When a GNSS displacement curve rises suddenly, the first question should not be “did the site move?” It should be: where did this value come from, and what level of conclusion can it support? A period may contain a usable result, a limited result, a processing failure or missing data. Similar-looking steps may come from physical deformation, the reference, an equipment event or a solution anomaly.
Reliable interpretation neither labels every anomaly as error nor treats a favorable quality status as proof of deformation. A stronger approach checks the data state, multiple quality indicators, reference stability, time-and-space patterns and independent observations in a fixed order, then states only what the available information supports.
This guide provides the interpretation sequence for one value or a segment of data. To turn that sequence into project clauses, see How Should GNSS Deformation Monitoring Results Be Accepted?.
The short answer: ask five questions before interpreting a displacement
Distinguish three levels of conclusion
| Level of conclusion | Defensible wording | What is still missing |
|---|---|---|
| Usable data | This period has the observation and quality evidence required by the project | It has not yet shown that a change is physical displacement |
| Credible change signal | The change is consistent across time or direction, with no evident data or reference problem | A consequential decision may still require field or independent confirmation |
| Confirmed physical deformation | GNSS agrees with a field event, control survey or other independent evidence | Continue tracking extent, rate and mechanism |
Conversely, “insufficient evidence” does not mean “no deformation.” It means the current data cannot responsibly support either an affirmative or negative conclusion.
1. Establish the data state before guessing from line shape
Before interpreting a plot, determine whether the point is supported by observations. Using the same line across a missing period creates false visual continuity and may imply evidence that does not exist.
You should be able to determine:
- whether the period is usable, limited, failed or missing;
- whether all source observations have arrived and whether the result may be reprocessed;
- whether the plot uses the same time and coordinate frame as the report;
- whether equipment, antenna, monument or processing version changed.
The JCGM vocabulary describes a measurement result not as an isolated number, but as values attributed to a measurand together with other relevant information; it is generally accompanied by uncertainty information [1].
2. Use multiple quality checks—not a single status indicator
Across repeated reviews of de-identified multi-station measurements, RST has found periods where one quality indicator looked favorable while other evidence showed a marginal or unstable result. The lesson is not that quality indicators are useless. Each one sees only one side of the problem.
| Evidence | Main question it answers | Evidence to combine with it |
|---|---|---|
| Solution state | Was a particular type of position solution formed? | Internal dispersion, amount of valid observation data and continuity |
| Dispersion within the observation period | Did this result converge or drift internally? | Site conditions, time scale and long-term repeatability |
| Amount of valid observation data | Was the result formed from sufficient information? | Data quality, systematic effects and reference stability |
| Relationship among components | In which direction is the change concentrated? | Expected structural motion, geometry and environment |
| Consistency across periods | Is the change instantaneous, persistent or recurring? | Gaps, equipment events and periodic environmental effects |
Project decisions use a validated combination of these checks, with the original quality states retained for review.
3. A quiet monitoring point does not prove a stable reference
A relative position describes how one point moves with respect to another. If the reference moves, monitoring points tied to it may all show a change in the opposite direction. If the reference moves with the monitored body, the result may partly cancel the actual deformation.
When a result is unusual, check at the same time:
- construction activity, accidental contact, maintenance or antenna replacement at both monitoring and reference points;
- changes in obstruction, reflectors, vegetation or snow;
- whether other points using the same reference show a similar signal;
- whether an independent control point, redundant reference or field record is available.
“Several points moved together” may indicate a common-reference problem or real regional deformation. Do not automatically filter common motion before physical and independent checks. IGS requirements for station records, equipment metadata and timely change updates exist so that coordinate changes can be interpreted against site events [2].
4. Plot shape can form a hypothesis, not identify a cause by itself
| Appearance | Possible explanations | Next useful check |
|---|---|---|
| One isolated jump | Temporary solution anomaly, incomplete data or a short physical event | Quality evidence, source observation and other sensors |
| Step followed by a new level | Physical step, monument or antenna event, reference change | Event records, components and other points |
| Recurrence at a similar time | Periodic environment or periodic forcing | Multi-day phase, direction and environmental record |
| Similar change at several points | Regional deformation, common reference or common environment | Independent reference and field information |
| Step at the edge of a gap | Different observation or quality states on either side of the outage | Check the observations and quality evidence on both sides |
| One dominant component | Structural motion or directional error | Expected movement direction and component quality |
GNSS coordinate time series can contain outliers, steps, seasonal signals and other periodic components. The MIDAS research by Blewitt and colleagues likewise shows why long-term velocity estimation needs methods resistant to outliers and undetected steps [3]. No trend estimator, however, replaces data-state and event records.
5. The consequence of a decision determines the evidence required
For routine observation, a credible change signal may first trigger a review. If the result could close a road, stop work, cause evacuation or drive another consequential engineering decision, stronger evidence is appropriate.
A useful escalation sequence is:
- preserve source data and its original quality state instead of overwriting it with post-processing;
- recheck the reference, equipment and site events;
- recompute or compare another period, component, nearby point or independent reference;
- obtain control survey, field inspection or another sensor;
- record “confirmed,” “pending confirmation” or “insufficient evidence” and the basis.
This is not a reason to delay safety action. It aligns response speed with evidence strength. A conservative precaution can be taken while the technical conclusion is still explicitly marked as under review.
From data interpretation to project acceptance
This guide provides a data-interpretation order: when a reader sees a GNSS displacement, it shows what to check first and how far the conclusion can go. The project document described by the acceptance guide turns that order into acceptance requirements for availability, acceptance periods, decision authority and delivery format.
That acceptance document records versioned quality thresholds, observation periods, weights, model coefficients and automated decision rules, preserving the link between site observations, delivered states and each later decision.
GNSS monitoring series
This article addresses interpretation of individual results; the adjacent articles define project acceptance and outage handling.
Previous: GNSS monitoring acceptance | Series 6 of 7 | Next: Near-real-time GNSS data gaps
References
- Joint Committee for Guides in Metrology. VIM3, 2.9 Measurement result and 2.26 Measurement uncertainty. https://jcgm.bipm.org/vim/en/2.9.html
- International GNSS Service. Station Operator Resources and Site Log Manager. https://www.igs.org/station-resources/
- Blewitt, G., et al. (2016). MIDAS robust trend estimator for accurate GPS station velocities without step detection. Journal of Geophysical Research: Solid Earth. https://doi.org/10.1002/2015JB012552
Frequently asked questions
Q: If only one point moves, must it be a false signal?
No. Local deformation can affect only one point. But a single point places more weight on checking its installation, reflection environment, quality evidence and site events; “the other points did not move” is not enough to reject it.
Q: Does a smooth plot mean good data quality?
Not necessarily. Smoothness may come from stable observations, processing or excessive averaging. Check data state and source evidence before judging the appearance.
Q: If several points change together, is regional deformation confirmed?
The evidence is stronger, but common reference, common processing and common environmental effects still need to be excluded. An independent reference or field evidence is needed before upgrading to a physical-deformation conclusion.
Q: Should low-quality data be deleted?
Not first. Preserve the value, state and reason so it can be reprocessed and audited. The displayed trend may exclude it, but must show the gap or limited state clearly.
Need help interpreting a GNSS displacement series or defining reviewable data states? Contact RST with a de-identified plot, event times and any available cross-measurement information.