Survey Quality Control Tools

Define the question before looking for a green result

A quality check needs a stated object, reference and decision rule. A file can be well formed yet use the wrong grid. A measured point can match a design identifier yet refer to another physical feature. A fitted model can have small residuals while predicting poorly beyond its controls. Begin by writing the question in ordinary language: can this table be read, did this object move, how far is this observation from design, or does this run close on known control?

Choose the corresponding tool after that question is clear. Do not combine unrelated success indicators into one general accuracy score. Preserve separate findings when only part of a check can be evaluated. A missing height may still permit a horizontal comparison, but it cannot support a vertical pass. A report that retains an unresolved condition is often more useful than a clean summary produced by discarding the difficult observations.

Check file structure before evaluating positions

Use the point file checker to identify invalid coordinates, missing required values, duplicate identifiers and layout mistakes. Confirm the header, delimiter, axis order and units against the export specification. Plotting the interpreted points can reveal a suspect row or a mirrored pattern, but the picture is only a clue. A structural check does not have the external control, observation history or uncertainty model required to certify the positions.

Duplicate review adds a separate set of questions. The same identifier can refer to inconsistent observations; the same location can legitimately carry several feature codes. Nearby points depend on the threshold supplied by the reviewer. Keep candidate groups and reasons for any action. When a filtered export is needed, save the unselected records as well so that omitted data remains accountable. Removing a warning from view must not remove the underlying source from the audit.

Compare confirmed design and measured points

For a design/as-built comparison, make sure both sources represent the same horizontal reference, height datum, units and intended revision. Establish correspondence before calculating differences. Repeated IDs require explicit resolution; an unmatched point must not disappear merely because it lacks a partner. If identifiers differ, confirm the paired records using reliable project evidence instead of selecting a convenient match solely by proximity.

The comparison uses measured minus design differences. Horizontal deviation is the magnitude formed from the Easting and Northing differences; height difference retains its sign. Supply separate horizontal and vertical tolerances from the applicable project requirement. A boundary equal to the stated tolerance is treated inclusively. Review the signed components as well as the magnitude, because a consistent direction across many points may suggest a reference problem that individual pass/fail labels would obscure.

Use a surface when no design point exists at the observation

A measured location often lies between the vertices of a design TIN. In that case a same-name point comparison is the wrong model. Select the design surface, verify its units and axes, and interpolate within its original supported triangle at the measured position. The surface comparison tool reports the resulting design height and the measured-minus-design difference. The original connectivity matters; reconstructing triangles from a vertex list can produce a different surface.

A point outside coverage, in a gap or in an unresolved overlap remains a distinct status. The nearest vertex height is not an acceptable substitute for interpolation unless a separate specified method actually calls for it. Keep coverage and exception counts alongside the tolerance summary. A high or low point can inform a construction decision, but a collection of height differences is not a computed earthwork volume and should not be labelled as one.

Preserve original closure before any adjustment

For a level run, the relevant checks include sight classification, back- and foresight totals, turning-point continuity and agreement with a known endpoint. Intermediate sights contribute elevations but do not belong in the route rise/fall sum. If an adjustment is justified, choose the supported setup or distance allocation explicitly and retain the original misclosure. An open run without endpoint control cannot demonstrate closure merely because a final elevation is available.

For a traverse, inspect directions, horizontal distances, coordinate increments and original positional closure. Bowditch distribution is an optional adjustment of those increments, not a replacement for evaluating the observed closure. The tool uses supplied final directions; it does not undertake a full angle-network adjustment or invent the survey's acceptance standard. Record whether you are reviewing an unadjusted result or an adjusted coordinate schedule so the two are not confused later.

Keep fitting controls and independent checks separate

A two-dimensional common-control fit estimates translation and rotation, with optional scale, from the controls assigned to the fit. Reserve suitable points as independent checks when possible. Review each residual, the spread of controls and whether application points lie beyond that spread. A small fitted residual alone can be misleading when there are very few controls, a poor spatial arrangement or a model that is unsuitable for the relationship between the two datasets.

Excluding a control requires a reason, and a parameter file should retain the source basis and model version. Do not repeatedly remove observations until only an attractive residual remains. The present model is equal weight and planar; it does not propagate a full control covariance model, apply a vertical transformation or certify a national datum operation. Heights and descriptions remain associated with their original records when plan coordinates are transformed.

Deliver evidence that another reviewer can follow

A useful review package includes the untouched source, interpreted column mapping, reference description, tool settings, exception list and complete numerical output. Distinguish an intentionally synthetic test from real project observations. When a report table is paginated, save the full export rather than a screenshot of its first page. Reconcile source, compared, unmatched and excluded counts so that every record has a documented destination.

For a suspect result, first reproduce it with a small non-sensitive example and an independent calculation. Identify the exact row, expected quantity and convention before reporting a bug. Formatting, rounding and unit labels can explain apparent discrepancies without changing the underlying result, but that possibility should be tested rather than assumed. These tools support a transparent review process; the responsible surveyor or project authority still determines the applicable controls, tolerances, acceptance rules and any corrective field action.

Questions and answers