Few things derail a mapping project faster than a coordinate reference system nobody agreed on. The symptoms are familiar: a boundary layer sitting a hundred metres east of the aerial imagery, a GPS survey that will not line up with a site plan, grid references that make sense to exactly one person in the room. None of that is a software fault. It is a decision taken too late, or not taken at all. For UK work the choice usually comes down to three candidates — British National Grid, ETRS89 and WGS 84 — and a handful of questions worth asking before anyone touches the data.

What you are actually choosing

A coordinate reference system is two things bolted together: a datum, which fixes a model of the Earth and its relationship to real places, and a coordinate system, which decides whether positions are written as degrees of latitude and longitude or as metres on a flat grid. Software makes this look like a dropdown menu. It is not. Change the datum and every coordinate shifts; change the projection and distances and shapes change with it.

In Britain, the datum is the argument. The country sits between a legacy national datum built to fit these islands and the modern European datum that satellite positioning delivers. Both are correct. They are simply correct about different things.

The three candidates, and when each one wins

British National Grid — EPSG:27700

Built on the OSGB36 datum, this transverse Mercator projection gives you eastings and northings in metres. It underlies Ordnance Survey mapping, most planning and land work, and the great majority of statutory and utility datasets in Great Britain. If the finished product goes to a council, a planning consultant, a utility records team or anyone working from OS mapping, British National Grid is usually the right answer. It is also the one people get wrong most often, because grid coordinates are only as good as the transformation used to reach them.

ETRS89 — EPSG:4258, or projected as EPSG:25830

ETRS89 is the European Terrestrial Reference System, fixed to the stable part of the Eurasian plate. GNSS observations in the UK are computed in it, and OS Net, the national active network, delivers coordinates in a realisation of ETRS89. It was aligned with WGS 84 in 1989, but the two have drifted apart since: ETRS89 sits still with Europe, while WGS 84 travels with the global reference frame. The gap is now close to a metre and grows by a couple of centimetres a year. Use ETRS89 when accuracy and long-term stability matter — control networks, monitoring, large infrastructure, anything that has to be defensible in ten years. In practice you keep it as geographic ETRS89 or project it to UTM zone 30N.

WGS 84 — EPSG:4326, or Web Mercator as EPSG:3857

WGS 84 is the default of consumer GNSS receivers, of GeoJSON, and of most web mapping stacks, which makes it excellent for exchange and for storing raw satellite positions. It is a poor choice for measuring anything, because its coordinates are degrees, and a degree of longitude near Cornwall covers a very different distance from one near Orkney. Web Mercator is a display projection: fine for tiles, useless for areas and distances. Note too that a raw GNSS fix in WGS 84 and an OS Net position in ETRS89 sit roughly a metre apart. For a walking route that is nothing; for a boundary dispute or a pipeline tie-in it is the whole argument.

Where the mistakes actually happen

Projections get the blame; transformations do the damage. Moving between ETRS89 and OSGB36 is not a simple shift, because OSGB36 was fitted to the UK as a whole and its relationship to satellite positioning varies from place to place. Ordnance Survey publishes a grid of shifts for exactly this purpose — and software will happily offer you something faster and worse if you let it.

• OSTN15 paired with OSGM15 is the accurate route between ETRS89 and British National Grid, and the one to insist on. If your GIS or CAD package does not have it installed, install it.
• Helmert-only transformations are a rough approximation. They can leave you a metre or more out, and the error is not the same everywhere in the country.
• A null transformation — treating OSGB36 and ETRS89 as the same thing — typically throws data tens to hundreds of metres off. It also looks plausible on screen, which is what makes it dangerous.
• Heights are a separate decision. Ordnance Datum Newlyn is not ellipsoidal height, and in Britain the two differ by tens of metres. Converting between them needs a geoid model, not a projection.

Pick the system the final consumer expects, and get there with the best transformation available. Everything else is detail.

A decision path that works

1. Start with the destination, not the source. Who opens the file, and in what software? A planning submission means British National Grid. A web map means WGS 84 for storage and Web Mercator for display. A monitoring programme means ETRS89.
2. Collect in the frame you measured in. Do not reproject observations on the way in. Store ETRS89 GNSS positions as ETRS89, and note the epoch if the work is precise.
3. Convert once, at the end, with the right transformation. Repeated round-trips between datums accumulate error and destroy any chance of tracing where a number came from.
4. Record the EPSG code in the file metadata, the filename, or both. A shapefile without a .prj file, or a CSV of eastings and northings with no header, is a future problem with your name on it.
5. Check a known point. Before delivery, take one benchmark or OS feature with published coordinates and confirm your output matches within the tolerance the project needs.

Traps worth knowing before you export

Axis order catches people out constantly. EPSG:4326 is defined as latitude first, longitude second, while almost every piece of software treats it as x then y. A swapped pair does not produce a small error; it usually puts your data somewhere entirely different. If something looks wildly wrong rather than slightly wrong, suspect this first.

Grid distance is not ground distance. British National Grid is a projection, so a metre measured on the ground is not a metre on the grid everywhere; the difference depends on where you are and how high you are, and it grows towards the edges of the projection. For site engineering, that scale factor is worth calculating rather than ignoring.

Northern Ireland is a different case again. It uses the Irish Grid on the TM75 datum, not British National Grid, so any dataset covering the whole UK needs both. And if you build a local site grid for a construction job, remember it is a convenience, not a standard: publish a documented transformation to a national system alongside it.

The short version

Use British National Grid (EPSG:27700) whenever the output feeds British planning, land, utility or OS-based work, and make sure OSTN15 is doing the conversion. Use ETRS89 when you need survey-grade positions that stay valid, and expect it to be the datum your GNSS work arrives in. Use WGS 84 for storing and exchanging global data, and Web Mercator only for drawing tiles on a screen. Write the code down, convert once, and check one known point before you send anything out.

Photo: Larry Hyler / Pexels