New publication: Putting toyon in the TARDIS

Recently, the Yoder Lab has been thinking a lot about the conditions plant populations need to flower and produce fruit — that is, to reproduce and create the next generation. We hit on the idea of using images from iNaturalist records of Joshua trees (Yucca brevifolia and Y. jaegeriana) to train machine-learning models relating flowering to annual weather variation, and then used a trained model of Joshua tree flowering to reconstruct more than 120 years of Joshua tree flowering trends. That was pretty cool, but it’s really only the start: iNaturalist has records for many, many more species beyond our favorite spiky desert monocots. Last week the first taste of that was officially published online at The American Journal of Botany — a student project that adapted the methods built for Joshua tree to study population health in another iconic California plant, toyon.

Toyon, Heteromeles arbutifolia, is a distinctive and ecologically important shrub in California chaparral and oak woodlands, from north of the Bay Area down into Baja California. It has shiny, evergreen leaves with toothed edges and produces clusters of bright red, berry-like fruits, so it’s long been known as “California holly” and “Christmas berry.” You’ll hear people say toyon is the “holly” for which the Hollywood Hills were named, but that turns out not to be true. Even without that origin story though, toyon is a recognizable floral neighbor throughout California, lining hiking trails in Griffith Park and up in the Sierras, and planted in hedges across Los Angeles.

So when undergrad researcher Daniel Dakduk was looking for a species to focus on as a followup to the Joshua tree flowering study, toyon was quickly at the top of the list. It’s well represented in iNaturalist, and as with Joshua trees, its flowers and fruits are highly visible in photos attached to iNaturalist records — and it doesn’t grow alongside other species that look so similar as to risk a lot of misidentifications. Daniel reviewed and annotated thousands of records of toyon, and worked through the process of training Bayesian Additive Regression Tree (BART) models of toyon flowering in relation to temperature, drought stress, and precipitation. A trained model could then translate weather records into predictions of toyon flowering, which varies in intensity from year to year.

Those predictions let us reconstruct how toyon populations responded to past weather conditions, going back to the year 1900. That time-travel aspect of the analysis inspired what we’re now calling it: Temporal Analysis of Reproduction Distributed in Space, or TARDIS. That is, we analyze flowering activity (“reproduction”) over multiple years (a “temporal” perspective) and across all the different locations where the species grows (“distributed in space”). That analysis gives us a sense for how toyon populations experienced year-to-year variation in weather — and the longer-term trend of changing climate — from the turn of the 20th century to the present.

We found that toyon flowers at fairly high frequency in most years. Comparing the average intensity of flowering over 1901-1929 and 1991-2020, we found that toyon flowering intensity increased a little bit over that period. Increasing flowering intensity was not clearly associated with more northerly latitudes or higher elevations — as we’d expect given that climate change is generally moving suitable climates for many species to higher latitudes, and uphill. On the one hand, that’s consistent with other studies showing that suitable conditions for toyon populations have been pretty stable as climate changes. It also suggests, though, that the species may not have much opportunity to migrate north as climate change continues.

There’s another sign like that in the TARDIS results: in addition to average flowering intensity, we calculated how much flowering varies year to year in the early and more recent periods, and we found that predicted flowering intensity has become more variable since the early 20th century. Ecological theory has suggested that species’ geographic ranges are often defined not by sharp cutoffs in population health — points at which conditions stop supporting any reproduction and population replacement — but by decreasing consistency of population health. We do find particularly strong increases in the variability of predicted flowering in the south, where we’d expect climate change to impact toyon populations first.

These effects are all comparatively subtle, and our overall conclusion is that toyon populations have been resilient to the climate change they’ve seen so far. Toyon faces more threats than simple climate change, like expanding development in much of its range, and growing wildfire risks from the intersection of climate change and human activity. By putting toyon through the TARDIS, we’ve learned a bit more about how to weigh those risks, so we can ensure it remains a part of California’s landscapes for a long time to come.

For more detail, check out the full paper online at The American Journal of Botany, where it’s part of an upcoming Special Issue, “Plant Resilience and Conservation for a Changing Climate”, or you can also read it as a preprint on bioRxiv. And if TARDIS looks like it might be useful for your own project, get in touch!