Flag for Taxon: Pacific Crab Apple (Malus fusca)
| Flagger | Content Author | Content | Reason | Flag Created | Resolved by | Resolution |
|---|---|---|---|---|---|---|
| bouteloua | Pacific Crab Apple (Malus fusca) |
does this naturalize outside of the west coast? hundreds of observations outside of the range on BONAP |
Oct. 14, 2019 18:13:37 +0000 | Not Resolved |
Comments
I went through 60 of the photos. One was Malus fusca in range. Some were Malus x domestica. Most were cultivated trees that showed no reason one would actually call it M. fusca, though I have no idea what they were. Some were reported from locations (Colorado, Utah) where I'm sure west-of-the-Cascades Malus fusca cannot survive. A few I couldn't say definitively were NOT M. fusca, but I think none of the ones out of range are M. fusca. This is worse than Phleum pratense confusion!
Any Malus fusca outside the range from SE Alaska to the North California Coast Range and west of the Cascade Crest is out of range for Malus fusca per http://www.pnwherbaria.org/data/results.php?DisplayAs=WebPage&ExcludeCultivated=Y&GroupBy=ungrouped&SortBy=Year&SortOrder=DESC&SearchAllHerbaria=Y&QueryCount=1&IncludeSynonyms1=Y&Genus1=Malus&Species1=fusca&Zoom=4&Lat=55&Lng=-135&PolygonCount=0 and http://ucjeps.berkeley.edu/eflora/eflora_display.php?tid=32578. All out of range observations should be reduced to Malus unless it demonstrates one of two features Malus fusca feature, fruit that is oblong-cylindrical or a leaf that is lobed at mid section with two distinctive lobes. Any observation that is in a urban location and out of range should be checked "no" in the Research Grade Qualification of "Organism is wild". My copy of Sunset New Western Garden Book states that there are some 200 named varieties of crabapple. I agree with @sedgequeen 's remarks
And to clarify, there are other Malus, e.g. cultivated M. sieboldii, that have lobed leaves.
I wonder if this is a function of the early concentration of observations in California to train the algorithm, which then assigned many crabapple- or non-crab-apple-like things to M. fusca. Pretty sure that's also why so many New Zealand plants are suggested as IDs for Pacific Northwest observations.
yep!
we're at 448 observations that are probably incorrect
I've gone through a few of the many in that identify query (and all of the Research Grade ones). I disagreed with Malus fusca at genus or family level - I don't really know what the things really are, but the ones I've disagreed with definitely don't look like Pacific Crab Apple and/or are well outside its expected range.
To be clear, I have no real expertise on crab apples in general, it's just that Malus fusca is pretty common in Southeast Alaska (which is the area I focus on identifying things), so I've ended up IDing a lot of them.
Wow! I wouldn't have guessed so many incorrect identifications for this species would be possible. If this is a function of the software, perhaps it is a matter to put before one of the webmasters?
thanks @gwark!
@kurtsteinbach they're pretty aware of the issue. some users even keep lists of the known "overconfident" IDs, like https://www.inaturalist.org/lists/1533061-Computer-Vision-Traps
I spent a little more time on these today and finished going through non-North American observations (there weren't so many of them - but I felt better about the map without those really far out there points).
I'll try to keep working through the observations from Eastern North America. Since I really don't know what most of these are (but they don't look like Malus fusca), so far I've mostly been disagreeing at Rosaceae (unless I'm suspicious that it may not even be in the rose family, in which case I go to dicots), with a note saying:
"This could be a crab apple (I'm not sure), but I am confident it's not a Pacific Crab Apple (which has a relatively narrow expected range along the west coast of North America)"
In the case that I'm suspicious it's not even in the rose family, I'll start by saying "I don't know what this is, but I'm confident it's not..."
Even using copy/paste it's a bit of a tedious process and I would welcome any other suggestions (is it even worth leaving a note at all?)
Some people get a bit feisty when you "downgrade" their observation without a reason, so if it's not too horrible to copy/paste I would suggest continuing to that. You could make it as generic as possible (the "I don't know what this is, but I'm confident it's not Malus fusca." one) so that it applies to all observations and not just Malus/Rosaceae. If you don't mind that a small proportion of folks might snap at you a bit, then identifying without a reason is fine too.
Thanks for your good work @gwark
yeah awesome work everyone. it went from over 500 misidentifications outside the known range to just a handful in a couple days!
This kind of data quality control hopefully helps train the next computer vision models not to be so confident with this suggestion.
Looks like we're back over 200 suspect IDs...
Another case of computer vision gone wild?
http://bonap.net/MapGallery/County/Malus%20fusca.png
@ajwright @gwark @carexobnupta @mikepatterson @kurtsteinbach @sedgequeen can you help / weigh in?
Link to help reidentify:
https://www.inaturalist.org/observations/identify?quality_grade=needs_id%2Cresearch%2Ccasual¬_in_place=6%2C10%2C14%2C46%2C7085&taxon_id=77948