Every GIS layer carries a coordinate reference system, whether anyone has looked at it or not. The CRS defines how a point on the curved surface of the Earth is written down as a pair of numbers. Get it right and layers from different sources line up, buffers are the size you asked for and area statistics are correct. Get it wrong and the errors are often silent: the map looks fine, but a 500 m buffer is really 420 m, or a land-cover total is off by several percent.

This guide explains the concepts you actually need, a practical way to choose a CRS, and the mistakes we see most often when we take over projects from other teams.

The building blocks: ellipsoid, datum and projection

Three ideas sit behind every CRS.

• Ellipsoid. A mathematical model of the Earth's shape, slightly flattened at the poles. For example, GRS 80 and WGS 84 are the two most common and differ by only a fraction of a millimetre.
• Datum. The ellipsoid plus a definition of where it sits relative to the real Earth. Older local datums, such as Everest 1830-based systems used across South Asia, were fitted to one country and can be offset from WGS 84 by hundreds of metres.
• Projection. The mathematical rule for flattening the curved surface onto a plane. However, every projection distorts something: area, shape, distance or direction. Therefore, you choose which distortion matters least for your purpose.

Geographic vs projected coordinate systems

A geographic CRS stores positions as latitude and longitude in degrees. WGS 84 (EPSG:4326) is the one most people use, because GPS receivers and most web services work in it. The problem for analysis is that a degree is not a fixed distance. A degree of latitude is about 111 km everywhere, but a degree of longitude is 111 km at the equator, about 95 km at Lahore and about 75 km at Zürich. As a result, measuring length or area in degrees gives meaningless results.

A projected CRS flattens the Earth and stores coordinates in metres (or feet). Distances, buffers and areas calculated in a suitable projected CRS are reliable, provided the study area is within the zone the projection was designed for.

The systems you will meet most often

Universal Transverse Mercator (UTM)

UTM divides the world into 60 zones, each 6° of longitude wide, numbered eastward from 180° W. Each zone has its own transverse Mercator projection with a false easting of 500,000 m and a scale factor of 0.9996 on the central meridian. Within a zone, however, scale distortion stays below about one part in a thousand, which is accurate enough for almost all planning and environmental work.

The limitation is the zone edge. Pakistan, for example, spans zones 41 to 44. In contrast, a national analysis forced into a single UTM zone will have growing distortion towards the far edges.

National grids

For this reason, many countries define their own projected system, tuned to their territory and linked to a national datum. Switzerland's LV95 and Great Britain's OSGB36 National Grid are familiar examples. Government, engineering and cadastral clients usually require the national grid, and it is often more accurate locally than UTM.

Why the CRS matters for analysis

• Buffers and distances. A 500 m buffer is only 500 m in a projected CRS measured in metres.
• Area. Similarly, land-cover totals and parcel areas need a low-distortion or equal-area projection. Web Mercator overstates area by about 38 percent at the latitude of Lahore and by more than double at the latitude of Zürich.
• Density analysis. Kernel density search radii are distances, so they inherit any distortion in the CRS.
• Network analysis. Likewise, route lengths, travel times and slope-based costs depend on consistent metric coordinates.
• Data integration. CAD drawings, surveys, GPS points and imagery only line up when their systems are known and transformed correctly.

How to choose a CRS for your project

1. Look at the size and shape of the study area

For a city, district or project site, use the local UTM zone or the national grid. For a whole country that crosses several UTM zones, use the national projection if one exists, or an equal-area projection centred on the country, such as a custom Albers or Lambert Azimuthal Equal Area.

2. Match the projection to the measurement that matters

Equal-area projections preserve area and are the right choice for land-cover statistics and density per square kilometre. Conformal projections, such as UTM and Lambert Conformal Conic, preserve local shape and angles, which suits engineering and navigation. In contrast, equidistant projections preserve distance from one point or along certain lines.

3. Follow client and national standards

If the client's engineers, surveyors or GIS department work in a specific system, deliver in that system. Agreeing this on day one avoids costly reprojection at the end, especially where CAD is involved.

4. Separate storage, analysis and display

A workflow we use often: keep source data in WGS 84 for exchange, run all analysis in the project's projected CRS, and publish to the web in Web Mercator. In other words, each stage uses the system that suits it.

5. Write it down

Record the CRS name and EPSG code in the project metadata and in file names where helpful. Six months later, nobody will remember which of three similar UTM variants the team used.

Case example: accurate distances in network and density analysis

In our bike route network analysis, OpenStreetMap roads arrive in WGS 84. Before building the network, we projected all layers to a metric CRS so that segment lengths, slope from the elevation model and terrain-aware travel costs were all in true metres. In the employment density mapping of Dübendorf, all layers were processed in a metric projected system, so the kernel density search radius represented a real walking-scale distance. Without those steps, both analyses would have looked plausible and been quietly wrong.

Common CRS mistakes to avoid

• Defining instead of projecting. "Define Projection" in ArcGIS, or "Assign Projection" in QGIS, only changes the label. Use it when the label is missing or wrong. To convert coordinates from one system to another, use "Project" or "Reproject Layer".
• Measuring in degrees. Consequently, lengths and areas calculated in EPSG:4326 are not meaningful in metres.
• Ignoring the datum transformation. Moving between an old local datum and WGS 84 needs the right transformation. However, picking the default can leave layers tens to hundreds of metres apart.
• Trusting on-the-fly display. Also, desktop GIS reprojects layers for display, which hides mismatches. Check the CRS of each layer before running analysis.
• CAD files with no coordinate system. Drawings in a local site grid must be georeferenced or transformed before they can be combined with GIS data.
• Repeated raster reprojection. In addition, each resampling step degrades the data a little. Reproject rasters once, with the right resampling method.

A quick checklist before you start analysis

1. Check the CRS of every input layer, and confirm that it is correct, not just present.
2. Choose one projected working CRS suited to the area and purpose.
3. Reproject all inputs into it, using the correct datum transformation.
4. Run a sanity check: measure a known distance, such as a road between two junctions, and compare it with a trusted source.
5. Record the CRS and transformation in the project notes.

Coordinate systems are not glamorous, but they sit underneath every number a GIS produces. If you are integrating CAD, survey and GIS data, or need a second opinion on an inherited dataset, our GIS and infrastructure and engineering GIS teams can sort out the projections before they cause problems.