Text from pilot report on long-term bee monitoring with glycol traps
Below is the text associated with the report mentioned in the previous email....the complete report contains quite a number of tables, figures, individual site reports, the original protocol, and responses to the questionnaires sent to each site.
Comments are welcome
DRAFT – 2/27/11
RESULTS OF A PILOT NATIVE BEE MONITORING SURVEY PROGRAM USING ARRAYS OF COLORED CUP TRAPS FILLED WITH PROPYLENE GLYCOL
A JOINT USDA FOREST SERVICE / USGS PROJECT
Sam Droege (sdroege@...) and Jim Guldin (jguldin@...)
This report documents the results of an initial evaluation of the logistical and biological practicability of using an inexpensive array of cup traps as long-term monitoring stations for bees.
Much has been written about the status of native bee populations in North America both in the popular and the scientific press. The primary theme in many of those reports is that native bee populations have declined. However, to date, such declines have been quantified over large geographic areas only for bumblebees (Cameron et al. 2011). The success of that single paper in measuring those declines is not due to the existence of a statistically sensitive set of surveys but due to the rapid collapse to zero or near zero in some populations of bumblebees. This collapse has created a signal so strong that simple comparisons of capture ratios between current bumblebee surveys and past accumulated bumblebee specimens in museums readily revealed changes despite significant statistical issues associated with using museum specimens as measures of past bee population size.
Rather than simply documenting the post mortem of a possibly fatal collapse of populations, an ongoing monitoring program should provide warning of trends and patterns of change in time to reverse those declines. Realistically, such a monitoring program needs to be inexpensive, deployable across almost any region and habitat in the U.S. and provide a statistical signal of change with as low a variance as possible while providing an unbiased reflection of change in bee populations.
For the past 10 years researchers have been testing techniques that could meet the above criteria. An evolution has occurred from traditional netting to the use of pan traps. Pan traps are colored (yellow, blue, white) bowls, cups, or pans that are filled with a liquid with low surface tension. Bees are attracted to the color of the trap and drown in the liquid and afterwards are removed, processed, and identified. Since traps are passive the influence of the skill of the observer (such as in netting bees) is removed and traps can be tended by anyone. The recent use of propylene glycol (not toxic to wildlife) as a long-term slowly evaporating liquid in traps now permits the deployment of such traps for weeks rather than days. Many researchers, research groups, and citizen volunteers contributed to this evolution and has led to the techniques piloted below by a group of USDA Forest Service Experimental Forests and Ranges in 2010. The acknowledgements at the end list a number of people who have made significant contributions.
This report details a piloting of glycol trapping arrays for Native Bee populations, summarizes the results, discusses the experiences and outcomes of those results, and makes recommendations for further use of this technique.
Twelve USDA Forest Service EFRs were initially chosen (Table 1, Figure 1) to act as pilots. Candidate sites were identified through the USDA FS EFR Working Group, a standing committee of EFR scientists that deal with issues of national coordination of EFRs for purposes such as this study. The sites were specifically designed to cover as broad an ecological range as possible and included sites in the states of Minnesota, Maine, Texas, Georgia, Colorado, Idaho, California, Ohio, Utah, Oregon, and the territory of Puerto Rico.
At each site a circular array of 9 traps (Figure 4) was established. The location of each site was up to the Station and locations varied from natural sites, sites associated with weather stations, to sites that were in lawn areas near headquarters. Sites used locally available propylene glycol which ranged in formulation from RV grade, automotive, plumbing, and commercial. Spacing of traps varied among stations and each station developed their own system for tending traps, minimizing trap disturbance, and dealing with local issues (e.g., Figure 4 demonstrates a local solution to protection from downpours). Blue Dawn dish detergent was added to the glycol to cut surface tension and glycol was added periodically to traps to account for evaporation of the water fraction.
Specimens were strained out of these traps approximately every two weeks, placed in whirl paks and shipped to USGS Patuxent Wildlife Research Center for processing and identification. The appendix contains the full protocol and recommendations for changes to processing, traps, and protocols are made in the recommendations section.
Bee specimens were sent to the USGS Native Bee Inventory and Monitoring Lab (BIML) in triple wrapped whirl paks in which the glycol had been drained but the specimens remained damp. At the lab the specimen’s bags were assigned a unique 4 digit number (Table 2), clearly labeled, and placed in the freezer until they could be processed. The 4-digit code was used to track the collection event’s progress through the system and different colors were assigned to collection events based on whether they were in the freezer, washed, pinned, labeled, or identified to genus. All specimens were pinned and labeled and provided with a unique individual 6-digit number. Specimens will ultimately be dispersed back to the BIML synoptic collection, to the Smithsonian, back to the sites, and to specialists. A commitment was made to identify all specimens to genus, but not to species. All Bombus will be identified to species as will select other groups and sites, depending on time and desire.
Table 3 presents totals and averages for each genus across the stations.
Ninety seven separate collections of bees were made by the network of 11 sites (Table 3) with no loss of collection events or sites during the period. These collections resulted in 524 individual genera-by-collection event detections across those runs, and 3587 total bees collected.
Among the sites, SFAEF was the most diligent (and had a long bee activity season) with 14 runs. Interestingly VFEF showed the greatest number of overall detections of genera; an indication of relatively high captures of bees and to some degree the evenness and diversity of bees captured. SDEF, despite the unusual preponderance of honeybees in their captures, ended up with the highest number of genera per run (one factor may be that they only had 5 runs and didn't hit the tail ends of the season where diversity and counts would become lower).
MEF collected the most individuals with very high numbers from the genus Lasioglossum (mostly the Dialictus sub group) and Bombus, but note their rather low ranking in terms of genera per run. This is likely a northern continental effect as continentally the number of genera present in bee communities decreases from the desert SW (the diversity capital for Bees in North America) to the north and east. This pattern may also be evident from the SDEF's high total of genera (19) despite only 5 runs total.
IITF is a standout in its low totals. This is likely due to the generally lower density of bees present on the Caribbean islands. We have been twice now to Guantanamo Bay on Cuba doing bee bowl surveys and hand collecting. Capture rates there have been about as slow as we have ever seen. Discussions with others who have collected on Puerto Rico, Bahamas, and elsewhere in the Caribbean also back this up, as do numbers of species present (90 on Cuba vs 400 plus in Maryland alone). But it is also possible that the IITF’s clever plexiglass rain shade IITF staff deployed may influence the number of bees getting into the traps (Figure 4).
Honeybees, Apis mellifera, are among the most ubiquitous of genera and were found at all sites except GLEES and MEF. SDEF stands out among sites with a very high proportional catch of honeybees; this is likely due to the close proximity of nearby tended hives and perhaps a lack of floral resources. Overall in this study, there are clearly sufficient numbers of honeybees captured across all these sites to consider that this type of system has potential as an independent means of tracking changes in honeybee populations. A network of glycol traps deployed across North America could act as an additional source of status information to the hive counts that are collected by USDA each year. Of possible interest here is that these sites would mix honeybees coming from kept hives with wild colonies and would be a different representation of the status of honeybees than the traditional tracking of hives. Since specimens are collected as part of the survey process, these individuals would be readily available for molecular analyses that could help define commercial, wild, and Africanized populations.
Bombus is a genus in the same family as honeybees and is our only truly colonial native genus except in the extreme Southwest where native colonial stingless bees sometimes venture northward. Bumblebees were captured in all locations except IITF (Bombus does not occur on Caribbean Islands). This genus and the genus Lasioglossum were by far the most commonly caught groups, making up nearly one third of all captures. Species diversity was great within this group with most sites hosting several species. We are in the process of finishing up the identification of all the Bombus caught to species will provide more details separately. Because of recent widespread documented declines, this is another priority group for USDA and other agencies in terms of tracking their conservation and status. A system of glycol traps holds promise for looking at long-term site trends.
The genus Lasioglossum was the only genus found at every location. This genus contains a large number of species (292 names are listed as occurring in the United States) and it was clear from the captures that many species were involved. It is also true that identification to species is difficult within this group, and at this point, we don't have plans to identify to species all the specimens found.
Halictus is another common Halictidae genus that would be expected at all sites given enough sampling time. They are generalist species and remain common in any open site, whether disturbed or not. They are reasonably easy to identify and we may take the time later this year to do so.
The other commonly caught genera (Megachile, Hylaeus, and Melissodes) all are abundant enough in captures that they too would eventually show up at all of the sites. Spring genera such as Andrena, Osmia, Nomada, and Hoplitis are under-represented in this pilot survey but almost certainly that was simply due to a late start at most sites this Spring.
A lot of the potential power in this type of system will be in documenting trends and patterns in individual species numbers. Despite that obvious attraction for recording species level information, it is also true that identifying specimens to species greatly increases the time burden associated with collecting that level of information as compared to quick, genus level counts. The costs and benefits of making this change are outlined in the discussion and recommendations sections.
A matrix of data from individual 2-week collection events with rows as genera and columns as collection events was submitted to Correspondence Analysis (CA) using the PAST software program (http://folk.uio.no/ohammer/past/). CA is an exploratory data analysis technique that places collection events and genera in a mathematically 2-dimensional space based on similarities among sites. The closer any two collection events plot near each other the more similar their community composition as measured by CA and, in this case, measure by genera totals. It is used here in an exploratory way to look at these preliminary results to get an indication of how similar/dissimilar things may be (at the genus level). Note that a more complete dataset and more complete alignment of events with dates run would be warranted for future analyses.
Three CA ordinations were undertaken. Figure 5 ordinates all collection events. Figure 6 illustrates an ordination with sites that were collected after June. In this analysis collection events run prior to June were removed because not all sites started their collections until July. Some of the collection events from SDEF and IITF stood out and were isolated far to the right. Both sites had a mix of genera and numbers of individuals that were clearly different from the other sites. In Figure 7 SDEF and IITF localities were removed as well as pre-July sites.
All three figures illustrate that collection events appear to be clustering primarily by location rather than time of year, an indication of consistency of community patterns and captures, at least at the level of genera. Within each of these figures, but particularly the last one where the most collection events were removed, there is a general patterning of eastern sites clustering to the right of the diagram and western sites to the left, though with plenty of overlap.
Figure 8 shows genera plotted in that same ordination space. Thinking of the x-axis as a West to East ordination of plots, the pattern among genera is consistent with eastern sites being associated with the genera Ceratina, Melitoma, Svastra, Augochloropsis, Xylocopa, and Augochlora and western sites associated with the genera Dufourea, Hoplitis, Osmia, Bombus, Diadasia, and Coelioxys.
SUMMARY OF SITE COSTS
The amount of glycol varied greatly across sites from a low of 5L to a high of 24L. Some of reason for this variation was climatological with evaporation rates being clearly higher at some sites. Additionally some of the loss is surely due to the preparation of the glycol used. RV grade glycol is already diluted and the instructions mailed out with this study had everyone diluting their mix by half, so in some cases the mix in traps had a high proportion of water. Water evaporates at over an order of magnitude higher rate than glycol. If we can demonstrate that catch is not affected, it may be worth using undiluted glycol in low rainfall high evaporation locations. Losses will occur with rain dilution, but if cups are topped off with 100% glycol it may ultimately be cheaper. It may also be worth testing type of rain screens used in Puerto Rico to see if they have an impact on capture rates as this should also greatly decrease the glycol needed, though increasing trap costs. One final reason for high rates on some sites may simply be due to the fact that the questionnaire did not mention whether to report the undiluted or diluted rate of glycol use. Note that no specimens appeared to have deteriorated or rotted in any of the various formulations and rainfall dilutions of glycol. Our guess is that, at times, the percent glycol in some of these traps was <25%.
Most sites did not report animal damage and when it was reported it appeared to be either random or due to rodents. In the recommendations we lay out thoughts for improving trap design that would decrease or eliminate rodent problems. At SDEF the array was in a fenced exclosure which was breached by some animal, possibly a bear (or possibly a vandal, though the forest thought this less likely) and all the traps were emptied, but with minimal tooth marks. Likely the need for fencing will have to be made on a site by site basis. Locations within existing enclosures may be useful trap localities. It should be noted that cattle and glycol trap arrays are incompatible simply due to trampling.
Shipping costs varied by location and by carrier. Typical costs for FedEx shipping was approximately $5.00 per monthly shipment. However, if a padded envelope was used and sent regular mail (we have never lost a shipment this way) costs decreased to $1.50/month.
MEF estimated their total costs were $540.00 per summer and included technician times of 9 hours per month, travel costs to the site and shipping.
Most sites thought the burden of running the traps was low, but several mentioned that compensation for some of the costs of shipping, glycol, or technician time would be appreciated. Again, suggestions regarding minimizing costs are made in the recommendations section.
The BIML lab contributed the cost of the traps, shipping those traps, coordinating the project, processing and identifying specimens, databasing specimens and writing this report.
What does this pilot tell us about the potential of this technique to survey native bees?
Before we can answer that question we have to define a successful monitoring program. Here we define success in three ways: 1) a monitoring program that can provide a statistically defensible and biologically meaningful measure of change in the native bee communities of the United States, 2) a program whose long-term cost is sustainable to the funding agencies, and 3) a sample design for which participation is not burdensome to those who collect and process the data.
Based on the experience of the participating stations we feel that with some modification of the protocols and perhaps a small amount of support from the parent agency, establishing monitoring locations at research stations, refuges, visitor centers, park headquarters, and elsewhere is sustainable and would require relatively minor work from participating groups. Suggested modifications are listed in the recommendations section.
Cost for coordinating a network of monitoring locations could vary from low (small number of sites, identification only to genus) to significant (many sites, identification to species). The recommendations section lists several options.
Biological meaningfulness is partially a product of the needs of the participants, the needs of the funding agencies, and the limitations of the statistical design. Below are listed the biologically meaningful products that will or possibly could be produced by a system of glycol traps
1. A list of the common and some of the uncommon species that occur in the area surrounding the traps. Over time that list will grow as dispersing bees from the surrounding landscape intersect with the traps. It’s unclear how large a sampling region such trap results are representative of and that would be useful future research. This also presumes that species ID is at some point achieved for all the specimens captured.
2. Estimates of change in bee populations. Over time, particularly over long periods of time, it will be possible to obtain statistically valid estimates of change in total number of bees, totals by genus, and totals by some of the common species, if species level data are collected.
3. Specimens. In order to identify bees even to genus, specimens need to be washed and dried. At minimum all specimens will be archived in petri-dishes and available to the site and to researchers for work on taxonomy, morphology, DNA analysis, and as educational materials.
REGIONAL AND NATIONAL LEVELS
1. Estimates of change in bee populations. In the attached manuscript (LeBuhn et al.) now in journal review we document, in detail, using all the long-term bee datasets available, the behavior and abilities of surveys of bees to track changes under different scenarios. While glycol arrays have not been explicitly tested, they are members of the class of surveys called pan traps, where bees are trapped in some sort of after being lured there by the colors of the traps. The statistical section of this document, as well as the attached manuscript (LeBuhn et al.), document the likely statistical limits of this technique. At minimum, trends can be assessed at the level of total number of bees and total numbers by genus with high statistical power and, if species level data are taken, trends in total number of species, common species and species guilds can be assessed.
2. Comparative data on bee communities. Particularly if species level data are collected, it will be possible to associate environmental variables among sites, evaluate similarity among sites, create evaluations of habitat associations for each group; develop trends by ecoregion, by habitat, by guild; identification of regions/ecosystems with greater than average declines and greater than average increases. A good example of the types of data an evaluations that can occur comes from the North American Breeding Bird Survey http://www.mbr-pwrc.usgs.gov/bbs/bbs.html and http://www.pwrc.usgs.gov/library/bibs.cfm.
3. Specimen data. All specimens from this system will be archived and even if not identified to the species level will be available into the future for work similar to what was mentioned under the site level information.
Creation of a system that produces statistically defensible data requires philosophical decisions about what is and is not important to us in terms of the types of statements we would like to make about the meaning of any trend data produced by our system of, in this case, bee sampling sites. This is perhaps difficult for most of us to wrap our minds around as it is not integrated into our primary education and its reliance on probabilities requires us to be more philosophical in our thinking than we are used to.
One way to think about bias in this context is to think about all the possible ways that the numbers of bees collected from glycol trap arrays may NOT reflect the real numbers of bees in the area or more importantly may NOT reflect real trends in populations. Realistically, there is no known way to directly assess the real trends or the real numbers of bees in the area, consequently, we must use indirect evidence to assess the situation. Furthermore traditional bias correcting techniques such as mark-recapture techniques are impossible to implement at a level of reasonable cost and sampling for bees nor would they be likely to be sampling the same populations that are being trapped in glycol traps.
There have been a number of studies looking at comparisons of captures across bee survey techniques (see LeBuhn et al. manuscript for many examples). From these we know that colored traps tend to capture bees in different proportions than either malaise samples or netting samples (the 2 most common alternative techniques). Since there is no ability to reference to a “true” population abundance, the differences among trapping techniques remains difficult to assess and only the most dramatic differences are clearly interpretable as important or significant. From these comparisons it is clear that with enough time or sampling locations, most species will end up in a bowl trap. The genus least likely to do so appears to be the genus Colletes which only uncommonly enter in bowl traps. Other common genera enter bowl traps regularly.
What landscape does a bowl represent? Clearly bees that live immediately adjacent to traps are part of the sampling universe. But how far does that universe extend? What sort of probability function describes a bees probability of ending up in a trap if it is located at various distances from the trap? How many of the trapped species are bees that are dispersing rather than bees that are gathering pollen and nesting locally? How does the probability of a given bee ending up in a given bowl vary with its age, sex, nest location, availability of floral and nectar resources, time of day, weather, relationship with the physical and floral landscape surround the traps? All good questions that have not been directly addressed.
While such biasing factors are likely influences in what does and does not enter into a trap on any particular day, in the long run such factors are likely to even out, becoming a component of the variance of counts rather than something that fundamentally changes trend or paints a very different picture of the common species located at a site leading to highly erroneous estimates of how populations are changing at the site. Research is warranted into these factors, but unless bias is very large and the ability to predict and correct that bias is accurate, it is unlikely that that any resulting correction factors will greatly improve the resulting trend estimates.
The monitoring site itself has a great deal of importance in the interpretation of both the sites’ trends and the meaning of an amalgamation of trends across all these sites. Trends of bees at any particular site are likely best interpreted as reflecting both populations in the immediate vicinity of the site and what is going on in the surrounding landscapes as these sites are not isolated, but exist within the context of the region’s bees. Thus, if trends are desired for a forest, a park, or any landscape, more than 1 monitoring site will be required. The number and distribution of those sites will change based on the types of data required and it is best to sit down with a statistician to discuss such matters. A good place to start is the Monitoring for Managers Web site (http://www.pwrc.usgs.gov/monmanual/) .
For regional and national estimates of change the interpretation of change has to account for where the sites are located. In this sort of network it is unlikely that true randomness or systematic sampling can be achieved. Thus large scale trends have to be interpreted as trends in the collection of sites chosen and those sites may or may not completely reflect what is going on with bee populations. That said, it should be possible to interpret large scale trends as reflecting “protected” areas, agricultural areas or urban areas, particularly if coordinators attempt to evenly cover sampling across the United States and include representative habitats or locations. Options on how to implement a network of sites is presented in the recommendations section.
For an individual site running glycol traps, expenses fall into 3 main categories: labor costs for tending, mailing, and correspondence associated with trap arrays; cost of glycol for the traps; and cost of fencing, if required.
SITE LEVEL PROTOCOLS
Location. Clearly costs will be lowered if location of a monitoring site is near a location such as a headquarters, visitor’s center, weather station or someplace that is visited regularly. Locations need to be in the open. Locating traps in a mown area is probably the most efficient and bees from the surrounding unmown landscape appear to visit such arrays regularly with interference with growing vegetation minimal. A video showing how and where to establish a set of glycol traps is planned.
Trap Design. There seemed to be little problem with the trap holders and they certainly were inexpensive to create. Because traps are expected to periodically be lost, destroyed, and the paint fades with time, it makes sense to have them be relatively inexpensive to replace. A schedule of replacement every 2 months seems reasonable. A more substantial (but still inexpensive) aluminum or heavy plastic cup holder is being sought that will minimize rodent damage and provide more substantial resistance to incidental damage from ungulates and other disturbances. One possibility is to replace the PVC rings used to hold the cups with lengths of PVC pipe that are longer than the cups. The current inexpensive plastic cups would be placed inside these holders.
Number of Traps in an Array. At this point all sites are collecting bees and while more traps will certainly yield more bees and this will increase the catch of rare and uncommon species it is difficult to say whether this increase would yield a more defensible monitoring dataset. Nine traps seem reasonable in that any trap lost during a trapping time period is balanced by eight other traps and dropping the number of traps to six or three would cause any trap’s loss to take on more significance.
Glycol. As mentioned in the discussion section there is likely to be an advantage in either a commercial or centralize set of internet sources of glycol that can be sent out to all sites. It minimizes a site’s investments and decreases potential (but unknown) problems with formulations affecting trap capture. To save costs glycol in traps should be reused unless it appears to be highly polluted and darkened by captures. We would like to suggest that uniform, unaltered, and undiluted glycol be made available to sites from a central distribution center, either within the government or from a bulk supplier. While each site generally was able to find glycol it was of varying dilutions, had additives, colorants, and sometimes took significant staff time to simply locate.
Collection of Specimens. Participants had relatively little to say about the process of straining specimens except to note that is was messy and took some time. A video showing options for straining and preparing specimens for mailing would be useful and is planned.
Labeling. The current labeling system, in general, worked well. However, in a few cases, labels deteriorated in the bags and it became difficult to figure out some sample dates and locations. There was a problem at a few sites where pens were used instead of pencils and the glycol erased the information. We are suggesting that labels be sent to each site each year and that they be made out of cardstock.
Mailing. While it makes sense for a new site to mail specimens after the first couple of collections as a double check of their practices, for ongoing sites there is no particular need for mailing specimens over short intervals. None of the specimens shipped were destroyed or had deteriorated. Our suggestion at this point is to do only 2 mailings, one in the middle of the season so that the main office can get a start on processing and one at the end of the season. Mailings should be done through regular U.S mail to minimize costs and a video is planned that explicitly shows how to drain and wrap bags so they do not leak ….which makes people who handle packages nervous.
In the BIML lab, efficiency can be gained by processing an entire batch of specimens from one location at once. After washing and drying, specimens can either be placed in drying dishes for future processing or be placed on sorting sheets and processed to genus immediately. The tag from the bag and an assigned 4 digit number then stays with the processed specimens.
Processing to the genus level
• It takes approximately 3 hours for 3 interns to wash and dry 12 collection events.
• It takes about 15 minutes for someone very skilled at identifying species to genus to go through a collection event of specimens, id to genus, and enter into a database.
• Specimens would be returned to a petri dish for archiving
• It takes about 3 minutes per collection event to database a site’s information and assign it a 4 digit code
• It takes about an hour per collection event to sort out and pin bee specimens
• It takes about half an hour per collection event to produce, print, cut, and label specimens
• It takes 2 hours per collection event to ID all the specimens to species (with a few left as unknowns)
• It takes about 15 minutes per collection event to enter the data into a database and check it
• It costs about $5.00 for pins and boxes per collection event
Proposed National Glycol Trap Monitoring System
In our minds, a glycol trapping system would provide the most cost effective and biological feasible monitoring system for native bees to address whether the United States is experiencing pollinator decline. Such a system would have the following characteristics
Nine traps of 3 colors
Trap arrays would be located at sites where a sponsoring site can easily tend them as part of other duties (e.g., near their field office, at a weather station, at a visitor’s center)
Fifty sites are recruited from 3 landscape categories
o Protected Natural Areas (including non-plantation forestry operations)
Additional networks can be created for individual agencies or individual sites.
Alaska and Island Ecosystems such as Puerto Rico, Hawaii, and other island territories can be added but it would likely make more biological sense to treat these with separate sampling systems since they are disconnected or distant from the coterminous states.
Similar sites in Canada and Mexico would increase the coverage to all of North America
Sites are chosen to provide systematic coverage with perhaps some (xxxxxx) by physiographic region.
Processing would take place at a set of processing centers
• Each processing center has someone who has been trained and has a verified ability to identify specimens to the genus level.
• All processing centers would archive specimens if species level ID is not made.
• A joint public database would be maintained.
Sam Droege sdroege@...
w 301-497-5840 h 301-390-7759 fax 301-497-5624
USGS Patuxent Wildlife Research Center
BARC-EAST, BLDG 308, RM 124 10300 Balt. Ave., Beltsville, MD 20705
Walking to Work
Today, it's the obsidian
ice on the sidewalk
with its milk white bubbles
popping under my shoes
that pleases me, and upon it
a lump of old snow
with a trail like a comet,
probably falling in love,
all the way to the corner.
- Ted Kooser
P Bees are not optional.