Georeferencing

Spatial reference is what a coordinate means. Georeferencing is how a picture acquires coordinates. Under fit-for-purpose, every boundary you own traces back to this step — so its errors are your parcels' errors.

Two different words

Term	Means
Spatial reference	The CRS a dataset's coordinates are expressed in. A property. Lesson 0008.
Georeferencing	The act of assigning real-world coordinates to something that had none — a scanned cadastral sheet, a drone frame, a raw satellite image.
Orthorectification	Georeferencing plus removing terrain and camera-tilt displacement, using a terrain model. The pixels themselves get resampled.

The third row is the one people conflate with the second, and the distinction is not pedantic: in orthorectification, cell resampling is performed on the raster and the result may have a different coordinate system and different dimensions from the input.¹ A merely georeferenced aerial image still leans — tall objects and hillsides displace outward from the centre of the frame, and boundaries traced from it inherit that lean.

Why an FFP project cares

Lesson 0002 established that boundaries come from imagery rather than field survey. That makes the imagery your measuring instrument. A non-orthorectified image in hilly terrain can displace a boundary by more than the accuracy class you promised the donor — and unlike a datum error, it varies across the frame, so it cannot be corrected by a constant shift afterwards.

How georeferencing actually works

Identify ground control points (GCPs) Features visible in both the image and a source of truth: road junctions, building corners, survey marks. Spread them across the frame, not clustered in the middle.
Choose a transformation order Zero-order shift needs 1 point; first-order affine needs 3; second-order 6; third-order 10 non-correlated points.² Affine shifts, scales and rotates the raster.
Add more points than the minimum More than three introduces residuals distributed across all points — but you want that, because with exactly three, one bad control point silently warps everything.²
Read the RMS error The residual between where your control points landed and where they should be. It is a measure of internal consistency — not of truth.
Write the result Either into the file's own header, or beside it as a world file: for a raw image with an affine transformation, the parameters are written to a world file.¹

The RMS trap

A low RMS does not mean the image is correctly placed. It means your control points are mutually consistent with the transformation you chose. Pile on a high-order polynomial and RMS drops beautifully while the image between the points warps like a sheet in the wind — lower-order polynomials tend to produce random-type error, higher-order ones extrapolation error.² Prefer the lowest order that fits, and keep a few points out of the fit to check against.

Where this lands in a land project

Scanned cadastral sheets. Historic index maps, georeferenced to be traced or compared. They carry the survey conventions and errors of their era — plus paper shrinkage, which is why a first-order affine often will not fit an old sheet.

Aerial and satellite imagery. The base for FFP capture. Ask the supplier three questions: is it orthorectified, against what terrain model, and what positional accuracy is claimed. "50 cm imagery" describes pixel size, not accuracy — a 50 cm pixel can still sit three metres off the ground truth.

Drone flights. Increasingly common for community mapping. Without ground control they can be internally beautiful and absolutely displaced; the GCP survey is the part that makes them admissible.

Every geometry in your spatial_unit table (lesson 0007) records capture_method and accuracy_class. This lesson is where those values come from — and why "traced from imagery" is meaningless unless you also record which imagery, at what accuracy, georeferenced how.

Drill 1 — Is this good enough?

For each situation: acceptable for FFP boundary capture, or a problem you must raise?

Drill 2 — Recall

Your primary source

Esri, Understanding Raster Georeferencing (ArcUser) — short, visual, and it covers control point placement, transformation order and RMS honestly:
esri.com — Understanding Raster Georeferencing (PDF)

Then the reference version: Fundamentals of georeferencing a raster dataset.

Ask your teacher. Good follow-ups: "what should the imagery clause in a procurement spec say?", "how do I georeference historic sheets that will not fit?", "how do I record imagery provenance per parcel?"