geographical accuracy (Graham Seaman)

Raphael Mankin raph at
Wed Feb 21 19:08:37 GMT 2007

Your OS grid reference is probably only accurate to 100m. An error of
20m sounds like about 1 arcsecond, so you are about as close as it is
reasonable to get. But if you are testing round trip converseions,
x/y->lat/long->x/y, you should be able to get within about 1m.

If you want to validate the code check it against John Snyder's 'Map
Projections - a working manual': the definitive work on the subject.

The parameters you need are:
TM projection
Spheroid: Airy 1830,  a: 6377563.4,     b: 6356256.9
Phi0: 49.0
Lambda0: -2,0
Scale factor: 0.9996012717
e0: 400000
n0: -100000

It is curious that the true origin of the *British* national grid is
actually in northern France.

On Wed, 2007-02-21 at 15:35 +0000, michael wrote:
> On Wed, 21 Feb 2007, Peter Corlett wrote:
> > On Wed, Feb 21, 2007 at 12:20:31PM +0000, michael wrote:
> >> Postcodes in central London are much much smaller than post codes in the
> >> Scottish highlands. So, if the coordinates are derived form postcode data,
> >> you would expect that the Scottish coordinates would but much less
> >> accurate than the London ones...
> >
> > Well, a postcode covers multiple addresses, and you need the house number to
> > uniquely identify a "delivery point" (or a letterbox as we normally call
> > them).
> >
> > With a postcode only, a lookup is going to get you the co-ordinates of the
> > first delivery point. Thus, from the point of the other delivery points,
> > there is an error, but that's because it wasn't actually their address being
> > looked up in the first place.
> >
> The OP was unsure of the source of the coordinate data, one possibility 
> was that it was derived from postcode data.  If he looked at 10 remote 
> rural locations and found them all accurate to 10 - 20 meters its unlikely 
> that he has hit the postcode centroid every time
> --
> Michael
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Michael John Lush PhD			Tel:44-20-7679-5027
> Nomenclature Bioinformatician		Fax:44-20-7387-3496
> HUGO Gene Nomenclature Committee	Email:  nome at
> The Galton Laboratory
> University College London, UK
> URL:
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

More information about the mailing list