Georeferencing a scanned policies map for design work

Georeferencing a planning map is what turns a scanned policies map from a picture into something you can measure a distance on. Until it is georeferenced, a scan has no spatial reference at all: the pixels carry no coordinates, the scale printed on the sheet has been changed by the scanning, and the paper it came from has stretched unevenly with humidity. Measuring a setback from an unreferenced scan and putting the number in a design statement is how a scheme ends up two metres inside a boundary it was supposed to avoid.

The work is quick once you know the method. The part people skip is stating how accurate the result actually is.

Why a scanned sheet cannot be measured

Three separate errors stack up in any scan of a paper plan.

Paper distortion. Drawing sheets absorb moisture and stretch differently along and across the grain, so a sheet that was accurate when printed is no longer uniformly to scale.

Scanner geometry. A flatbed or roller scanner introduces skew, and a large sheet scanned in sections and stitched carries a discontinuity at every join.

Reproduction history. A policies map extract downloaded as a PDF may itself be a scan of a print of a plot at a different scale, with each generation adding error.

On top of that, many policies maps are diagrammatic in places. A settlement boundary drawn with a thick line at 1:10,000 covers several metres on the ground, and the line thickness is not a measurement. Where a boundary line's exact position changes the answer, no amount of georeferencing will resolve it, and the source data is the only honest route.

Choosing control points

Georeferencing works by matching identifiable features on the scan to the same features on a base with known coordinates. The quality of the result depends almost entirely on the points you choose.

Good control points share three qualities. They are sharp, so a road junction centre or a building corner rather than a field edge. They are stable, so a bridge parapet rather than a hedge line or a tree canopy. And they are unambiguous on both images, which rules out anything that has been redeveloped since the map was drawn.

Distribute them. Four points clustered in one corner will fit that corner beautifully and leave the opposite corner badly wrong. Aim for points near each edge of the useful area plus one or two in the middle.

A first-order transformation needs a minimum of three or four points and preserves straight lines, which suits a scan with simple skew. Higher-order polynomials fit local distortion better and can produce nonsense away from the control points, bending straight roads into curves. Start with the simplest transformation that gives an acceptable fit and only increase the order if the residuals justify it.

Avoid using the map's own grid ticks as your only control unless you are confident the grid was plotted accurately. It often was. It sometimes was not.

Working in British National Grid

English planning data is published in British National Grid, the projected coordinate system used for Ordnance Survey mapping, which is based on a datum specific to Great Britain. Web mapping and satellite imagery use a global system instead.

That difference matters more than it looks. Converting between the British datum and a global one is not a simple shift, because the historic survey the British grid rests on does not match a modern global reference exactly across the country. Ordnance Survey publishes a transformation for accurate conversion, and using a simple approximation instead introduces an error of the order of a metre or two, varying by location.

So do the work in British National Grid throughout. Bring the base data in as British National Grid, set the project coordinate system to it, and georeference the scan into it directly. Where you have to use satellite imagery as the base because nothing else covers the area, note that the imagery is in a different system and that any conversion contributes to the error.

One practical check. British National Grid eastings and northings for England are six-figure numbers in metres. If a coordinate readout shows degrees, or numbers in the millions, the project is in the wrong system.

Checking the fit across the whole sheet

The software will report a root mean square error in the units of the target system, so in metres. Read it, and then do not trust it on its own.

Root mean square error measures how well the transformation fits the control points you gave it. It says nothing about the areas between them. A scan can report a 1.2 metre error and be five metres out in a corner where no control point was placed.

So check visually, away from the control points. Overlay a known linear feature such as a railway, a motorway or a river and follow it across the whole extent. Check a building block in each quarter of the sheet. Where the fit degrades at one edge, either add control points there or restrict your use of the result to the area that fits.

The same principle applies to any mapped constraint whose exact line matters. The approach described in checking public rights of way on the definitive map is the same discipline: the statutory or source record governs, and the raster is an illustration of it.

Recording the accuracy you achieved

Write four lines into the project file and, where the result appears in a submission, into the drawing notes.

The source of the scan, with its title, the authority, the document date and where it was obtained. The base data used for control, with its source and date. The transformation type and order, and the number of control points. The reported root mean square error in metres, plus a note of any area where the visual fit was poorer.

That record does two things. It lets someone else reproduce the work, and it stops a measurement taken from a georeferenced raster being treated as a survey dimension by the next person who opens the file.

If the accuracy is not good enough for the decision you are making, say so rather than rounding it away. An honest statement that the boundary position is uncertain to within three metres is more useful to a client than a confident number that is wrong.

Presenting a georeferenced extract

Where a georeferenced extract goes into an application, label it plainly as a georeferenced extract of a named document, with the accuracy stated, and not as a survey.

Do not scale dimensions off it into a planning statement. Take dimensions from the measured survey and use the georeferenced raster to show the relationship between the site and the constraint. Officers accept that distinction readily and they notice when it has been blurred.

Check the licensing position for any base mapping you reproduce. Ordnance Survey mapping is licensed rather than freely reusable, and the terms that apply to a council's own published extract are not necessarily the terms that apply to your submission.

Where the extract supports a technical document, keep it consistent with the constraint mapping in that document. A Flood Risk Assessment that shows a flood zone boundary in one position and a georeferenced policies map extract that shows it in another gives a consultee an easy question to ask. The same applies to ecological layers, where the distinction between habitat inventories and statutory designations is set out in reading habitat inventories and classification data.

When to ask the council for the source data

Georeference the scan for design work and orientation. Ask for the source data whenever the answer changes the scheme.

The situations that always justify the request: whether the site is inside or outside a settlement boundary, whether a parcel falls within an allocation, whether the red line crosses a conservation area boundary, and whether land is within a designated area that triggers a different policy test.

Most authorities will supply the relevant layer as a spatial data file on request, and several publish their policies map as an interactive map with downloadable data. National designation datasets are also published centrally, which is often quicker than asking. Where you do ask, ask in writing and keep the reply, because a boundary position confirmed by the authority is worth considerably more in a negotiation than one you derived yourself.

And if the authority cannot tell you where its own boundary runs, that is useful information too. It usually means the point is arguable, which changes how you present the scheme.

The index of report types by category sets out which documents depend on accurate constraint mapping and what each is expected to contain.

This is general guidance, not legal advice. Check the position with your local planning authority or a qualified professional before relying on it. PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval.