Ordnance Survey Mapping Errors

Two Fallacies:

”The UK’s Ordnance Survey maps are reliably accurate”

”If the map says that’s where it is, then that’s where it is!”

The above statements have been the accepted wisdom for decades, but unfortunately they are incorrect, particularly for OS maps in the 1:50K Landranger series. The following clear and unequivocal evidence reveals that this mapping merely depicts the approximate position of many map features with errors of up to 65 metres, and cannot be considered accurate enough to support very precise micro-navigation or very precise orienteering. This is a bold claim, but the following irrefutable facts speak for themselves. Although OS maps in the 1:25K Explorer series are considerably better, they still show lessor errors in exactly the same places.

The evidence presented here is based on 2012 Landranger (1:50k) and 2014 Explorer (1:25k) Ordnance Survey digital mapping of the Yorkshire Dales, specifically in the area of the ‘Yorkshire 3 Peaks’ (Y3P), Whernside, Ingleborough and Penyghent. Because of the controversial nature of the conclusions, and to prevent pointless challenges, the methodology will be explained before the evidence itself is presented.

The author has walked the circuit of the Y3P on more than two dozen occasions in the past decade, the route followed being the same on each occasion except for a variation at only one point of the route, in that either of two options were used for the route between Nether Lodge and the Ribblehead viaduct; one is the conventional route heading westward from Nether Lodge along the farm road then northwards along the B6479; the other route avoids road-walking by heading northwards from Nether Lodge along the Ribble Way, then usually westwards across Runscar. These are shown clearly in Annex A.

Walking such a route carrying a GPS Data Logger provides a track which may be superimposed onto the digital mapping to provide evidence of the route taken. For many years I have carried such a data logger, pictured above, on all mountain and ski excursions, usually set for an update interval of 5 seconds, to provide a date-time-stamped log of exactly where I have been. All of the snips of digital maps in this article show numerous GPS tracks of exactly the same walked route, all superimposed one on top of the other. From the startling precision with which they almost precisely overlap each other, often almost merging into just one trace, it is clear that GPS tracks must be more consistently accurate than many users have been led to believe.

On one of these Y3P trips, proof was obtained that the GPS was definitely complying with the OSGB36 datum, by assuming that Trig Points will always be shown in their correct position on maps, even if nothing else is. There are three widely dispersed Trig Points on the route, on the summits of Whernside, Ingleborough and Penyghent. By walking slowly in a straight line directly towards each of these Trig Points, then walking slowly away from it in another precise compass direction, turning round and walking slowly back to the Trig Point in the reverse direction, then repeating this out-and-back procedure in a couple of other arbitrary compass directions, you end up with a precise indication of where the GPS system records the position of the Trig Point. Here are the same results displayed on both 1:50K Landranger mapping (upper 3 snips) and 1:25K Explorer mapping (lower 3 snips).

Whernside	Ingleborough	Penyghent

This proves two things beyond any doubt;

• firstly, that the GPS receivers were definitely set to the correct OSGB36 datum, and
• secondly, that the GPS network agrees with OS mapping about the position of Trig Points.

Mapping Errors

One good example of a mapping error evident on the 1:50K Landranger maps (but curiously not on the 1:25K Explorer maps), is that the Trig Point on the summit of Ingleborough is shown to be collocated with the summit ‘Viewpoint’, thus:

1:50K Landranger	1:25K Explorer

In reality these two features are anything but collocated, as you can clearly see from the following three photographs of the Trig Point, the Summit Cairn and the Viewpoint

The depiction on OS maps of footpaths, railway lines, streams, bridges and roads is unfortunately much less accurate than it is for Trig Points. Starting with footpaths, a good example is the approach to Whernside from the east, after leaving Ribblehead northwards on the footpath to the east of the railway line and then crossing the line at an aqueduct. The only footpath on the route has been there for decades and is illustrated here, looking down the path from west to east:

In the table below are shown snips from 1:50K Landranger and 1:25K Explorer mapping depicting this footpath. In the left hand column is the raw mapping. In the right hand column 17 GPS tracks recorded on 17 different occasions have been added. The upper 2 snips are of 1:50K Landranger mapping and the lower two are of 1:25K Explorer mapping. Since the 17 tracks have almost merged into one, they must indicate the true position of this footpath. This very old and very obvious footpath is clearly not depicted accurately at all, particularly on 1:50K mapping, and it is just one example out of countless many.

Eastern approach to Whernside	Track superimposed on the map

The slight track divergence, upper left, was due to the flagged path having sunk through the years making it a water-course in wet weather. On one occasion it had frozen overnight after a day of rain, making the flags unsafe, so a parallel route was taken off the path.

Now consider the depiction of the Settle-Carlisle railway line, which has been in the same position since the 1870’s. The following snips are of the line running northwards from Ribblehead towards the aqueduct and the tunnel.

1:50K Landranger mapping	1:25K Explorer mapping
1:50K Landranger mapping	1:25K Explorer mapping

If OS maps were indeed completely accurate, this would provide conclusive proof that I must have been trespassing on the Settle-Carlisle railway line by walking along the bed of the track on each of 17 occasions recorded over a period of several years, whereas in fact I was following the footpath to the east of it. The inescapable conclusion is that neither Landranger nor Explorer maps always show railway lines in their correct position, occasionally showing a significant mapping error.

At the aqueduct, the stream crosses just to the south of the footpath, which means that the stream is shown on the 1:50K Landranger map to the north-east of its correct position at the aqueduct, with a mapping error of 45 metres, proving that neither streams nor bridges are necessarily shown on 1:50K OS maps in their true position. The distance between the Aqueduct and the tunnel mouth also differs noticeably between the two map series; both cannot be right. Here are some photographs of this aqueduct:

The following snips are of 12 superimposed tracks recorded on 12 widely separated dates, showing the route along the farm road from Nether Lodge to the B6479, and then northwards towards Ribblehead

1:50K Landranger mapping	1:25K Explorer mapping

Next is a comparison of the same G-Trek II trace taken 17^th April 2015, shown overlain both on 1:50K Landranger mapping and on Google Earth.

1:50K Landranger mapping	Google Earth mapping

The position of the farm road is obviously inaccurate with a mapping error of up to 65m in the case of the Landranger map, and since I walked along the right-hand (eastern verge) of the road northwards to Ribblehead on each of these 12 occasions, it proves that even the B6479 is shown with a noticeable mapping error. This is what the map is attempting to portray as the sharp bend with the greatest mapping error:

Readers who are sceptical may very easily check the facts for themselves. Just follow my Y3P route recording your track on your own GPS receiver, then download the overlay you obtain onto digital mapping and confirm that your own track shows exactly the same mapping errors in exactly the same places as the multiple overlays that I have relied on to justify the conclusions presented here. A sceptical colleague who accepted this challenge found the same discrepancies, which he said he would have otherwise dismissed as GPS errors.

Although a circumnavigation of the Y3P has been used to illustrate the problem, similar discrepancies have been found almost everywhere else in the UK where routes have been walked by the author carrying a GPS Data Logger; many of these routes have been walked at least twice, and on each of these occasions the GPS tracks overlaid one another almost precisely, proving that the mapping errors that had been revealed were genuine and not due to random or other GPS errors. The problem must therefore be quite widespread, making it a national issue. Nipping one weak counterargument in the bud, it is obvious from the maps that the Data Logger would have enjoyed an uninterrupted view of nearly the entire overhead hemisphere throughout the Y3P route, with no GPS satellites shaded by obstructions.

In Annex B is appended the Ordnance Survey’s explanation of (or excuse for) these mapping errors, followed by an analysis of why and where their arguments seem flawed.

If the depiction of footpaths, railway lines, streams, bridges and roads is so inaccurate, what confidence can we have that contours, crags, re-entrants and any other features are shown in their correct position, particularly in maps in the Landranger series? What effect might this have on the teaching and assessment of micro-navigation on ML and WGL courses? What effect might it have on orienteering?

Conclusions and Advice:

Modern GPS Receivers, ‘if’ correctly set up, ‘if’ correctly used, ‘if’ in full working order, ‘if’ without flat batteries and ‘if’ enjoying an uninterrupted line of site to the overhead hemisphere, are highly accurate, much more so than OS mapping. Although there are undoubtedly a lot of ‘ifs’ to manage, this is not too difficult, but best practice if you are seriously relying on them for navigation is to carry at least two, both with fresh batteries.

Disappointingly and unexpectedly, the Ordnance Survey’s Landranger mapping is not accurate, all-too-often only showing the approximate position of features. This conclusion had not been sought and has come as quite a surprise.

You are warmly invited to counter these conclusions if you can, by producing concrete evidence to refute the methodology, the arguments or the data, and sending full details to the author at mel.aac@blue-roan.org. Be warned though, that the author is a one-time radar system design team member and radar systems instructor and examiner at postgraduate level, which is indeed completely relevant to this issue since the ‘accuracy’ of both radars and GPS depends on exactly the same technology, in that the accuracies of both depend upon their abilities to measure the very minutest differences in the times of arrival of radio signals; the rest is straightforward digital processing, supported by the incontrovertible laws of physics and mathematics. Your counterarguments therefore need to be scientifically sound. I freely admit that there is one obvious counterargument that I have ignored, namely that these errors do not matter to the map user, i.e., the ’So What!’ retort, but this is a highly subjective area in which agreement is unlikely. The point surely is that if such significant errors are obvious to anyone capable of walking anywhere with a cheap easily obtainable GPS receiver, then the Ordnance Survey should at least be more openly acknowledging the existence of the problem and the true extent of the discrepancies, even if unwilling to do anything about it by correcting them.

Annex A:

The full route taken round the Yorkshire 3 Peaks is shown in the snip below. Each overlay is shown as a line either in blue, red, or green, as follows:

Blue: All 17 blue tracks were recorded taking the route variant west from Nether Lodge and then north along the B6479. Dates: 14 May 07, 06 Oct 07, 12 Apr 08, 20 Jun 08, 27 Jun 08, 10 Jul 09, 11 Dec 09, 11 Jun 10, 20 Jun 11, 06 Nov 11, 12 Mar 12, 04 May 13, 01 Dec 13, 17 May 14, 04 Oct 14, 17 Apr 15 and 07 May 15.

Red: All 5 red tracks were recorded taking the route variant north from Nether Lodge along the Ribble Way and then across the Hawes road and along Runscar Scar, (where there is no obvious footpath, resulting in minor route variations). Dates: 23 Apr 04, 18 Jun 04, 14 Oct 05, 14 Jul 06 and 06 Dec 06.

Green: The green track was recorded 19 Apr 13.

Annex B:

It is worth quoting verbatim from the official Ordnance Survey's website:

”If we were just using GPS, that would be the end of the story. However, we are often trying to relate our position to features on the map, which causes additional problems and errors, because GPS and Ordnance Survey maps use different models for the earth and coordinate systems”

”This is especially the case with height because the model that relates the global coordinate system that GPS uses, WGS84, to map height above sea level is very poor. WGS84 is a modern coordinate system, but maps in Britain are based on the National Grid, which uses an older system known as OSGB36.”

”Although it was surveyed very well for its day, OSGB36 has distortions in it, which have to be taken into account when comparing with GPS-derived coordinates. GPS receivers can correct for the large main offset between the two systems, and it is vitally important to set the receiver's datum (the definition of the coordinate system used) to OSGB36 so that it does so.”

”Please refer to your receiver’s manual for information on how to get your receiver to work with Ordnance Survey maps. Smaller level distortions are still present which could add up to 5m errors when using GPS with Ordnance Survey maps. Most receivers can cope with a variety of coordinate systems for different countries.”

”On top of the nationwide errors in OSGB36, individual features on the map may only have been surveyed to a local accuracy of 7m (for 1:25,000 scale maps) and some features, such as boulders, may only be shown schematically. Of course, this last problem affects anyone trying to position themselves on the map, not just GPS users.”

Conclusions:

1. The position errors are vanishingly small at the locations of the 3 widely-spaced Trig Points, casting doubt on any claim that there is any significant datum error in the general area of the Y3P or in the GPS receivers’ interpretation of OSGB36.
2. The causes of the errors that the Ordnance Survey are referring to above cannot explain the evidence, in that some of the mapping is very accurate indeed, but with inaccurate mapping being seen just a kilometre away and at a similar altitude, an example being the railway line and footpath running north from Ribblehead. Datum errors cannot change over such a small distance on relatively level ground.
3. The extent of the local errors is significantly greater than the 7m that Ordnance Survey acknowledges.

Return to the top of this page, or to the complete Index, or to this section's Index.

This webpage was first web-published in 2010 and was last updated 16^th May 2018.