Fit a site localization
Fit a shift, rotation and optional scale from your check shots on known control, and read the per-point residuals. Solved and reported, not applied.
On this page
Measured coordinates rarely match a site's published control exactly. The difference is usually systematic: a shift, a small rotation, sometimes a scale. Site localization fits that difference from your check shots on known points and shows how well each control point agrees with the fit.
Before you start
- A project with known coordinates to occupy: imported design points, or points collected as the Control Point type.
- At least one check shot against them, recorded with Tools → Check a known point. The Site localization row stays greyed out until the first check shot exists.
- A coordinate system set for the project, so the fit is solved on the published grid. Without one, it's solved on a local projection and says so.
Each check shot is one control pair: the known coordinate and the position you measured while standing on it. Stakeout records are never used. They describe where a stake landed, not the coordinate system.
Solve it
- Record check shots on as many control points as you can, spread around the site. See Check a known point.
- Go to Tools → Site localization. The row shows RMS 0.012 m once a localization is saved, or Not solved if you have check shots but no saved fit.
- The fit solves immediately. Read the headline, then the Control list.
- Adjust the fit: hold or drop points, change the options. Every change re-solves on the spot, and nothing is written yet.
- Tap Save to keep the fit, or Close to leave without saving.
To discard a saved fit, open the sheet and tap Remove saved localization.
Reading the screen
Headline
Residuals against held control shows the horizontal RMS in metres. Below it are the vertical RMS with the vertical model, and a line such as 4 points held · redundancy 5. Redundancy is how many more measurements you have than the fit needs. At zero or below, the residuals are zero by construction and prove nothing.
Control
One row per check shot, showing the control point's name, its horizontal residual, and ΔN, ΔE and ΔU. Residuals are known − fitted, so a positive ΔN means the known coordinate lies north of where the fit put your measurement.
Tap H or V to hold a point in the horizontal or vertical fit, or leave it out. A point left out still gets a residual and is labelled independent check. Because nothing was tuned to it, that residual is the best evidence the fit is right.
Parameters
| Row | What it shows |
|---|---|
| Rotation | Degrees, or not solved when the control can't support a rotation. |
| Scale | held at 1, or the solved scale in ppm. |
| Shift | East and north shift in metres. |
| Vertical | The vertical shift and the height basis (msl or ellipsoidal). |
| Vertical slope | East and north tilt in mm/km, with an inclined plane only. |
| Frame | The EPSG code of the grid it was solved on, or local TM. |
The footer gives the centre of the control and how far it extends from there. Inside that radius the fit interpolates. Beyond it, it extrapolates, and any rotation or scale error grows with distance.
Options
- Hold scale at 1 is on by default. Over a handful of points, a free scale absorbs fix noise as a ppm term that then multiplies out across the whole site. Turn it off only if you know the control is on a ground-distance grid, and check the result against a distance you know.
- Vertical can be None, Constant shift (the default) or Inclined plane. An inclined plane tilts to absorb a geoid slope. It needs at least 4 held heights; with fewer, it falls back to a constant shift and says why.
Heights are only compared like with like: orthometric with orthometric, or ellipsoidal with ellipsoidal. If no pair has comparable heights, there's no vertical model.
How much control you need
| Held points | What you get |
|---|---|
| 1 | A shift only. No rotation, no scale, nothing to check it against. |
| 2 or more | A rotation too, if the points are spread well beyond the fix noise (at least 1 m, and at least six times the receivers' average accuracy). Otherwise it stays a shift and says so. |
| 4 or more | The app also looks for a blunder: it refits without each point in turn to see whether one disagrees with the rest. |
Findings
Below the options, Findings list what to know before you trust the fit. Examples:
- … looks like a blunder. One point misses by far more than the rest agree to. This is usually a wrong coordinate on the control sheet, the wrong monument or the wrong mark. Exclude it and re-solve.
- No redundancy — the residuals prove nothing. Hold another point, or leave one out as an independent check.
- Control is effectively a straight line. The fit is barely constrained across the line. Add control off to one side.
- Residual RMS … is large for the fix quality. What's left isn't a shift, rotation or scale. Check the base coordinate, the antenna heights, and whether the control is on the datum you think it is.
- Scale departs … ppm from unity. Far more than a grid factor. Suspect metres mixed with US survey feet, or a blunder.
- Solved on a local Transverse Mercator. No coordinate system is set, so the parameters aren't a grid transformation a deliverable could name.
If nothing can be fitted, for example because every point is excluded, a Can't solve section explains why.
Where it's reported
The saved localization appears in the QA report, with horizontal and vertical RMS, rotation, scale and the control held.