Positional Accuracy Improvement (PAI) Programme

Positional Accuracy Improvement (PAI)

As part of an ongoing commitment to data quality, Ordnance Survey (OS) undertook a national Positional Accuracy Improvement (PAI) programme between 2001 and 2006. This programme improved the absolute accuracy (absolute accuracy is the position of features in relation to the OS National Grid) of all maps covering the 1:2500 scale rural mapping. The PAI programme has resulted in a more consistent accuracy standard of mapping for rural areas.

The reason for undertaking such a programme was that the majority of 1:2500 scale mapping was based on overhaul mapping from the 1950s. The overhaul process took the existing County Series mapping, each based on an individual projection, and converted it to the National Grid. There were limitations in that conversion process that resulted in ± 2.8 metres root mean square error (RMSE) absolute accuracy. However, the relative accuracy of features (distances between features close to each other) was very good, ± 1.2 metres RMSE, and is retained.

Details of the PAI programme

Work started in April 2001 on the two elements to the national programme; one covering rural towns, the other covering the remaining rural areas. The rural towns completed in December 2004 and the programme for all other 1:2500 scale rural areas completed in March 2006.

The PAI programme has delivered:

The re-survey of rural towns to an absolute accuracy of ± 0.4 metres root mean square error (RMSE); and
all other 1:2500 scale rural areas to an overall absolute accuracy of ± 1.1 metres RMSE.

The completion of the PAI Programme has resulted in considerable improvements to the levels of absolute and relative accuracy within the 1:2500 mapbase:

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

A relative accuracy of ± 0.40m RMSE, a normal distribution of errors and a maximum error

of 0.80m must exist. This standard is applied to any contiguous area of data

Example of relative accuracy

If the distance between two well-defined points of detail 60.0m 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 59.2m and 60.8m

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

areas)

A relative accuracy of ±1.00m RMSE, a normal distribution of errors and a maximum error

of 1.90m must exist. This standard is applied to any contiguous area of data

Example of relative accuracy

If the distance between two well-defined points of detail 100.0m 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 98.1m and 101.9m

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

Absolute accuracy (outside of built-up areas within defined rural towns and 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 ± 1.90m;
99% of points should be in error by no more than ± 2.40m; and
no point should be in error by more than 3.0m