MGRS ↔ Latitude/Longitude Converter

The five parts of an MGRS reference

A complete MGRS string at 1-metre precision has 15 characters (excluding optional spaces). The Empire State Building in canonical form reads:

18T  WL   85628   11322
│ │  ││   │       │
│ │  ││   │       └ northing in the square (5 digits = 1 m precision)
│ │  ││   └ easting in the square (5 digits = 1 m precision)
│ │  │└ row letter (A–V skipping I, O)
│ │  └ column letter (A–Z skipping I, O — but only 8 per zone)
│ └ UTM band letter (C–X skipping I, O)
└ UTM zone number (1–60)

The zone and band are pulled directly from UTM. The two-letter square code identifies which 100-km cell of the zone the point is in; column letters cycle every 100 km of easting, row letters cycle every 100 km of northing. Inside that 100 km × 100 km cell, the digits locate the point. Five digits each = 1 m; four = 10 m; three = 100 m; two = 1 km; one = 10 km; no digits at all = just "somewhere in the 100-km square."

Precision is a digit count, not a separate field

The single biggest design feature of MGRS is that you encode the precision into the string by choosing how many digits to give. Half the digits are the easting; the other half are the northing. The split is always equal, so the total digit count is always even. Trim from the right (low-precision suffix) to coarsen.

Two practical consequences:

• You can radio a coarser reference in fewer syllables. A grid square (no digits) is two letters. A 1-km cell is two letters plus four digits. Voice procedure in NATO communications relies on this brevity.
• Coarser MGRS is not approximate — it's the actual cell."18T WL 85 11" isn't "somewhere near the Empire State Building"; it's the entire 1-km × 1-km square that contains it. Other points in that same square share the same MGRS at 1-km precision. That's correct, not lossy.

How MGRS ↔ lat/lon math works

Three layers stack on top of each other:

1. Lat/lon → UTM via Snyder PP 1395 §8 (the same path the UTM converter uses).
2. UTM easting → column letter. Divide the UTM easting by 100,000 m; subtract one to get the column index; look up the letter in one of three 8-character alphabets (zone 1 uses A-H, zone 2 uses J-R, zone 3 uses S-Z, then cycles).
3. UTM northing → row letter. Take the northing modulo 2,000,000 m; divide by 100,000 m; look up the letter in the appropriate 20-character alphabet (odd zones start at A on the equator; even zones start at F).

The inverse is harder than the forward direction. MGRS's row letter is ambiguous on its own — the same letter recurs every 2,000,000 m (about 18° of latitude). The decoder uses the latitude band letter as a coarse anchor: it converts the band center to UTM northing, then snaps the candidate northing to the cycle nearest that anchor. The math is exact whenever the band letter and row letter are consistent; the only failure mode is a malformed string with a band-row mismatch.

Ten worked examples — both directions

The Greenwich row is the cleanest tell of how MGRS encodes a point near a zone boundary. Greenwich Royal Observatory at longitude 0.0005°W is just inside zone 30, with UTM easting 708,287 m. Dividing by 100,000 gives column index 7-1 = 6; in zone 30 (set index 2, alphabet S-Z), letter index 6 is Y. Cross the prime meridian and you'd be in zone 31 — set index 0, alphabet A-H — with column letter changing to C (zone 31's western edge). Behaviour at zone boundaries is the single most common source of confusion in MGRS workflows.

Misconceptions worth getting straight

"MGRS is just UTM in shorthand"

Closely related but not identical. MGRS strips the explicit hundreds-of-thousands digit from the UTM easting and northing, replacing them with the two-letter square code. It also encodes precision into the digit count, which UTM doesn't. Going MGRS → UTM is lossless; the only thing you gain in UTM is explicit hemisphere from the band letter.

"Five digits each means 5-digit accuracy"

No — five digits each means 1-metre cellprecision. The cell is 1 m × 1 m. Inside that cell the point can be anywhere; the MGRS string doesn't say where within the 1 m cell the point sits. For sub-metre work, MGRS is the wrong representation; use UTM with full-resolution easting / northing (the underlying math has sub-millimetre precision).

"The row letter tells you the hemisphere"

Not directly. The latitude bandletter (C-M = South, N-X = North) tells you the hemisphere; the row letter cycles independently every 2,000,000 m. The decoder uses the band letter to disambiguate row-letter cycles, but only the band tells you which side of the equator you're on.

"You can radio ‘WL 856 113’ and people will know where you are"

Only within the same zone. The 100-km square code WL exists in every UTM zone — the column letter cycles every 3 zones, so most zones have some WL square. The complete reference always includes the zone-and-band. Skipping it ruins interpretation.

"MGRS is global — works everywhere"

Almost. UTM-based MGRS covers 80°S to 84°N. The poles use UPS-based MGRS with the single letters A (SW polar), B (SE), Y (NW), Z (NE) instead of a zone number. Modern military and USGS implementations support both; this converter supports only the UTM portion. Don't use UTM-MGRS for an Antarctic research station above 80°S.

When to use MGRS, when not to

How to verify an MGRS reference

Three sanity checks any MGRS result should pass:

1. Total digit count is even. Each digit position contributes equally to easting and northing. An odd digit count is a typo or a copy-paste error.
2. Column letter belongs to the zone's alphabet.Zones cycle through three 8-letter alphabets (A-H, J-R, S-Z). A column letter outside its zone's alphabet is a wrong zone or wrong square.
3. Round-trip the conversion. Convert MGRS → lat/lon → MGRS at the same precision and confirm it matches. Any discrepancy points to a malformed input.

How this converter is built

Encoding and decoding run entirely in the browser via src/lib/coords/mgrs.ts, layered on top of src/lib/coords/utm.ts (USGS Snyder PP 1395). The input is auto-detected as MGRS or lat/lon by a regex: a string matching <digits><band><column><row><digits> goes through the MGRS path; anything else is parsed as lat/lon. The deep report below the map shows the grid-square breakdown, the equivalent UTM, the EPSG code for the underlying zone, the same point in all six common notations, the nearest place (Mapbox v6, server-side proxy with no-store cache), and the elevation (USGS 3DEP for US, OpenTopoData SRTM30m elsewhere) — per CLAUDE.md §19.2 and §19.4. No coordinates are ever logged or retained.