11 How is the positional accuracy of spatial data measured?
11.1 Chrisman (1991, pp169-170) discussed the issue of how the positional accuracy of spatial data is measured in some depth. In brief some of the points to note from his paper are:
11.1.1 The geometrical accuracy of maps has been a concern since long before this data was computerised. He states that the ‘US National Map Accuracy Standard (Bureau of Budget 1947) considered the accuracy of well defined points’ as the sole measure of a map.’
11.1.2 A well-defined point is one that has no attribute ambiguity and can act as control for positional measurement e.g. a right-angled road intersection. Much of the data in a GIS does not ‘fit the restrictive definition of well-defined points…the uncertainty in positioning an ill-defined object must be added on to the error in the well-defined points.’
11.1.3 The accuracy of a product should be tested. This may be with either other features in a GIS database that have been captured to a greater degree of accuracy or via survey with e.g. a GPS unit.
11.1.4 He stated that most US Agencies infer their products would have passed the positional accuracy test, based on their ‘compliance with certain specified procedures and equipment.’ As such many of their products have probably not been actually tested.
11.1.5 Chrisman also briefly discusses the American Society of Photogrammetry and Remote Sensing’s (ASPRS) Accuracy Standards for Large-Scale Maps. This standard uses the concept of well-defined features as well and specifies acceptable levels of error, measured as root-mean-square deviation from zero.
11.2 Chrisman later re-visited this issue and again found that most testing methods are restricted to the use of well-defined points (Chrisman, 1997, pp118-121). Fisher (1991) also discusses both the US National Map Accuracy Standard and the ASPRS Accuracy Standards for Large-Scale Maps. He provides two tables describing the error allowable under both standards (Fisher, 1991, p179).
11.3 The concept of using a well-defined point to define map and spatial data accuracy is also used in Australia. Some examples of specific map accuracy statements made by some Australian organisations with regards to their products are:
11.3.1 Division of National Mapping, Sheet SD52-14, Edition 1, 1:250,000:
11.3.1.1 ‘The average accuracy of this map ±100 meters in the horizontal position of well defined detail and ±20 meters in elevation.’
11.3.2 Division of National Mapping, Sheet 5650, Edition 2, 1:100,000:
11.3.2.1 ‘The average accuracy of this map ±25 meters in the horizontal position of well defined detail and ±5 meters in elevation.’
11.3.3 Australian Surveying and Land Information Group (AUSLIG), Sheet 5073, Edition 4, 1:100,000:
11.3.3.1 ‘Horizontal Accuracy: ±50 meters.’
11.3.3.2 ‘Vertical Accuracy: ±10 meters.’
11.3.4 Royal Australian Survey Corps, Sheet 5072 Edition 2, 1:100,000:
11.3.4.1 ‘Horizontal: 90% of well defined detail within ±25 meters of true position.’
11.3.4.2 ‘Vertical: 90% of elevations within ±10 meters except in areas of dense vegetation where this may not be achieved.’
11.3.5 AUSLIG, Topo250K Metadata, 1:250,000:
11.3.5.1 According to the Data Quality Statement file ‘D5204HG.DQS’ that is delivered with AUSLIG’s vector data for SD5204, AUSLIG’s Topo250K series of vector spatial data sets have been derived from the media used to produce the 1:250,000 series of maps that cover Australia. AUSLIG used stringent Quality Assurance methods, procedures and testing. Statistical sampling procedures in accordance with Australian Standard AS1199 – 1988 were used to check the resultant datasets. The stated error rates are achieved with a 99% confidence.
11.3.5.2 "TOPO-250K data complies with the following statement of planimetric accuracy: “The summation of errors from all sources results in data with a standard deviation of 100 meters for well defined points.”’
11.3.5.3 This statement is further qualified by an exclusion of liability statement where AUSLIG does not warrant that the data is free from errors or omissions.
11.4 The US Federal Geographic Data Committee (FGDC) has also released the Geospatial Positioning Accuracy Standards (GPAS) in 1998. These standards include separate sections for Geodetic Networks and for Spatial Data Accuracy. There are also sections in draft form for Engineering, Construction and Facilities Management as well as for Navigation Charts and Hydrographic Surveys (FGDC, 1998).
11.5 The National Standard for Spatial Data Accuracy component of the GPAS also uses the concept of well-defined points to test for error. The preferred test for positional accuracy is to test the data against an independent source that is of higher accuracy. The standards further define the minimum number of points to test (i.e. twenty) and the preferred arrangement of these points within the dataset. The standards state (FGDC, 1998):
11.5.1 ‘The NSSDA uses root-mean-square error (RMSE) to estimate positional accuracy. RMSE is the square root of the average of the set of squared differences between dataset coordinate values and coordinate values from an independent source of higher accuracy for identical points.’
11.5.2 ‘Accuracy is reported in ground distances at the 95% confidence level. Accuracy reported at the 95% confidence level means that 95% of the positions in the dataset will have an error with respect to true ground position that is equal to or smaller than the reported accuracy value. The reported accuracy value reflects all uncertainties, including those introduced by geodetic control coordinates, compilation, and final computation of ground coordinate values in the product.’
11.6 As can be seen the accuracy of spatial data is determined in several ways, most relying on the concept of a ‘well-defined’ point. Most statements recognise that there is uncertainty inherent in spatial data and that no absolute statement of accuracy can be realistically applied to a given sample of spatial data.
11.7 False Precision and Accuracy An additional factor to be aware of is that of False Precision and Accuracy. Many GIS users, especially novice GIS users are unaware of the issues involved in spatial data accuracy, error and uncertainty, and assume that their data is absolute. They often report levels of accuracy that are unattainable with their source data. An example is a person quoting one-meter accuracy for data collected using their handheld GPS unit measuring GPS data that has Selective Availability enabled. This level of accuracy and precision is often quoted because the GPS unit displays coordinates to the nearest meter (Foote and Huebner, 1995).
To make suggestions or corrections to
this web page, please contact posacc@geoinnovations.com.au.
The information on this web page is © Copyright
1999 B.S.Bannerman.
All commercial rights reserved.
B.S.Bannerman materials published on the Internet are protected by copyright law. Apart from fair dealing for the purposes of private study, research, criticism or review, as permitted under the Copyright Act 1968, no part may be reproduced or reused for any commercial purposes whatsoever.