Diario de cappaert

In our projects in western OR/WA, we can generally ID the "typical" larger species (Lasioglossum sensu stricto), for which there is a well regarded key. But the little "weak-veined" (sensu lato) Lasioglossum are a problem. One, because they represent a large fraction of our collection from prairie habitats. So we need IDs to uncover much of the pollinator diversity. Two, because there is no reliable route to IDs. You would need to consult a range of keys for the different subgenera to get close. But none of these cover Pacific Northwest species. (There is a project underway, by Joe Engler and the Oregon Bee Atlas, to develop a partial key).

So in the meantime, we need at a minimum to define taxa--species or morphospecies--that might improve the resolution of pollination networks. My attempt at this is 100 hours devoted to a pretty crude beginning. That is how taxonomy proceeds--for me at least. I have a key, and a set of supporting trait documents.

The key - Weak-veined Lasioglossum of the Prairie Pollinator Project- identifies 14 taxa, a mix of named species (mainly from barcoded specimens), and morphospecies. It is likely that the key fails to ID some species (lumping), or confounds variations that are in fact one species (splitting) . But it is a place to start.

In the process of creating the key, I needed to educate myself on the traits that typically differentiate species. These came from DiscoverLife, and various publications of Jason Gibbs and his various co-authors. So I have documents like these:

Subgenus Guide. Categories for little Lasios.
Propodeums of Lasios. An effort to simplify characters of the most important diagnostic feature.
Comparison of dialictus species in the key. Useful if you can't hold a sight picture of many things at once.

I have spent weeks of my life in the DiscoverLife Andrena key. It is a great resource but far short of a slam dunk for ID of the 521 Andrena species in the guide. One of the limitations is the lag in scoring. Ideally, each species in the key is associated with trait conditions. Species X has long fovea, a ridged pygidial plate, etc, for dozens of traits. But for rarer taxa, these traits are unrecorded, or scored for a wide range. E.g., species Y has moderate OR long OR very long fovea. This leads to many false positives. The solution is for a highly qualified taxonomist with a vast collection to spend months ironing this out. Or more likely, to chip away at it as resources are available.

A second limitation is the user's level of knowledge about how traits are defined. I can help with this by adding explanatory trait images to the DL key. But these images are small, and often insufficient to get the user to a place of confidence when deciding on trait conditions.

So I created a page: Andrena deconstructed: A photo guide to the characters of the DiscoverLife Andrena key

Anatomy discussed:

The key: Identikit key: Select Andrena Examples
This is NOT a useful diagnostic resource. It is a proof-of-concept experiment with Identikit software, a possible alternative to the xml-based interactive keys of DiscoverLife. I've populated it with an arbitrary set of 7 taxa, and a subset of characters used in the DiscoverLife Andrena key.

---

Background:
Andrena is kind of a worst case for making definite IDs, because there are so many species, which often differ only in subtle ways. There are a couple of resources that you might use to ID a specimen.

You can consult dichotomous keys in the primary literature, which are spread among many hard-to-access publications (that may be out of date), and rife with complicated terminology and comparisons to exemplars that you do not have in a collection. This is the conservative path, but impractical for most.

Or you can use the key in DiscoverLife, which is far friendlier to non-specialists. Criticially, it is interactive, so that a diagnosis does not hinge on a correct call for each step of a dichotomous key. The design favors plain English, and descriptive illustrations and photos. In most cases, species pages have specimen photos and detailed descriptions from primary literature, and distribution maps. The DL Andrena key does have flaws, to be expected in a resource attempting to accurately discriminate among >500 US species. But at the moment, it is probably the best first step to get at least close to ID.

However, the software behind the DL keys is suboptimal. To get a sense of what I mean, compare the
Osmia key from DL,
to a much more elegant interface in the Osmia key from ExoticBeeID*. The latter is cleaner, more efficient and intuitive, and offers trait and specimen images at hi res. A down side: it uses Lucid software, which is expensive. In my sample Andrena key, I've used Identikit, which is open source and free. A long term goal would be to move keys like DL's to this kind of platform.

The DL authors - I've discussed with Sam Droege and Clare Maffei - are amenable to a new kind of deployment. There are a couple of impediments. The most obvious is that re-formatting a key with 520 taxa and 100 traits would be a crazy amount of work. A second problem is about the long term - the current xml-based platform is somewhat future-proofed, maintained by USGS, on a stable server space. Identikit software is written in javascript and deployed on github. What would be needed to assure its performance as software evolves and bugs come up? I don't know, but I am scouting for info on this.

If you have read this far (all 3 of you), my current take on this: Keys like Identikit or Lucid will win out over older formats, sooner or later. I can generate such keys, and I likely will for a bee genus without the mountain of data in the DL Andrena key. I am also interested in using this software for any other taxon where I have enough information and imagery to do a credible job. I've done this for the draft Ceratina of the Pacific Northwest. A reach, but worth contemplating: Andrena subgenus, perhaps based on the subset represented in OR/WA specimens in hand (collections of Quamash EcoResearch and Institute for Applied Ecology), and in other material I can get from the Eastern Ecological Science Center (Sam Droege, Clare Maffei) or other projects. Stay tuned.

---

• Neither key is anywhere near complete--there are ~150 species of Osmia in the US. Exotic bee ID is limited in scope, with 29 species. DL is getting closer to comprehensive in a draft version of the key, with 131 species, which incorporates western species.

The complexity of structures, textures, and colors on the surface of an insect are of obvious interest just for their elegance and beauty. Of course you think about that as you also explore these features for the more practical objective of making an ID. Two considerations for entomologists, which are irrelevant to the insect; of course they know they are beautiful, and they ID each other without reference to measuring wing venation and the like.

Another thing to think about, easy to forget when looking at dead specimens, is the fact that in the living insect, most of the tiny details are in the service of some function. So a friend asked: Why are the eyes of hoverflies hairy? I can't find an answer. Could it just be that hairiness is incidental to the formation of the eye? It seems not; some syrphid eyes are absolutely bare, others are carpeted with hair.

My own main concern is bees, in particular Andrena. In keys, you are asked to look at the pygidial plate, a flat surface that might be smooth, or patterned in many ways. Having recently discovered this trait, I have paid close attention to its diagnostic value. But as much as I like to think of myself as curious ... it had not occurred to me to wonder about the adaptive value of this feature. Until it was pointed out to me (thank you WIll Peterman) that the pygidial plate is a digging/smoothing tool, used by ground-nesting bees to form and water-proof brood cells. As far as the specific advantages of the different forms of pygidial plates ... a new thing to consider.

At the end of several rainless months, when fields are nearly devoid of flowers, you expect bees to be equally scarce. But I am seeing lots of bees on the few open flowers (Mentha, Lontodon(?), Madia). An occasional Apis or Ceratina. And lots of Lasioglossum. I collected a few dozen of these (large ones, olympiae? and dialictus), and they were invariably male.

What kind of sense does that make? The odds that any of the males I saw will mate on that day is zero. Do they wait around?

BTW, the dense and highly plumose clypeal hairs on the dialictus are striking. Presumably that is a dialictus male thing.

The obvious advantage of an interactive key is that just one character might get you close to an ID. You could diagnose Megachile pugnata or Halictus ligatus either by walking through a long set of traits--or get to species if you know about the projections from the cheek. A couple of Osmia shortcuts:

In DiscoverLife you use the "has" option in the menu, choosing Sparser centrally on frons, with stiff, thick, straight hairs above and below angled towards the center of the frons, forming a basket. This gets me to 3 possibles for Oregon (out of 96). The bad news is that when I then use the "difference" option for those three, only one of these (O. ednae: clypeus bulging) can be distiguished from the others. So you have to go to details beyond the key; however narrowing the Osmia list to 2 or 3 is huge (for me).

Even better, again using the "has" option and choosing Head, hypostomal area, hairs
Strongly curved, mostly restricted to a row along edge of hypostomal area nearest gena- Forming basket, I get to two choices (O. densa, O. clauca), distinguishable by hair color of the scopae and clypeus.

Any field has its daunting terminology. In taxonomy, there is no way around it. A crucial set of traits for bees: surface sculpturing. It has just been pointed out to me that there is glossary of terms, with images: Surface Sculpturing. The source publication is referenced. The terms are particularly useful for ants (shucks!), but still useful the next time you want to know what areolate-rugose means.

Caveat: the following includes free-association speculation not based on careful reading of literature. This is a journal; I get to do this.

If you are from the east, the WIllamette Valley climate is an adjustment. In Michigan or Connecticut it actually rains in the summer, so you can find many flower-visiting insects on any decent day from April to November. Late summer is particularly rich, as goldenrod and aster flower in dense stands. By contrast, western Oregon is dry from mid-June until mid-fall, and nectar and pollen resources fall to near-zero in most habitats. Any non-honeybee attempting to forage in August is wasting its time.

Unless it lives in the city. In lots of urban areas, floral resources abound. Where I live (Corvallis), most yards are dead grass, but on any block there will be one or two properties jammed with late summer flowers and an abundance of bees and flies. I've seen Halictus, Megachile, Triepeolus, Agapostemon, Melissodes, Bombus, Ceratina, and honeybees. Plus syrphids including Eristalis, Eupeodes, Myathropa, and Syritta. And cabbage, alfalfa, skipper, and gray hairstreak butterflies. There has to be a powerful selective advantage for insects that are active for two or three months beyond the natural wildflower season. For bees this would include larger or better-provisioned broods. For some syrphids, it could mean additional generations. But of course it is also complicated. Consider possible cases.

Halictus ligatus as a species exhibits some plasticity in the colony cycle, which can be multivoltine, with pauses during dry conditions (in southern Florida). Could either feature appear in bees in Oregon's Mediterranean climate? If so, that might amplify reproduction. But what would happen to the sex ratio during the colony cycle? I note that where I am, males are predominant (but not exclusive) in September; these might have minimal effect on populations.

Osmia lignaria of the Willamette Valley is the subject of a just-out publication examining the (impressive) rate of brood cell production over the month or so that nests are active. The life cycle is probably of fixed length; O. lignaria will enter diapause regardless of an extended availability of floral resources. If so, effects of an extended floral season might be indirect, favoring competitors or cuckoo bee enemies.

Eristalis tenax is observable in the Willamette Valley in any month of the year--a longer activity window than other common Eristalis. Scanning iNaturalist observations, it is pretty well represented on wildflowers--esp yarrow and Queen Anne's lace--that occur in dry conditions late in the season. But in a city, these resources are dwarfed by garden landscape plants that are also exploited by E. tenax, which enjoys multiple generations. There is at least the potential that their populations would be significantly augmented by late summer landscape plants, favoring this species over others.

There is some literature on this question - E.g., Temporal dynamics influenced by global change: Bee community phenology in urban, agricultural, and natural landscapes, which reviews fairly scant literature. But urban landscapes deserve consideration, as mini-experiments in adaptation, and as habitats that are expanding with increasing interest in native plant landscaping.

Along these lines, I have the pet idea that it would be really interesting to study pollinators in botanical gardens, which have diverse kinds of flowers in a large area, well defined and generally isolated from other irrigated landscapes.

Below: some August/September flower visitors in Corvallis.

I may eventually be good at diagnosis of at least the easier genera. In the meantime, I can appear wherever a species announces itself with a single, unique trait. E.g., Halictus ligatus and Megachile pugnata are unmistakeable for conspicuous genal processes. Another case: In DiscoverLife there are 96 possibles for Oregon Osmia. Osmia inurbana is the only answer for Middle of rim with distinct, narrow, acutely angled, triangular notch in Oregon.

This field season I indulged a curiosity about Andrena astragali, an obligate associate of death camas. I looked pretty hard for this species in stands of abundant Toxicoscordion venenosum in the Willamette Valley near Corvallis. My main finding was that T. venenosum was intensively visited by the dronefly Eristalis hirta; A. astragali was not to be found--until I uncovered just one site (Thanks Tom Kaye) with a lively population of A. astragali. Today I was looking at insects collected from Fisher Butte, a site with acres of death camas and swarms of visiting droneflies, but no history (we have 4 years of data) of A. astragali. But we had one:

One of the things I like about astragali is that the ID is straightforward. DiscoverLife narrows to this species pretty quickly with simple characters, which is an unnecessary step if you know the host plant is death camas.

I had a similar-looking Andrena, one of few (in our study sites) with complete tergal hair bands:

But this was collected from Potentilla gracilis, an argument against astragali. In the key, I get to Andrena auricoma. In the description I am told This small western species is very similar to A, astragali but can be separated from the latter by its smaller size, rhomboidal and feebly emarginate labral process, and its shinier terga. Check, check, check. Gratifying, until someone tells me I am wrong.