Extract from: Marhofn 316.18, May 2016
Author: Alan Dawson
Date: May 2016
Surveying And Mapping Standards

Surveying and mapping standards

Heights on OS maps are derived from three types of survey: a) triangulation and levelling, b) aerial photography and photogrammetry, c) differential GNSS (Global Navigation Satellite System). These methods have different degrees of accuracy. There are many variables, but as a general guide, heights of flush brackets on triangulation pillars are accurate to within 0.5m, other spot heights derived from ground surveys are usually accurate to within 1-2m, heights from aerial surveys are accurate to within 3-4m, and heights from GNSS surveys are accurate to within 0.05-0.1m. This is no secret. OS has published a blog on the subject. A fourth survey method, using LIDAR, is reckoned to be less accurate than GNSS surveying but more accurate than the other methods. I am not aware that it is used for OS mapping.

OS maps have their faults but for most purposes they are superbly useful and accurate. However, it is obvious that numerous hills have different heights on different scales and editions of OS maps. This is not surprising when they have different sources - co-operation between OS departments has historically not been good. OS are now trying to make their maps more consistent, though not necessarily more accurate. For example, when I informed them that Ladylea Hill is 609m, which is the ground survey height still shown on some OS mapping, they responded that:

'the various height models used in OS mapping products have different levels of precision and accuracy and this sometimes leads to small discrepancies. The value of 609 in the 1:250000 scale product will be flagged for update to 610. Hill heights and contours have nearly always been derived from aerial surveys. This is the case for Ladylea Hill and the difference between 609m and 610m is within the accuracy tolerance of the method used to capture the height with the notable exception of heights at trig pillars which come from ground based survey methods.'

OS are unlikely to know which of their spot heights is more accurate. It is common for spot heights to refer to points other than summits, and typographic errors are not unusual. We all make mistakes, including OS staff. For hill lists based on relative height, value judgements are required for col interpolation, as most cols do not have spot heights. Anyone compiling or revising a hill list has to deal with these basic issues. The main options are to use skill, judgement and experience to analyse OS maps at various scales or to survey hills using GNSS.

Most of my surveys have not been ratified by OS. I do not usually follow in full the OS specifications for its contractors for several reasons:

  1. I don't need to. I do not work for OS or have a contract with them.
  2. OS is not an authority on hill lists and does not claim to be. My main motivation for surveying is to make hill lists more accurate. There is no reason for OS to be interested in this.
  3. OS mapping standards are nowhere near accurate enough for my purposes. Heights obtained from my GNSS surveys are far more accurate than those from aerial surveys. My standards are necessarily high but not higher than they need to be for the task.
  4. I use the OS standards that are relevant and necessary: the OS geoid model, the OSTN02 national grid transformation, and Rinex data for OS Net base stations.
  5. OS validation of my survey results would not prove that they gave the correct hill or col height, so would be largely meaningless. All OS could do is validate that the GNSS data I collected is accurate to within 5-10cm. They can not validate that I was at the highest point of a hill, or at the critical point of a col, or whether I followed certain protocols. OS do not know whether I have removed a cairn to locate the highest point underneath. I could submit photographs, but they can be misleading. Many hills have multiple cairns and alternative summits.
  6. Hill heights are an understandably low priority for OS. I have the impression that correcting spot heights on maps is a chore that OS could do without. One email I received said that 'We do not update spot heights except for when significant change occurs that changes the elevation of the surface that they lie on, in which cases spot heights are deleted.' For example, when I notified them that Gragareth was the highest point in Lancashire (100m east of the trig pillar), rather than Green Hill as shown by OS maps, they said: 'after consideration we have concluded that the change does not impact our maps and is not a change to a prominent heights (sic) such as a Munro so there is no need for us to process the data.'
  7. OS do not want my survey data. In one email they said: 'I regret to inform you that Ordnance Survey collects its own height data and does not use survey information from individual members of the public, so we would not use your height surveys in our data, despite the lengths you have gone to in ensuring reliability.' However, this is not entirely true, as a different member of staff asserted that 'cases where Ordnance Survey will likely want to reprocess the data are where the result is close (+/-0.1m) to a mountain classification threshold'. Most of my results are not within 0.1m of a threshold, so OS would have no reason to reprocess the data.
  8. Every measurement involves sources of possible error. On some surveys, e.g. bare rock summits, this margin for error is very small. On others it is substantial, especially on summits with large cairns, where the margin for error can be up to a metre unless the cairn is dismantled. Cols are very variable in nature and often have 10-20cm of uncertainty, sometimes more. Many summits and cols are vegetated, and many cols are within a large area of peat hags. There is a limit to the accuracy that can be obtained from a survey of a heathery or tussocky summit unless the vegetation is removed. The results are still valid, but the margin for error is greater. In order for OS to accept GNSS data for its maps, it specifies a minimum data collection time that is not necessary for hill surveying, except in rare cases. A long survey time can do little to increase the accuracy of a result if there are other sources of potential error that are greater.
  9. I can get useful information from my survey equipment about the probable accuracy of the final result, e.g. the number of satellites available and the figures for VDOP (vertical dilution of precision) and CQ (coordinate quality). I can therefore vary data collection time by taking these into account, along with the importance of the hill, the probability of the result being marginal, and the margin for error on the ground. Rigidly following the same recommended OS specifications would be a considerable waste of my time and a probable cause of hypothermia.
  10. When processing survey data to obtain final results, there are several indications of the accuracy and reliability of the data: the number of OS base stations used, the solution type used, average variation of figures from different base stations etc. On rare occasions when the data quality has been poor for a specific point, due to atmospheric conditions, low satellite coverage or signal interference from trees, then I have not used the data.
  11. If I obtain a result that is within 10cm of a critical height or drop threshold, I go back and resurvey that point with a longer data collection time, so I have surveyed some summits and cols twice. I find this a more useful and time-effective strategy than surveying a much smaller number of hills for an unnecessarily long time.

So much for the words, what about the data? As I am diligent and pedantic, I am naturally interested in validation and replication as useful techniques for confirming the accuracy of my survey methods and data. To date I have used six different methods of data verification.

  • Validation against the OS fundamental benchmark (FBM) in Glen Almond. OS height is 213.365m, survey result was 213.356m, difference 0.009m.
  • Validation against flush brackets on OS trig pillars. Heights of these are not as precise as FBMs but they provide a useful and regular means of ensuring that my equipment is in good working order, as any significant inaccuracy would be obvious.
  • Validation by replication. I have so far repeated surveys of 17 points - ten summits and seven cols. The average difference between the pairs of points is 0.037m. Average survey time was 32 minutes. These findings do not prove accuracy beyond doubt, but they do show consistency and give a strong indication of accuracy.
  • Comparison with results from G&J Surveys for the same summit:
Hill AD mins G&J mins AD height G&J height Difference
Beinn Bhreac 16 60 912.41 912.44 0.03
Cairnie Hill 60 60 228.83 228.85 0.02
Carn na Caim South Top 30 126 914.56 914.61 0.05
Creag na Caillich 25 128 914.31 914.30 0.01
Creag na Caillich (resurvey) 50   914.28   0.02
The Fara 22 120 911.32 911.36 0.04

Having the same summit surveyed by different teams on different occasions with different equipment giving results consistently within 5cm or less of each other is a strong indication of accuracy. Data collection time made little difference to these results. Conditions were good for all my surveys apart from Carn na Caim South Top, where satellite coverage was low and weather poor. G&J Surveys also resurveyed the col for this hill. They obtained a result that was 0.1m lower than mine, which increased the drop from 30.3m to 30.4m. This G&J result is likely to be more accurate than mine, as they had better overhead conditions and spent more time establishing the optimum col location amongst all the peat hags.

  • Validation of two summit heights by OS reprocessing of my survey data. The OS height for Creag an Dail Bheag was 0.004m higher than my result, and 0.001m higher for Carn Liath. OS assessment was 'the height accuracy of the survey is in the region of +-0.05m'. This was useful to know, though of course OS could not validate whether my surveys took place on the highest point of either hill.
  • Validation using a method suggested by OS. After some persistence, I was able to obtain a list of twelve criteria required by OS for acceptance of GNSS survey data. One of these is that 'Quality check on the processed baselines should be via coordinate recovery of the OS Net stations from an MCE (minimally constrained estimation) adjustment holding the nearest OS Net station fixed to its accepted ETRS89 coordinates. OS Net coordinates should be recovered to better than 0.1m in all dimensions for the solution to be acceptable.' What OS are essentially asking for is a guarantee that results are accurate to within 0.1m. That is fine, as it is about the same standard I expect from my survey results. It is 30-40 times more accurate than the 3-4m accuracy that OS and other organisations accept for heights derived from aerial surveys.

Providing this guarantee using the method recommended by OS is time-consuming and fiddly. Very few of my survey results are of interest to OS, so it would be pointless for me to apply this process to most of my data, even where all eleven other criteria are met. However, it is another form of validation, so I applied this OS method to assess the level of accuracy for selected results. For example, I was able to calculate that the 609.0m height of Ben Aslak is accurate to within 0.034m, validated against seven OS Net base stations.

I am happy to help OS make their maps more accurate, but I have found it difficult to engage their interest in heights of hills. It has been easier to do that by drawing attention to errors or discrepancies on OS maps than by submitting survey data. For example, they have corrected mapped heights for Maoile Lunndaidh, Sgurr Dubh Mor, Sgurr Dubh an Da Bheinn, Mynydd Gethin and others because of information I have supplied to them about their own maps. OS mapping of Maoile Lunndaidh is now correct, but another survey could still be useful if the cairn was removed, in order to be certain of the highest point.

OS has a right to define and apply whatever policies, standards and methods it wishes for its maps. It is a commercially-driven limited company that has a right to be wrong if it wishes, as long as it does not wilfully mislead purchasers of its products.

Similarly, I have a right to use whatever policies, standards and methods I see fit for my hill lists. I am not infallible and if I make a mistake I will acknowledge that, though I try to make sure that it does not happen very often. I do not have a right to impose my standards on any other list or organisation.

One of the things I always liked about hill walking is its sheer simplicity. Not much equipment needed, no permission necessary, no schedules or deadlines, no hassles or disputes, no corporate crap, just put one foot in front of the other and keep on doing that. Things are more complicated now with all this data and all these hill categories, and I am as responsible for that as anyone. I get satisfaction from the process of surveying and making hill lists more accurate, but I still like to keep things as simple as possible. I like to be able to pack a bag and head off at short notice when conditions are favourable. I can carry all my own gear, move quickly and get the job done well by following my own methods and backing my own judgement. I can do a lot on my own because I have access to fantastic technology, because I am interested and experienced and because I take care, in a diligent, scientific and accurate but simple way. Well, relatively simple.