UNCERTAINTY_NOT_SPECIFIED - AtlasOfLivingAustralia/ala-dataquality Wiki

Description

Uncertainty in metres was not supplied with the record.

Related fields

Implications and caveats

Severity: Warning

The uncertainty in metres value is defined as the radius (from the coordinates) of the smallest circle the location could fit within.

If this value is not supplied, it is difficult to know whether the occurrence happened exactly at the coordinates given, within 5 metres of the coordinates, within 50 metres of the coordinates or within an greater area. The level of accuracy may be important for various purposes, including the ongoing monitoring of an organism or determining the habitat of the organism.

This article provides information on the point radius method of georeferencing and estimating uncertainty: https://www.tandfonline.com/doi/pdf/10.1080/13658810412331280211

Where this value is not provided, or a very high value is provided, it indicates the accuracy of the coordinates may not be high.

Data custodian recommendations

If possible, provide a value for coordinate uncertainty in metres. This may be able to be determined from the methodology used to calculate the coordinates.

Code reference

ALAOccurrenceIssue.UNCERTAINTY_NOT_SPECIFIED

Discussion


Name UNCERTAINTY_NOT_SPECIFIED
Code 27
Description Uncertainty was not supplied with the record
Severity/Verification/Metric Warning
Failure implication See discussion below
Revision Date Extensively revised 28 October 2011

How should we represent spatial uncertainty of occurrence records when precision and/or extent of location are not supplied?

Occurrence locations are typically represented using the point–radius method with latitude/longitude (DwC:decimalLatitude/DwC:decimalLongitude) being the centre point of the location, and coordinate uncertainty (DwC:coordinateUncertaintyInMeters) representing the radius of a circle that covers the whole location, including the extent of the location and ALL errors and uncertainties associated with determining the coordinates.

DwC:coordinateUncertaintyInMeters is exactly equivalent to Maximum Error Distance used in some georeferencing tools (J. Wieczorek in litt.).


In a typical Australian situation coordinateUncertaintyInMeters would often be calculated by adding the four following terms:

See example.


In Australia there has been considerable confusion regarding spatial precision and uncertainty, with the terms sometimes used interchangeably. DwC:coordinateUncertaintyInMeters has often been interpreted to be just the extent of the location, rather than the extent plus all other errors and uncertainties associated with determining the coordinates e.g. georeferencing method; DwC:coordinatePrecision and DwC:geodeticDatum. This confusion probably relates to there being no Darwin Core term for the actual extent of the location, and that the Darwin Core description of coordinateUncertaintyInMeters does not explicitly mention that is it is inclusive of all uncertainty metrics: "The horizontal distance (in meters) from the given decimalLatitude and decimalLongitude describing the smallest circle containing the whole of the Location. Leave the value empty if the uncertainty is unknown, cannot be estimated, or is not applicable (because there are no coordinates). Zero is not a valid value for this term".


With regard to spatial uncertainty, it is proposed that

  1. Where recorded the extent of the location should be explicitly entered in a term new called locationExtent. Examples of location extent are the many millions of observational occurrences contained in the BA Bird Atlas data collected from 2ha, 500m and 5km survey plot locations, and historically from 10’ and one degree location grids. It is proposed that ALA nominated locationExtent be added as a term to the Darwin Core standard.

  2. Where provided DwC:coordinatePrecision and DwC:coordinateUncertaintyInMeters (the overall metric) be also be saved as verbatim values in fields verbatimCoordinateUncertaintyInMeters and verbatimCoordonatePrecision. This will ensure consumers are aware of the original resolution of the data.

  3. Where coordinates are generalized for the ALA Sensitive Data Service (typically by dropping decimal places from the coordinates, i.e. affecting the coordinatePrecision) then the Dwc:dataGeneralizations field be appropriately annotated; and DwC:coordinatePrecision and DwC:coordinateUncertaintyInMeters be adjusted thus:

Where verbatimCoordinateUncertaintyInMeters is explicitly available then:

DwC:coordinateUncertaintyInMeters = verbatimCoordinateUncertaintyInMeters and where coordinates are generalised by ALA then: DwC:coordinateUncertaintyInMeters = verbatimCoordinateUncertaintyInMeters + additional uncertainty from rounding coordinates, DwC:coordinatePrecision would be altered to relect the adjusted number of decimals.

Where DwC:coordinateUncertaintyInMeters is NOT available then:

DwC:coordinateUncertaintyInMeters is calculated by adding four nominal terms to represent overall uncertainty. The four nominal terms are (i) coordinate precision: the decimal representation of the precision of the coordinates, converted to metres (if not available label as ‘Unknown’ and substitute 150 metres); (ii) locationExtent: i.e. extent of the location (where location extent is not available label as ‘Unknown’ and substitute 1000 metres); (iii) potential datum shift: where the datum (DwC:geodeticDatum) is not available label as ‘Unknown’ and add an additional 200 metres. Do this also if the coordinates have not been projected to GDA 94 or WGS 84 coordinate system. (iv) geocode uncertainty : i.e. GPS accuracy or map scale uncertainty, an additional 150 metres

Therefore default DwC:coordinateUncertaintyInMetere = 150+1000+200+150 = 1500 metres

  1. Presently where uncertainty is specified some ALA spatial tools plot an uncertainty circle, while for visual clarity no circle is drawn when that field has an ‘Unknown’ value. Where this occurs there should be an explicit visual and textual indication that the location is in fact quite general and does not represent a point location.

It may be preferable that when the circle is shown it should represent the overall coordinateUncertaintyInMeters, and also be plotted when terms are ‘Unknown’. It would be worth conducting trials to obtain an optimal solution, perhaps with different shading/outline symbology where the terms are ‘Unknown’, e.g. if unknown a pale translucent circle with dotted outline could be plotted, while for specified values a darker circle with solid outline could be plotted.

Note: It may be possible to assign more appropriate substitute values for certain data sets, for example data from ALA ‘Share a Sighting’ tool could use a default coordinate precision of 0.00001 degrees or 2 metres, with the datum error as 0, and with the Coordinate uncertainty (extent) being defined by the user. Also, there is a trend to increased spatial precision through time, e.g. records in 1800's are generally much less precisely located than those geocoded today using GPS or Google maps.


A detailed explanation of each of the terms follows:

Overall Coordinate Uncertainty

Coordinate uncertainty (DwC:coordinateUncertaintyInMeters) represents the radius of a circle that covers the whole location, including the extent of the location and ALL errors and uncertainties associated with determining the coordinates. DwC:coordinateUncertaintyInMeters is exactly equivalent to Maximum Error Distance used in some georeferencing tools (J. Wieczorek in litt.). It is proposed that where altered the original values be displayed as verbatimCoordinateUncertiantyInMeters.

Uncertainty associated with geocoding method

These uncertainties generally relate to the method (DwC:georeferenceProtocol) used to determine the latitude and longitude. Generally coordinates determined from GPS are quite precise (5-10 metres), especially if the Horizontal Dilution of Precision (HDOP) is also recorded. Coordinates determined from modern satellite maps, for example, Google Maps/Earth are also quite precise (although beware of satellite image rectification issues). On the other hand coordinates determined from say 1:1,000,000 maps or from gazetteers presenting coordinates to one minute are much less precise. Where this values is unknown, generally substitute about 150 meters i.e. the approximate average error of reading geocode from a 1:100,000 to 1:250,000 map sheet.

Coordinate precision

The precision of the location is defined by DwC:coordinatePrecision ‘A decimal representation of the precision of the coordinates given in the decimalLatitude and decimalLongitude. Examples: "0.00001" (normal GPS limit for decimal degrees), "0.000278" (nearest second), "0.01667" (nearest minute), "1.0" (nearest degree)’.

Where coordinate precision is ‘Unknown’, generally substitute 0.001 degrees or about 150 metres.

Sometimes the coordinates are generalised to protect sensitive species locations. Common methods are to (i) round or truncate coordinates to fewer decimal places, (ii) grid the data or (iii) randomly move the location. Where altered the original values be displayed as verbatimCoordinatePrecsion. Any generalisations need to be factored into the overall coordinateUncertaintyInMeters.

It is important to note that the number of decimal places in the coordinates cannot be relied upon to explicitly indicate the precision of the coordinates as they are often (unknowingly) rounded, truncated and imprecisely transcribed.

Note that the conversion of coordinate precision to metres varies with latitude; see Guide to Best Practices for Georeferencing page 27.

Extent of the location: locationExtent

Where ‘Unknown’ generally substitute 1000 metres

The extent of the location is not presently defined in the Darwin Core standard, although is often used in the calculation of the overall spatial uncertainty, i.e. DwC:coordinateUncertaintyInMeters.

The extent of a location varies greatly, for example, it may be an actual point within a metre or two of the coordinate; it could be a general location such as Black Mountain Reserve which is about 3.2 km across (locationExtent of about 1600 metres); or it may be a standardised survey plot like the 2ha areas used by Birds Australia (locationExtent of 100 metres).

Geodetic datum and spatial reference system

Where ‘Unknown’ generally substitute 200 metres.

For a latitude/longitude coordinate to be precisely located you must also know the spatial reference system under which it was collected: DwC:geodeticDatum. The current global standard is WGS84 (used by GPS, Google Maps), with the Australian standard GDA94 being virtually identical. Coordinates collected under older Australian standards: AGD 1966 and 1984 are in a different physical location to those of the modern standards, and so may be mis-mapped by 200 metres if not accounted for (see http://www.icsm.gov.au/gda/mapsgda.pdf ). This potential inaccuracy is called datum shift. A fuller definition of the Spatial Reference System may be contained in DwC:verbatimSRS, but this is generally not needed if the datum is available.

References

Chapman, A.D and J. Wieczorek (eds). 2006. Guide to Best Practices for Georeferencing. Copenhagen: Global Biodiversity Information Facility. Available at http://manisnet.org/GeorefGuide.html

TDWG. 2009. Darwin Core Terms: A quick reference guide. Aaialable at http://rs.tdwg.org/dwc/terms/

Wieczorek, J. 2007 Georeferencing Calculator, Version 070228. Available at http://manisnet.org/gc.html, with guidelines at http://manisnet.org/GeorefGuide.html


Examples from Georeferencing Calculator, http://manisnet.org/gc.html for Charleville Qld (midway NS across Australia). MED is Maximum Error Distance aka coordinateUncertaintyInMeters (CUIM)

Assuming the location is an exact point:

Scenario Geocoding method Uncertainty from geocoding Latitude Longitude Datum Datum Shift (metres) Coordinate precision (degrees) Coordinate precision (metres) Extent of Location MED(m) CUIM presentation (m)
Google Earth (always WGS84) Google Earth 10m accuracy 10 -26.40066 146.24122 WGS84 0 0.00001 1.5 0 11.5 10
Google Earth (always WGS84), and generalised to 2 decimal places Google Earth 10m accuracy 10 -26.4 146.24 WGS84 0 0.01 1500 0 1510 1500
GPS, datum recorded GPS 10m accuracy 10 -26.40066 146.24122 WGS84 0 0.00001 1.5 0 11.5 10
GPS, datum recorded, and generalised to 2 decimal places GPS 10m accuracy 10 -26.4 146.24 WGS84 0 0.01 1500 0 1510 1500
GPS, datum unknown GPS 10m accuracy 10 26.40066 146.24122 ? 200 0.00001 1.5 0 211.5 200
GPS, datum unknown, and generalised to 2 decimal places GPS 10m accuracy 10 -26.4 146.24 ? 200 0.01 1500 0 1710 1500
Bare coordinates 5 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.40066 146.24122 ? 200 0.00001 1.5 0 1201.5 1200
Bare coordinates 4 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.4006 146.2412 ? 200 0.0001 15 0 1215 1200
Bare coordinates 3 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.4 146.241 ? 200 0.001 150 0 1350 1500
Bare coordinates 2 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.4 146.24 ? 200 0.01 1500 0 2700 3000
Bare coordinates 1 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.4 146.2 ? 200 0.1 15000 0 16200 15000
Bare coordinates 1 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26 146 ? 200 0 150000 0 151200 150000

Assuming the location has an extent of 1000 metres (radius):

Scenario Geocoding method Uncertainty from geocoding Latitude Longitude Datum Datum Shift (metres) Coordinate precision (degrees) Coordinate precision (metres) Extent of Location MED(m) CUIM presentation (m)
Google Earth (always WGS84) Google Earth 10m accuracy 10 -26.40066 146.24122 WGS84 0 0.00001 1.5 1000 1011.5 1000
Google Earth (always WGS84), and generalised to 2 decimal places Google Earth 10m accuracy 10 -26.4 146.24 WGS84 0 0.01 1500 1000 2510 2500
GPS, datum recorded GPS 10m accuracy 10 -26.40066 146.24122 WGS84 0 0.00001 1.5 1000 1011.5 1000
GPS, datum recorded, and generalised to 2 decimal places GPS 10m accuracy 10 -26.4 146.24 WGS84 0 0.01 1500 1000 2510 2500
GPS, datum unknown GPS 10m accuracy 10 26.40066 146.24122 ? 200 0.00001 1.5 1000 1211.5 1200
GPS, datum unknown, and generalised to 2 decimal places GPS 10m accuracy 10 -26.4 146.24 ? 200 0.01 1500 1000 2710 3000
Bare coordinates 5 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.40066 146.24122 ? 200 0.00001 1.5 1000 2201.5 2000
Bare coordinates 4 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.4006 146.2412 ? 200 0.0001 15 1000 2215 2000
Bare coordinates 3 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.4 146.241 ? 200 0.001 150 1000 2350 2500
Bare coordinates 2 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.4 146.24 ? 200 0.01 1500 1000 3700 4000
Bare coordinates 1 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26.4 146.2 ? 200 0.1 15000 1000 17200 15000
Bare coordinates 1 decimals, assume geocoded from 1:1,000,000 map 1:1,000,000 map 1000 -26 146 ? 200 0 150000 1000 152200 150000

Note: CUD of earlier versions has been abandoned in favour of recording verbatimCoordinateUncertaintyInMeters and verbatimCoordinateArecision.

Replaces

uncertaintyNotSpecified