CRS Errors That Can Ruin a GIS Analysis

Coordinate reference system problems cause more quiet damage in GIS projects than almost anything else. A layer ends up in the wrong place, a buffer comes out the wrong size, an area total is off by a few percent, and the project carries on because nothing returned an error.

This post covers what a CRS is, why it matters and seven specific errors, each with a way to recognise it and a way to fix it.

What is a CRS?

A coordinate reference system (CRS) defines how coordinates relate to locations on the Earth's surface. A pair of numbers such as 28.6139, 77.2090 means little until you know which system it belongs to.

There are two main kinds.

• A geographic CRS uses latitude and longitude, which are angles measured in degrees. EPSG:4326 (WGS 84) is the one you will meet most often.
• A projected CRS converts the curved Earth to a flat plane and usually uses metres. EPSG:32643 (WGS 84 / UTM zone 43N) is an example.

An EPSG code is a standard number that identifies a particular CRS, so EPSG:4326 and EPSG:32643 are labels for two different systems. An EPSG code tells you which CRS a dataset uses. It does not tell you which one you should use for your analysis.

A datum sits underneath both. It defines the size and shape of the model of the Earth and how that model is positioned, and different datums can give slightly different coordinates for the same place.

Why a CRS matters

Every flat map distorts something: distance, area, shape or direction. Different projected systems are designed to keep different properties under control, over different parts of the world. A CRS that works well for measuring in one region can be poor in another, and a geographic CRS cannot give you metres at all.

When layers use different systems and are combined without care, they do not line up. When measurements are made in the wrong system, the numbers are wrong without looking wrong.

How a CRS error shows up

• A layer sits far from where it belongs, or near the coordinates 0, 0.
• Two layers of the same place do not overlap, or are offset by a fixed distance.
• A buffer or distance comes out the wrong size.
• A layer does not appear at all.
• A point lands in the wrong country or in the sea.

Seven CRS errors

1. A layer with no CRS

A dataset arrives with no CRS defined, so the software does not know what the coordinates mean. In QGIS it may show up in the wrong place, or the software may guess. For a shapefile, the .prj file next to it holds the CRS, and a missing .prj is a common cause.

How to spot it: the layer properties show no CRS (in QGIS, look under the Source tab), or the layer lands far from where it should.

The fix: find out what the CRS should be from the data provider or the metadata, then assign it. Do not guess.

2. Assigning a CRS when you should reproject

These two actions sound alike and do different things.

• Assigning (defining) a CRS changes the label only. The coordinates stay as they are.
• Reprojecting (transforming) recalculates every coordinate into a new CRS.

If a layer is in the wrong place and you assign a different CRS to "fix" it, you have only changed the label. The coordinates are the same and the layer is still wrong. If a layer has a missing label and you reproject it, the software has nothing to transform from.

How to spot it: a layer that is still in the wrong place after you changed its CRS.

The fix: assign only when the label is missing or wrong and you know what it should be. Reproject whenever you need the same data in a different CRS. In QGIS, the Reproject layer tool in the Processing Toolbox does this, as does exporting a layer with a new CRS.

3. Measuring in degrees

A geographic CRS uses angles. A difference of 0.001 degrees does not cover the same ground distance everywhere, so lengths, areas and buffers measured in EPSG:4326 are unreliable.

How to spot it: a buffer distance such as 0.05 in the tool settings, or a shape that looks stretched.

The fix: reproject to a projected CRS in metres, suited to your study area, before you measure.

4. The wrong UTM zone

UTM divides the world into 60 zones, each 6 degrees of longitude wide, numbered from 1 to 60 going east. A zone that does not cover your study area adds distortion that grows with distance from the zone's central line.

The zone number follows from the longitude: zone = floor((longitude + 180) / 6) + 1. Delhi, at about 77 degrees east, falls in zone 43. For WGS 84, northern zones use EPSG codes 32601 to 32660 and southern zones use 32701 to 32760, so zone 43N is EPSG:32643 and zone 43S is EPSG:32743.

How to spot it: the EPSG code does not match the location of your study area.

The fix: pick the zone from your study area. If your data crosses several zones, UTM is not a good choice, and another projected CRS may suit better.

5. Using EPSG:3857 for measurements

EPSG:3857 (WGS 84 / Pseudo-Mercator) is the projection behind most web maps. Its units are approximately metres, which makes it tempting for distance and area. Its scale distortion grows away from the equator, so measurements in it can be far off at higher latitudes.

How to spot it: distances or areas that look too large for a place far from the equator.

The fix: use EPSG:3857 for display and tiles. For measurements, choose a CRS suited to your region, such as the right UTM zone.

6. Mixing datums

Different datums can give slightly different coordinates for the same place, sometimes by several metres or more. Layers that look like they share a CRS can differ in datum, and a transformation between them may need a specific method.

How to spot it: a small, steady offset between two layers that are otherwise aligned.

The fix: check the full CRS definition, not only its name. When two datums are involved, let the software apply the appropriate transformation, and compare a few known points after reprojecting.

7. Swapped latitude and longitude

Coordinates are written as latitude, longitude by many people and as longitude, latitude (x, y) by many tools. GeoJSON and most GIS software use x, y, which means longitude first. A swap places a point somewhere else entirely. A point meant for Delhi (28.6139 N, 77.2090 E) with the values reversed lands far away.

How to spot it: a point that appears in the wrong country or in the sea.

The fix: check the axis order first, and confirm that the x column holds longitude and the y column holds latitude.

A practical example: a layer that will not line up

A project has a village boundary layer and a recent satellite image. Overlaid, the boundaries sit about 40 km off. This is how to diagnose it.

1. Read the CRS of the layer in QGIS, on the Source tab of the layer properties. Do the same for the image.
2. Check the project CRS in the bottom-right corner of the QGIS window.
3. Compare the coordinates. Values like 715980, 3167205 are projected and use metres. Values like 77.2, 28.6 are degrees.
4. Decide what is wrong. A missing or wrong label means assign. A correct label in a CRS you do not want means reproject.
5. Fix it once and save the result as a new file in a chosen CRS, so the next person does not repeat the diagnosis.

Reprojecting rasters

Rasters need a second decision when you reproject, which is how to compute the new pixel values. This is called resampling. Use nearest neighbour for categorical rasters such as land-cover classes, since an average of two class codes is meaningless. Bilinear or cubic resampling is usually better for continuous values such as elevation or temperature. In QGIS, the Warp (reproject) tool lets you choose the resampling method.

A short CRS workflow

1. Read the CRS of every layer before doing anything else.
2. Fix missing or wrong labels, but only when you know the correct CRS.
3. Pick one projected CRS suited to your study area.
4. Reproject all layers into it before measuring or overlaying.
5. Check that layers overlap where they should.
6. Save outputs in a stated CRS and record it in your notes.

Key takeaways

• A CRS defines how coordinates relate to places on Earth. Numbers mean little without one.
• Geographic CRSs use degrees. Projected CRSs usually use metres. Measure in the second kind.
• An EPSG code identifies a CRS. It does not choose one for you.
• Assigning changes the label, and reprojecting changes the coordinates.
• Pick the right UTM zone for your area, and use EPSG:3857 for display, not measurement.
• Check datums and axis order when points land somewhere unexpected.
• Choose the resampling method to suit the raster when you reproject.