adjusted coordinates against the historical record is precisely how movement is detected.

Mechanism 3: Re-observation and monitoring cycles

Redundancy and adjustment guarantee accuracy at the moment of measurement. To hold accuracy over time, the network must be re-observed and compared against itself — this is the maintenance loop that most clients underestimate.

A maintenance re-survey repeats a subset (or all) of the original observations using the same instruments and procedures, then re-adjusts the network. The new coordinates are compared mark-by-mark against the previous adjusted set. A mark that shifts by more than the combined uncertainty of the two surveys has genuinely moved; one that stays within noise is confirmed stable.

The discipline that makes this work is keeping the reference frame separate and stable. For deformation monitoring of a pit wall or dam, ISS establishes monitoring control on competent ground outside the zone of influence, so the reference marks themselves are not moving. If the reference moves with the structure, the survey reports zero deformation while the wall is failing — one of the most dangerous outcomes in the discipline.

⚠️ Watch out: Re-observing with a different instrument, prism, or procedure than the original survey introduces systematic differences that masquerade as movement. Maintenance surveys must replicate the original method — same reflector constant, same EDM, same observation scheme — so the only thing that changes between epochs is the ground.

How the national datum keeps the network honest

A maintained control network does not float in isolation. Wherever possible it is tied to Australia's national framework so it stays consistent with mapping, GNSS, neighbouring projects and statutory boundaries.

• GDA2020 (horizontal). Marks are connected to the datum through Geoscience Australia's permanent CORS network (AUSCORS / AUSPOS processing) or to nearby state survey marks. Because GDA2020 is fixed to the plate at epoch 2020.0, marks tied to it remain mutually consistent even as the continent drifts beneath global systems.
• MGA2020 (projected grid). Site coordinates are usually expressed in the Map Grid of Australia 2020, applying the correct zone, scale factor and grid convergence so distances on the grid reconcile with measured ground distances.
• AHD (vertical). Heights are referenced to the Australian Height Datum through connection to state benchmarks or, increasingly, the AUSGeoid2020 model applied to GNSS ellipsoidal heights.

Tying maintenance surveys back to CORS data is also a powerful independent check: if your local primary marks all shift by the same vector relative to the national frame, that points to a datum or processing issue rather than real ground movement. The national datum, re-checked at each epoch, is the external truth against which a network proves it has not quietly drifted.

Equipment and procedures that protect accuracy

The instruments and field discipline used in maintenance directly determine the accuracy that can be held.

Equally important is forced-centring and consistent targets. Pillar-mounted forced-centring removes the ±1-2 mm of setup error that otherwise reappears at every visit, which matters enormously when you are trying to detect 3 mm of real movement. Calibration certificates current within 12 months, identical prism constants between surveys, and weather records to correct EDM for temperature and pressure all close off systematic error that would otherwise pollute the comparison between epochs.

Cost of maintaining a control network

Maintenance is a fraction of establishment cost and a tiny fraction of the cost of a control failure.

The economics are decisive. Re-surveying a project because months of work was set out from a moved or undetected-bad mark routinely costs five to ten times the price of the control. On a mine or major build, an undetected control error that propagates into pit design, structural steel or a tie-in can cost six figures in rework — against a maintenance visit measured in low thousands.

Common mistakes that quietly destroy accuracy

Treating the network as permanent

The most common error is assuming that because marks are physically present, they are still accurate. Intact does not mean unmoved. Without scheduled re-observation, a network's real accuracy is unknown — and unknown accuracy is functionally no accuracy at all.

Avoid it: Set a re-observation interval matched to the site's activity and soil, and hold every mark to its adjusted history.

Relying on single-baseline RTK as control

Picking up a "control point" with a single RTK shot gives a position with no redundancy and no check. One bad fix, one mis-keyed antenna height, and the error is invisible.

Avoid it: Establish and maintain control with redundant static GNSS and total-station ties, then use RTK only for downstream set-out from that control.

Mixing datums and epochs

Combining GDA94 and GDA2020 coordinates, or marks from different adjustment epochs, introduces metre-level and millimetre-level inconsistencies respectively.

Avoid it: Record the datum, epoch and adjustment date for every mark, and transform — never just relabel — when bringing legacy control forward.

Frequently asked questions

How often should a control network be re-observed to maintain accuracy?

It depends on the site's activity and ground. ISS re-observes primary control quarterly on active mines, monthly on tailings landforms and during high-rise construction, weekly on tunnelling projects, and annually on stable permanent plant. The principle is simple: re-observe often enough that movement is caught before it propagates into downstream work.

What is least-squares adjustment and why does it keep a network accurate?

Least-squares is the mathematical method that resolves the small contradictions in an over-measured network by finding the most statistically probable coordinates for every mark, weighting each observation by its precision. It distributes random error fairly, flags blunders, and produces a defensible uncertainty for every point — and re-running it at each maintenance epoch is how movement is detected.

Does switching from GDA94 to GDA2020 affect my existing control?

Yes — significantly. GDA2020 coordinates differ from GDA94 by about 1.8 metres because the datum accounts for plate motion since 1994. Existing control must be formally transformed to GDA2020, not simply relabelled, and any new survey work mixing the two frames without transformation will carry that 1.8 m offset.

How do you tell real ground movement from measurement error?

By comparing the movement against the combined uncertainty of the two surveys. A shift larger than that combined uncertainty is statistically real; a smaller one is noise. This is only reliable when the maintenance survey replicates the original instruments and procedures, and when forced-centring removes setup error between epochs.

Can laser scanning or drones maintain a control network?

Not on their own — they consume control rather than create it. A FARO or Leica scanner and a CASA Part 101-compliant drone survey both register to, and are checked against, the precise control network. The high-accuracy backbone is still established and maintained by GNSS, total station and levelling; scanning and UAV data inherit their accuracy from that maintained framework.

Talk to us about maintaining your control

A control network only delivers the accuracy it was commissioned to provide if it is maintained — re-observed, re-adjusted, tied back to GDA2020 and AHD, and proven against its own history. Left unchecked, it decays silently and takes the accuracy of everything built on it down with it.

Industrial Spatial Solutions designs, establishes and maintains survey control networks to ICSM SP1 standards for mining, construction and industrial clients across Australia. We set re-observation schedules matched to your site, run rigorous least-squares adjustments, and give you documented per-mark uncertainty so you always know your control is fit for purpose. Call us on 0407 057 015 to discuss a maintenance programme or request a quote for your project.

Industrial Spatial Solutions — control established, accuracy maintained, decisions trusted.

Related reading: How to establish a survey control network, What is dimensional control, How to interpret deformation monitoring data