Shifting Plant Phenology and Annotations

[chase-prairie]

Hey all!

I’m getting winter cabin fever and stressed, so I’m spending my downtime annotating early spring flowers. I highly recommend doing this right before bed because I keep ending up dreaming of spring flowers instead of other …eventful things happening in America right now.

In case you haven’t played with phenology before, as plant observations are annotated, their species page produces charts that show when flowers bud, bloom, and fruit/seed and when leaves bud, green, color, and fall. This is really handy for folks like me, looking to schedule when to check out particular sites and go pollinator photographing, or for people collecting seeds for restoration and conservation.

Here’s a great demonstration how phenology charts work and here’s a guide on how to annotate.

As I’m working through these, I’ve been curious about two things:

1 - Can/should iNat UI track phenology changes?

iNat UI currently shows an average pattern of phenology over the past 20+ years, but part of what we’re seeing with climate change is shifting phenology, with plants often blooming earlier or at generally weird times. Since we know this to be the case, and phenology weirdness varies from year-to-year, does iNaturalist have a means of showing us this data? If you were interested in this, how would you like to see that data displayed? I’m picturing something like this, but am all ears for other ideas.

This is a very clumsy shoop forgive me! And yes, I know I can make a feature request, but this is the discussion first so we can see if anyone can find a better way to do it. I know other projects like Budburst (rip ) track phenology too, but it seems a shame to not tap this resource with all the data we have here. Also, while being able to download this data off the backend is cool on its own, making it visible to casual users seems like a neat way to motivate tracking and annotating.

2 - “Re-blooms”? “Remontancy”? “Climate weirdos”? What do you call ‘em and have you seen a project with them?

My favorite climate phenology weirdness is early spring plants reblooming in fall. The classic example most people notice here is lilacs, but I’m quite partial to our confused native friends, like this very confused Viola pedata I saw in Spring Green, WI and a recent Claytonia virginica someone saw. What do you call them? I have seen some people use “remontant”, a term coming from roses and raspberries that produce multiple rounds of flowers, but this doesn’t seem quite the same as “blooms in the usual spring time, sometimes will do a weird thing in fall depending on conditions.”

What do you call these rebloomers? Has anyone see a Project on these? Searching around hasn’t revealed anything, but maybe I just don’t know the right word to use.

[spiphany]

@chase-prairie

Has anyone see a Project on these?

I have one, though I’ve sort of been neglecting it because I realized I wasn’t certain how to best structure it. So I am basically leaving it up to users and you’re welcome to add observations: https://www.inaturalist.org/projects/time-is-out-of-joint

[rupertclayton]

@chase-prairie

1 - Can/should iNat UI track phenology changes?

You actually can use the experimental “Compare” UI to visualize phenology across years (to some extent).

Here is a Compare search that graphs the number of observations of Malacothamnus nuttallii by month for the years 2021–2025 (all phenology).

And here is the same Compare search restricted to only flowering observations.

These examples indicate some of the challenges in using iNat data in this way:

• For a lot of species, the proportion of observations with phenology data is very low.
• Once we divide up data in phenology groups (e.g. flowering vs. not) and also time buckets (year-by-year, or even groups of multiple years), the number of observations gets much smaller and so variations are much harder to interpret.
• iNat usage has grown substantially over time, so when we compare observation counts from past years with more recent years we need to take that into account.
• iNat observations have a huge peak each April with the City Nature Challenge (and a smaller one in October with the Great Southern Bioblitz)
• The data on these graphs are also vulnerable to fluctuations based on hard-to-predict factors. For M. nutallii we might look at the graph above and conclude “Wow, there was some kind of superbloom in June–December 2021, and again in April and May 2022. But this species has really had trouble flowering since then.” What we’d be missing is that @keirmorse used iNat data to inform a bunch of his research on Malacothamnus, and as part of that made a major effort to add IDs and annotations to observations of the genus several years back. The supposed lack of flowering observations in earlier or later years probably just reflects a change in Keir’s research priorities!

As identifiers, we can mitigate some of those limitations by working to add more comprehensive phenology annotations (if it interests us).

Researchers interested in working with iNat phenology data can start by pitching in to improve those annotations. Ultimately, though, they’re likely to want to download a dataset and apply a bunch of manipulations in order to try to derive more value from the data.

For example, if you ensured that (nearly) all Malacothamnus nuttallii observations were given a phenology annotation for Flowering/Flower Buds/Fruits/No Flowers, you could then use the downloaded dataset to examine what percentage of observations in a given week were assessed to be flowering, which might control for the variable number of observations. You could compare the shape of the flowering percentage curve from one year to the next (maybe smoothing across 3 or 5 weeks might give less noisy data). You could compare flowering dates across related taxa. You could try to correlate early or late flowering with average (or minimum) winter temperatures, or with the prior season’s rainfall.

So, lots of ways to use this type of phenology data, but I think most of the analysis and visualization is going to happen outside of iNat!

[pisum]

@chase-prairie

how would you like to see that data displayed?

i don’t think it’s wise to try to jam all that functionality into the taxon page’s graph panel. it’s better to have a separate page / tool for custom or more complicated visualizations. i personally think comparison of years works best on a calendar heatmap or spiralized calendar heatmap that displays multiple years.

[xris]

@chase-prairie

This is really handy for folks like me, looking to schedule when to check out particular sites and go pollinator photographing

I’ve used this technique to hunt for pollen-specialist bees. Two years ago, I created two “Bee Quest” Projects, one for the Blueberry Digger Bee and the other for Pickerelweed Pollen Specialists. Both include observations, or links to observations, of their pollen host plants, to see where they’ve also been observed, and when they are in bloom.

Last year, it paid off. Through 2024, there had been only eight New York City observations of Dufourea novaeangliae, the Pickerelweed Shortface Bee, all on the eastern banks of the Lake in Brooklyn’s Prospect Park. Pickerelweed is conspicuous in bloom, and lots of folks like to upload observations of it to iNat. I watched for observations of the plant in bloom in Prospect Park, and checked it out. I found my first Dufourea there last July, along Prospect Lake where it had been observed by others, adding two observations.

Now that I knew what it looked like in life, I hoped to find it elsewhere in NYC. In August, I found it on the banks of the eastern lobe of Harlem Meer in Central Park, adding three more observations. These are the first, and so far only, iNat observation of this species in Manhattan (New York County). This confirmed for me that the approach can work, and that iNat can be used as a tool for it.

These two locations are 9.5 miles from each other, much greater than the expected foraging distance for this medium-small bee. This gives me hope that it’s already present in other locations around NYC where pickerelweed can be found.

[keirmorse]

@rupertclayton

The data on these graphs are also vulnerable to fluctuations based on hard-to-predict factors. For M. nutallii we might look at the graph above and conclude “Wow, there was some kind of superbloom in June–December 2021, and again in April and May 2022. But this species has really had trouble flowering since then.” What we’d be missing is that @keirmorse used iNat data to inform a bunch of his research on Malacothamnus, and as part of that made a major effort to add IDs and annotations to observations of the genus several years back. The supposed lack of flowering observations in earlier or later years probably just reflects a change in Keir’s research priorities!

Yep. I annotated everything at the time for my treatment to best represent what the phenology of each taxon was and compare where useful. I probably haven’t annotated any Malacothamnus phenology since. With this data, I also was not looking for variation between years. The weather in the years since iNat started have been quite variable, at least in California. So, I just looked at the full dataset to show what to expect whatever the weather does. I can also note that many Malacothamnus can bloom most of the year but it may only be a random branch or a few flowers off season, which can lead to misleading phenology data due to observer bias. While a population of 1000s may have only a few flowers on one plant outside of the peak season, an iNat observer may be much more likely to photograph that one plant out of context of what the majority of the population is doing. And, if that plant is right on the side of a popular trail, 20 people may photograph it, which can lead to strange peaks in the phenology data.

[pisum]

@pisum

i personally think comparison of years works best on a calendar heatmap

i made a page for this:
page: https://jumear.github.io/stirfry/iNat_calendar_heatmap
code: https://github.com/jumear/stirfry/blob/master/iNat_calendar_heatmap.html

as an example, this would visualize American Robin observations in Texas between 2019-01-01 and 2025-12-31: https://jumear.github.io/stirfry/iNat_calendar_heatmap?taxon_id=12727&place_id=18&d1=2019-01-01&d2=2025-12-31

it might be useful for a use case like this, too: https://forum.inaturalist.org/t/enhance-observation-calendar-contrast-differentiate-shades-for-high-volume-days/54271.