Is measuring biodiversity possible on iNat?

[conorflynn]

I’m trying to measure pollinator biodiversity using iNaturalist, but I’m having doubts as to whether this is even possible for a non-expert.

Over the last five years I’ve photographed insects on flowers in my area, with the goal of creating a summary of total pollinator species per plant species per month. I’m using an Observation Field to associate the plant species with the insect Observation: https://www.inaturalist.org/observations?place_id=1823&user_id=conorflynn&verifiable=any&field:Nectar%20%2F%20Pollen%20delivering%20plant=

I knew from the beginning that this would not be a perfect research study because I’m not systematically sampling the flowers with equal effort and equal time, and I knew I wouldn’t be able to ID every insect, due to the quality of my photographs.

But I hadn’t considered that the level of taxonomic ID could bias the total species counts. Even if I could ID every photograph to Genus, some Genera have dozens of species while others only have a few. Even worse, some taxa have more people helping ID and therefore have more detailed IDs, whereas other taxa have less interest and are stuck at Family level IDs.

I think this makes comparing pollinator diversity difficult or impossible on iNaturalist. At best I could say something about how many different Genera or Families visit a particular species of plant – but I’m not sure how important genera- of family-level biodiversity is??

[fffffffff]

You of course can use the data, but iNat can’t be a single source unless you focus solely on what your iNat observations show for that question. Are you going to write an article on that or?

[kentross]

I think it would be helpful to control for the number of observations and the concentration of observations from individual observers. More observations would likely positively bias your estimates since you would likely observe more species with more observations, and I think a higher concentration of observations from individual observers would also positively bias your estimates since people are probably more likely to observe things that are new to them.

[kevinfaccenda]

One concept with may be useful to you is that of a “morphospecies”, i.e. individuals which appear similar and are likely part of the same species, even if you don’t know exactly what that species is called. This concept is often used in ecological studies when it would be too complicated to identify everything to the species level or species are unsubscribed.

I imagine you could implement something like this using the observation fields too.

[kevintoo]

You also need to consider what I call photographer/photography bias as well. It is far easier to get a identifiable photo of a largish pollinator on a flower that supports longer visit times than it is to photograph very small pollinators or pollinators on very small short visit flowers.

[pfau_tarleton]

If you’re wanting to draw meaningful conclusions, then I think you’ve identified several issues of study design that will prevent you from doing that. It sounds like you’re doing this just for fun, rather than as a scientist planning on publishing the results. The lack of systematic sampling may be an even bigger limiting factor than not being able to identify to species. Even if you can’t use them for your intended purpose, I’m sure you learned a lot along the way–and I’m sure the observations will be useful to someone in the future.

[kgrebennikov]

The results of such a “measurement” can hardly be relevant or of scientific significance. Anyway, what people pay attention to more often and what lives (grows) in some place are two distinctly different sets of species. Finding out the actual diversity requires, firstly, a specific research design to get results that are at least approximate to reality.

In the case of pollinators, however, the “measurement” will be very far from reality. Insects are a group with great diversity. And it is not uncommon that species of a genus or an entire family may be identified reliably by only a few people in the world, or even one person. Sometimes such people don’t even exist currently. But even if they do, they may not visit this site. In the case of Hymenoptera or Diptrera, this is a usual situation, alas. In addition, observations cover mostly large and highly visible species. Although, plants are not pollinated by them alone.

So, this method can only show something near to reality for highly visible and relatively easily identifiable groups. Like trees or birds, maybe. But certainly not most of the insects.

[conorflynn]

iNaturalist and the iNat forum are such great communities. Thank you to everyone who considered my question and replied here.

I’m not a professional research scientist -this is just a hobby.

However, I want to push back a little. Although this may be “citizen science”, many of the problems, complexities, and caveats mentioned here are the same issues inherent to “real” science. Part of the method of science is identifying and working with these problems; just because there are problems in the data, doesn’t mean that any analysis is meaningless. (I especially want to thank @kevinfaccenda here, for his suggestion of “morphospecies”, and @kevintoo for listing some more sources of potential bias to consider.)

Through this project, I’ve become more curious about what is meant by “biodiversity”. How is anyone able to measure it? As @kgrebennikov points out, there may be many insect taxa that cannot be identified by anyone in the world, certainly not from iNat photos.

So how do scientists study or compare biodiversity?

[bugsorbust]

Arachnologist and professional outreach/community scientist here who is studying spider diversity.

To echo some of the sentiment above, without a meaningful sampling method, you cannot get statistically significant results.

Alternatively, you can design a protocol (I suggest doing a heavy lit review to see other people’s methods) for gathering data on pollinator species and have images be a part of that (for example, I upload quite a few of my spider samples to iNat because it helps me as a reference guide and it contributes to the species lists in the state and local areas), but this is something many researchers spend years working on to write a single article- and it requires a lot of time and training (but worth it imo if you have the means and motivation). As you pointed out too, it’s hard to ID a lot of things from just images. As an arachnologist, it’s almost impossible to identify most of the species I work with without a microscope and looking at their specific anatomy (mostly genitalia). This is often the case for a lot of pollinator species.

Often times when biodiversity is discussed in terms of research, there are a variety of ways biodiversity is measured. Some common examples include Simpson index, Shannon-Wiener index, etc. But again, these measurements of diversity require specific sampling protocols as to not skew data and to make sure its significant. There are a lot of stats videos on YouTube you can watch that reference how these measurements are calculated and what kind of protocol designs work for them.

One suggestion I have is to make an iNat project where you can create your own community science based species list. You can even outline specific locations using google maps and uploading it as a place in iNaturalist. Any observations added to the species list will not be able to go into any statistical analyses, but its still a really important list! It can be used by all levels of scientists and could be noteworthy news for the local community and potentially beyond depending on how you want the news to reach people (show the power of community science!).

You could also look into any local bio-blitzes or research past ones from anywhere in the world. Bioblitzes, especially during the ongoing pandemic, often utilize iNat to catalog all of the observations. At the end, you’re able to say, “We found x many species within xyz range!” It may not be statistically significant but again, good for showing the power of community science and being able to give non-researchers experience and exposure to the local flora and fauna. There’s also the opportunity for something like, “This researcher’s paper shows there are x number of bee species in this area. Our community science list found x many of those species!”

One thing I would like to mention just in case is to remember that a lot of knowledge about biodiversity in specific areas is because there was someone there looking for those species. i.e. it might seem like Green Bay, WI has a high diversity of spiders because there are a lot of iNat observations there, but its likely because there are multiple arachnologists here looking for species and recording them whereas a lot of places aren’t being actively studied.

Anyways, that’s my two cents. If you have any questions over any of this, I’m happy to answer them.

[kgrebennikov]

A scientific approach (at least as I understand it) would look like this:

0. Defining a list of ecological niches in the territory under study (in the case of pollinators, a list of known insect-pollinated plants in that territory).

1 (hypothesis). Making a list of species that may inhabit these niches (associated with them and known or likely to be detected in the area).

2 (method). Choosing methods to test the hypothesis.

3 (experiment). Application of methods in the wild.

4 (Analysis). Comparison of the results of the experiment and the hypothesis.

This is a very approximate scheme, but even it clarifies it that research cannot be replaced by a set of casual photographs.

[sedgequeen]

The data you’ve collected about pollinators and their flowers does have value. It inevitably has problems, but you can learn things from your data. Right now, you’re working with the raw data, trying to find out if they suggest ideas about relationships. A very early step in the scientific process, but an essential one.

If I were using your data, the first thing I would do would be to put it in a spreadsheet, with columns for pollinator genus, species (if known), family, maybe even order and plant genus, species (if known), family. I’d add columns for insect color (bee-mimic? dark? etc.) and flower color, maybe shape. And date. Maybe rough habitat classification like garden, meadow, forest, marsh. Then I’d enter my observations, one per row. Then, count them up, in different combinations. (Sorting a spreadsheet is great for that.)

Then, start looking for trends. Do flies seem more common or diverse on some plant families than others? Are certain species seen only in spring? Do insects colored like bees seem to visit different plants than others? Are you aware of environmental factors that may influence results, like location in forest or field? Do some flowers have a greater diversity of taxa than others and do you think the comparison is legitimate or complicated by other factors?

Don’t be worried about the different number of species in different genera or how identifiable the species are – you can lump them into larger groups. It is legitimate to compare Genus A with species B (that’s in another genus), for example, just as long as you know what you’re doing (and report it, if you write this up).

A simple statistical test called the Chi-Square test may help you figure out if the differences you see are big or not. Don’t take the results too seriously – you’re probably violating the test’s assumptions in eight or ten different ways – but at this early stage the test may help.

What data to you wish you had? Would you like more observations on certain kinds of flowers? In different habitats? What complexities do you wish you’d taken into consideration? Now is time to plan for next year’s observations.

[engelhard]

I love your post and recommendations: that’s the spirit. Yes there are issues with the data - but there are many questions you could address, with some creativity and of course, applying some caution when interpreting the results.