Today iNaturalist released the Geomodel, which will be replacing the previous method used to weight and label taxa that are “Seen Nearby” in Computer Vision suggestions.
On the website (and soon in the mobile apps) you’ll see “Expected Nearby” rather than “Seen Nearby”, to make it clear that the label is based on a predictive model. Note that even if you see “Seen Nearby” in the iNaturalist mobile apps, the results are coming from the Geomodel now.
Please read the in-depth blog post. If you have any questions, please ask them on the blog post page to keep discussion in one place. This forum post is mostly just an announcement. :-)
As always thank you to everyone who’s contributed to iNat. This has been many months in the making and we’re excited to use the Geomodel to improve iNaturalist and Seek going forward.
Thanks, exciting. Just finished reading the page, I have a few -probably asinine- questions popping into my mind (the first one being: “given enough species ranges included in the model, and with the addition of supra-specific feedback loops… could ‘biogeographic provinces’ slowly emerge and in turn influence the suggestions lists?”).
Will have to re-read several times.
I made a comment on the blog post, but it would seem natural that there (also) ought to be a whole new topic or subtopic here on the Forum where we could have detailed exchanges about the Geomodel.
I hear you. I think for now let’s keep it to the blog post while the announcement is still fresh, to capture the initial deluge of questions and Scott (or others who are familiar with how it works) aren’t answering questions in two places. Maybe open things up here on Monday?
I think it will be nice to compile some FAQ based on “the initial deluge of questions”. Of course, if there will be enough questions.
BTW, can link to the blog post be added to geomodel predictions page?
UPD: Also, about " Learn more about the Geomodel here." on the taxon page. It doesn’t look like link to specie geomodel, but to some general explanation of the concept. Of course it is only first-time stop, but anyway.
Is it possible to move the page to new tab of the taxon page? or even better, as new overlay on the observations map?
Exited to see the model in action!
In the past days, it gave me some headaches to translate it properly (thanks for the screenshots, @kueda!) and I still need to see how it integrates into the general interface before I submit all strings - so I am eager to discover it!
I don’t know if this was added after your post or not, but there is such a link, behind the word “here” in this sentence: “You can read more about how the Geomodel is made here.”
The lack of climate data is an issue, I can find a bunch of humid lowland insects “expected” in the Taklamakan desert because it is a low elevation, and I find honeybees and giant hornets “expected” in the middle of the ocean. The reason elevation matters is because it affects climate, climate is what really matters, and elevation is only one factor that affects climate
I’m going to close this, as the conversation on the blog post has died down. If people want to discuss the Geomodel, feel free to start a topic in General, but note that iNat staff may not be monitoring it and answering questions about it there. If you think there’s a bug with the Geomodel, please make a new topic in Bug Reports.
