I’ve recently noticed that multiple users have marked a lot of my wild/escaped cultivar flower observations as “not wild”. A lot of my observations that are being tagged are pretty old, so I didn’t notice that none of them were eligible for research grade until recently. If I’m posting a close-up photo of a flower with basically no image of the background and context it is growing in, it seems ridiculous to assume that it’s a cultivated plant if said plant species also is frequently seen as a wild and/or escaped cultivar in the region. I frequently tag actual cultivated plants as such, so it is quite irritating when my wild or escaped plants are tagged incorrectly. I made this post just to rant, but I’m also honestly curious as to what people’s reasoning is for how they think they can tell a deliberate cultivar from an escaped cultivar from limited photos— obviously certain plants such as some flowers and trees that are non-native can definitely be determined as cultivated especially if you can see the substrate they’re growing in, but all of my photos have been very ambiguous.
If the photo is ambiguous, add an observation note saying “Closeup of an escaped whatever.” Admittedly, not everyone will read it, but it should help.
Science is all about “show me the data and let me judge for myself”. From a scientist’s standpoint, if I can’t glance at the photo and see habitat (or observation notes) to tell me it’s an escape and it’s a commonly cultivated plant, it’s best to assume it is cultivated. If I’m going to use iNat data in an analysis, it is better to lose a few good data points by eliminating a few non-cultivated escapes where you can’t really tell whether they are cultivated or not than to include a bunch of cultivated things in a dataset that is supposed to be of wild stuff.
In addition to including a note, you might also consider taking a more distant photo that shows some context – say, the pavement crack out of which the plant is growing – and makes the non-cultivated status of the specimen less ambiguous. I find it helps to use this as the first image to discourage people from reflexively checking the “not wild” box just based on the thumbnail.
I think this is a side-effect of new users overloading iNat with observations of cultivated plants that are not marked as such, so some people will take up the task to “clean up” the data. Truly feral plants are probably an exception and unless specifically noted as such are likely to get marked cultivated along with the rest of the “noise” generated by all the truly cultivated stuff. You may also be interested in some of the suggestions in this thread: How to handle escaped flora
Also, if someone uses Computer Vision (CV) to add an ID to the observation (including the observer), my understanding is that the system will automatically mark the observation cultivated if at least 90% of observations of that taxon “nearby” are also marked cultivated. (Someone please correct me if that is no longer the case.) Have you checked to see if it is the system or specific user(s) voting cultivated?
[EDIT: see corrections and clarifications below!]
The scale of the problem of the opposite situation (cultivated plants marked as wild) is so vast that I think it’s inevitable that some observations of frequently-cultivated plants that were growing feral will get tagged as cultivated, particularly if there is no zoomed-out shot putting them in context.
Someone doing as you do and actually marking cultivated observations correctly is actually a rarity, in my experience.
I think it’s 80% and at least 10 observations total in the area (so starting at 8 out of 10 marked cultivated), and as far as I can tell it happens also when you manually enter the ID without CV because I’ve seen it happening to at least one of mine. That was a self-inflicted injury because I had previously marked all the cultivated ones while doing IDs pushing them to over 80% so when I posted my wild one iNat thumbed it down and I had to go check the casuals for a while to rescue any wild observations.
Right, it’s a separate algorithm from the CV so it doesn’t matter how the IDs were made
When I see a plant that I suspect is planted, I will ask the user “Planted” and if they respond yes, I’ll mark it as captive/cultivated. If they are unsure and I am not certain that it’s planted I’ll leave it as wild.
This happens to me a lot, so to avoid this I usually don’t observe plants that I have confirmed to be planted in person. One of my examples includes Liriope, which is usually planted, but deep in a preserve woodlands it is pretty apparent that this would be an escapee(bushwacking 10 feet to get close to it). However it is currently sitting at Casual grade.
Genus Liriope (Liliturfs) from Massapequa Preserve, Massapequa, NY, US on October 17, 2020 at 01:05 PM by Robert Levy · iNaturalist
I don’t think a plant should be marked captive/cultivated solely because of the species. Pretty much every species can escape with very few exceptions, and location means a lot most of the time. A Narcissus deep in the woods is a lot less likely to be planted than a Narcissus in a garden bed.
With that said, garden escapees are amazing to the observer but unfortunately skew distribution maps for researchers.
I regularly encounter Narcissus and other nonnative or out-of-range species “deep in the woods” (or similar) that were actually deliberately planted decades ago at a now-vanished homesite. Sometimes the evidence is quite subtle – a foundation stone and the fact that all of the introduced species are planted in perfect rows, for example. I would not consider distance from current human habitation or disturbance to be a good enough metric to decide.
I never knew that, thans for the tip! I’ll be on the lookout for anything that looks out of place near potential escapees.
Yes, along with lilacs, daylilies, and apple trees, all great indicators that you are in the area of an old homesite (in eastern USA, at least)!
In Greenville, a new trail recently opened – only open to the public outside of school hours because it traverses a school property. Although both sides of the trail are currently woodland, one side parallels a chain link fence in disrepair, with several Japanese privet, evenly spaced, just in front of the fence line. Unlike glossy privet and Chinese privet (which are invasive), Japanese privet almost never escapes here. My guess is that the school’s open lawn once reached all the way to the fence, with the Japanese privet planted as a screen; but the woodland since encroached from the ravine on the other side.
I see daylilies spread by seed so I don’t know if that is a good example
Lycoris radiata red spider lily aka hurricane lily is a good one in southeast USA. Especially if they are in a straight line.
The common daylily in the wild in my area (New England in the US), Hemerocallis fulva, doesn’t set seed, as far as I know. It spreads by root rhizomes that are often pushed along roads or rivers by road maintenance or flooding/ice scour.
Daffodils do self-seed, both in garden settings and where naturalizing in the wild. I have a few patches around my yard that popped up from seed coming over from the neighbor’s lot. I never planted any myself. There are large populations of them in the Smokies, for example, and Daffodil Flats in Linville Gorge. Sure, they may have started with a few plants at a homesite now long gone, but they are reestablishing from seed and spreading on their own so by iNat’s definitions can be considered wild in some settings.
I think it comes back to the issue of the app automatically marking everything as Wild unless you tell it otherwise. Personally, when I submit a photo and it’s marked Wild, it’s because I think it’s wild and I’ve made a thoughtful assessment of the surroundings. And I also get annoyed when someone with zero information about context just swoops in and disagrees with my assessment without any explanation or reasoning. It’s frustrating to have to add a comment saying “I think this is wild” when I’ve already marked the observation as Wild when I submitted it. It seems that those who go through observations Captive-voting things give essentially zero weight to the fact that the observation was submitted with “Wild” selected.
IMHO, the automatic “Wild” selection provided on the app has cheapened the value of the observer’s “Wild” assessment to the point that most identifiers view it as meaningless. I’ve suggested in a feature request forcing users to manually click the “Wild” button in at least some contexts, but some users appear to view this as too great a burden on their observation workflow. As long as the app continues to automatically call everything Wild unless otherwise specified, identifiers/annotators will continue to ignore this form of input from the observers, and the number of false-positive and false-negative Wild votes will remain where it is now.
Another feature request that I know is out there somewhere is a revamping of the notification system. I’d like to get a notification if someone Casual-grades one of my observations, so I can reply or fix it. Currently, you could have a thousand observations go Casual because of someone else’s votes and you’d never get notified at all.
I think both those feature changes would alleviate some a lot of the frustration. In the meantime, I do a monthly check through my Casual observations and downvote any incorrect annotations.
I agree with this in principle, but I don’t think it’s a reason for identifiers to “assume cultivated” in ambiguous cases. If I’m a researcher using iNat as a data source, I’m going to need to comb through the data myself before using it and make my own assessment about whether the IDs, Wild/Captive, dates, and locations are correct. I’d rather be given a dataset that’s over-inclusive and make my own decisions as a researcher about which points to throw out than be given a dataset that’s under-inclusive because someone else elected to make those choices for me.
It’s not just old homesteads. Some misguided souls decided to make a beautiful subalpine meadow near Lake Tahoe “prettier” by planting daffodils and other bulbs. Guerilla gardening is fine in urban landscapes, but please discourage people from distributing non-native plants in the wild.
Technically iNat does not mark observations as “wild” by default, nor are you marking your observations as wild unless you explicitly click the corresponding DQA box. You are merely not marking them “not wild”. This is a minor distinction, but a not-insignificant one, because a positive affirmation that something is wild is different than a default baseline status in which only observations that deviate (“captive/cultivated”) are marked.
Requiring observers to actively mark observations as wild may seem tempting for plants, but it would make little sense for many other organisms, from mushrooms to insects, which are only rarely observed in a captive state.
Inat marks all observations as “Verifiable” unless there is a downvote in the DQA(includes captive/cultivated), there is no media/location, an observation is flagged, or the organism is identified as “Human”(there might be others but I don’t know).
For some reason Inat demarcates “Verifiable” and “Wild” as two different categories, see Observations · iNaturalist and Observations · iNaturalist differ by around 14 million observations.
When you vote “Captive/Cultivated” in the DQA the observation always go to Casual grade but there are Casual grade observations in “Wild”. I wonder if the “Wild” category is every observation not marked Captive/Cultivated.