How to Georeference a Historical Map in QGIS

Georeferencing gives a scanned map a modern location

To georeference a historical map in QGIS, load a modern reference layer, open the scanned map in Georeferencer, place matching ground control points on features visible in both maps, choose a target coordinate reference system and suitable transformation, inspect residuals, then export and verify the aligned raster. Keep the untouched scan and save the control points so the result can be reviewed or rebuilt.

Georeferencing does not make an old map “correct.” It creates a mathematical relationship between image pixels and geographic coordinates. Historical surveying, paper stretch, scanning, cartographic choices, and changed landscapes remain part of the source. The aim is a documented alignment good enough for the question—not cosmetic perfection achieved by kneading history like pizza dough.

What you need before opening QGIS

Gather four things:

1. A high-quality scan. Use the least compressed image available. Keep an untouched archival copy.
2. A reliable reference layer. This might be current official mapping, an orthophoto, or another georeferenced source appropriate to the project.
3. A target coordinate reference system (CRS). Select one suited to the project's location and intended measurements.
4. Features visible in both sources. Stable road intersections, building corners, bridge abutments, monuments, or parcel corners can become ground control points.

Confirm the scan's usage rights before publishing it or a derivative. A map's age does not by itself settle copyright in every jurisdiction, and a repository's download page may carry an item-specific rights statement.

Use the current QGIS Georeferencer documentation alongside this workflow. Interface wording can change between releases; the official manual documents the installed generation more precisely than a screenshot preserved in an elderly tutorial.

Step 1: Inspect the historical map as a source

Read the title, date, publisher, survey notes, scale, orientation, legend, projection information, marginal grids, and edition. Look for folds, tears, scan rotation, missing edges, or photographs taken at an angle.

Ask what kind of map it is. A surveyed topographic sheet, a fire-insurance plan, a tourist map, and a hand-drawn memory map carry different geometric expectations. Decorative maps may communicate relationships without maintaining consistent distance or direction. No transformation can recover precision the original never claimed.

Record the map's identifier and repository. If the date is uncertain, write “circa” or a date range rather than upgrading a catalog estimate into a birthday.

Step 2: Set up a clean QGIS project

Open QGIS and create a new project. Add the modern reference layer to the main map canvas. Set the project CRS deliberately rather than accepting an unfamiliar default.

A CRS defines how coordinates relate to the Earth. A geographic CRS uses angular coordinates; a projected CRS represents the curved Earth on a flat plane and is often more convenient for local distance and area. Choose a projected CRS commonly used for your region when accurate local measurements matter. If uncertain, consult the relevant mapping authority or GIS specialist.

Save the project in a folder containing subfolders such as:

• source for the untouched scan
• control-points for GCP files
• output for transformed rasters
• notes for provenance and decisions

Do not overwrite the source image. Reproducibility is much easier when “final_final2_reallyfinal.tif” is not the entire data-management plan.

Step 3: Open the scan in Georeferencer

Open QGIS's Georeferencer and load the historical raster. Depending on the QGIS release and interface configuration, the tool may appear in the Layer or Raster area or through search. Confirm the scan displays at the expected orientation.

In the Georeferencer window, the image initially has pixel coordinates rather than meaningful map coordinates. Your control points will connect selected pixels to known locations in the reference layer.

Step 4: Choose strong ground control points

A ground control point, or GCP, is a feature whose location can be identified in both the historical image and the georeferenced reference layer. Add a point on the old map, then use From map canvas to select the same place on the modern map, or enter known coordinates with the correct CRS.

Good GCPs are:

• Small enough to click consistently
• Stable across the two dates
• Clearly identifiable in both sources
• Distributed across the whole map
• Independent of one another

Avoid relying only on riverbanks, shorelines, vegetation edges, or road centers that may have moved. Be cautious with church symbols whose icon is offset from the building, labels placed for readability, and intersections rebuilt into roundabouts.

Place points near corners and edges as well as the center. A tight cluster can align one neighborhood beautifully while allowing the outer map to swing away like an unsecured gate.

Add more points than the mathematical minimum needed by the chosen transformation. Extra well-distributed points allow residual checking and make one mistaken match easier to detect. Quality beats quantity: twenty vague bends do not outrank eight stable intersections.

Step 5: Choose the transformation deliberately

Open Transformation Settings. QGIS offers several algorithms because maps have different geometry and distortion.

Linear positioning is appropriate when a good raster already has a known CRS and mainly lacks location and scale information. In current QGIS documentation, it creates a world file rather than transforming pixels and does not handle the range of distortions other methods do.

Helmert supports translation, rotation, and uniform scaling. It can suit a relatively accurate map or plan that needs those broad adjustments without local warping.

Polynomial transformations can handle increasing levels of global distortion. A first-order polynomial, often called affine, allows translation, rotation, scale change, and skew while keeping straight lines straight. Higher orders bend the image more and require more control information; use them only when the source distortion and project purpose justify that flexibility.

Projective transformation can help with a flat map photographed at an angle, where perspective distortion is the main problem.

Thin plate spline can fit local distortions closely. That is useful for warped paper or irregular sources, but it may create substantial local deformation and poor behavior outside the control-point network.

Do not select the algorithm solely because it produces the smallest displayed error. Choose the simplest transformation consistent with the source and test how it behaves on independent features.

Step 6: Set output, CRS, and resampling

Choose the target CRS used by your project. Select an output filename that includes the source identifier and a version, not merely oldmap.tif.

For most scanned maps, a resampling method such as nearest neighbor preserves original pixel values and hard edges, while bilinear or cubic methods may produce a visually smoother display by calculating new pixel values. A smooth image is not necessarily a more accurate one. Keep the original scan so any resampling remains reversible at the project level.

Enable options to save the GCP points and generate a report if available. The QGIS report can record transformation parameters, residuals, and the control-point list. Add your own notes about why points were selected or rejected; software cannot document that judgment for you.

Step 7: Read residuals without worshipping one number

After enough GCPs are present, QGIS displays residual information that shows disagreement between a point and the fitted transformation. A large residual can indicate:

• The wrong feature was matched
• The click landed imprecisely
• The feature moved over time
• The historical map is locally distorted
• The selected transformation is unsuitable

Zoom in and inspect any outlier. Do not delete a point merely to improve the summary number. First determine whether the point is wrong or whether it reveals genuine distortion. Record exclusions and reasons.

Residual units and interpretation depend on the CRS, output, transformation, and software display. Rather than announcing that a result is “accurate to X meters” from one residual statistic, validate against independent checkpoints not used to fit the model.

Step 8: Run the transformation and verify the output

Start georeferencing and load the output into the main project. Adjust transparency so the historical map can be compared with the modern reference.

Inspect:

• The center, corners, and edges
• Several independent known locations
• Roads or boundaries across their full length
• Areas between control points
• Places beyond the outermost points
• Text and symbols for unreasonable stretching

Misalignment that changes gradually may suggest the transformation is too rigid. Wavy or folded distortion may mean a flexible method is overfitting sparse points. A strong fit near every GCP but implausible shapes between them is not a victory.

If revising, return to the GCPs, change one decision at a time, rerun to a new output, and compare versions. Save screenshots or reports that show why the final version was selected.

Step 9: Document the result

Create a metadata note containing:

• Historical-map title, creator, date, edition, repository, and identifier
• Scan source, resolution if known, and rights statement
• QGIS version
• Reference layer and access date
• Project and target CRS
• Transformation and resampling methods
• GCP file and number of included or excluded points
• Validation method
• Known areas of poor fit
• Creator and processing date

This note is what turns an image that “looks lined up” into a research object another person can evaluate.

If the raster supports a larger community heritage map, publish the uncertainty in reader-facing language. For example: “Building locations were interpreted from a georeferenced 1890s plan; alignment is strongest near surviving street intersections and weaker along the altered riverfront.”

Common problems and fixes

The output appears in the wrong part of the world

Check the coordinate order, source assigned to entered coordinates, and target CRS. Longitude and latitude may have been reversed, or coordinates may have been entered under the wrong CRS.

The center aligns but edges do not

Spread GCPs closer to the edges, inspect the source for distortion, and test an appropriate transformation. Do not add fictional precision to features that genuinely moved.

The image is extremely warped

Review for a mismatched point first. Then consider whether a highly flexible transformation is overfitting. Compare with a simpler method and validate away from the GCPs.

Modern features no longer match

They may have changed. Use older stable landmarks, parcel corners, monuments, or grid intersections where defensible. Treat historical change as evidence, not automatically as GIS error.

The map has no reliable control points

Do not force it. Present the scan unaligned, use an approximate locator inset, or map interpreted features separately with explicit uncertainty. A decorative or memory map can remain valuable without wearing borrowed coordinates.

Georeferencing is interpretation

Every GCP says, “I believe these two marks represent the same place.” Every transformation says what kinds of distortion you permit. Those are historical and cartographic judgments as much as technical settings.

Where a map includes community-held or sensitive places, geometric accuracy does not automatically authorize public precision. Follow the project's ethical mapping rules and generalize or restrict locations when needed. A map can be technically exact and ethically lost.

The best result therefore has three qualities: it fits the evidence, it explains its method, and it remains modest about what the old map can prove.