Adjusting National Grid coordinates for ley-hunting

When searching by computer for long-distance alinements (e.g. leys, dragon lines, fire leynes) it is necessary to allow for the fact that a geodesic does not in general project onto a straight line in the National Grid. We could apply the required correction to each line as it is found. A simpler and more efficient method is to adjust the site coordinates before beginning the search, as explained below.

We know that for ley-hunting the Earth’s polar flattening can be ignored. More precisely: the Ordnance Survey give us a Transverse Mercator projection of the spheroid onto a plane; if we project the plane by Transverse Mercator onto a perfect sphere of suitable radius, then the image of Britain on the sphere is negligibly different from Britain on the spheroid. This fact was used to derive the formulae in the 1977 notes. It is found empirically that a good value for the sphere’s radius R is such that

1/6R2=4.092×10−15m2

This quantity often appears in the formulae.

The Transverse Mercator projection of the sphere does not in general send geodesics into straight lines. But there is a projection of the sphere that does so, namely the Central projection defined as follows: let O be the centre of the sphere, T a plane not through O; then the image of a point P on the sphere is the intersection of OP and T.

Before beginning the computer search, we can convert the National Grid coordinates of each site into coordinates in such a plane T; then geodesics will become (for practical purposes) straight lines. It is natural to choose T as tangent to the sphere at a point in the centre of Britain. We shall use here a point on the central meridian, having NG coordinates E=40km, N=50km (the point NZ 00).

Let (E,N) be NG coordinates and let ,φ) be longitude and latitude on the sphere, with λ measured from the central meridian of the projection (2° W). The Transverse Mercator projection of the sphere is given by

E=Rtanh1(sinλ cosφ)+E0
N=Rtan1(secλ tanφ)+N1
(1)

where E0=40km and N1 is some constant. Let the plane T touch the sphere on the central meridian; let the point of contact have NG coordinates (E0,N0) and geographical coordinates (0,φ0). In T take the origin at that point, X eastwards, Y northwards. Then the Central projection of the sphere is given by

X=Rsinλ cosφ /W
Y=R (sinφ co0cosλ cosφ si0) /W
(2)

where W=sinφ si0+cosλ cosφ co0.

We can now express (X,Y) in terms of (E,N). The N1 drops out and we get simply

X=Rsinh((EE0)/R) sec((NN0)/R)
Y=Rtan((NN0)/R).
(3)

As in the 1977 notes, put D=EE0; also put M=NN0 and k=1/6R2. For ley-hunting we can ignore terms in 1/R4, so that (3) becomes

X=D (1+k (D2+3M2))
Y=M (1+2kM2).
(4)

With a unit of 100 km we shall have

D=E4,M=N5.

In the tangent plane T it will be convenient to use coordinates (E,N′) given by

E=X+4,N=Y+5,

so that the point of contact has coordinates (4, 5) in both planes.

In practice we can regard the mapping (4) as an adjustment to the NG coordinates given by

EE=kD (D2+3M2)
NN=2kM3
(5)

where k=4.092×105.

The following examples confirm that the adjustment (5) does indeed transform geodesics into straight lines with sufficient accuracy for ley-hunting.

Example 1 : Greenwich meridian

Take the geodesic to be the Greenwich meridian, with points every 1° from 50° to 60°. From the OS Projection Tables the NG coordinates of the points are found to be:

LatEN
505.43315410 0.13081909
515.40319840 1.24253656
525.37281173 2.35442150
535.34200325 3.46647263
545.31078226 4.57868844
555.27915821 5.69106723
565.24714067 6.80360710
575.21473934 7.91630596
585.18196404 9.02916151
595.1488247310.14217125
605.1153315011.25533252

Let us apply the adjustment (5) and see how far the resulting points are from the best-fitting straight line (calculated by least squares). The result is:

LatENError (m)
505.43744576 0.12137123-1.029
515.40574347 1.23819496-0.243
525.37409713 2.35290610+0.141
535.34248958 3.46617748+0.316
545.31090298 4.57868232+0.409
555.27931885 5.69109424+0.486
565.24771808 6.80408727+0.548
575.21608094 7.91833581+0.539
585.18438715 9.03451466+0.345
595.15261587 10.15329895-0.200
605.12074577 11.27536417-1.311

The adjusted points are contained in a strip of width 1.9 metres, versus 484 metres for the unadjusted points.

Example 2: the St Michael line

One version of John Michell’s famous line is got by drawing a geodesic from St Michael’s Mount to Bury St Edmunds Cathedral. Following Bob Forrest’s notes, take three points A, B, C on this line: A and B near the points where the geodesic differs most from a straight line drawn on the map, and C near the point where the geodesic crosses that line.

(No attempt is made here to find the theoretically exact coordinates of those 3 points, but A, B, C are accurately on the geodesic as calculated from the O.S. tables.) As shown by Bob Forrest, the distance on the ground between geodesic and straight line is about 18 metres at A, 3 metres at B.

SiteEN
St M M 1.5146 0.2986
A 2.708371420.94302948
B 4.629482791.97955893
C 5.291547852.33680477
Bury St E5.8559 2.6414

As in Example 1, apply the adjustment (5) and then calculate the best-fitting line. The distance offline is cut down to about 50 mm.

SiteENError (m)
St M M 1.507227920.29009553-0.027
A 2.705673500.93756472+0.014
C 4.630197981.97730377+0.053
B 5.292760552.33525889+0.011
Bury St E5.857429002.64032619-0.052

Michael Behrend
18 June 1986