I think the point is that people are uploading older data, not obtained using smart phones, into iNat. These data are imprecise yet useful, and will be confined to the dustbin if treated as secondhand citizens, even though it should be up to the end-users of the data to set the parameters for whatever precision is needed.
Why make iNat less useful?
Records from the archives, historical data is a separate issue to most iNat obs - which require, were you there, did YOU see this. We have a rough guideline of within the last 100 years. Which excludes 1832. For iNat. With a few exceptions allowed.
Part of the issue here is the term Research Grade. Apparently some interpret that designation to mean that if an observation is RG it should be of use to their particular research question. And if it isnāt, it should be made Casual or something other than RG. As has been said many times, itās up to the researcher to decide what records are of use to them. Itās not the responsibility of the observer to make sure their observations meet the researcherās demanding criteria. That said, we should all be striving to post observations that have the best associated data we can provide.
I have to admit that part of my feeling on this is actually more personal than research-based. I like the idea of working towards finding every species previously found in my āareaā (probably unattainable, but hey, itās nice to have something to aim for!) - and if the data is cluttered up by people happening to plunk their random āsomewhere in Victoria or SAā pinpoint for an observation of a whale in the Grampians, that bugs me. I realise this probably isnāt a good reason for objecting (or a particularly probable use case - around Alice Springs might be more of an issue if theyāre covering all of Australia!), but if there were some way to automatically exclude observations with such huge accuracy circles (over 100/500km?) from searches, observation lists etc, I would be happier. After all, itās not so much a point as a circleā¦
I use ābigā accuracy circles as appropriate when either my GPS location is clearly wrong or (for some reason) no location was recorded - but I think the biggest Iāve ever used (when I stopped at a few places along a road and all locations for the day were either the campground I started from or nothing) was a couple of kilometres. Based on the āobservation trackā of my hikes (far from linear), I typically add an accuracy circle of 100-250m to all observations because my phone GPS doesnāt seem particularly accurate, whatever people claim for GPS locations (if anyone ever claims they can go to the metre, theyāre (a) using better equipment than me, and/or (b) in a better location, and/or (c) kidding themselves).
Observations with accuracy > a certain value (I think itās the 0.2 x 0.2 bounding box for obscuration?) arenāt mapped on general maps in iNat, so I donāt think this is a huge concern.
You can already exclude observations with any accuracy value via search URLS, see:
While they arenāt visible on the map in explore, observations with very large accuracy radii do still show up in any geographic search (and on the checklist for any standard place) that happens to contain the pin, no matter how small the bounding box/place/etc. is and how much of the accuracy circle lies outside it
To me, the point is, who has the right to determine which observation a researcher can use? Obviously, iNaturalist data are limited in dozens of ways, but given what has been uploaded, who gets to choose what location accuracy we keep available for researchers? I insist that only the researcher gets to limit the data.
Keep in mind that iNaturalist is a citizen scientist project. Such projects are always inefficient. The data isnāt perfect and canāt be expected to be perfect. We should, of course, try to make iNaturalist data the best we can in terms of the data we ourselves upload and of educating other iNatters. We canāt make other observers and their observations perfect, though.
I am not a fan of this idea and strongly second @sedgequeen on that. If you do not feel comfortable to ID or use observations with large accuracy circles, skip them. They can still be useful to others.
I have several observations that have comparably large accuracy circles and I still think they can be useful.
e.g. this one, which might be the only photograph of a life specimen of this species, but was taken somewhere in the desert before I had a smartphone ( and even with smartphone I learned not to trust them to much in such environments, judging from my experience in the Andes with smartphones and waaaay off GPS data sometimes). The accuracy circle.is 50km here, but would this one have been casual it might have never been IDed, which would be a pitty I think
https://www.inaturalist.org/observations/38000576
I have several similar cases from Southern Africa or the South Americas
I think at this point the consensus is that 25-50km is a pretty reasonable uncertainty for exactly situations like youāve described.
As for larger circles of hundreds or thousands of km, I think we all agree that most of these are errors and that for many of them the centre of the circle is accurate. There are also cases where itās obviously inaccurate and some of those cases should be different topics (related to issues with using the default Google locations and glitches like this one etc.) and I think generally observations with very high uncertainty are more likely to have highly inaccurate locations than ones with small uncertainties. In any case Iām not a fan of justifications for them because there are evidently errors involved in one way or another that obviously ideally would be fixed. Hopefully they can be addressed at the source with whatever glitches or user habits that lead to them.
But an observation being Casual is also an indication of something wrong for those aware enough and the only effective method in cases where no comment was left to make a notification, or the notification was missed. I think everyone should check their verifiable=false observations occasionally for things they didnāt intend to be in there. It doesnāt really need to be checked very often but thereās a good chance youāll find a couple DQA votes that you disagree with. Some of these involve judgment calls that are just as subjective or arbitrary as choosing a particular radius as a cutoff (e.g. recent evidence of organism). And Iāll reiterate that researchers arenāt prevented from using Casual observations if thereās value to them; one personās trash is anotherās treasure.
Maybe there could be a kind of iNat health checkup that the user can opt into (or opt out of), that pings you once a month, or once a quarter, etc. Just to make sure that everything is still good.
Like the Google account security check. Is this still the best email for you? Is this still your phone number? Etc.
Like mileage-related checks on a vehicle. With āmileageā in this context being a certain number of IDs or observations since your last āiNat Health Checkupā. Are there tags that you havenāt responded to? Are there maverick IDs that you accidentally created?
I donāt know anything about programming, but I guess iNat doesnāt have the resources right now to create something like this, and maybe they never will have the resources. But at least describing the idea is the first step in bringing it to fruition. Although it probably has been suggested in countless comments and feature requests.
Well, I do also have some observations with larger then 50km accuracy circles, especially from that trip through the desert but also from primary rain forest and still think they are as eaqually valid as the one I linked to.
I do think putting a cap on those accuracy circles just creates an incentive to actually create not only imprecise put plain incorrect data.. if you do not exactly know the location but want your observation to not go casual? Just make the circle smaller and donāt care about whether the right location is still included or not.. I am not convinced this would be the better outcome.
There are a bunch of habitats where 50km might not be large enough when gps maybe is not an option all the time, like oceans, deserts, rainforests.
GPS has always been a great option for me in deserts and (on) oceans, and itās the only thing I use even in less remote areas. I make observations in Airplane Mode to prevent cell tower signals from confusing and degrading location accuracy derived from GPS.
Yeah there would be some of this happening. So if a limit were chosen a good starting point for deciding where to put it would be based on how far it makes sense to not know the location. For example somewhere between 2-3,000 km is likeā¦ you remember it was in the USA but donāt remember which coast, you remember it was in Africa but donāt remember if it was northern or southern Africa, you remember your cruise was to Antarctica but donāt remember where around Antarctica, you remember you were on a remote Pacific island but not which oneā¦ All of those sound fairly implausible to me but Iām sure theyāre possible. Itās kind of funny to me that an observation is supposed to be within a quite restricted time frame (a specific date, which is also very easy to forget; and less than ~100 years from the present) but thereās no restriction on geographical uncertainty.
I suppose Iāve had days where I wasnāt sure what planet I was on.
Iām trying to remember from my museum days if I ever saw lat-long coordinates on a specimen record that also included a plus/minus error distance. I donāt recall that being a thing, and this was in the days of GPS Selective Availability. Lots of specimen records had no lat-long, maybe just a location description which might not have been very precise. Township-Range-Section was common. And rarely UTM coordinates.
This is where adding notes to oneās iNat record is important. A person could always describe the location, even if itās general, which would be helpful when assessing the usefulness of lat-long coordinates.
Yeah many observations are missing an uncertainty value (thread here) and thatās also not ideal. Many phones and camera donāt save it in the metadata and itās such an easy step to skip doing manually. I guess the web/app uploaders could force you to enter something before uploading, but who knows how accurate the value would be from the average user.
Some of the old scanned slides Iām planning to upload will have a time like āAugustā. I think they could be useful so Iāll select a reasonable time to report, e.g. August 15. I wish I could add a measure of temporal accuracy, but I canāt except that I will put a comment in notes. Sigh.
Somewhere on the forum is a good post - with a reminder to check things like Maverick on your own obs. Cannot remember what the title is.
I had a batch of photos I couldnāt find the originals for, but I at least had a dateā¦.so I guesstimated the time of the observation based on when I knew I was probably outside plus lighting, and then put a note with the observation that observation time was within 3 hours of what I noted (and the reason why).
I know people would sometimes like to know the time, but I figure theyāre lucky to get an accurate date. I donāt bother shifting the cameraās date between standard time and day light saving because itās not easy to do and Iād forget anyway. I suppose iNaturalist could come up with a way to report an accuracy time, but I think Iām going to enter 00 01 (just past midnight) which any researcher will recognize as impossible.
You can delete the time text in the time/date box and it will get entered as the just the date with no time information.