A scanned paper map is just a picture: its pixels have no relationship to real-world coordinates. Georeferencing builds that relationship by matching points on the scan, such as road junctions, river bends, or survey markers, to the same features in a georeferenced basemap, then warping the image so it sits correctly over modern data. Once aligned, the old map can be overlaid with satellite imagery, digitised into vectors, and used to study how landscapes, towns, and rivers have changed over decades.

Good control points are the foundation. Choose features that are stable over time and easy to identify in both the scan and the reference data, such as church locations, old railway stations, or sharp river confluences, and spread them evenly across the whole map rather than clustering them in one corner. Avoid features that have moved, like coastlines or river channels, and avoid vague features such as forest edges, which shift with interpretation. Six to ten well-chosen points beat twenty sloppy ones.

Open the Georeferencer from the Raster menu, load your scanned image, and set the target coordinate reference system to match your project, usually the local UTM zone or the national grid. Click a control point on the scan, then click the same location on the map canvas or enter its coordinates manually, and repeat until you have points spread across the image. The Georeferencer shows a residual error for each point, which tells you how far the warped position misses the target; investigate any point with a much larger error than the rest, since it is probably misidentified.

For transformation, Thin Plate Spline is forgiving for old maps with local distortions, because it bends the image to fit all points exactly, but it can distort areas beyond your control points, so it suits maps where you only care about the covered area. For maps that are already fairly accurate, a first- or second-order polynomial preserves straight lines better and behaves more sensibly outside the point cloud. Choose the resampling method, usually cubic or nearest neighbour for scanned maps, set the output raster location, and run the transformation.

After georeferencing, load the result over a modern basemap and inspect it at several zoom levels, paying attention to edges and corners where errors grow. A quick check is to measure the distance between a few well-known features on the old map and their modern positions; errors of a few metres are excellent for a century-old map, while errors of hundreds of metres suggest a misidentified point or a wrong coordinate system. Record the residual errors and the transformation used in your project notes, because that documentation is what makes your work reproducible.

With the map aligned, you can digitise historical features as vectors, such as old field boundaries, roads, or settlements, and compare them with modern layers to quantify change. Historians use this to reconstruct lost landscapes, ecologists to find former wetlands, and planners to understand why a road follows a certain line. A georeferenced map is no longer a picture in an archive; it is a dataset, and that transformation is one of the most satisfying skills in practical GIS.

If your archive holds a series of maps from different decades, georeference them all to the same coordinate system and you can animate change over time, watching a town grow, a river shift, or a forest shrink in sequence. The comparison is most persuasive when each map carries its own accuracy note, because a viewer can then judge whether a small shift is real change or just the error of an old survey. Museums and local history groups often welcome such projects, and they may hold additional maps you have not seen, so sharing an early result can unlock more source material. What begins as a technical exercise in the Georeferencer can grow into a public history resource that outlasts the original assignment.