Geomodel issue: often observed plant no more suggested

A couple of days ago, I submitted an observation of Lepidium graminifolium which was found near that place several times already, and observations at RG do exist here. In the past, the correct identification used to be suggested by CV.
But when I prepared this observation for upload, CV suggestions for each photo (on the web site, I let CV suggest an id for each photo before I combine them into one observation) showed other species only, and I was no more sure about the exact species, so I used a different species of the correct genus.
The issue can still be reproduced by opening the observation, and requesting a CV suggestion there: only when you use the “not expected here” feature, the correct species will show up.

This fits with my post from a few days ago:

Anyone else finding that the change to “expected nearby” is actually making identification suggestions less accurate?

1 Like

I’ve noticed the same thing for one of our local oddities, an introduced liverwort growing on campus. Most observations are in greenhouses and gardens, where it came in as a nursery weed with potted plants. We have a nice patch of it along the creek running through campus and quite a few observations, all RG, so it used to come up as the top suggestion. Not any more - for most of the observations it is still the first suggestion when checking for not expected nearby, but for one of them even that fails completely and it doesn’t come up at all. However, it is a somewhat unusual occurrence and therefore I chalked it up to being truly an unexpected find in this location (except, we know and have confirmed that it has somehow managed to establish itself).


The situation here is also full of “oddities”. It is one of the warmest places of Germany, mild winters, warm summers, dry sun exposed rocks, low rainfall; river Rhine nearby, and big cities like the Frankfurt area just a little farther up the river or its contributories.
During the very dry summer of 2022, I enjoyed iNatting in the dry river bank, and found many exotic plants. And the “sand dunes of Mainz” nearby are an area of deep dry sand with its specific fauna and flora, with many elements found in east european steppe regions.


The CV often fails when used on fuzzy or distant subjects, which is probably why it didn’t suggest the correct species in your example.

In cases where the CV doesn’t suggest the species you think it is, the best thing to do is to enter at a less specific level (Lepidium sp) if you aren’t sure of the exact species.

1 Like

This is a different issue. The old CV used to know that a given species had been recorded nearby and could therefore offer it as an option. But the new CV sometimes excludes species that are well documented if they’re at the edge of the model’s interpretation of their geographic / elevational range so it’ll never offer it as an option no matter how clear the photo is. You have to turn off the “seen nearby” option, which is usually helpful because it helps keep continental fauna mismatches from being suggested. So we’re in a bit of a catch. I can understand the staff and developers have been ignoring these complaints (here and on the blog post about the geomodel) because these represent a minority use case, but this issue is frustrating for the rare times it does show up.

The last few dozen posts over on this thread offer a number of other examples of this issue:


This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.