Category: Data Quality
Element: ATTRIBUTE ACCURACY
Definition of Element:
Attribute accuracy is an assessment of the reliability of values assigned to features in the dataset in relation to their true 'real world' values.
The attribute accuracy generally includes:
A precise attribute accuracy assessment may not always be possible. In these cases an intuitive estimate of the expected attribute accuracy or the likely maximum error based on previous experience is acceptable. In many cases this may be "just a feel" for the data but it is important to state this.
Allowable content:
Free text. If no answer is possible, use one of the following Not Known, Not Documented or Not Relevant.
Format Rules: Sentences are preferred.
Field Type: Text
Length of Field: Maximum 1000 characters.
Other comments: Nil.
Example(s)
Example 1: [Status Report on Water Quality in the Tweed, Brunswick, Richmond and Clarence Rivers]
Water quality attributes were measured at site and others were analysed from samples sent to selected laboratories.
On site testing of the water quality was done using a Horiba U10 Water Checker (Japanese) and the Yeocal Dissolved Oxygen meter (Australian) and was done according to the US Standard Methods for the testing of water and waste water.
The selected laboratories were Richmond Pathology Services, Ensight Australian Water Technologies and Tweed Council Water Laboratories. The performance of these laboratories were tested against control laboratories which were the Department of Health, Division of Analytical Laboratories, Lidcombe and EPA Laboratory. All laboratories are NATA registered except for the Tweed Council Laboratory. The laboratory analyses were done according to methods described in the procedures manual for each laboratory and are generally based on US Standard Methods.
Example 2: [Forest Types]
The attribute of this data set is the forest type which is defined as any group of tree-dominated stands which possess a general similarity in composition and character. The classification in forest type was never intended as an ecological classification of forest vegetation in NSW. The forest type classification is based on economic and forest management considerations rather than on purely scientific ones. Considerations in using forest type classification were:
There are 192 forest types identified and described by the Forestry Commission of NSW. Each type is given a distinctive number as a numerical reference.
These 192 types are divided into three major groups. Within each of these groups the types are furthercombined into assemblages of related types, called "leagues:. Whilst the leagues must be recognised as being essentially artificial groupings, they in most cases serve to unite those types which are most closely related to each other. The numbers given to forest types run ordinal through each league, with a gap in the series between most leagues so that, should further types be subsequently needed, these can be fitted into the existing framework of leagues and major groups.
The Aerial Photograph Interpreters do ground checks on forest types to check and correct identification. Their route is recorded on the aerial photographs together with the check points. The resource reports record where indicator species are present and any discrepancies that exist between the aerial photograph and the ground.
Example 3: [Soil Landscapes of the Michelago 1:100 000 Sheet]
The attributes associated with the soil landscape maps are: the primary attribute, the soil landscape and the associated attribute, the soil materials.
The soil landscapes include Geology, Topography, Vegetation, Community land use, Existing erosion/land degradation, Included soil landscapes. Each of these can be taken out and shown in a derivative map. The soil landscapes are individualised by the recognisable topographies and soils which can be mapped into the same unit.
The soil materials is a categorical attribute stated in the map legends, and detailing the soil description (it is not mapped and consist of soil morphological characteristics and laboratory tests results). The detailed description is recorded in the report that accompanies the soil landscape map sheet.
The attribute accuracy, as the closeness to field reality, is unknown considering the fuzzy distribution and the composition heterogeneity of the primary attribute, soil landscape.
Example 4: [Brisbane Region Street and Address Network - Version 3.3
Street address numbers are sourced either from Local Government Authorities or field verification. Street classification consists of 8 categories. These are:
This street classification methodology does not conform to any national standard. One way indicators are stored for each street centreline as a "0" or "null" for bi-directional travel and "1" for positive direction travel.
The accuracy of the attribute information is estimated at 99%. Street names were originally sourced from the DCDB. The accuracy (ie. spelling and correct location) of these street names were then validated against Local Government Authority databases along with selected field verification. Checks for null entries appearing in the Road Name and Road Type attribute fields were also carried out.