[Taxacom] Data quality of aggregated datasets

Richard Pyle deepreef at bishopmuseum.org
Thu May 9 12:58:02 CDT 2013


Arthur Chapman wrote me off-list to explain this issue.  Apparently, some GPS units provide a "precision" value in the neighborhood of a few m; but that's not a reflection of accuracy.  I suspect this is a feature used by some GPS manufacturers to make the end-user feel better about the numbers.   Most GPS units that I have used typically give an error in the neighborhood of 10-25m (when connected to multiple satellites) -- which is more along the lines of the actual accuracy of most (relatively inexpensive) GPS devices in most parts of the world (at least the parts that I go to).  And, of course, as someone already mentioned, there are circumstances where public GPS numbers are (deliberately) less accurate than this.

I suspect that Doug Yanega's approach is probably the best (i.e., when in doubt, "round up").  Perhaps something on the order of 50-100m radius would be appropriate to apply to numbers derived from an unknown GPS unit.  That, at least, gets you to within an order of magnitude or so of the actual accuracy, and is plenty accurate enough for most biodiversity analyses -- especially when the accuracy of retrospectively georeferenced locations is often on the scale of km.

Aloha,
Rich

> -----Original Message-----
> From: Karl Magnacca [mailto:kmagnacca at wesleyan.edu]
> Sent: Wednesday, May 08, 2013 10:02 PM
> To: taxacom at mailman.nhm.ku.edu
> Cc: Richard Pyle
> Subject: Re: [Taxacom] Data quality of aggregated datasets
> 
> > You should *always* capture the radius -- even for GPS coordinates.
> > You cannot know in advance what the information will be useful for,
> > and not all GPS devices record with the same accuracy.  A GPS with
> > +/- 100m may be no different from one with +/- 10m if you are doing
> > ecological analysis; but it could make a huge difference if your goal
> > is to return to the same tree or coral head or whatever.  I think it's
> > just silly to throw away information when it's easy to capture.  It's
> > also silly to think you know in advance what information will, or will
> > not, be useful to future researchers.
> 
> The problem is that the supposed accuracy reported by the instrument
> seems to have little relation to the actual value, whatever it is.
> As you said, it's cryptic - I've had a unit record points that seem to be
> repeatably correct within <3m while reporting an error radius of 15m (though
> that could be an example of false precision from using the same unit), and
> just yesterday it was telling me the accuracy was 7m while simultaneously
> shifting over 50m as I was standing still.
> 
> Also, since it's not recorded by the unit (at least any that I've
> used) and changes by the second, recording it requires taking it down in a
> notebook and maintaining it partly separate in a file.
> Not a huge burden, but when it's of dubious quality, it hardly seems worth it.
> 
> Karl





More information about the Taxacom mailing list