Positional accuracy improvement programme

<<< FAQs and comments | OS MasterMap FAQs >>>

General FAQs

How does PAI affect my organisation?

Due to the improved accuracy of the 1:2500 scale mapping, you will now see that features compared before and after data has been through the PAI programme may have been moved. If you have associated your own assets to fit in with our 1:2500 scale mapping, you may find that they do not fit with the new positionally-improved data.

Return to menu

In discussions on PAI, I have heard there are different measures of accuracy. Please explain.

When a 1:2500 scale accuracy mapped area is surveyed to improve or revise the map content there are a number of accuracies that must be met, and are in order of precedence:

1. Geometric fidelity

The principle of geometric fidelity is that any real-world alignment or shape must be accurately reflected in the data to the required specification, for example:

• detail that is square on the ground must be represented as square in the data, and shapes must be accurate;
• alignments that are straight in real life must be represented as straight lines within the data;
• lines of sight that pass through points on the ground should pass through the map positions of the corresponding points; and
• adjacent features should be in sympathy with each other as regards alignment, distance apart and orientation.

2. Relative accuracy

Relative accuracy is a measure of the positional consistency of a data point in relation to other near points of detail. Relative accuracy compares the scaled distance between features measured from the map data with distances measured between the same features on the ground.

3. Absolute accuracy

Absolute accuracy is a measure that indicates how closely the coordinates of a point in the map dataset agree with the real coordinates of the same point on the ground in the British National Grid reference system.

Conformity

The Positional accuracy improvement programme deals with improving the absolute accuracy from current 1:2500 scale overhaul accuracies to either 1:1250 scale resurvey standards (for the built-up areas of defined rural towns) or 1:2500 scale resurvey standards (the remaining rural areas). However, any surveyed changes must also conform to the geometric fidelity and relative accuracy of those standards of survey. The conformity statements below define these accuracy levels.

Note that certain types of feature, such as road centrelines, vegetation and landform limits and underground features, are surveyed to a lesser degree of accuracy. Centrelines and vegetation and landform limits are subjective and are not surveyed while underground features may have been supplied by third parties.

The absolute and relative accuracies are measurable and definitive, and statements of the expected errors of these by survey scale are given. Geometric fidelity cannot be closely defined and is a matter for subjective judgement. The guideline is that the detail must be acceptable in terms of geometric fidelity when plotted or displayed at a scale no larger than the storage scale.

Relative accuracy (of built-up areas within defined rural towns)

A relative accuracy of ± 0.40 m root mean square error (rmse), a normal distribution of errors and a maximum error of 0.80 m must exist. This standard is applied to any contiguous area of data.

Example of relative accuracy

If the distances between two well defined points of detail 60.0 m apart were measured in the real world, there would be an expectation that this distance would be represented in the map data by a scaled distance of between 59.2 m and 60.8 m.

Relative accuracy (outside of built-up areas within defined rural towns plus other rural areas)

A relative accuracy of ±1.00 m rmse, a normal distribution of errors and a maximum error of 1.90 m must exist. This standard is applied to any contiguous area of data.

Example of relative accuracy

If the distances between two well defined points of detail 100.0 m apart were measured in the real world, there would be an expectation that this distance would be represented in the map data by a scaled distance of between 98.1 m and 101.9 m.

Absolute accuracy (built-up areas in defined rural towns)

An absolute accuracy of ±0.4 m rmse and a normal distribution of errors must exist. This standard is applied to any contiguous area of data as per the following:

• 95% of points should be in error by no more than ± 0.7 m;
• 99% of points should be in error by no more than ± 0.9 m; and
• no point should be in error by more than 1.2 m.

Absolute accuracy (outside of built-up areas within defined rural towns plus other rural areas)

An absolute accuracy of ±1.10 m rmse and a normal distribution of errors must exist. This standard is applied to any contiguous area of data as per the following:

• 95% of points should be in error by no more than ± 1.90 m;
• 99% of points should be in error by no more than ± 2.40 m; and
• no point should be in error by more than 3.0 m.

Return to menu

Apart from a different absolute accuracy level what other differences are there between the PAI of rural towns and the PAI of other rural areas?

Survey tolerances

Ordnance Survey map data aims to portray an accurate survey of the features which comprise the detailed topography, so that each feature is correctly depicted in terms of shape, size, orientation and position relative to all the surrounding features. However, the accuracy levels chosen at the basic scales of survey restrict the absolute achievement of this objective.
For that reason information is selected="selected" for inclusion so that display and graphic plotting of the data gives optimum results at the accuracy of survey. This requires there to be differing tolerances for data capture of similar features at the different accuracy survey levels.

Constraints imposed by survey tolerances

Juts and porches which are an integral part of a building, bay windows and recesses are shown when their smallest dimension is not less than:
1.0 m – Resurvey of rural towns for PAI
2.0 m – PAI of other rural areas
Nevertheless, smaller juts and projections are shown when they abut onto a public thoroughfare or they carry height information.

Rivers, streams and drains are shown at their true scale width or by a single line where their width is less than:
1.0 m – Resurvey of rural towns for PAI
2.0 m – PAI of other rural areas

Where the central alignment of an unmade path is less than 1 m (Resurvey of a rural town for PAI), 2 m (PAI of other rural areas) from an adjacent building, fence, hedge or wall, the central alignment is shown at that minimum distance away from the feature.

The symbol used for square pylons and similar features is shown oriented and true to scale, except where the feature is below the minimum size, in which case the minimum sized symbol is used. The minimum sizes are:
3 m² area – Resurvey of rural town for PAI
6 m² area – PAI of other rural areas

Where the limits of cliffs and slopes are coextensive with features such as fences, the limits of cliff or slope are offset by approximately 0.5 m at source scale for clarity.
Where areas of slope or cliff taper at their extremities, they cease to be shown when the minimum horizontal width is less than:
1.0 m – Resurvey of rural towns for PAI
2.0 m – PAI of other rural areas

(Extracted from the Ordnance Survey Land-Line User Guide v.4)

Return to menu

What time and resource will we need to cope with moving our data to the new map positions?

This is impossible for Ordnance Survey to estimate, since only you know the quantity and quality of your own data. Ordnance Survey is, however, working with customers and solution providers, not only to assist them but also to build up knowledge of best practice. As the programme advances these cases will be added to our web site, which should help individuals to calculate the scale of the task. Research is also underway to draw on experience from around the world, where other countries have undertaken similar improvements to their data. This too will be shared with customers online.

Return to menu

How is data matched at the external edges to positional accuracy improved blocks?

Whatever the size of the block that has undergone positional accuracy improvement (PAI), there will be an external edge to it and, during the course of the (PAI) programme, this may sometimes abut a 1:2500 scale map tile that has yet to be improved. Where this occurs, there will inevitably be some discrepancies between detail across tile edges.

At the external edge, any improved data (where the position has changed and the feature extends beyond the edge) is improved into the adjacent 1:2500 map tile (collar tile) if it is currently unimproved. This prevents movement again at the tile edge of the improved map when its neighbour is subsequently positional accuracy improved. As a consequence this results in small positional shifts on the unimproved tile to certain lines of detail.

This feathering of changes into the collar tile are restricted to the point where decreasing shifts for PAI no longer need to be made to still meet relative accuracy standards or to where revision can be taken to a logical conclusion. The geometric fidelity of features is retained at map tile edges.

If the adjacent map tile (collar tile) to a PAI block edge is a 1:1250 scale tile with a content indicator of A, or a 1:2500 scale tile with a content indicator of R, the data is already at or better than the accuracy of the PAI tile so will not be subjected to edge adjustment in terms of PAI. However there may be alterations at the map edge due to real world change (RWC) to the data.

Adjacent map tiles that are at 1:10000 scale will have the result of PAI changes feathered into the edge detail of these tiles. That is any accuracy improved or revised map data that crosses a PAI block edge into a 1:10 000 scale accuracy DMU will be continued to 1:2500 scale specification until the change can be stopped or blended in at a suitable point. Polygons are completed across the map tile edge.

However the unimproved map tile will not be part of the reissue of PAI map tiles to Land-Line customers and a link file is not created, although the map may be supplied as part of an organisation’s next normal supply depending on its supply regime. Such shifts, however, are transitory and will be resolved when the tile affected is processed through the PAI programme to bring it into line with its neighbour. OS MasterMap customers, however, receive all topography changes as part of the change-only update process.

Overall, the edgematch issue is comparatively small; it is estimated that, nationally, less than 8,500 of the 620,000, 1:2500 scale Land-Line map tile edges will be affected at any one time. That said, an organisation will need to recognise that these discrepancies between detail across tile edges will occur during the course of the PAI programme. Data presentation will, as a result, be affected during this period, which could impact on business services (internal and external) that organisations provide.

Link files and data transformation

Link files are only effective for the map area they cover so do not include any feathered data in the collar tiles to a PAI block. Each link file is titled by the south-west corner of the National Grid square that it covers.

In setting up transformation software, organisations will need to decide how they wish their data to be presented at the edge of the improved area. Much of this will depend on:

• the options in the way in which data can be handled at tile edges by the transformation software;
• the accuracy of the current data;
• the quality, quantity and type of an organisation’s data assets;
• the quality assurance processes that are needed to inspect the data after transformation;
• the amount of time the data, or a portion of it, can be withheld from general use;
• the use to which the data is put;
• the correlation between the layers of data;
• whether there are different business rules for each data layer; and
• whether different layers require different transformation algorithms.

Return to menu

Only 1:2500 scale data is being addressed under PAI but what happens to those tiles which have a mix of scales (Dual Accuracy)?

A map tile can contain a mixture of content indicators to show that it contains a mix of accuracies to the map data. For example a content indicator of AB shows that the map tile contains some map detail at ±0.4m accuracy and some at ±2.8m accuracy.

Content indicators and map tile formats of Land-Line data

All maps that contain a Content Indicator of B will be inspected during the PAI programme and any data not meeting the PAI accuracy standards will be adjusted. After PAI a map tile will have its content indicator changed from B to R or from AB to AR. Rural towns will have their content indicator changed from B to A or from BR to AR.

There are approximately:

• 9,000 500m² (1:1250 format) tiles with a content indicator of B (±2.8m accuracy), which will naturally be subject to positional accuracy improvement.
• 900 500m² (1:1250 format) tiles with a content indicator of AB (some data at ±0.4m accuracy and some at ±2.8m accuracy). Here only the detail captured to 2.8m will be improved to 1.1m as the 0.4m accuracy detail already meets accuracy standards.
• 2,500 1km² (1:2500 format) tiles with a content indicator of AB (some data at ±0.4m accuracy and some at ±2.8m accuracy) and the same rule applies as above.
• 152,500 1km² (1:2500 format) tiles with a content indicator of B.

<<< FAQs and comments | OS MasterMap FAQs >>>