I started this project in May 2021 because I noticed numerous misidentifications in iNaturalist for species in the Malva genus. Also, I was frustrated because the iNat Computer Vision (CV) algorithm was no help in figuring out the proper ID for most new Malva observations - it mostly suggested M. parviflora. Since the CV software was influenced by thousands of incorrect Malva IDs, the solution was to correct the misidentified observations. Now that thousands of the incorrect IDs have been corrected, the CV software is giving much more accurate species suggestions.
After working on this project for over two years, identifying 11,005 Malva observations in California, and reviewing countless others, I've summarized my findings as follows:
Correct IDs: 7,218
Incorrect IDs: 3,663
Total ID'd to Species: 10,881
Total ID'd to Genus (No Flowers/Fruits): 3,969
Unconfirmed: 2,514
Actions I've taken:
- Created this umbrella project and subprojects to manage the work.
- Researched identification resources and shared them with other iNat users, especially when ID errors were found.
- Reviewed nearly 17,000 Malva observations in California (both Research and Needs ID status) dating from 2007 to the present date.
- Corrected errors I found during the review, including those that had been set to Research Grade by previous reviewers. Tagged Malva experts when discrepancies were found.
- Collected metrics on misidentifications in Excel for reporting purposes.
- This is a work in progress... There are still 2,514 observations that haven't been confirmed and more are added every day.
While doing this work, I've made the following observations:
-
In addition to the above errors, there are approximately 3,969 foliage-only observations of Malvas in iNat for California that had mostly been ID'd as M. parviflora. It is my understanding that foliage-only Malva observations should not be ID'd to species (except M. assurgentiflora, M. arborea, and M. subovata), so I reverted them to genus level and marked them "No, it's as good as can be" so they'd become Research Grade and get removed from the queue.
-
The majority of Malva observations in California have been ID'd as M. parviflora. Thousands of observations, including the foliage-only ones, had been reviewed and set to Research Grade, so the CV model was suggesting M. parviflora for almost all new Malva observations, which perpetuated the problem.
-
Almost all the M. sylvestris observations had been misidentified (most of them are actually M. multiflora). Hundreds of them had been reviewed and set to Research Grade, so I'm sure these errors were causing the CV model to suggest incorrect species. All these misidentifications have been corrected.
-
"Cretan Tree Mallow" was set as the common name for M. multiflora in iNat when I started this project. I'm guessing people thought it must be a tree like M. arborea, so they avoided using it. In May 2022, I contacted Tony Iwane and asked, "How are common names determined in iNaturalist?" Here was his reply; I followed his advice:
They're not "determined", really, they can be added by any user (go to a taxon's page, click on Taxonomy, scroll down and click on "Add a name"). Site curators and staff can edit the order of preference for common names. In this case, I'd recommend flagging the taxon and suggesting that "cretan mallow" be the default name for the taxon and explain why."
There seem to be a few errors in Jepson eFlora:
A project of this size cannot be managed alone. Many people have contributed their time and expertise to this effort:
Updated 2024-01-09. I missed about a thousand M. parviflora observations in southern CA in my first pass, which I'm still in the process of reviewing.
Updated 2024-01-14. Finished reviewing M. parviflora observations and made a second pass through all the M. nicaeensis observations. A large percentage were actually M. multiflora.
