I’ve found that there are a lot of observations with locations which were set from typing in a state, country, or county, which creates a specific point for the observation though the accuracy circle in extremely large. For example, there are 66 verifiable observations from various users at the exact point in rural Kansas, due to selecting “United States” for a generic location upon upload. However, these observations get categorized as being specifically in Montgomery County, KS. There are even more locations set to a state level. 336 verifiable observations are pinpointed to a single spot on the edge of Lake Jordan in North Carolina due to selecting North Carolina as the location, but the observations are all categorized as being in Chatham County.
What might be the appropriate way to handle these observations, as their accuracy circles technically make most of them accurately located, but it causes inaccuracies on species maps and checklists, with research grade observations like this putting species in places they surely aren’t.
As you mention, there is nothing technically wrong with these observations, although with such a big accuracy circle they are of limited scientific value. You could add a polite comment to let the observer know that if they can narrow down the geographic area more precisely, their observations will be more useful for science.
I’ve also seen observations where a generic location is selected and where the accuracy circle does not encompass most of that geographic space. In cases where there is no doubt that the accuracy circle doesn’t include the true location, those can be marked as location accurate: no.
I’ve seen a similar issue with herbarium specimens that all map to a point in central Oregon because all the label says is “Oregon.” Sigh.
I think it’s appropriate to mark “No” for “Is location accurate?” for these center-of-the-US records. Just as locations with latitude and longtitude both as 0 are marked as having inaccurate locations.
I have to say that I’ve sometimes also taken the approach of voting “No” on the DQA question “Is location accurate?” for observations with unreasonably large accuracy circles. I only tend to bother when I’m interested in the distribution of this species and if it seems unlikely that the observer will return to correct a very implausible location.
If the accuracy circle is semi-reasonable (e.g. encompassing all of Yosemite National Park), then I’ll just ask the observer if they’re able to narrow that at all. If the accuracy is unhelpfully large and the center point is way outside the documented distribution, that’s when I take a more activist approach in the DQA. I’d rather that one observation with very dubious accuracy doesn’t confuse future observers (or show up misleadingly in GBIF).
If the original observer happened to return and tell me “Yes, I’m positive I saw this within 158 km of that exact spot,” then I’d withdraw my DQA vote. I can also be outvoted, of course!
There are lots of records from natural history collections with very vague/general points/locations on GBIF though. I had to work through a bunch during my MSc work. They are often older records where the location on the tag/label just says something like “Puerto Rico” or “Kansas”. Despite that lack of spatial accuracy, the records definitely can still be useful, so I would urge users not to use the DQA to make a low-precision record Casual just because the record doesn’t seem useful at the time. As long as the accuracy is set, that will transfer to GBIF so it wouldn’t be misleading and will allow data users to make their own determinations. In short, I think that the data user should get to determine what
is for their end use.
iNat helps prevent some potential confusion from these types of points by not mapping points with very large accuracy circles so that they don’t show on distribution maps and confuse users.