What Happens When a GIS File Has No CRS?

When a GIS file has no coordinate reference system, its geometry still contains numbers, but software cannot reliably know where those numbers belong on the Earth, which units they use or how they should be transformed. The layer may appear in the wrong place, remain invisible at the current extent or be interpreted as though it used the project's CRS.

Do not reproject the file until its original CRS has been identified. First recover and assign the correct definition; only then transform the coordinates if another CRS is needed.

Missing metadata does not mean the data was created without a CRS

Most real-world geographic coordinates were produced in some reference system, even when its definition was later lost, omitted or stored separately from the geometry.

A Shapefile may be missing its .prj sidecar. A CSV may contain eastings and northings without a data dictionary. A database export may preserve numeric geometry but drop the spatial reference identifier. In each case, the coordinates still reflect the source system; the file no longer carries enough evidence to interpret them.

There are exceptions: a digitised drawing, game map or image coordinate space may never have been tied to the Earth. Such data needs georeferencing from control points, not a guessed EPSG code.

What coordinate values can tell you

Ranges and patterns are useful clues:

x between −180 and 180 and y between −90 and 90 may indicate longitude and latitude in degrees;
values in the hundreds of thousands and millions may indicate a metre-based national grid or UTM CRS;
coordinates around ±20 million are common near the limits of Web Mercator;
small values with a local origin may indicate an engineering, image or site grid;
swapped latitude and longitude can place otherwise valid geographic coordinates in the wrong region.

None of these patterns identifies a CRS uniquely. Many UTM zones share the same easting and northing ranges, several datums can underlie the same projection, and a pair such as 500000, 5200000 can refer to many locations because every UTM zone resets its grid around a separate central meridian.

Coordinate values are evidence, not a self-describing CRS.

Why trial-and-error assignment is dangerous

Assigning a CRS tells software what existing numbers mean. It does not convert them. If the assigned definition is wrong, the layer becomes confidently mislabelled.

Trying several labels until the layer appears over the expected country may catch a gross error, but it cannot reliably distinguish closely related systems: two candidate datums may differ by tens of metres and still look identical on a small-scale basemap. Reprojecting the mislabelled layer then generates a new set of coordinates from a false source, making the error harder to trace.

Visual alignment is a verification step, not the primary evidence.

An evidence-led recovery process

1. Preserve the original

Work on a copy or create a new version. Do not overwrite the coordinate values or sidecars while testing candidates. Record the file checksum if the dataset is important enough to require a formal audit trail.

2. Look for provenance before inspecting geometry

Ask the producer, recover the export settings and search for a readme, data dictionary, project file or neighbouring layer from the same workflow. Check the source agency's standard CRSs and the period in which the data was produced.

This evidence is stronger than a coordinate-range guess because it can identify the datum, zone, axis order and epoch together.

3. Inspect coordinates and extent

Sample several records, not just the first point. Note minima, maxima, decimal precision and the expected geography. For multipart or global data, check whether the geometry crosses the antimeridian or uses a wrapped longitude convention such as 0–360°.

4. Build a short list of candidates

Use the geography, units, source organisation and coordinate ranges to eliminate impossible definitions. Check the official area of use for each remaining CRS. Do not treat the fact that a CRS covers the country as proof that the producer used it.

5. Test known locations

Assign each candidate in a non-destructive copy and compare several identifiable features with authoritative control data. Use points spread across the extent: a candidate can align near one location and diverge elsewhere.

6. State the confidence

If one definition is supported by provenance and control points, assign it and document the evidence. If several remain plausible, preserve the uncertainty. A file with an inferred CRS may be suitable for a rough illustration but not for cadastral, legal or precision overlay work.

Assign the source CRS before transforming

Once the original definition is established, attach it to the coordinates. The geometry should not move merely because its correct CRS was assigned; software now knows how to interpret the same numbers.

Next, overlay it with trusted reference data and examine residual differences at the scale appropriate to the task. Only after that validation should the layer be reprojected into a working or publication CRS.

This order matters:

unknown coordinates → identify source CRS → assign source CRS → validate → transform if needed

Skipping the identification step asks software to perform a transformation with an unknown starting point. No algorithm can recover information that has not been supplied.

File-format differences

Some formats make CRS loss more likely than others.

Shapefile: the CRS commonly lives in a separate .prj file; moving only .shp, .shx and .dbf loses it.
GeoPackage: the CRS definitions and layer identifiers are stored inside the database, making the package more self-contained.
GeoJSON: RFC 7946 uses WGS 84 longitude and latitude, but older or non-conforming exports may not follow that rule.
CSV: there is no universal geometry or CRS metadata unless a companion specification supplies it.
GeoTIFF: CRS tags can be embedded, though damaged or stripped metadata can still leave a raster unreferenced.

The format changes where you look for the evidence; it does not change the need to establish the source meaning.

When the responsible answer is “do not use it”

If the CRS cannot be identified with enough confidence, precise distance, area, overlay and boundary decisions are not defensible. The layer may still support exploratory work if its uncertainty is explicit and the positional tolerance is broad.

A sound workflow should not turn uncertainty into a confident-looking guess. Keep the original file, the candidate interpretation, the comparison evidence and the limitation together so that another person can review the judgement. A reusable spatial result should preserve uncertainty rather than laundering it into a clean-looking layer.

References

Assigning a CRS vs Reprojecting Data. Focused explanation of the operational distinction.
RFC 7946 — The GeoJSON Format. The current GeoJSON coordinate reference convention.
EPSG Geodetic Parameter Dataset. Authoritative CRS definitions and areas of use.