coordinatePrecision - Darwin Core Hour Input Form 2/8/2017 3:10:55

[iDigBioBot]

A user submitted this information via the Darwin Core Hour webform:
Timestamp: 2/8/2017 3:10:55
Please provide a topic of interest: Acceptable values for coordinatePrecision
Are you capable of and interested in participating: No
Who else would you recommend to participate in the presentation: @tucotuco @ArthurChapman @dagendresen
What resources can you point to: http://rs.tdwg.org/dwc/terms/index.htm#coordinatePrecision
Your name: @cgendreau
Your email:

[cgendreau]

I would like to have some precisions on different aspect of the term coordinatePrecision

1. Does this term represent the precision of the physical instrument (e.g. GPS) used to "read" the decimalLatitude and decimalLongitude ? If yes, does it represent the manufacturer declared value or the value read on the unit when the event occurred ?
2. Can this term be calculated afterward ?
3. Does this term correlate with coordinateUncertaintyInMeters ?

[ArthurChapman]

I will leave others to answer some of this, but I will address one part of this. The accuracy as given on a GPS instrument is not a true accuracy - it is the precision of the instrument. Basically, it is telling you that repeated measurements of the instrument are within x meters of each other. This bears no relationship to accuracy in the ground. The accuracy (really precision) reported by the instrument may be 2m, but actually be 7 meters from the real point as determined by a differential GPS. Currently (as of December 2016), the GPS signal in space (in the USA) will provide a "worst case" pseudorange accuracy of 7.8 meters at a 95% confidence level. This will vary in other parts of the world.

So, in summary, the value given on your GPS receiver for accuracy is the precision of the instrument itself and has nothing to do with accuracy on the ground. It DOES NOT correlate with coordinateUncertaintInMeters.

[dagendresen]

Remember that we also have many specimens collected in an earlier time when GPS was not yet available. I believe the very first handheld GPS was introduced around 1989. And before 2000 there was a military use restriction for accurate satellite signals. Many of the geographic coordinates for museum specimens since around the second world war were recorded using the MGRS reference system, and still today many use different UTM grid systems to record geographic coordinates for new specimens collected. Earlier collections have only place names given and no original coordinates.

Based on the number of digits in MGRS coordinates reported, the Norwegian GBIF node normally calculates the appropriate coordinateUncertaintyInMeters for Norwegian datasets before publishing in GBIF. 10 numerical digits (5 digits in each "tuple") equals a grid cell size precision of 1 meter, 8 numerical digits equal 10 meter precision etc. We have also some routines to estimate the coordinateUncertaintyInMeters based on the size of a reported administrative unit such as the size of a municipality or a reported named place.

So yes, we do calculate coordinateUncertaintyInMeters in many situations when it is not provided from a handheld GPS device.

I understand the preference for using coordinatePrecision (given in decimal-degree form) for consistency together with decimalLatitude and decimalLongitude. However, the coordinateUncertaintyInMeters is more intuitive for people working in the field - and in particular because many people (at least in Norway) actually still use UTM tuples (given in meters northing and easting) in the field.

Because the earth is round a given decimal-degree coordinatePrecision value will correspond to a different number of meters in the north-south direction as in the east-west direction. For locations further north (or south), a given number for coordinatePrecision (in decimal-degree form) corresponds to less actual meters in the east-west direction than in the north-south direction. In the north-south direction the number of meters per a given decimal-degree coordinatePrecision will remain the same when going north.

So, coordinatePrecision does not correlate perfectly well to coordinateUncertaintyInMeters if you consider both latitude and longitude direction.

I vote anyway for making interpretations between coordinatePrecision and coordinateUncertaintyInMeters. How about simply making the calculation for the north-south direction and simply accept that a coordinate precision circle in a UTM reference system (or simply drawn "directly on the ground") will have an elliptic shape?

[pzermoglio]

See #29

[tucotuco]

In answer to the original question, "[What are] acceptable values for coordinatePrecision?" here is a table showing the English language equivalent and the coordinatePrecision value. Any missing precisions should be possible to interpolate.

coordinate precision	coordinatePrecision
nearest degree	1
nearest half degree	0.5
nearest quarter of a degree	0.25
nearest tenth of a degree	0.1
nearest hundredth of a degree	0.01
nearest thousandth of a degree	0.001
degree to four decimal places	0.0001
*degree to five decimal places	0.00001
nearest 10 minutes	0.1666667
nearest minute	0.0166667
nearest tenth of a minute	0.0016667
nearest hundredth of a minute	0.0001667
*nearest thousandth of a minute	0.0000167
nearest 10 seconds	0.0027778
nearest second	0.0002778
nearest tenth of a second	0.0000278
*nearest hundredth of a second	0.0000028

* these three levels of precision are commonly found on commercial GPS devices in the three modes: decimal degrees, degrees decimal minutes, and degrees minutes seconds.

There is a nice treatment of accuracy and precision, including references to GPS capabilities, here.

[tucotuco]

I'd like to follow up on the enumerated questions from @cgendreau from the first comment and hopefully clarify what has been a mix of concepts.

Question 1: "Does this term [coordinatePrecision] represent the precision of the physical instrument (e.g. GPS) used to "read" the decimalLatitude and decimalLongitude ? If yes, does it represent the manufacturer declared value or the value read on the unit when the event occurred?"

Response to Question 1: Strictly speaking, no. The coordinatePrecision is a numerical value to represent the precision of the coordinates provided in decimalLatitude and decimalLongitude, with values such as those given in the table above. It may or may not have anything to do with a GPS, and almost certainly has nothing to do with the accuracy of the GPS, even if a GPS was used to determine the coordinates. Though poor coordinate precision can affect accuracy, GPS accuracy is a distinct measure. Both accuracy and precision contribute to coordinateUncertaintyInMeters. The distinction between these concepts, and how they contribute to uncertainty is taught in georeferencing workshops and is not trivial. These concepts are covered in the Georeferencing Quick Reference Guide, The point-radius method for georeferencing locality descriptions and
calculating associated uncertainty, and the Guide to Best Practices for Georeferencing. The Georeferencing Calculator can be used to explore the effects of precision and accuracy on 'coordinateUncertaintyInMeters`.

Manufacturer's tend not to declare an accuracy for a GPS unit, because accuracy is more dependent on environment and satellite configuration at the place and time of the Event than anything inherent in the device. Some GPS units do declare "accuracy" in their interfaces, but this is a theoretical accuracy based on current conditions, not a measured accuracy.

Question 2: "Can this term [coordinatePrecision] be calculated afterward?"

Response to Question 2: Strictly speaking, no. However, one can infer the minimum coordinatePrecision of a coordinate pair, or the coordinatePrecision for the coordinates in a data set, if they are consistent. For example, if the decimalLatitude and decimalLongitude values in an entire data set are given to two decimal places and all of the decimal parts are multiples of a quarter degree (.00, .25, .50, .75), one could infer that the coordinatePrecision for each record was to the nearest quarter of a degree, or 0.25. Note that this inference could be much weaker for a single coordinate pair.

Question 3: "Does this term correlate with coordinateUncertaintyInMeters?"

Response to Question 3: The coordinatePrecision contributes to the coordinateUncertaintyInMeters in ways described earlier in this comment and in previous comments. The contribution of coordinatePrecision to coordinateUncertaintyInMeters is location dependent and calculable (see references cited above).