The Nipplewort-Imposter

This plant has been bugging me since we first noticed it last year and all the folks with plant ID apps on their phones helpfully identified it as Nipplewort (Lapsana communis) introduced from Europe. I checked it out on the computer at home, and the USDA database had a report of it only in one of the more southern counties in NC but not where we were hiking. I subsequently tried to find it in my field guides and came up empty-handed. Even Weakley's Flora just mentions it as being uncommon in NC but doesn't deem it important enough to provide a description.

That's odd, I thought, especially since it seemed to be growing everywhere along trails and roadsides in the woods. Maybe the plant we saw was actually something else? Like a rookie, I had missed to get a picture of the leaves though, so I had to wait a year for it to pop up again before I could investigate further. So here we are, one year later, and I think I finally have figured it out: It appears to be the native Allegheny Hawkweed (Hieracium paniculatum), and all the plant ID apps out there don't know anything about it. Instead, they frequently misidentify it as nipplewort.

Further digging into herbarium records confirms that the hawkweed should be a lot more common around here than the introduced nipplewort. Both species occur in the mountains of western North Carolina with the nipplewort's range overlapping with that of the hawkweed. But where Hieracium paniculatum is reported as "common in the mountains" in "dry to mesic forests and woodlands, especially where disturbed, such as logging roads, trailsides, wooded borders," Lapsana communis is listed as "very rare." It was collected in 1938 in Haywood County, 1995 in Madison County, 2002 in Swain County, 2005 in Ashe County, and 2010 in Buncombe County. That's it for NC herbarium collections.


Range comparison of nipplewort (purple) and Allegheny hawkweed (pink and purple). The map was colored in using data from Vascular Plants of North Carolina. The plant drawings are from the illustrated flora by Britton & Brown, published in 1913 (now public domain), and were downloaded from the USDA PLANTS database.

The best way to tell the plants apart seems to be by the leaves. Nipplewort's leaves are variable along the plant but the lower leaves have a large terminal leaflet and one to four small side leaflets and regularly toothed margins. Allegheny Hawkweed produces leaves only on the stem with basal leaves usually lacking, and they are thin and elongated by comparison and have an irregularly toothed to entire margin.

It seems the iNat computer vision algorithm is not the only one that gets this one wrong. (On all the plants I checked, it never actually offered what I think is the correct ID.) Other apps, like Picture This, misidentify the plant as well. It seems to be a self-perpetuating ID error - people identify the plant as nipplewort because that is what the app suggests, which then teaches the app to identify this as nipplewort again on similar photos. I suspect a lot of the "nippleworts" out there on iNat are actually Hieracium paniculatum, at least here in the Southern Appalachians. Please feel free to comment on this or help fix some of the incorrect IDs!

Posted on September 12, 2020 10:09 PM by annkatrinrose annkatrinrose

Observations

Photos / Sounds

What

Panicled Hawkweed (Hieracium paniculatum)

Observer

annkatrinrose

Date

September 8, 2019

Description

This plant I believe gets misidentified as nipplewort by the computer vision algorithm. This first time we found it, I missed to take pictures of the leaves, which help tell the difference.

Photos / Sounds

What

Panicled Hawkweed (Hieracium paniculatum)

Observer

annkatrinrose

Date

August 21, 2020

Description

Taking a closer look at the panicled hawkweed this year. I'm pretty sure that's what it is. We get into a lot of arguments about this plant on hikes because several phone ID apps tend to call it nipplewort.

For the seeds, see this observation:
https://www.inaturalist.org/observations/59353366

Photos / Sounds

What

Panicled Hawkweed (Hieracium paniculatum)

Observer

annkatrinrose

Date

September 4, 2020

Description

Exact same plant as this but gone to seed.

Photos / Sounds

What

Panicled Hawkweed (Hieracium paniculatum)

Observer

annkatrinrose

Date

August 30, 2020

Description

Panicled hawkweed - again pretty sure that's what it is and not nipplewort based on the leaves. There were flowering and fruiting plants in this location.

Comments

Thank you for this analysis. I agree with your assessment. Hieracium paniculatum is also common in New England, but rarely identified as such. As you said, I think the algorithms missed this because there weren't enough records of H. paniculatum to meet the threshold for the initial algorithm training, and so the algorithm keeps offering Lapsana communis as the next best option, and the errors get compounded.

Posted by tsn over 3 years ago

Thank you for this! Sounds like we all have a mission to go make some Hieracium paniculatum observations to retrain iNat.

Posted by atlasendured over 3 years ago

I see - so basically since it wasn't included in the "training package" it's like trying to key something out with a key that doesn't include that species. You end up somewhere (in this case nipplewort most commonly) but never at the right thing because it just wasn't considered when the key was built. One thing we can easily do maybe is to go through the nipplewort observations for the eastern US and correct those that are obviously something else. Not an easy feat as I think a lot of them may be missing a good look at the leaves. But if we alert enough observers to the problem, maybe over time we can reduce the false nippleworts and gain some good Hieracium paniculatum observations to retrain the algorithm.

Posted by annkatrinrose over 3 years ago

Ah neat! Good writeup! I recently discovered Hieracium paniculatum in Flora of Virginia and I think based on foliage I've gotten mine correct, mostly... I'll definitely have to check though!

Posted by wildlander over 3 years ago

This is now included in the newest Computer Vision Model! If you go to the species page and check the "About" section, it says the Computer Vision Model now knows about this taxon. Thanks everyone who helped identify a sufficient number of these to make that happen.

Posted by annkatrinrose almost 3 years ago

Add a Comment

Sign In or Sign Up to add comments