Georeferencing a wartime aerial photograph without ground control points

Anyone who has pulled a 1946 sortie print from the Historic England or NCAP catalogue knows the problem before they've even opened the file. There's no grid reference scribbled on the back, no surveyed control point, nothing but a grainy panchromatic frame and a best guess at the flight line. You need the image to sit correctly against a modern base map, and you need it to hold up to scrutiny if it ends up in a conservation area appraisal or a listed building setting assessment. Ground control points in the proper sense, points whose coordinates were fixed by survey at the time of the photograph, essentially don't exist for wartime material. So you work with what the landscape itself gives you.

Finding tie points the landscape still remembers

Georeferencing wartime prints means working with tie points instead: features visible in both the old print and the current base map that you're confident haven't moved. Church towers are the classic choice, assuming the church survived the Baedeker raids or simply wasn't rebuilt on a different footprint. Road junctions are usually reliable, though watch for realignments, especially anywhere a bypass or roundabout has gone in since. Field boundary corners work well in pastoral areas where the hedgerow or wall pattern has clearly survived from the period; a 1960s field amalgamation erases the original enclosure and takes those corners with it.

Churchyard boundaries, bridge abutments, and the corners of farm buildings with masonry walls tend to be the most stable of all. Avoid anything organic: tree canopies shift, pond edges migrate, and a field gate is not where it was in 1944. For a typical parish-scale area you want a dozen or more tie points spread across the frame, not clustered in one corner, because wartime sorties are rarely flown dead level and scale varies from the centre of the print to the edges.

Why the warp matters more than the points

Picking good tie points solves half the problem. The other half is the transformation you apply to them, and this is where a lot of informal georeferencing in GIS software goes wrong. A simple first-order (affine) fit assumes uniform scale and rotation across the whole image, which is a reasonable assumption for a modern orthophoto and a bad one for a 1940s oblique or near-vertical print scanned off a contact copy. Lens distortion, aircraft tilt, and relief displacement over anything but flat ground all mean the error grows as you move away from the principal point. A second-order polynomial or a proper rubber-sheet (spline) transformation, anchored on your tie points, handles this much better, provided you have enough points to constrain it. Too few, and the spline will happily produce confident-looking nonsense in the gaps between them.

Print shrinkage and fiducial mark damage add a further layer. Original negatives measured against their fiducial marks correct for this; a photocopy of a photocopy, which is what a lot of local archive holdings actually are, does not. If your source print has visible fiducial marks or a frame border, use them to check the image hasn't been non-uniformly stretched before you trust any tie point near the edges.

Once the warp is applied, check it against a feature you deliberately held back from the tie point set rather than one that built the fit. If a hedge line that should run straight across a parish boundary kinks at the seam between two sorties, the transformation needs more points nearer that seam, not a global tweak.

This is fiddly, archive-dependent work, and doing it properly for every sortie covering a conservation area review takes real time, time most heritage teams would rather spend on the assessment than on image rectification. Historic Landscape Change carries out this orthorectification against a recent high-res pass as a standing part of the service, putting a wartime sortie and a current image on the same base so the change between them becomes something you can measure.

If you've got a sortie reference and a conservation area boundary, that's enough to start seeing what's changed since the war.