Bugs on Flowers: A Survey of Too Many Observation Field Options

General
1

Generally speaking, I’m a plant person, or, more accurately, I thought I was a plant person until I got too deep in the pollinator-supporting-plants sauce. In academic publications, we know some things about which (colloquial) bugs visit which particular flowers, but wouldn’t it be cool if we could use iNat to add to that knowledge base? We do that!

Folks have been doing that work for a long time, dating back to at least to 2012 with the “Pollinates” observation field and projects like the excellent 2013 Moth & Butterfly Project and 2016 Pollinator Associations Project. Because so many different people have all tried to document these relationships, there are a bunch of different Observation Fields in use to record interactions. To understand what’s going on with each, I’ve done some survey work to see what’s currently in use.

Observation Field Text input Creation Date Worldwide Usage (Research Grade observations)
Interaction-> Visited Flower of taxon 8/12/15 520,287
Nectar / Pollen delivering plant taxon 6/15/16 189,804
Name of Associated Plant taxon… ish 12/7/17 170,777
Associated species with names lookup taxon 4/5/13 146,968
Nectar Plant taxon 8/9/13 106,926
Host plant Taxon 3/13/13 96,413
Feeding on taxon 10/12/14 29,755
Flower Plant Name (What plant was the pollinator visiting) Text 6/6/17 15,042
Pollinates taxon 10/4/12 12,430
Plant association taxon 5/15/19 11,313
Flower or plant association Text 6/22/17 3,000
Insect on flower=Yes y/n 12/21/22 2,554
Associated plant name text 3/2/16 1,564
Visiting flower of: taxon 12/10/21 1,490
Plant species observed Taxon 4/2/15 655

That’s a lot! And these are just the top ones that hit my interest in the Midwest US. I’m sure there are plenty of less-used tags. Part of the issue is that a lot of these overlap, but aren’t synonymous. A visit to a flower doesn’t necessarily indicate pollination nor nectar acquisition. Do you mean Host Plant like a milkweed for a monarch, or just any flower a bug could snack on? Associated plant species can often, but not always, indicate herbivory rather than flower visitation. Sometimes, they’re just standing on a leaf! Should that count for flower visitations? (no.) Add in that user choice can ignore the stated purpose of a OF, like how any butterfly photographed flying near a flower is often tagged as “Nectar Plant”. Add that all up, and you end up with some messy catergories.

Due to specific projects using different OF, there’s also a species bias depending on the OF you’re looking at. Here’s a quick survey of the Research Grades in each Observation Field:

Observation Field Worldwide Obs Bees Wasps Beetles True Bugs Flies Butterflies Skippers Moths Hummingbirds Ants
Interaction-> Visited Flower of 520,287 163,289 10,526 22,164 2,609 34,708 88,839 21,092 16,840 3,786 823
Nectar / Pollen delivering plant 189,804 51,028 3,630 3,588 252 9,156 51,801 8,486 3,131 3,172 129
Name of Associated Plant 170,777 25,153 5,690 10,760 4,395 8,738 15,643 2,905 4,633 173 623
Associated species with names lookup 146,968 20,465 1,537 6,247 3,117 5,795 20,532 4,626 3,146 343 535
Nectar Plant 106,926 18,984 1,393 2,214 298 4,796 41,185 8,640 8,681 868 135
Host plant 96,413 566 74 4,107 1,680 5,195 7,458 326 8,974 21 82
Feeding on 29,755 2,804 95 2,532 356 899 3,004 389 1,133 300 126
Flower Plant Name (What plant was the pollinator visiting) 15,042 8,644 33 548 62 644 1,231 181 218 28 33
Pollinates 12,430 2,643 55 303 60 468 2,953 443 318 684 25
Plant association 11,313 2,496 70 383 123 616 662 120 218 32 43
Flower or plant association 3,000 847 115 94 39 180 237 73 65 2 20
Insect on flower=Yes 2,554 439 31 141 4 232 455 97 71 0 1
Associated plant name 1,564 165 9 210 79 51 198 40 62 2 5
Visiting flower of: 1,490 176 13 76 12 199 132 21 21 176 5
Plant species observed 655 544 1 3 4 5 9 1 2 6 0

Or even better, as percentages, here’s this table. For each column, consider that, for example, 54% of all bee observations linked to flowers are in Interaction→ Visited Flower of:

Observation Field Bees Wasps Beetles True Bugs Flies Butterflies Skippers Moths Hummingbirds Ants
Interaction-> Visited Flower of 54.75% 45.23% 41.53% 19.93% 48.42% 37.91% 44.46% 35.44% 39.47% 31.84%
Nectar / Pollen delivering plant 17.11% 15.60% 6.72% 1.93% 12.77% 22.11% 17.89% 6.59% 33.07% 4.99%
Name of Associated Plant 8.43% 24.45% 20.16% 33.58% 12.19% 6.68% 6.12% 9.75% 1.80% 24.10%
Associated species with names lookup 6.86% 6.60% 11.71% 23.81% 8.08% 8.76% 9.75% 6.62% 3.58% 20.70%
Nectar Plant 6.37% 5.99% 4.15% 2.28% 6.69% 17.57% 18.21% 18.27% 9.05% 5.22%
Host plant 0.19% 0.32% 7.70% 12.83% 7.25% 3.18% 0.69% 18.89% 0.22% 3.17%
Feeding on 0.94% 0.41% 4.74% 2.72% 1.25% 1.28% 0.82% 2.38% 3.13% 4.87%
Flower Plant Name (What plant was the pollinator visiting) 2.90% 0.14% 1.03% 0.47% 0.90% 0.53% 0.38% 0.46% 0.29% 1.28%
Pollinates 0.89% 0.24% 0.57% 0.46% 0.65% 1.26% 0.93% 0.67% 7.13% 0.97%
Plant association 0.84% 0.30% 0.72% 0.94% 0.86% 0.28% 0.25% 0.46% 0.33% 1.66%
Flower or plant association 0.28% 0.49% 0.18% 0.30% 0.25% 0.10% 0.15% 0.14% 0.02% 0.77%
Insect on flower=Yes 0.15% 0.13% 0.26% 0.03% 0.32% 0.19% 0.20% 0.15% 0.00% 0.04%
Associated plant name 0.06% 0.04% 0.39% 0.60% 0.07% 0.08% 0.08% 0.13% 0.02% 0.19%
Visiting flower of: 0.06% 0.06% 0.14% 0.09% 0.28% 0.06% 0.04% 0.04% 1.83% 0.19%
Plant species observed 0.18% 0.00% 0.01% 0.03% 0.01% 0.00% 0.00% 0.00% 0.06% 0.00%

So what should we do?

(edited for clarity) iNat seems to consider Observation Fields something of a lost cause not worth revisiting, so presumably if there will be a better-curated tool in the future, it would be more like Annotations, which has taken the place of some OF data. I’m not super worried about any change coming from devs, tbh. I am most interested in the data we have already at hand, just waiting to be utilized.

Personally, in order to make all this amazing community data useful, I am consolidating bug-on-flower data in one field. “Interaction → Visited Flower Of:” personally seems most accurate and neutral for who is visiting flowers, and I am utilizing @Megachile ‘s amazing Universal Metadata Tool (Beta) to retag existing Research Grade observations in my region. By bringing the flower visitation documentation in other OFs into one coherrent tag has allowed me to treat this as a dataset for my professional research as a pollinator-supporting plant mix designer.

The example view of my IDing setup with @Megachile ‘s tool:

Screenshot 2025-12-03 at 8.50.41 PM
Screenshot 2025-12-03 at 8.50.41 PM2886×1692 637 KB

I hope this longwinded explanation of some bug-on-flower observation fields is useful for someone out there. It took me months to understand what’s out there and I’m so excited for what we can do with this data together!

For more deets, and the URLs I used to generate this data which you can apply to your region, check out:

https://docs.google.com/spreadsheets/d/1bP0_Sc_-DjGg6R88PiWSikwww5r__yj2QfTcY0Hc22M/edit?usp=sharing

PS: If you’re a fellow academic interested in publishing on this topic, reach out! I have a few things in progress…

37 Likes
Expand "Reviewed" to Status = "Reviewed", "Watch", "Ignore", "Unreviewed" (none)
Auto-link associated observations with observation fields when duplicating an observation
2

Brilliant dive into OF data. Thanks for this amazing effort.

5 Likes
3

in addition to ‘Host Plant’, there’s also ‘Host Plant ID’, which is currently being used for over 400,000 observations. The description for that one does state that it’s meant to be used for ‘Host of a herbivore or parasite species’, and that’s generally what I use it for (galls, fungal diseases, mistletoes, etc), but from a quick scan there are definitely also a lot of pollinator records that are currently using it

7 Likes
4

Ooooh boy, thank you. Looks like I’m gonna be retagging in there for a while!

2 Likes
5

Excellent post. I would like to mention that I started a Pollinator Project for my interest area of Chironomidae. While not generally viewed as pollinators. I have started trying to document flower interactions of them on iNaturalist.

https://www.inaturalist.org/projects/chironomid-flower-interactions

4 Likes
6

really? that’s news to me.

it may be worth noting that GloBI aggregates interactions from iNat and other sources. it does a decent job of linking one taxon to another, but it’s not going to provide additional dimensions like place, etc.

7 Likes
7

I personally use Name of Associated Plant the most, I’m nearing 4000 observations with that field. I do wish observation fields were curated and merged to prevent the extreme amount of duplicated fields that cover the same topic. I know quite a few power observers and identifiers that don’t bother with the feature at all (because of all the duplicates and joke fields) which is a shame.

7 Likes
8

very cool! can’t wait to tag all my bugs

2 Likes
9

Congrats and thank you for all your tagging work! This is a great example of a challenging tag to data-wrangle. Because it’s not explicitly about the flower, a large quantity of Name of Associated Plant is either herbivory, whether that’s caterpillars, leaf miners, rusts, etc, or it’s the plant the colloquial bug is standing on. When I download the raw data, there’s no way to tell the difference between a flower visit and a place to rest. That’s why it’s one of the OF I’ve been retagging over the past few months.

4 Likes
10

What’re you referring to?

1 Like
11

I do wish observation fields were curated and merged to prevent the extreme amount of duplicated fields that cover the same topic.

Me too. The worst is when you think you’ve found the perfect observation field for what you need, but then the description says it’s specifically for one type of organism, like badgers or leafhoppers. So then you have to create a generic version with an awkward name, like “Wingspan 2”. This is one of the reasons there are so many duplicates.

7 Likes
12

The fact that Observation Fields is a banned topic for feature requests, the slow-walk addition of Annotations as an alternative for some observation fields (I assume flower visitation will never be one), and the general sense I’ve seen searching the forums that there’s a resignation that OFs are and will always be a chaotic mess. The general sense I’ve encountered, reading tons of old forum threads trying to learn how OF work and who has been trying to resolve these issues, is that OF exist, but they are not a priority for dev improvement or organization. Presumably, if there would be any improvement made, it would be via a different, new route. I am personally imagining something like Traditional Projects vs Umbrella/Collection Projects, but who knows.

What I wouldn’t give for even an OF UI that lets you cleanly access them, instead of URL strings you need to hunt down from three different threads from six years ago OR by clicking the Observation Field links on tagged observation (what this guy is for me). Alas I know it wont happen. That’s why I’m slowly building out my own OF url builder.

edit to add important note: Yes, every time I read or write OF, I am thinking about an OnlyFans but for bugs.

8 Likes
Annotations - happy to help
13

There was some interest in improving how iNaturalist deals with interactions, but it seems to fall into the “hard-to-do” category.

There are also ways of capturing interactions that don’t depend on OFs, such as through projects: https://www.inaturalist.org/projects/interactions-linked

5 Likes
15

I can see it might be hard to find the relevant thoughts shared by staff yeah but I don’t think there’s anything “cagey” about what has been said, so was confused by that characterization.

Next to observation fields, I added this text to the pinned feature request topic you linked to: “(individual fields are user-controlled and there are no plans to develop these further)”

Maybe, like Guides, there should be a banner at the top of the Observation Fields homepage to indicate that as well, if that’s still the plan:

https://www.inaturalist.org/guides
https://www.inaturalist.org/observation_fields

image
image1066×249 92.3 KB

3 Likes
16

I initially felt the same way as you. That iNat should step in and standardize things. Make a choice and then do a data conversion to normalize the interaction data. I still sometimes feel that way tbh. But…

There are many projects and customizations that likely rely on these user defined fields. Even though many of them might be perceived as synonymous they are often used for slightly different reasons. I’d worry about several things when trying to consolidate several synonymous fields into “Interaction → Visited Flower Of”…

  1. Pinging every owner of every observation (and all the IDers) when the redundant field is added to each observation. (unless there’s now a way to shut off the notifications through the meta-data tool… there didn’t use to be).

  2. Storing all the taxon data on each observation twice. The Gerald observation is slow to load (probably due to the massive amount of meta-data that the UI needs to interpret).

  3. Choosing the wrong field to consolidate into. Visited flower of is pretty specific. It suggests that a caterpillar isn’t eating a leaf, for instance. In most cases this data is inherently hierarchical. If I wanted to try to consolidate the data, i’d choose a very broad category such as “found on”. This does’t say anything about what one taxon is doing on another taxon. It just annotates the fact that the one was observed on the other. Once the “found on” taxon is captured, a simple name/value pair (with options limited via a drop-down) could be used to describe what the one taxon was doing on the other (sample actions… pollinating, nectaring, eating, mating, etc). The added benefit of this is that there woudn’t be redundant taxon data stored with several observation fields of type taxon stored on each observation (even when they are often redundant). People would rightly point out that this solution isn’t perfect. An insect could be found sitting on one plant while pollinating another… but I still think designing a solution that satisfies the 90% while failing the 10% is worth considering.

1 Like
17

I must have explained myself poorly, I apologize! I’m not removing existing tags, merely adding an additional tag. I am never advocating for removing or replacing others, because it will break so many local projects who chose to create their own OFs.

Yes, I can’t stop it from notifying people. believe me, I wish I could. I’ve met strangers on trails who complain about it when they realize who I am. Alternatively, it has also gotten me invitations to speak at others academic labs and also gotten many people to message me to learn more and add more OFs to their own obs. So some good, some bad. I used to add all relevant OFs, so that each observation would be picked up by whatever the relevant project was, but that was making the notification spam worse. I had to pick just one, and given Interaction→ Visited is already the top one in use and it is appropriately specific, it seemed like the one to go for. To be clear, I am only tagging things that actually are visits to flowers. I am intentionally excluding herbivory, because that’s what makes the other tags useless as floral datasets.

So that leaves us down to “it’s annoying to have the same plant tagged twice on my observation.” Do you view that as a pressing issue? Is this website not intended to have people add detail to your observations? I’m a little confused by the argument. And of course, if it really bothers you, you can remove OF permissions from non-curators.

From my perspective:

Either I download every field, and carefully edit my local, private spreadsheets of 50,000 observations for literally weeks. Every time I do additional data downloads, I have to make the same data cleanup effort. I cry.

OR

I do the editing/tagging work on the website itself, saving not only my own time, but giving everyone else access to cleaner data they can get themselves. I don’t have to do duplicate work, I am able to click one link to see every observation of species visiting Gentiana andrewsii flowers, and so can you!

What would you suggest as an alternative? The extension of “it’s annoying/laggy if there’s too much metadata” is that OF should be limited to only one or two per obs, which doesn’t seem useful.

Part of what drove me to do this work (in addition to tagging ~20,000 previously untagged obs in my region), is that pollinator visitation to flowers is data that has been a bit obscure. I’ve literally had it treated by fellow scientists as proprietary information that they refuse to share. Hoarding spreadsheets, so weird! iNat is all about data transparency and access. In this imperfect system, I’m just trying my best to make this accessible to more folks.

If you have the solution that fixes what we have now with tools that are currently available, please do share!

7 Likes
18

Can you provide an example of why you would look at your local private spreadsheet of 50,000 observations. When you look at this spreadsheet, what are you looking for? I realize each time you look at it, you might be looking for something different… but what is one example of why you would look at the spreadsheet?

19

Sure! I don’t mind disclosing:

My city currently has a street side plant height limit of 6 inches. We are trying to lobby to change that height. In order to do that, I am building a dataset that relates height to number of pollinator species supported. I am adding iNat research ID bugs-on-flower obs to existing historical data on the subject to best count how many pollinator species each plant species is documented as supporting.

In order to do detailed analysis and publish this, I need a static data set downloaded locally and standardized.

1 Like
20

If you need the dataset to be static anyway, why not merge the observation fields after extracting the data?

21

Because, as both you and I indicated above, they can’t be merged after extraction without including herbivory and bugs-sitting-on-leaves.