annkatrinrose's Journal

Every now and then, someone will revive the discussion on the iNaturalist forum about the supposed uselessness of the male/female annotation option for plants. They aren't common but they are out there! Botanically speaking, these plant species are dioecious meaning male and female reproductive structures are present on separate individuals. So let me present a compelling case of using the male and female annotation for plants for citizen science.

Meet the fairywand, Chamaelirium luteum, also known as devil's bit, false unicorn, blazing-star, and helonias (all very confusing common names as they are shared with other plants). Its range extends over much of Eastern North America all the way from parts of New England and Canada down to Florida and west to Arkansas. A long-lived perennial, it takes about 5-7 years for a seed-grown plant to reach flowering age. The plants grow as basal rosettes of leaves with about 10-20% of the population bolting with tall flower stalks blooming in spring. As you might suspect already, this species is dioecious with male plants making staminate flowers only and female plants making pistillate flowers only. Interestingly, these male and female flowering plants are not evenly distributed but show a generally male-biased sex ratio. Why is that?

Chamaelirium luteum morphology, based on Britton & Brown's Illustrated Flora (public domain) and iNaturalist observation photo of a male (left) and female (right) plant (Milo Pyne, Creative Commons).

The sex distribution in this species was studied by Thomas Meagher at Duke University, NC, and summarized in several papers published in the early 1980s. The first study reported on the spatial distribution of males and females at four different study sites in North Carolina over the span of six years and found that plants of each sex tend to cluster together in same sex groupings with males usually closer to each other while females tend to grow more isolated (PDF). The second part of the study then took a closer look at the sex ratios at different developmental stages from seedlings to adults. While propagated seedlings started out with near equal representation of both sexes, flowering plants were found to show a significant bias with on average 3.5x as many males as females (PDF).

A third study classified the plants into three groups, juveniles of unknown sex, males, and females, and took a closer look at the possible explanations for the observed gender bias. It found differences in mortality rates with seedlings having the highest mortality rate, followed by juveniles and then females before males. Males started to flower at a younger age compared to females, and females grew consistently taller flower stalks with more cauline leaves then males, followed by a reduction in plant size the following year suggesting a higher reproductive cost of flowering and fruiting for females. While males were able to flower in subsequent years with some individuals blooming for every year of the study, females flowered at most every other year. (Unfortunately, this article is not freely available as a PDF but you can get it from Wiley or JSTOR.)

Having read about the sex ratio bias in this species, I was curious whether that was something that would show up in iNaturalist observations as well. So I set out on a little annotation project, marking male and female plants in the Southern Appalachian Mountains (to limit time investment). Overall, iNaturalist observations show a good overlap with the range map for this species on BONAP (the outlier in New Hampshire is a rosette of leaves that looks to be something else misidentified as this species).

Observations of Chamaelirium luteum recorded on iNaturalist (left) compared against the distribution map from BONAP. The yellow shaded area represents the "Southern Appalachian Mountains" place boundary on iNaturalist.

There are no sex chromosomes so the only way to distinguish between male and female plants is by flower morphology. Fortunately, this is fairly straight-forward on typical iNaturalist pictures due to the sexual dimorphism of the flowers as shown in the images below.

Comparison between male (left) and female (right) flowers of Chamaelirium luteum, using a collage of Creative Commons licensed pictures from iNaturalist observations. Male flowers can be recognized by the presence of anthers atop six stamens, while female flowers have a three-parted stigma leading to a swelling ovary that eventually turns into the seed capsule. Both male and female flowers have six white, linear petals surrounding their reproductive parts.

So what's the verdict? I think the case is pretty clear: iNaturalist observations support that there is a sex ratio bias towards more males than females. As reported in the papers above, peak bloom time was in May, which also coincided with peak observation time on iNaturalist. I only annotated flowering and fruiting plants (155 or ~69% of 226 RG observations in May; most of the remaining plants showed rosettes of leaves with no flowers) that were clearly identifiable as either male (128 observations or ~83%) or female (9 observations or less than 1%) for a bias of a whopping 14x more males than females observed at peak bloom. I was undecided about the sex on 18 (~12%) flowering observations, mostly due to the photos being too far away or too blurry to see the necessary details, but even assuming that these plants are females the gender bias is still pretty clear.

Phenology (A) and sex (B) distribution data for Chamaelirium luteum observations in the Southern Appalachian Mountains.

And now that I've successfully procrastinated until midnight, I've got to get ready for final exams tomorrow... o_o (Fortunately, I'm the instructor, not a student, haha!)

While we're finishing out fall semester, I'm starting to plan what to do with my students in spring. I'd love to incorporate iNaturalist into class somehow, but spring is a challenging season for observations (winter weather keeping a lot of plants and critters dormant until shortly before finals). So I've been thinking about annotation projects that can be done with already existing observations that would also work as a homework assignment or snow day activity. So far, I've come up with this list of guidelines for such projects:

• make use of observations already uploaded to iNat
• suitable for beginners (no steep learning curve)
• don't require students to make identifications
• have a reasonable time limit (e.g. 3 hours to gather data)
• could be split up to work as group projects for student teams
• obtain results for discussion and further analysis

I like to try things out before using them in class, so I figured I would give the assignment to myself and pretend to be a student. Since I plan on focusing on phenology annotations in my Botany course, I wanted to pick something as a "pilot project" that isn't necessarily plant-focused to avoid basically doing the homework for my students. I want them to get curious about plants and design a cool project that they can own! Therefore, I picked the monarch butterfly for a proof-of-concept study and decided to annotate life stages. I think this is very similar to looking at plant phenology as it should be just as easy for a student to recognize e.g. a plant in bloom as it would be to tell butterflies and caterpillars apart. Additionally, it is a very iconic and easily recognized creature that gets ID'd very quickly and accurately by the community.

To demonstrate their understanding of the scientific method, students should start with a question to provide the motivation for collecting data. In my case of the monarchs, I've been curious for a while how their appearance here in the mountains compares to other places off the mountains. I rarely see any before mid summer. Why is that? My hypothesis was that the monarchs coming up from Mexico in spring avoid the colder mountains at first, but then later show up here to breed over the summer before the fall migration back south. Based on this, I would predict a lack of adult monarchs in the mountains in spring, but a bump of caterpillars for late summer and a bump for the adults migrating through in fall. The objective was to annotate life stages throughout the year across several locations to be able to test this hypothesis and compare the data between mountain and non-mountain areas.

I started with annotating life stages (egg, larva = caterpillar, pupa = chrysalis, adult = butterfly) for NC observations and indeed found a low number of butterflies in spring compared to summer and fall. Not quite content yet, I expanded west to include three more states crossing the migration path around the same latitude - Tennessee, Arkansas, and Oklahoma. I decided to stick with the same latitude since decreasing day-length is a major trigger for the migration in fall. Adding annotations for these took a couple of hours, but I think for a team of four students this would still be within the scope of time investment for a lab class project by splitting up the work to have each student tackle annotations for one state. I then took screenshots of the graphs for the annual distribution of life stages for all four states and used PowerPoint to create a summary of the results.

Monarch life cycle stages observed across the year for the states of Oklahoma, Arkansas, Tennessee, and North Carolina. NC is further split into Mountains, Piedmont, and Coastal Plain. Graphs on top of each other were adjusted in height to match the scales on the Y-axis for comparison across states and NC regions. (Note: I did not pay attention to scale the graphs for the regions to the states as well.) Vertical lines indicate peak observation times of butterflies in April, August, and October.

So the big question: Did this result in something interesting to discuss in class? You bet! Look at that neat bump of migrating butterflies in April in OK and AR that's basically missing in TN and NC. The 'central flyway hypothesis' for spring migration confirmed! Then there's a bump in adults showing up in August in NC, TN, and to some extent even AR, that's completely missing in OK, lending support to the idea of an eastward migration of butterflies into and across the Southern Appalachian Mountains in summer. Numbers peak again in October, consistent with the fall migration south.

Students could search for literature to expand the discussion, e.g. comparison with publications based on citizen science data from Journey North (two flyways revealed), or the study by Miller et al. that found monarchs crossing the Appalachians to the east coast in early July north of the area annotated here. One follow-up question to ask could be whether the bump in August in NC/TN represents monarchs raised locally or coming from the west or south or maybe even north. Students could suggest experiments that could be done, informed by the studies listed above and the Monarch Watch tagging program.

Additionally, there's the opportunity to discuss observer bias in citizen science projects. E.g. one would expect more monarchs migrating in fall through OK and AR than NC, yet the peak is higher in NC. How come? Are there more observers in NC than OK? Closer inspection of the observations in question reveals that there are a good number of fall roosts with dozens or even hundreds of butterflies observed in the Midwest, but basically none of those in NC. Therefore, the number of observations underestimates the total number of butterflies being observed in Oklahoma during October. Further, eggs and caterpillars are likely under-observed in general due to the difficulty in finding these to make observations in the first place.

One thing that might prove challenging from an instructor perspective is the difficulty of keeping track of annotations to confirm that students are doing their homework, and providing quality control for those from the teacher side. Unlike for identifications with leader boards for species and locations, there are no stats available for annotations. I haven't found a way yet to search for annotations made by a particular user/student. Maybe there is a URL hack for this, but I haven't been able to figure it out yet. Currently, I'm thinking of having the students summarize their results in the form of a term paper listing exactly what they did on iNat in the methods so I can follow up on it. There is also no way currently to correct a wrong annotation (there's a feature request) short of messaging the observer and/or annotator, as I think those are the only two people who could remove an annotation made in error. I don't think iNat notifies you when someone annotates your observations, so a lot of errors are probably not even getting noticed.

Overall, I'm pretty excited about the possibilities here and look forward to seeing what projects my students will come up with. I think the annotation features on iNat are still way underused and could be a neat way to have students engage with observations of interest to them and add value to iNaturalist by adding or improving data.