Telesonix jamesii in a rock crevice

Citizen Science Boosts Models of Rare Plant Distribution

“Didn’t expect that, did ya?!” Mason Heberling

You can listen to this page as an audio file (English only)

There is often a lot of noise about the potential for citizen science projects, but does this reflect hype or reality? Andrew Gaier and Julian Resasco studied the contribution in a paper titled “Does adding community science observations to museum records improve distribution modeling of a rare endemic plant?” published in Ecosphere. In a rare loss for Betteridge’s Law of Headlines, the answer is “yes”.

Scientists seeking to understand and protect endangered species often develop models that predict where a species is most likely to be found based on factors like climate, terrain, and habitat. These species distribution models require large amounts of data on where the species has actually been observed. Building accurate models can be challenging for rare species with limited observation records.

However, the rise of citizen science platforms provides new opportunities. Apps and websites like iNaturalist allow amateur naturalists to document plant and animal observations. While often considered lower quality than professional surveys, these citizen science records can significantly boost the data available for rare and endangered species.

A small purple-flowered plant hides in a rock crevice in Colorado.
Telesonix jamesii. Image: Gaier & Resasco 2023.

Gaier & Resasco investigated this approach for the Telesonix jamesii, James’s False Saxifrage, a rare high-altitude plant with few documented observations. It’s a small, up to 20 cm tall, saxifrage that grows in crevices in the Rocky Mountains of Colorado and New Mexico. Its natural history needs to be better known, and there is some uncertainty about its distribution.

The botanist developed five different models to predict the plant’s distribution, including statistical models and machine learning algorithms. Incorporating citizen science records from iNaturalist doubled the available data for the species.

The researchers found that a “random forest” machine learning model, trained on both professional and citizen science data, performed the best, with an accuracy of 98% (measured as the “area under curve” score). All the top models relied heavily on climate variables like average temperature and rainfall during the growing season, suggesting this plant is at risk in a warming climate.

Validation of the models against independent observations also showed the benefits of citizen science data. Models based only on professional records from museums and herbaria performed worse when checked against iNaturalist records than those that included iNaturalist data during training. Gaier & Resasco write in their article:

There is a wide potential usefulness for the data integration approach shown in this study toward the conservation and management of rare species. We have shown that community science data can be reliable and substantially increase the number of usable records leveraged for modeling distributions. Both museum records and iNaturalist records require examination to determine taxonomic and location reliability. The use of iNaturalist data improved model fits only slightly. Choice of modeling algorithm showed more variation in our results than choice of data source. Much of the information needed to accurately model T. jamesii distribution was already captured in the herbarium data. We therefore speculate that this framework may be more useful for a species with more iNaturalist observations in novel habitats. It is important to consider that this is a species-specific study, and greater insights could be gained through a multispecies approach. A potential next step would be to evaluate how this data integration approach differs between species displaying different patterns of rarity (Rabinowitz, 1981). Notwithstanding these caveats, the information obtained from our model projections can aid in the conservation of T. jamesii to support future targeted surveys (Williams et al., 2009), help identify populations most at risk, and predict how distributions may be affected by future climate change (Franklin, 2013).

Gaier & Resasco 2023


Gaier, A.G. and Resasco, J. (2023) “Does adding community science observations to museum records improve distribution modeling of a rare endemic plant?,” Ecosphere, 14(3). Available at:

Dale Maylea

Dale Maylea was a system for adding value to press releases. Then he was a manual algorithm for blogging any papers that Alun Salt thinks are interesting. Now he's an AI-assisted pen name. The idea being telling people about an interesting paper NOW beats telling people about an interesting paper at some time in the future, when there's time to sit down and take things slowly. We use the pen name to keep track of what is being written and how. You can read more about our relationship with AI.

Add comment

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Read this in your language

The Week in Botany

On Monday mornings we send out a newsletter of the links that have been catching the attention of our readers on Twitter and beyond. You can sign up to receive it below.

@BotanyOne on Mastodon

Loading Mastodon feed...