Part 3 of 3. Declines in bees in the Washington D.C. area

[Sam Droege]

All: 

This completes the third part of an investigation into both declines in bee populations in the Washington D.C. regions and an investigation into the data that could be used to document those changes. Let me know if you did not get the previous 2 sections.

Looking at Changes of More Common Species

In the previous two sections we looked at patterns of change in the bee community of the Washington D.C. region spread over 100 years of opportunistic sampling, comparing simple lists of the bee species detected during historic and recent time periods.   The results indicated that most species of bees found here historically are likely still here somewhere.  Differences in the presence/absence of bee species on the historic or recent lists appears associated with the arrival of non-native bee species in the region, the previously documented loss of a subset of bumble bees, the ebb and flow of some species at the edges of their ranges, and the stochastic loss of species due to the combination of naturally low populations and relatively poor sampling.

Using these changes in the lists of species to address bee community species changes is neither efficient nor that informative with most changes in populations remaining masked and creating an appearance of stability when that is certainly not the case.  With lists you are, in a sense, throwing out most of the data; the counts of specimens collected for each species.  In this section we will take some numeric risks and see if we can extract useful additional information about bee species population changes using the counts of species, but correcting them in hopes of diminishing some of the known and often severe biases outlined in previous sections. 

The first thing we will do is to trim the recent dataset to only specimens captured with a net.  The reason for this is that data from trapping and visual observations have very different probabilities of detection for many bee species ... both positive and negative.  In many ways trapping and visual surveys can only be used as secondary datasets similar to that in the last section with direct comparisons restricted to netting. 

Another problem in need of correction has to do with historic specimen loss.  We know both from common sense and the above sections that GBIF data (our sole source here of historic information) is incomplete and that the behaviors of both collectors and museums is such that many bees are discarded, some overlooked during collection, and there are a number of reasons that existing specimens may not end up in GBIF (hopefully this will be corrected over time). 

I take a two-fold approach to "correcting" the recent and historic datasets.  The first part is to only use the number of days an individual species was captured rather than the total number of species in the dataset.  This diminishes the impact of large numbers of an individual species being collected on a single day (there are many reasons this could occur).  The other fold is to divide the total number of days any individual bee was captured by the total number of days any bee was captured in that dataset.  The resulting proportion for an individual bee species represents the proportion of total days in which that bee was seen.  

Thus a high proportion means that this species was around a lot and a small proportion means that it was rarely or uncommonly captured.

We can then subtract the proportions of captures between the two time periods to see if the proportions have changed.  A large change indicates that during one of the two periods there were a lot of that species present and in the other only very few.  Thus the change value can be interpreted as a combination of abundance (measured by numbers of days) and degree of difference between the two time periods.

Finally, instead of pooling all the regional data together I split them into 3 independent groups:  District of Columbia, Montgomery County, and Prince George's County.  Having three separate groups allows us to look at the consistency of changes across locations.  Note: Ideally this would be extended and done across the continent... looking not as closely at the individual values from just one region, but at the pattern of positive or negative trends among species.

OK, so the results are presented in the attached spreadsheet.   

 

Three things are certain: death, taxes, and lost data. - Anonymous

If I were being generous, I would call the results "mixed."  I think for many of the reasons listed above you can see even clearer problems with the numbers and ratios of collected specimens.  In a steady state, about 50% of the species should be increasing and 50% should be decreasing.  With the data at hand we can see the pattern in Montgomery County is 86 species increasing and 89 species decreasing.  However DC (113 increasing and 51 decreasing) and Prince George's County (180 increasing and 63 decreasing) show a great departure from a 50:50 ratio and in an intuitively opposite (but not necessarily incorrect) way (far more species increasing than decreasing). 

We can inspect some of the species groups we already and independently know have dramatic changes and see if those are mirrored in these data.  For example, the bumble bee data appears to reflect much of what we know about recent changes and we will cover that below in its own section.  Recently introduced species do show the expected large increases (Anthidium manicatum, Anthidium oblongatum, Anthophora villosula, Halictus tectus, Hylaeus leptocephalus, Hylaeus punctatus, Megachile sculpturalis, Osmia taurus) no surprise there, but good to confirm. Four introduced species that have been around enough to have occurred regularly in the historic data also show uniform increases across all subregions (Andrena wilkella, Apis mellifera, Megachile rotundata, Osmia cornifrons). 

Since there is an indication that the data are pre-warmed to be more positive than negative it's worth looking at any species demonstrating declines in all 3 subregions as a possible indication that their declines have been grave enough to be detected despite the data's tide of biasing factors.  In order for that to happen the species must be common enough to have been detected in these two databases in all three subregions thus we are not dealing here with rare species here.  There are only 9 species that met these criteria ( Agapostemon sericeus, Andrena arabis, Bombus affinis, Bombus fervidus, Bombus pensylvanicus, Bombus vagans, Megachile latimanus, Melissodes dentiventris, Osmia georgica). Agapostemon sericeus, Andrena arabis, Melissodes dentiventris, and Osmia georgica are all still present in the area and it is unclear as to why they show consistent declines but given the large number of species involved there would be an expectation that some would show consistent declines simply by chance.  Megachile latimanus and Bombus vagans were both historically regular in the area but are clearly no longer present, their range contracted to the north and west into the mountains.  The other 3 are bumble bees (B. affinis B. fervidus, B. pensylvanicus) all have been associated with range-wide decline due to suspected introduced pathogens and will be discussed below in a bit more detail.

Bumble Bee Changes

Nature is what we see -

The Hill - the Afternoon -

Squirrel - Eclipse - the Bumble bee -

• - Excerpt from Emily Dickinson's poem number 668

Bumble bees deserve a slightly closer look both because there are lots of reports of declines in research papers and the media and because the Smithsonian's Natural History Museum (unlike other species) has entered data for all their bumble bees. So, we are in much better shape, data-wise, to look for changes in this easy to detect group. 

Since the number of bumble bees specimens present in the dataset were relatively greater, bumble bee data were treated separately.  In the historic dataset instead of calculating the number of dates each bumble bee species was detected divided by all the possible dates any bee was seen, the bumble bee value was calculated as the number of days a bumble bee species was seen out of any day in which a bumble bee was seen.  The result was that the number of collection days for each subregion for bumble bees was lower for bumble bees: Montgomery County (180 days for bumble bees/194 for all bees), Prince George's County (83/109), Washington D.C. (84/101).  The trajectories are unaffected but the degree of change becomes slightly higher because the divisor is slightly lower.

The region has records for 13 species of bumble bees.  Of those, four were only seen in the historic period (B. affinis, B. terricola, B. ternarius, B. fraternus). Bombus affinis has moved from regionally common to an endangered species and is now found in Maryland only in extremely small numbers on the Garrett Plateau at the western end of the state.  The other three only occurred in small numbers during the historic period and while B. terricola and B. fraternus appear to have been impacted by recent pathogen introductions, B. ternarius seems unaffected but was at the very southern edge of its range. 

Of the remaining bumble bee species four show mixed changes between the subregions (B. impatiens, B. bimaculatus, B. citrinus, B. auricomus). Two species show consistent increases across all regions (B. perplexus, B. griseocollis) and four show consistent declines (B. affinis, B. fervidus, B. pensylanicus, B. vagans).  Of the group of bumble bees consistently declining, all have previously been associated with losses of populations due to pathogen load using other data sets. As mentioned, B. affinis and B. vagans while once common in the region, have contracted to the mountains in the west; the other two remain regularly present in the region despite steep drops in numbers. 

We know that in the past B. pensylvanicus was one of the most common bumble bees of the region, perhaps even the most common.  The degree of decline has been large but the species still can be regularly found both in good quality rural habitat as well as native species rich environments in cities.  I find this species every year at the gardens at work and my house, for example.  Circling back to discussions of bias.  This species, with its nearly black wings and striking hair color patterns is easy to spot in the field. Thus, upon speculation, this species is likely collected at a greater rate than other bumble bee species and the degree of decline (while obviously severe) is partially masked by an increase in detectability within the recent database as well as in iNaturalist.  It remains regular in the state but should be watched for status changes, hopefully in the positive direction.

iNaturalist (accessed 9/26/2025) data for DC contains many records for bumble bee species during the past two decades: B. impatiens (1634), B. griseocollis (725), B. bimaculatus (235), B. auricomus (127), B. pensylvanicus (119), B. fervidus (23),  B. citrinus (22), B. perplexus (10).  These ratios appear to give a good sense of the order of abundance of species in DC but almost certainly are biased towards the less common species.  Populations of the first 3 species, in particular, are certainly many more times greater than they appear.  Observers are fatigued after taking pictures of a few of the common species and then pass on subsequent opportunities while looking for novel species to photograph such as the dark-winged bumble bees.

For bumble bees, the approach taken in this document of comparisons between historic and modern databases appears to capture and validate the large (and dramatic) changes these populations have undergone over the last few decades.

Increasers

Whereas 9 species were decreasing in all 3 subregions, 49 were increasing in all three.  That is suspicious.

Some Statistics

"Statistics should be like a drunk leaning on a lamppost; for

support rather than illumination." - Andrew Lang

I could not resist doing a simple Goodness of Fit Test on the data.  Here is the thinking...if we restrict the dataset to only those species that occurred in all 3 sub-regions we find there are 96 species meeting those criteria.  Of those species 9 show all declines in the subregions, 49 show all increases, 24 show 1 decreasing region and 2 increasing regions, and 14 show 2 decreasing regions and 1 increasing region.  In the mythical perfect world there should be 24 species in each category (thank you number 96 for being so easily divisible by four).  Unshockingly our data do not conform to the perfect world and we obtain a whopping chi-square of 39.583333 with a equally whopping probability that this situation is due to chance of p< 0.000000 (meaning that there are more zeros and a one somewhere in ridiculously small p land).

Well enough, but what if we eliminate the Lasioglossum and Ceratina species (id problems) as well as the introduced species leaving just native species?  Of those species 9 show all declines in the subregions, 28 show all increases, 23 show 1 decreasing region and 2 increasing regions, and 14 show 2 decreasing regions and 1 increasing region. This works out to 74 species available for analysis and the chi-square value dropping to 12.01 and the p-value increasing to 0.007339.  The species are still showing a very strong predominance of increasing records, and the new values provide suggestions that the data are still quite biased.

Another way to approach the issue of interpretation in the face of bias in the dataset is to order the species by their average change (across sub-regions) and see if patterns emerge. On the positive end of the spectrum are the introduced species and also quite a number of common species that one could conjecture are only increasing because they were not included as much in the historic record due to the behavior of the collectors and museums which would have discarded or passed over most of the specimens thus precluding them from showing up in the historic record.

Interestingly, a group of Megachilidae show up as strong increasers in the data set (Megachile mendica (24.9%), M. exilis (13.3%), M. rotundata (10.4%), Coelioxys sayi (9.9%), M. brevis (6.6%).  These species I think represent a Lespedeza cuneata effect.  I don't have any quantitative evidence for Maryland, but while this plant was introduced back in the late 1800s it appears to have greatly increased in the 1990s and later.  At Patuxent Research Refuge, where I have had a job for more than 40 years, almost any disturbed site now contains dense colonies of this invasive and has been that way the entire time I have worked there. Older employees blamed early biologists at the Center for introducing the species in their investigations of useful plants for wildlife and the farmer (this is also how we got autumn olive, Elaeagnus umbellata, with trunks two feet across). Lespedeza is highly attractive to Megachilidae and the species mentioned above use it commonly and when netting I almost always spend time netting on L. cuneata, if present. It is unclear whether the ubiquity of this plant has actually increased the populations of these Megachilids or simply made them easier to find.

On the negative side of things the top four declining species are all bumble bees (Bombus pensylvanicus (-21.0%), B. vagans (-18.7%), B. affinis (-14.5), B. bimaculatus (-7.6%).  In contrast the 5th most increasing species is B. griseocollis (21.7%).  Again, such changes mirror already documented patterns of decline in bumble bee species, but the increase in B. griseocollis is perhaps less documented (I will have to look into the literature on this...the focus is often on declining species and that is what sticks with me).

In summary, we can see that trying to quantify changes in bee populations using historic and modern sources of data is problematic.  It is clear that bias in collecting, preservation, and databasing contribute to these issues.  That said, large scale increases and decreases are documented but must be interpreted through a natural history and understanding of historic collecting patterns. 

Even though attempts were made to remove some of the bias by using numbers of days captured and using relative rather than absolute capture rates, analyses of change were clearly biased, creating the appearance of primarily population increases in bees in this particular example. If raw, rather than corrected, numbers were used the results would have been even more skewed toward increases.

Despite the very strong biases skewing data to portray species as increasing, the species that did show declines are worth looking at and considering as possible population crashes or disappearances.  Large negative changes that buck the push the system has to alter data in a positive trajectory are potentially very worrisome. 

We are confronted by a set of species that stand out from the remainder of the data by their large and negative changes. Bumble bees. They show the greatest declines and the average of their average subregion trajectories is -15.4% while the next four ranked declining species average just -3.2%. These results independently closely match the already known large declines in some bumble bee populations and indicate the usefulness in this approach for documenting the collapse of large common populations.     

The great increases in Bombus griseocollis are in some ways an interesting coda to the losses exhibited by the other bumble bees. It requires one to craft a more complex story of collecting bias by bumble bee species in order to explain the fact that you have both greatly decreasing and greatly increasing species in the dataset.  I believe the existence of both trajectories bolsters the validity of great declines and adds some confidence in our data for bumble bees.

While the capture of the declines and disappearances of bumble bees is useful it is also emblematic of the problem of monitoring bees (really any insect). These data are coarse and imprecise, capturing very large changes in very common species and documenting in most cases what was already known.  These data show some bumble bee populations have crashed. So what?  We knew that already.  While in some ways this analysis is seemingly unnecessary it does provide numbers to what could otherwise just be anecdotes and storytelling.  It helps justify and target the next stages of collecting data on bumble bee species on the presence of pathogens and the extent of bumble bee populations. 

In the best of all worlds, monitoring data should provide a warning of an impending problem, not just document ex post facto extinctions and collapses of common species.

What then of the other species; the bee species that are not common but could be disappearing?  I have little confidence in our ability to detect stand alone quantitative change using existing data from just the Washington D.C. area despite indications that some of the changes we have quantified are true in the sense of their trajectory.  We know that species that moved into the region like Megachile xylocopoides show increases (no collector would pass one of these species up...) and the data also track the regional retractions of conspicuous species ( e.g., Megachile latimanus).  We have also seen from LeCroy et al.s (2020. Decline of six Native Mason Bee Species following the arrival of an exotic congener.  Scientific Reports: 18745) work that careful use of historic and recent data provided valuable supporting evidence of shifts in populations of common and introduced Osmia and therefore could be used for other species groups.

These numbers nicely document the increase in now common introduced species but because of a clear bias that pushes estimates of change in a positive direction it is difficult to use the data for the bulk of the species as an objective measure of change.    

In many ways the comparisons of simple lists of species between historic and recent data, as was done in the first part of this document, provide the most insightful information on potential species losses.  In large part due to the greatly increased efforts and documentation associated with recent bee data we know that most species can still be found diminishing our concern that most bees have declined to the point of at least regional extinction.

As a reminder, here are the patterns revealed by the comparison of historic and recent lists of bee species: 1. Most bee species found historically can still currently be found.  2. Bee species that can no longer be found are all rare and their absence from the data is most easily explained by their low probabilities of detection and for some combined with range retraction from the region.  3. Some bumble bee populations have crashed but not all and a few have increased.  4. We clearly captured the increases in introduced bee species.   5. Information about the status of the remaining bee species could be interpretable if additional datasets were inspected for similar patterns from other regions where historic information was collected.

The Way Forward 

More time needs to be spent validating and entering existing specimen data into public databases from regions with extensive historic data. And. This also goes for recent collection data. Data collected but not publically available in many ways does not really exist. And. In the future, targeted collecting to determine current status needs to be done to determine if historic species still exist in residual natural areas. By doing this across the continent and looking for consistent patterns of species loss patterns are revealed.  Each geographic location studied becomes an independent replicate. Compiling patterns of loss/change across regions should allow firmer conclusions regarding the status of bees in a world of often sparse and erratically collected observations/collections of those bees. 

Secondarily, in areas that lack sufficient historic data, the creation of repeatable inventories of bees is needed so that future workers will have baselines for comparisons. A consideration for making some aspects of these new projects repeatable would be to use traps that will allow for more precise measures of changes in common species.

One thing is clear.  You cannot go back in time to collect bee data (perhaps we can do this in the future...just kidding).  If you cannot go back in time this then makes ANY historical information on bees invaluable.  Historic data are a non-renewable resource.  This fact elevates natural history collections out of the dark sea of museums as repositories of curios to one of the most valuable resources on the planet (we knew that already, but the proletariat needs educating). Many existing datasets remain undigitized and many drawers of historic bee specimens sit in museums without identifications, obvious places for improvement. 

Finally, keep in mind that the data and results spoken of here fail to document ongoing losses of bee populations that are hard to imagine.  As natural landscapes are converted to housing, roads, industry, agriculture, and what remains conformed to society's ideals of lawn-like environments with scattered trees and no bugs, bee hunters are forced to look for remnant populations of bees in a declining number of residual natural habitats. A tipping point will be reached at some point and those residual habitats will not be sufficient enough to sustain some bee species. As a society we must more explicitly and conscientiously identify and maintain housing and food in natural and man-dominated environments for bees and other species as our populations increase and land use intensifies.  

The Man Watching

I can tell by the way the trees beat, after

so many dull days, on my worried windowpanes

that a storm is coming,

and I hear the far-off fields say things

I can't bear without a friend,

I can't love without a sister

The storm, the shifter of shapes, drives on

across the woods and across time,

and the world looks as if it had no age:

the landscape like a line in the psalm book,

is seriousness and weight and eternity.

What we choose to fight is so tiny!

What fights us is so great!

If only we would let ourselves be dominated

as things do by some immense storm,

we would become strong too, and not need names.

When we win it's with small things,

and the triumph itself makes us small.

What is extraordinary and eternal

does not want to be bent by us.

I mean the Angel who appeared

to the wrestlers of the Old Testament:

when the wrestler's sinews

grew long like metal strings,

he felt them under his fingers

like chords of deep music.

Whoever was beaten by this Angel

(who often simply declined the fight)

went away proud and strengthened

and great from that harsh hand,

that kneaded him as if to change his shape.

Winning does not tempt that man.

This is how he grows: by being defeated, decisively,

by constantly greater beings.

      -Rainer Maria Rilke