Would GBIF think there are two observations of “bird X” by two different observers?
or is it sophisticated enough to deduce that a report on eBird is the same observation shown on iNaturalist and conclude there was actually only one observation?
(Just don’t want to skew the stats, basically)
Thanks
(Really, this must have already been a topic somewhere in these realms [in the forums of either site]…? Sorry if I missed it)
I am unsure you can or will get a definitive answer about GBIF’s behaviour or practices here. Externally it appears they simply load both records. But I’m not sure of that. Part of it may be that Ebird locations are a central point in that hotspot, versus what can be an exact gps on iNaturalist, so verifying what exactly is a duplicate, or how close geographically they have to be is something GBIF themselves would have to answer.
Whoa, ya… I visited their website (GBIF). Not sure what kind of academic I need to be to ‘get it’. Pretty intense was my initial thought.
They obviously need a pretty clever algorithm to gather records and then figure out which ones are duplicates/overlap.
I mean if you post your ultra rare ‘three-eyed ibis’ on three or four sites it might think they are duplicates, but if you post a flock of robins on three or four sites, it could be plausible that there are that many robins. Gee makes you think.
I was trying to at least find a list of which databases they scour…
I personally only submit birds to eBird because of the potential for duplication issues, because I think eBird is better suited for recording and organizing bird observations, and because I don’t feel like having to deal with the submission process for both websites.
A lot of people I know do submit birds to iNat, which means I fall lower in species comparisons (year list challenges for example), but I don’t care too much about that.
I see some value in posting birds when I don’t know the species (for example in locations I’m unfamiliar with, or recordings of calls) since iNaturalist has a good identification system. iNat also allows things like feathers, dead birds, bones, pellets, droppings, etc. that eBird can’t accomodate at this point.
No website is going to shun traffic, so it’s unreasonable for one to direct people to only load certain observations only once and to another site.
(for you and I at the very least) it depends on the altruistic behaviour of the user: first recognizing the potential for overlapping records, and then just deciding personally to only upload sightings once.
I’d like to hear from other users to know if this has occurred to them, if they contribute to more than one database, and whether they’ve participated in an ultra rare sighting and later noticed too many records of it.
Obviously GBIF knows there could be one birder (eg), belonging to two database sites, with two different user names, and similar observations. But how to reduce it down enough to recognize that for certain…
Can GBIF reduce the uncertainty down to exact time, location, and observer? Fairly technical exercise I think.
Finally, if there’s a link or list of databases that GBIF uses, it would be most interesting to share!
(I see on their site they do solicit database suggestions, but don’t seem to offer where they get their data… maybe I didn’t dig around enough.)
GBIF gets data from (as I write this) over 43,000 different sources. Some are static one time loads, like a museum collection inventory (or at least infrequently updated) others are updated frequently.
Its not simply a matter of dupication from individuals who may use multiple sites, its also from multiple individuals submitting the same thing. You are a birder in Canada, but I dont know where so hopefully the example makes sense. How many different people reported that famous Great Kiskadee in Rondeau this fall ? Let alone how many different folks report something that is not a one in a lifetime experience (likely), but rather simply rare, for instance the Black-throated Grey Warbler that visited a park near my home in Ontario this fall.
And then GBIF has to deal with that globally across 40000+ providers, with no time limit on when something may be submitted. Not easy.
Yes precisely… plus even ‘armchair birders’ might be motivated to go and see and log a lost winter-in-Ottawa Lazuli Bunting. Now suddenly there are a hundred occurrences logged! D’oh!
If you create a table it could be any of the following cases:
hundred different birds, one observer logs them.
hundred different birds, hundred observers one each.
One bird, one hundred observers log it.
One bird, one observer sees one hundred times (around the yard eg) and logs it each time.
And that doesn’t even consider possible cross-posting!
Hey thanks also for that link.
(Big Data is fascinating to me in general…)
I’m not even sure it is the intention or the role of GBIF to make the dataset they host ‘clean’. That’s an impossible metric to even define, let alone how it could be achieved.
Their focus does seem more on the aggregation, and then let users do with it as they see fit.
I don’t think it really matters. Having you add to both is no different than having two people go to the same ebird hotspot on the same day and add the same robin. eBird internally has mechanisms to estimate abundance and absense, and I am sure they have some algorithm or method to parse out the fact that some areas are visited more than others. iNaturalist is not meant to record absence and recording abundance is hit and miss too. Since GBIF imports from iNat as well as a bunch of other sources, I don’t think there is a way to infer number of individuals nor absence per se from GBIF. So really, it doesn’t matter. It would be neat if there were a way to link the sites so the iNat life list tracking and perhaps calendar would also include eBird sightings,but i suppose it would be a very difficult thing to code that only a subset of users would use (i wouldn’t, since i don’t really do ebird).
I know that’s not in the cards at least for this moment. I too still struggle with knowing whether to duplicate. I use both sites daily (or multiple times daily) and my not-so-strict personal rule of thumb is to upload all bird sightings to eBird and then if I get a photo of something I upload to iNat. This doesn’t make any particular “Data sense” but I enjoy mingling some bird obs with my other iNat obs and like to take advantage of CID on iNat where that’s not even close to an option on eBird, especially if I snap a shot of an accipiter sp. or something.
eBird doesn’t allow photo upload via the app so I almost never bother to upload photos anymore (unless I get a special visitor like the Western Tanager we had a few years ago and need to demonstrate the validity of the obs). I too, would be interested to know if there was an “official” position on this though. As fun as it all is, I do want to be contributing valuable not just “neato!” data.
Yes, exactly. User beware. Even if GBIF were just aggregating museum specimen data, there would still be the issue of duplicate specimens of the same collection in different museums, potentially being georeferenced differently by each one, etc.
That said, there is something to be said for the redundancy of posting observations in more than one place, “bet-hedging” against the unknown future of individual web sites and the Internet. The “duplicate” observations are valuable within the context of each site where they are posted. And it’s up to the users of aggregated data like GBIF to interpret appropriately.
Reporting on multiple sites can cause issues for some, but not for GBIF in their role as aggregator. While not addressing the eBird question, I help collect data for moth and butterfly atlases in Ontario, and from GBIF alone I can get duplicate/triplicate records of someone’s moth from iNat, BugGuide (BG nows submits some data to GBIF, but it suffers from poor location information), and LepSoc (who submit their seasonal summaries). We may end up with 3 or more instances of the same Imperial Moth for example (we also get data directly from BAMONA and other sources).
To add to the fun, some individuals have provided their own datasets to GBIF, possibly giving us four copies of a single record from GBIF. I am continually amazed at the energy levels some people have to share information across multiple platforms. Chris points out the difficultly in reconciling similar locations. Eg. BugGuide has only vague municipalities, iNat uses lat/long with some odd default location names, and LepSoc relies on user-submitted info that could be more precise that BugGuide, and subtly different from iNat. I have also found museum specimens reported in various locations, eg. the museum’s GBIF data, and LepSoc. It’s a challenge.