Chapter 06

Georeferencing

Assigning coordinates to images. Learn how to transform scanned paper maps and historical aerial photos into accurate spatial data.

At a Glance

Prereqs: Chapters 01, 03, 04 Time: 25 min read + 25 min practice Deliverable: 1-page georef report

Learning outcomes

  • Place well-distributed ground control points (GCPs) and justify your point choices.
  • Explain what an affine transformation can and cannot correct.
  • Interpret RMSE and residuals and identify when a point should be moved or removed.

Key terms

ground control point (GCP), checkpoint, residual, RMSE, affine transformation, resampling, world file

Stop & check

  1. Why is it risky to place all GCPs in one small area of the image?

    Answer: The model fits well locally but can warp badly elsewhere.

    Why: Transformations interpolate error; clustered points provide weak constraints across the full extent.

    Common misconception: "More points in one corner increases accuracy"; distribution matters more than density in one spot.

  2. Your RMSE is high. What is the first thing you should check?

    Answer: Individual point residuals (and whether one point is a mis-click).

    Why: One bad point can dominate error and distort the fit.

    Common misconception: "Use a higher-order polynomial"; that can overfit and make the map look worse.

Try it (5 minutes)

  1. Use the simulator below to align the scanned map with translation and rotation.
  2. Write one sentence: what distortion remains that translation/rotation cannot fix?

Lab (Two Tracks)

Both tracks produce the same deliverable: a short report (1 page) documenting points, error, and decisions.

Desktop GIS Track (ArcGIS Pro / QGIS)

Georeference a scanned map (or historical aerial) using 6-10 GCPs. Report the final RMSE, list 3 strongest points, and include 1 screenshot of the overlay.

Remote Sensing Track (Google Earth Engine)

Load a basemap image and create 6-10 validation checkpoints (as point features). Use them as a checklist to verify a georeferenced overlay and document any systematic offset you observe.

Common mistakes

  • Picking blurry features: choose sharp, stable targets (corners, intersections), not vegetation edges.
  • Clustered points: spread points across the full image extent.
  • Chasing a low RMSE only: validate with checkpoints and visual inspection, not just the single number.

Further reading: https://gistbok-ltb.ucgis.org/

What is Georeferencing?

Georeferencing is the process of defining the internal coordinate system of a digital map or aerial photo so it can be associated with a real-world coordinate system. Effectively, you are "nailing" a flat image onto the curved surface of the Earth.

🔑 Key Goal: To take data that doesn't "know" where it is and give it a spatial context so it can be used alongside other GIS layers.

Ground Control Points (GCPs)

To georeference an image, we use Ground Control Points. These are locations on the scanned map that can be accurately identified on a modern, georeferenced base map (like street intersections or building corners).

  • 1 Point: Can move the map (Translation).
  • 2 Points: Can move and rotate the map.
  • 3+ Points: Can scale, warp, and skew the map (Affine Transformation).

Interactive: Georeferencing Simulator

Drag the sliders to align the historical "Scanned Map" (Yellow) perfectly with the modern "Ghost Map" (Dotted Outline). Watch the RMSE (Error) drop as you get closer!

Target
📜 Scanned Data

Current RMSE: HIGH (Unknown)

Measuring Error: RMS

No georeferenced map is perfectly aligned. We measure the "goodness of fit" using Root Mean Square (RMSE) Error. This value represents the average distance between where you placed a point and where the mathematical model says it should be.

Pro Tip: High RMSE error? Check your points! A single "mis-clicked" point can ruin the whole transformation. Professionals often look for the Residual of each point to find the culprits.
🛑 Critical GIS: Counter-Mapping

Georeferencing is not just for old insurance maps. Indigenous groups worldwide are digitizing and georeferencing their ancestors' hand-drawn sketch maps or oral history sites. By bringing these into a modern GIS coordinate system, they can legally prove historical land occupancy and challenge government "empty land" claims in court. This practice is known as Counter-Mapping.

Summary of Big Ideas

  • Georeferencing links historical or non-spatial images to Earth's coordinates.
  • GCPs are the link between the unknown (raster) and the known (basemap).
  • Transformation Models (1st, 2nd, 3rd Order) determine how the image is warped.
  • World Files (.tfw, .jgw) store the coordinates and scaling data for georeferenced images.

Chapter 06 Checkpoint

1. What is the minimum number of Ground Control Points (GCPs) required for an Affine (1st Order) transformation?

1
3

2. If your RMS error is very high, what is the most likely cause?

One or more GCPs were placed inaccurately.
You used too many points.

📚 Chapter Glossary

Residual The error distance between where a Ground Control Point (GCP) was placed and where the transformation model predicts it should be.
RMS Error (Root Mean Square) A statistical measure of the overall accuracy of a georeferencing transformation. Lower is generally better, but must be validated visually.
Affine Transformation A first-order transformation that can scale, rotate, and translate an image but maintains straight lines and parallel relationships.
← Chapter 05: Digitizing Vector Data Chapter 07: Database Management →

BoK Alignment

Topics in the UCGIS GIS&T Body of Knowledge that support this chapter.