UTM converter

UTM slices the world into 60 strips of longitude, six degrees each, and flattens them one at a time. Inside a strip you get plain metres: an easting, a northing, and arithmetic that behaves. Measure a distance with Pythagoras and you’ll be right to a fraction of a metre per kilometre, which is why surveyors, the military and every GIS on earth lean on it.

The catch is that the strips never join up. Zone 17 and zone 18 are separate flat worlds, and a point near the border has an address in both. That’s the price of keeping distortion small: you can either have one grid for the planet with horrible distortion, or sixty tidy ones with seams. UTM picked seams.

Each zone gets its own central meridian down the middle, and that line is defined as exactly 500,000 m east, so nothing inside the zone ever carries a minus sign. The scale along that meridian is pinned at 0.9996, deliberately shy of 1. Towards the zone edges it creeps above 1. Splitting the error that way halves the worst case, which lands around one part in 2,500.

The letter after the zone number is the latitude band: 8° tall, running C at 80°S through to X in the far north, which is stretched to 12° so the system reaches 84°N without an extra band. I and O are skipped, since they read as 1 and 0. The band is a convenience for people reading grid references, and the hemisphere is what the maths needs, so the tool above carries both.

Every number in this table came out of the same code the converter runs, at the moment the page was built. Check your own work against a row near your latitude.

The last two rows are there to catch out a converter. Bergen sits at 5.3221°E, which the plain six-degree rule puts in zone 31. It comes out as zone 32, because Norway widened that one in the 1950s to keep its west coast and the shipping lanes off it on a single grid. Ny-Ålesund is the other trap. Above 72°N the Svalbard zones get reshuffled, so 11.9231°E lands in zone 33 instead of 32. Plenty of converters skip both rules and quietly hand you a position hundreds of kilometres out.

The projection runs as a truncated series on the WGS84 ellipsoid, the standard Snyder formulation out of USGS Professional Paper 1395. Within a zone it agrees with the exact solution to well under a millimetre, which is thousands of times finer than any GPS fix you’re likely to feed it. Round-tripping a point out to UTM and back lands within a millimetre too.

So the arithmetic isn’t your limit. Rounding is. An easting printed to the whole metre is a metre of slop before anything else happens, and that’s usually fine, because UTM is metres and people round to metres without thinking about it.

The format that bites is DMS. One second of latitude is about 30.9 m, so a position printed as whole seconds carries up to 15 m of error before it reaches this page. Convert that to UTM and you get a clean-looking easting and northing, to the metre, built on a number that was never that good. The tidiness is an illusion. Keep two decimals on your seconds and the problem goes away.

Two other things worth remembering. Scale factor: distances measured on the grid run short by 40 cm per kilometre along the central meridian and long by as much as 98 cm per kilometre out at a zone edge near the equator, and the panel above tells you which you’ve got. Convergence: grid north and true north differ by a degree or more away from the central meridian, more still at high latitude, so a bearing off a UTM grid needs correcting before it goes anywhere near a compass. London’s convergence is 2.249°. Bergen’s is -3.199°, and that far over from a central meridian it starts to matter.

Beyond 84°N and 80°S, UTM stops. The polar regions use UPS, a different projection with a different grid, and nothing here will produce a sensible answer up there. The band letter goes blank as a warning.