Geolocate Uncertainties

Hi all,
We use Geolocate through Specify a lot and we were wondering if you could tell us more about how the uncertainties are determined.
Firstly, we've become wary of some very precise-looking uncertainties that crop up frequently - e.g., 3036m.
Secondly, sometimes we can't work out why Geolocate plotted an uncertainty and whether to trust it. For example, Geolocate gives "White Mound, Oklahoma" a radius of 1807m. Does it have some information about the extent of the hill that we don't? Actually, 1807m might be another recurring uncertainty, so perhaps that's not the best example. 
Finally, we wonder if it is better to have consistency between all the georeferencers in the group, or whether it is better to be more particular about the uncertainty. For example, if a town has an uncertainty that seems too large or small, personal preference might be to resize it. However, if the difference is minor it might be better to keep it as is, so we all get the same result. Plus, its faster if we don't resize. If we had a better understanding of how the uncertainties are determined, we might favor one approach over the other. Currently, we try to be consistent by a) noting in a remarks field if we resized the Geolocate-determined uncertainty and b) by georeferencing localities with the same named-place at the same time. This issue is usually minor, since the differences are small, but any advice appreciated all the same!
Thanks, Una

Submitted by nelson on

Una,
Thanks for the question.  As you know, one of the key factors in determining uncertainty is the extent of the named place.  If we have a polygon for the named place available in our gazeteer, we use that for determining the extent.  Most placenames do not have associated polygons and therefore we have to make assumptions about the extent of the place based on the feature type (populated place, town, city, mountain, park etc.).  
So if you are 1) georeferencing a lot of simple localities (ie- named places without displacements or other modifiers)  and 2) the named places for these localities are of the same feature type and 3) those named places aren't associated with polygons, then the system uses an extent based on the feature type and you will see you the same uncertainties. To really improve things, we need to get a lot more polygons into the system, something we are actively working on.
Nelson

Submitted by nelson on

I also forgot to mention that if you want to know when geolocate is using a polygon and when it isn't, you can turn on the option to "generate error polygons".  If a polygon is returned in the error polygon box, then you know geolocate used a polygon.

Submitted by unaf on

Hi Nelson,
Great, thanks for the help - we'll keep an eye on the polygons.
Una

CAPTCHA
This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.
3 + 8 =
Solve this simple math problem and enter the result. E.g. for 1+3, enter 4.