What is georeferencing?
A scanned map shows roads, rivers and boundaries, but GIS does not know where on Earth it belongs. Georeferencing assigns geographic coordinates to the image or map so it can be positioned correctly in GIS. The introduction to GIS covers the steps in more detail.
How it works
You pick locations you can recognize both on the image and in a spatially referenced dataset. These are control points, such as:
- Road intersections
- Building corners
- Bridges
- Survey markers
A practical example
Say you have a scanned 1990 map of a village and want to compare it with current satellite imagery. You find features that still exist today, assign their coordinates, and georeference the old map.
Choosing a transformation and reading the error
Once you have control points, the software needs a rule for moving the image onto them. That rule is the transformation. Simple transformations need few points and keep the shape of the image. Flexible ones fit more points and can bend the image.
| Transformation | What it does | Fewest points |
|---|---|---|
| Linear | Shifts and scales the image, without rotating it | 2 |
| Helmert | Shifts, rotates and scales evenly, so shapes are kept | 2 |
| Polynomial 1 | Also allows shear, so straight lines stay straight | 3 |
| Polynomial 2 and 3 | Curved fits that can correct bending or warping | 6 and 10 |
| Projective | Corrects the perspective of a flat object photographed at an angle | 4 |
| Thin plate spline | Stretches the image locally so it passes through every point | Many, well spread |
Start with the simplest transformation that fits. A higher order can match the control points closely and still distort the image between them. The note on transformations in GIS covers the other kinds of transformation and resampling.
Residuals and RMSE
A residual is the distance between where the transformation puts a control point and where that point should be. Every point has one. The root mean square error, or RMSE, combines them: square each residual, take the average, then the square root. It is given in map units or in pixels.
Look at the residuals one by one. A point whose residual is much larger than the others is usually a mistake, such as the wrong feature picked on one of the images or a typing error in a coordinate. Check it, fix it or remove it, then calculate again. For image-to-image work, aim for an error of about one pixel of the reference image or less. For a scanned map, the acceptable error depends on the scale of the map and on what you will use the result for.
A low RMSE does not prove that the whole image is right. A thin plate spline passes through every point, so its residuals are zero by design. A high-order polynomial can show small residuals at the points and large errors between them. Spread the control points over the whole image, including the corners, and check the result against points that were not used in the fit.
What is digitization?
A paper map holds its information visually, while GIS needs it in a structured digital format. Digitization converts geographic features from an image, map or other source into digital GIS features.
What you can digitize
| Feature type | Examples |
|---|---|
| Points | Wells, trees, utility poles |
| Lines | Roads, rivers, pipelines |
| Polygons | Buildings, land parcels, lakes, agricultural fields |
A simple example
Load a scanned map of agricultural fields into QGIS, georeference it, and trace each field boundary by hand. Each traced field becomes a polygon, and you can add attributes such as field ID, crop type, area, owner and irrigation status.
Manual vs automated digitization
Digitization can be manual, semi-automated or automated, using image classification, machine learning, object detection or automated vectorization. For simple, high-accuracy boundaries, manual editing is still very common.
Georeferencing vs digitization
| Answers the question | |
|---|---|
| Georeferencing | Where does this image or map belong geographically? |
| Digitization | How do I turn the features on that map into GIS data? |
A common workflow is to georeference the old map, digitize its features, then analyze them.
Snapping and editing rules
When you digitize, your clicks are approximate. Snapping pulls the pointer to existing geometry within a set distance, so that a new feature meets an existing one exactly instead of just near it. You can snap to vertices, which are the corners of a feature, or to segments, which are any point along a line.
The snapping tolerance is the distance within which snapping happens, set in pixels or map units. A tolerance that is too small misses the point you aimed at. One that is too large grabs the wrong vertex. Zoom in before you click, and the tolerance in pixels covers less of the ground.
Topological editing goes a step further. When two features share a vertex and you move it, the neighbour moves with it, so shared boundaries stay shared. Together, these settings prevent the errors described in the next section: gaps between polygons, overlaps, and lines that stop short of a junction or run past it.
Good habits: edit a copy of the data, save edits regularly, and run a validity or topology check when you finish.
What is topology in GIS?
Topology describes the spatial relationships between geographic features. Two polygons share a boundary, a road crosses a river, a parcel touches another parcel. These look simple on a map, but GIS has to represent them correctly.
Why topology matters
In a parcel map, Parcel A should meet Parcel B exactly. If the two were digitized separately, you can end up with a gap, an overlap, a dangling line or an incorrectly connected feature.
Common topology errors
- Polygon overlaps
- Gaps
- Duplicate geometry
- Dangling lines
Topology is about relationships such as "Parcel A touches Parcel B", "Road A intersects Road B" and "River A flows through Polygon B". Getting them right keeps the data faithful to the real-world relationships you are modeling.
Popular tools
Both tools cover the whole workflow in this lesson.
| Name | What it is | Good for |
|---|---|---|
| QGIS | Free, open-source desktop GIS | Georeferencing scans, digitizing features and checking topology |
| ArcGIS Pro | Commercial GIS software from Esri | The same tasks, plus topology rules in a geodatabase |
Glossary
| Term | Meaning |
|---|---|
| Georeferencing | Giving an image or map real-world coordinates |
| Control point | A feature identifiable on both the image and a reference dataset |
| Digitization | Converting map or image features into digital GIS features |
| Vectorization | Converting raster features into vector geometry |
| Topology | The spatial relationships between features |
| Dangling line | A line that ends where it should connect to another |
Test yourself
Try to answer each question before opening the answer.
1. What are control points?
Locations you can recognize both on the image and in a spatially referenced dataset, such as road intersections, building corners, bridges and survey markers.
2. What is the difference between georeferencing and digitization?
Georeferencing places an image or map in the right location. Digitization turns the features on it into GIS data.
3. Which geometry type would you use for an agricultural field?
A polygon.
4. Name two common topology errors.
Any two of: polygon overlaps, gaps, duplicate geometry and dangling lines.
5. Why is manual digitization still common?
For simple, high-accuracy boundaries, tracing by hand is often the most reliable approach.
6. What attributes might you add to a traced field polygon?
For example field ID, crop type, area, owner and irrigation status.
Quick quiz
Pick an answer for each question to see if you got it right.