Lat/Long to OSGB36: UK Coordinate Transformation Guide

Every GNSS receiver, from a handheld unit to a survey-grade rover, ultimately reports its position as a latitude, longitude and ellipsoidal height. Yet almost every map, drawing and dataset used in Great Britain is referenced to the Ordnance Survey National Grid — eastings and northings in OSGB36, with heights above Ordnance Datum Newlyn (ODN). Moving rigorously between the two is not a single calculation but a chain of well-defined geodetic operations, and getting any link wrong introduces errors that range from a few centimetres to several metres.

This article sets out the complete transformation: the datums and reference ellipsoids involved, the difference between WGS84 and ETRS89, the Helmert seven-parameter datum transformation, the definitive OSTN15/OSGM15 method, and the Redfearn series that defines the Transverse Mercator projection. The formulae here are the published Ordnance Survey definitions and will reproduce National Grid coordinates exactly.

Key takeaways

GNSS gives you a position on a global ellipsoid (WGS84/ETRS89, GRS80 ellipsoid); the National Grid is defined on the Airy 1830 ellipsoid and the OSGB36 datum.
A Helmert seven-parameter transformation gets you to OSGB36 to about ±5 m — limited not by the maths but by real distortions in the OSGB36 triangulation network.
The definitive transformation is OSTN15, a gridded “rubber-sheet” shift that achieves ~0.1 m; vertical conversion to ODN uses the OSGM15 geoid model.
The final projection from geodetic coordinates to grid eastings and northings is a Transverse Mercator, evaluated with the Redfearn series.

The starting point: what a GNSS position actually is

A GNSS-derived position is a set of curvilinear coordinates — geodetic latitude (φ), geodetic longitude (λ) and ellipsoidal height (h) — referenced to a particular datum and its associated reference ellipsoid. The ellipsoid is a mathematical surface approximating the shape of the Earth; the datum fixes that ellipsoid’s position, orientation and scale relative to the physical planet.

It is essential to separate three things that are routinely conflated:

The datum (e.g. WGS84, ETRS89, OSGB36) — the realised reference frame.
The ellipsoid (e.g. GRS80, Airy 1830) — the shape on which φ, λ are defined.
The projection / coordinate system (e.g. the National Grid) — the flat, gridded representation.

You cannot move between National Grid and latitude/longitude without being explicit about all three.

WGS84 versus ETRS89 — a distinction that matters

Most users treat “WGS84” and “GPS coordinates” as synonymous, and for navigation that is fine. For survey work in Great Britain it is not.

WGS84 is a global datum maintained by the US and broadcast by GPS. It is tied to the International Terrestrial Reference Frame (ITRF) and therefore moves with the rotating, deforming Earth as a whole.
ETRS89 (European Terrestrial Reference System 1989) is fixed to the stable part of the Eurasian tectonic plate. It was defined to be coincident with ITRF/WGS84 at epoch 1989.0.

Because the Eurasian plate drifts north-east at roughly 2.5 cm per year, WGS84(ITRF) and ETRS89 have diverged by well over half a metre since 1989. In Great Britain, ETRS89 is the GNSS datum — it is the frame realised by the OS Net active station network, and it is the correct input to the OS transformations. Treating a raw handheld WGS84 position as ETRS89 introduces an error of roughly 0.6–0.7 m at present, growing by about 2.5 cm a year. The two ellipsoids, however, are effectively identical:

GRS80 (used by ETRS89): a = 6 378 137.000 m, 1/f = 298.257 222 101
WGS84 ellipsoid: a = 6 378 137.000 m, 1/f = 298.257 223 563

The flattening differs only in the eleventh significant figure — a sub-millimetre effect that can be ignored. So for the geometry below, ETRS89 and WGS84 share the GRS80/WGS84 ellipsoid; the difference between them is one of datum realisation and epoch, not ellipsoid shape.

The destination: OSGB36 and the Airy 1830 ellipsoid

The National Grid is defined on OSGB36, realised physically by the retriangulation of Great Britain carried out between 1936 and 1962 — the network of trig pillars. Its associated ellipsoid is Airy 1830:

a (semi-major axis) = 6 377 563.396 m
b (semi-minor axis) = 6 356 256.909 m
1/f = 299.324 964 6
e² = (a² − b²)/a² = 0.006 670 540 1

Two points follow. First, the Airy ellipsoid is both smaller and differently oriented than GRS80, so a datum transformation is unavoidable. Second — and this is the crux of UK geodesy — the OSGB36 network contains internal distortions of several metres across the country, an artefact of the observation and adjustment methods of the mid-20th century. No exact mathematical transformation can reconcile a clean modern GNSS frame with a distorted historical network. This single fact dictates the two-tier approach below.

Step 1 — Geodetic to 3D Cartesian coordinates

All rigorous datum transformations operate on geocentric Cartesian coordinates (X, Y, Z), not on latitude and longitude directly. Convert the ETRS89 geodetic coordinates to Cartesian on the GRS80 ellipsoid:

ν  = a / sqrt(1 − e² sin²φ)            (prime-vertical radius of curvature)

X  = (ν + h) cos φ cos λ
Y  = (ν + h) cos φ sin λ
Z  = ((1 − e²) ν + h) sin φ

Here a and e² are the GRS80 values, φ and λ are in radians, and h is the ellipsoidal height. The result is a point in a right-handed, Earth-centred, Earth-fixed frame with origin at the geocentre, Z towards the pole and X towards the Greenwich meridian.

Step 2 — The Helmert seven-parameter datum transformation

The classical way to change datum is the Helmert (similarity) transformation: three translations, three rotations and a scale factor relating the two Cartesian frames. The Ordnance Survey publishes a set of parameters taking ETRS89 to OSGB36 using the position-vector rotation convention:

Parameter	Symbol	Value
Translation X	tₓ	−446.448 m
Translation Y	t_y	+125.157 m
Translation Z	t_z	−542.060 m
Rotation X	rₓ	−0.1502″
Rotation Y	r_y	−0.2470″
Rotation Z	r_z	−0.8421″
Scale	s	+20.4894 ppm

The transformation is applied as:

[X']   [tx]   [ 1+s   −rz    ry  ] [X]
[Y'] = [ty] + [  rz   1+s   −rx  ] [Y]
[Z']   [tz]   [ −ry    rx   1+s  ] [Z]

with the rotations converted from arc-seconds to radians (multiply by π / 648 000) and the scale applied as a fraction, so 1 + s = 1.000 020 489 4 (20.4894 ppm = 20.4894 × 10⁻⁶). The sign convention is critical: these parameters are defined for the position-vector convention. If your software uses the coordinate-frame (Bursa–Wolf “reversed”) convention, the three rotation signs must be flipped, or you will introduce an error of the order of tens of metres.

Note that because the rotations and scale are tiny, this is mathematically a near-rigid shift dominated by the three translations.

Step 3 — Cartesian back to geodetic, on the Airy ellipsoid

Having moved into the OSGB36 frame, convert the transformed (X′, Y′, Z′) back to latitude and longitude — this time using the Airy 1830 constants. Longitude is exact:

λ = atan2(Y', X')

Latitude requires iteration (or a closed-form solution such as Bowring’s). The standard iteration is:

p   = sqrt(X'² + Y'²)
φ₀  = atan2(Z', p · (1 − e²))          (initial estimate)

repeat:
    ν   = a / sqrt(1 − e² sin²φ)
    φ   = atan2(Z' + e² ν sin φ, p)
until |Δφ| < tolerance

h = p / cos φ − ν

using the Airy a and e². Convergence is reached in three or four iterations to sub-millimetre level. You now have OSGB36 latitude and longitude on the Airy 1830 ellipsoid.

The accuracy of this Helmert route is approximately ±5 m. That residual is not numerical — it is the OSGB36 network distortion, which a seven-parameter similarity transformation cannot model. For navigation, GIS context, or metre-level work, this is adequate. For survey-grade work, it is not.

The definitive route: OSTN15 and OSGM15

To obtain National Grid coordinates that agree with OS mapping and the trig network to better than 0.1 m, the Helmert step is replaced by OSTN15 (the National Grid Transformation OSTN15 — released in 2016; the “15” denotes the 2015 model epoch). OSTN15 is not a set of parameters but a gridded rubber-sheet model: a regular 1 km grid covering Great Britain, each node carrying an easting shift (sₑ) and a northing shift (s_n). It encodes the actual, irregular relationship between the clean ETRS89 frame and the distorted OSGB36 National Grid, point by point.

The workflow differs from the Helmert route in an important way — OSTN15 operates in projected grid space, not on latitude/longitude:

Project the ETRS89 latitude and longitude through the Transverse Mercator projection (Step 4 below), but using the GRS80 ellipsoid with the National Grid projection parameters. This yields “ETRS89 National Grid” eastings and northings.
Look up the OSTN15 shifts at that location by bilinear interpolation of the four surrounding 1 km grid nodes.
Add the interpolated shifts to obtain the OSGB36 National Grid easting and northing.

E_OSGB36 = E_ETRS89 + se(E, N)
N_OSGB36 = N_ETRS89 + sn(E, N)

Because the shift to be interpolated depends on the OSGB36 position you are trying to find, the reverse direction (National Grid back to ETRS89) is iterative: take the OSGB36 coordinates as a first estimate of where to sample the grid, subtract the interpolated shift, and repeat until the position stabilises (a few iterations).

OSTN15 is the definitive ETRS89↔OSGB36 transformation. There is deliberately no exact algebraic alternative, because the thing being modelled — the historical network — is itself not algebraic.

Heights: OSGM15 and Ordnance Datum Newlyn

Horizontal position is only half the problem. GNSS delivers ellipsoidal height (h), the height above the GRS80 ellipsoid, which has no physical meaning — water does not flow according to ellipsoidal height. Engineering and mapping use orthometric height (H), the height above the geoid, realised in Great Britain as Ordnance Datum Newlyn, the mean sea level recorded at the Newlyn tide gauge in Cornwall between 1915 and 1921.

The two are related by the geoid–ellipsoid separation (N):

H = h − N

The separation N is supplied by the OSGM15 geoid model, distributed together with OSTN15 on the same grid. Across Great Britain N is around +50 m, varying by roughly ±5 m (some 10 m greater in the west than the east), so it is never negligible. Applying OSGM15 converts a GNSS ellipsoidal height directly to an ODN height to a few centimetres.

Step 4 — The Transverse Mercator projection (Redfearn series)

Whichever datum route is used, the final step from geodetic latitude/longitude to grid eastings and northings is a Transverse Mercator projection. The National Grid uses these defining parameters:

Ellipsoid: Airy 1830 (for the OSGB36 route) — a = 6 377 563.396 m, b = 6 356 256.909 m
True origin: φ₀ = 49° N, λ₀ = 2° W (i.e. −2°)
Scale factor on the central meridian: F₀ = 0.999 601 271 7
Easting of true origin: E₀ = 400 000 m
Northing of true origin: N₀ = −100 000 m

The OS evaluates the projection with the Redfearn series. First compute the auxiliary quantities (with e² for the chosen ellipsoid and n = (a − b)/(a + b)):

ν  = a F₀ (1 − e² sin²φ)^(−1/2)                  (transverse radius of curvature)
ρ  = a F₀ (1 − e²)(1 − e² sin²φ)^(−3/2)          (meridional radius of curvature)
η² = ν / ρ − 1

The meridional arc M (distance along the meridian from φ₀ to φ) is:

M = b F₀ [ (1 + n + 5/4 n² + 5/4 n³)(φ − φ₀)
         − (3n + 3n² + 21/8 n³) sin(φ − φ₀) cos(φ + φ₀)
         + (15/8 n² + 15/8 n³) sin(2(φ − φ₀)) cos(2(φ + φ₀))
         − (35/24 n³) sin(3(φ − φ₀)) cos(3(φ + φ₀)) ]

Then, with p = (λ − λ₀), the northing and easting are:

N = I + II·p² + III·p⁴ + IIIA·p⁶
E = E₀ + IV·p + V·p³ + VI·p⁵

I    = M + N₀
II   = (ν / 2) sin φ cos φ
III  = (ν / 24) sin φ cos³φ (5 − tan²φ + 9 η²)
IIIA = (ν / 720) sin φ cos⁵φ (61 − 58 tan²φ + tan⁴φ)
IV   = ν cos φ
V    = (ν / 6) cos³φ (ν/ρ − tan²φ)
VI   = (ν / 120) cos⁵φ (5 − 18 tan²φ + tan⁴φ + 14 η² − 58 tan²φ η²)

This reproduces National Grid eastings and northings to better than a millimetre across Great Britain.

Worked example (projection step)

For the OSGB36 geodetic coordinates φ = 52° 39′ 27.2531″ N, λ = 1° 43′ 4.5177″ E (the standard OS test point near Caister), the Redfearn series above yields:

Easting  E = 651 409.903 m
Northing N = 313 177.270 m

i.e. National Grid reference TG 51409 13177. This is the canonical Ordnance Survey worked example and is a useful unit test for any implementation.

The reverse projection (grid to geodetic)

To go from National Grid eastings and northings back to latitude and longitude, first find the foot-point latitude φ′ by iterating the meridional arc:

φ' = (N − N₀ − M) / (a F₀) + φ₀      ← compute M(φ'), repeat until (N − N₀ − M) is negligible

Then apply the VII–XIIA series (all evaluated at φ′, with d = E − E₀):

φ = φ' − VII·d² + VIII·d⁴ − IX·d⁶
λ = λ₀ + X·d − XI·d³ + XII·d⁵ − XIIA·d⁷

VII  = tan φ' / (2 ρ ν)
VIII = tan φ' / (24 ρ ν³) (5 + 3 tan²φ' + η² − 9 tan²φ' η²)
IX   = tan φ' / (720 ρ ν⁵) (61 + 90 tan²φ' + 45 tan⁴φ')
X    = sec φ' / ν
XI   = sec φ' / (6 ν³) (ν/ρ + 2 tan²φ')
XII  = sec φ' / (120 ν⁵) (5 + 28 tan²φ' + 24 tan⁴φ')
XIIA = sec φ' / (5040 ν⁷) (61 + 662 tan²φ' + 1320 tan⁴φ' + 720 tan⁶φ')

Putting the chain together

For a survey-grade ETRS89 GNSS position, the definitive sequence to OSGB36 National Grid and ODN is:

ETRS89 (φ, λ) → project through Transverse Mercator on the GRS80 ellipsoid → ETRS89 National Grid (E, N).
Interpolate OSTN15 shifts (bilinear, 1 km grid) and add → OSGB36 National Grid (E, N).
Interpolate OSGM15 separation N and apply H = h − N → ODN orthometric height.

The approximate Helmert sequence (metre-level only) is instead:

ETRS89 (φ, λ, h) → Cartesian (X, Y, Z) on GRS80.
Apply the seven-parameter Helmert transformation → OSGB36 Cartesian.
Cartesian → geodetic on Airy 1830.
Project through Transverse Mercator on Airy 1830 → OSGB36 National Grid (E, N).

Accuracy, in summary

Method	Horizontal accuracy	When to use
Helmert 7-parameter	~ ±5 m	Navigation, GIS context, metre-level work
OSTN15 + OSGM15	~ 0.1 m	All survey, engineering and mapping deliverables

The headline point for anyone commissioning or checking survey data is this: the limiting factor in UK coordinate transformation is not arithmetic precision but the historical OSGB36 network itself. A Helmert transformation is exact in the mathematical sense yet still ~5 m from true National Grid, because it cannot reproduce the network’s distortions. Only the gridded OSTN15/OSGM15 models, which were built specifically to absorb those distortions, deliver coordinates that agree with Ordnance Survey mapping and the national control network to the centimetre. For any deliverable that will be set out, overlaid, or relied upon against OS data, OSTN15 and OSGM15 are not optional — they are the definition of “correct”.

It is worth pausing on what that means for anyone commissioning survey work. Modern software will happily perform every step above at the press of a button — and that is precisely the trap. The arithmetic is automated; the judgement is not. Knowing that a handheld WGS84 position is not ETRS89, that a Helmert transformation cannot do better than five metres, that the rotation signs depend on the convention your software assumes, that OSTN15 must be applied for any deliverable set out against OS mapping, and — critically — being able to look at a coordinate and recognise when it is wrong: none of that comes from the tool. It comes from training, qualification and years of doing it.

This is the real difference between a qualified, experienced surveyor and an operator running a black box. The unqualified can produce numbers that look authoritative and are quietly metres adrift, because nothing in the workflow flagged the error and nothing in their background equipped them to catch it. A RICS-regulated surveyor brings the geodetic understanding to choose the right datum, the right transformation and the right vertical model for the job, to validate the result against independent control, and to stand behind the coordinates as correct. When you commission a survey, you are not buying a button-press — you are buying the expertise that knows when the button-press is lying to you. That distinction, more than any single deliverable, is what separates a professional survey from a cheap one.

References and further reading

Ordnance Survey, A Guide to Coordinate Systems in Great Britain — the primary reference for the Airy 1830 and GRS80 parameters, the ETRS89→OSGB36 Helmert parameters, the National Grid Transverse Mercator constants and the Redfearn projection series, including the worked example reproduced above.
Ordnance Survey, National Grid Transformation OSTN15 and the OSGM15 geoid model — the definitive ETRS89↔OSGB36 transformation and ETRS89→ODN height model.
Ordnance Survey, Accuracy of OS Net, OSTN15 and OSGM15 — the published 0.1 m (RMSE) agreement between OSTN15 and the OSGB36 control network.

From Latitude and Longitude to OSGB36: A Technical Guide to UK Coordinate Transformations