Combining community science and taxonomist expertise for large‐scale monitoring of insect pollinators: Perspective and insights from Abeilles citoyennes

While evidence of insect pollinator declines accumulates, little is known about the pollinator communities that are most vulnerable to population fluctuations and may require conservation actions. Among the main reasons for this lack of knowledge about the status and trends of native pollinators are the time, cost, and expertise required to collect and identify wild insect pollinators (bees, more specifically). Here, we discuss how leveraging the complementarity of community science and taxonomist expertise can help overcome these challenges and provide perspective and insights from launching the large‐scale monitoring program Abeilles citoyennes. The overall objective of this community science project is to monitor wild bee (Apoidea) and hover fly (Syrphidae) diversity in the province of Quebec, Canada, and study the effects of landscape composition on their communities. From 2019 to 2021, 131 volunteers collected insects at 161 sites across the province. A total of 13,558 bees and 2,486 hover flies were collected and identified to species. The project protocol and potential data uses are presented, along with a discussion of the benefits and challenges of using an expert‐assisted community science approach for pollinator monitoring and opportunities for improvement.


| INTRODUCTION
Insect pollinators are essential for the sexual reproduction of cultivated and wild plants, and thus are key to food security and ecosystem stability (Klein et al., 2007;Ollerton et al., 2011).In the past two decades, evidence of widespread insect declines has accumulated (Hallmann et al., 2017;Lister & Garcia, 2018;S anchez-Bayo & Wyckhuys, 2019;Seibold et al., 2019), including for wild bees and other insect pollinators (Biesmeijer et al., 2006;Cameron & Sadd, 2020;LeCroy et al., 2020;Powney et al., 2019;Soroye et al., 2020).Identified drivers of these declines include agricultural intensification, pesticide use, habitat loss, climate change, invasive species, and pathogens (Cameron & Sadd, 2020;Goulson et al., 2015;Potts et al., 2016).However, trends of declines can be quite heterogeneous (Biesmeijer et al., 2006;Powney et al., 2019) and most studies to date have focused on quantifying changes in the status of native pollinators either at local scales or for a limited number of taxa (Cameron & Sadd, 2020;Saunders et al., 2019).As such, historical baseline data on insect pollinator populations at larger scales (that are both biologically and geographically relevant) are rarely available (Woodard et al., 2020).This lack of data limits the identification of pollinator communities that require conservation actions as well as those that need special attention because of the extended services they provide to ecosystem function and crop production (Rondeau et al., 2022;Woodard et al., 2020).
In North America, concern over pollinator declines has given rise to discussions about the best course of actions to monitor the status and trends of native bees, including the development of regional and national monitoring programs (Lebuhn et al., 2013;MacPhail, Gibson, & Colla, 2020;Tepedino & Portman, 2021;Woodard et al., 2020).A key goal when launching a monitoring program is to gather information on temporal and spatial trends on species assemblages in the most cost-effective way, while remaining scientifically rigorous.An efficient monitoring program should be simple, reproducible in space and time, costeffective, and provide sufficient sample sizes to detect declines over a short period if they are occurring (Lebuhn et al., 2013).One way to achieve these goals is by involving volunteers, often non-experts, in data collection through community science (CS; also commonly known as citizen science, but see Eitzel et al. (2017)) projects (Bonney et al., 2009;MacPhail & Colla, 2020).Indeed, CS has gained in popularity and recognition as a valuable tool for monitoring biodiversity (MacPhail & Colla, 2020;Mason & Arathi, 2019).While birds and butterflies dominate as study groups in CS (Basile et al., 2021;Devictor et al., 2010;Dickinson et al., 2010;Flockhart et al., 2019;Government of Canada, 2023;Prudic et al., 2017), a number of CS projects have recently been launched to document the distribution of bumblebees (Beckham & Atkinson, 2017;Lye et al., 2012;MacPhail, Gibson, Hatfield, et al., 2020;Suzuki-Ohno et al., 2017) or the abundance (Ganzevoort & van den Born, 2021) or assemblages of other wild bee species (Le Féon et al., 2016;Mason & Arathi, 2019) from data collected by targeted groups of trained citizens in schools or public and private gardens.In recent years, CS data have been used successfully to track changes in insect populations (Forister et al., 2021;Soroye et al., 2020;Wilson et al., 2021) and represent a promising avenue for assessing population trends in insect pollinators.
Distribution data for specific taxonomic groups (e.g., birds, butterflies) can be collected relatively easily through volunteers.However, for many invertebrates, identification challenges can make collecting taxonomic data at the species level difficult (e.g., through photos of live specimens, as is often the case in CS projects) (Le Féon et al., 2016;Stafford et al., 2010).Bees (Apoidea), for instance, are a very diverse group for which identification to species can be particularly challenging and requires taxonomic expertise (Kremen et al., 2011;Le Féon et al., 2016).While data collection by volunteers alone can be used for bee monitoring at the genus, superfamily, or morphospecies levels (Kremen et al., 2011;Le Féon et al., 2016;Mason & Arathi, 2019), identification to species level is essential to detect the effects of environmental variables on bee richness (Le Féon et al., 2016) or to monitor species at risk.Even for bee taxa that are relatively easier to identify to the species level (i.e., bumblebees), accuracy of species identification by volunteers is typically low and identification of submitted records need to be verified by experts (Falk et al., 2019;MacPhail, Gibson, Hatfield, et al., 2020).Moreover, habitat requirements of bee species vary substantially within genera and morphogroups, limiting the relevance of biodiversity data collected at these levels for informing conservation actions (Mason & Arathi, 2019).As such, combining the benefits of data collection by community scientists with the help of expert taxonomists for the identification of insect pollinator specimens appears to be a sensible approach to reach both efficiency and reliability of collected data (Le Féon et al., 2016).
In Canada, insect pollinator diversity is relatively well documented, but the lack of long-term data halts the assessment of population trends and species conservation status (Sheffield, 2018).One exception are bumblebees (Bombus spp.), several species of which have been shown to be declining in the province of Ontario and western Canada (Colla & Dumesh, 2010;Colla & Packer, 2008;Colla & Ratti, 2010;Macphail et al., 2019).In Quebec, short-term surveys have been conducted in agricultural (Gervais et al., 2017(Gervais et al., , 2018;;Moisan-Deserres et al., 2014;Moisan-DeSerres et al., 2015;Rondeau et al., 2022;Slupik et al., 2022) and urban (MacInnis et al., 2023;Martins et al., 2017;Normandin et al., 2017) areas, but more work is needed to reliably assess trajectories of wild bee communities over time and to forecast their resilience to ongoing and future perturbations.
This article seeks to describe the Abeilles citoyennes (abeillescitoyennes.ca) project, a pollinator monitoring initiative that was launched in 2019 in Quebec (Canada), as a case study leveraging the complementarity of CS and taxonomist expertise for large-scale monitoring of insect pollinators.The project objectives, protocol, and potential data uses are presented, along with a discussion of the benefits and challenges of using an expert-assisted CS approach for pollinator monitoring and opportunities for improvement.Insights and recommendations for each of the challenges described are also discussed.

| Project objectives and study area
The overall objective of Abeilles citoyennes is to combine CS with taxonomist expertise to monitor the abundance and diversity of wild insect pollinators (bees and hover flies) and study the effects of landscape composition on pollinator species assemblages along an agricultural gradient across the southern half of the province of Quebec, in Canada.Our aim is to monitor insect pollinator communities in the long term (>10 years) to assess trends in population fluctuations and community changes.Finally, by using a CS approach, the project also aims at improving knowledge and fostering positive attitudes towards wild insect pollinators among participants, as well as raising public awareness about the importance of pollinating insects.
Quebec is the largest province of Canada, covering 1,542,056 km 2 in area.Because of the agricultural focus of the project, we concentrated our recruitment efforts in rural areas.As such, and to keep sample sizes within our processing and identification capacities, all administrative regions of Quebec (i.e., geographical divisions used to organize the delivery of provincial government services; Figure 2) were targeted, but the Northern ecozones (Taiga shield and Arctic) and the top two major cities of the province, Montreal and Quebec City, were excluded.

| Recruitment
Abeilles citoyennes was first launched at a smaller scale in 2019 (pilot year) and continued at full capacity in 2020-2021.Initially, a small number of participants (n = 21), mainly agricultural producers and acquaintances of the research team, was recruited to test the data collection protocol and online tools.Following feedback from the participants, the material was adjusted before the next data collection season (2020).
Volunteer recruitment for the 2020 and 2021 seasons was achieved via multiple strategies, including online social media (Facebook and Twitter), radio interviews, presentations at conferences, and other communication channels.Two categories of participants were targeted: (1) agricultural producers, notably growers of highly pollinator dependent crops and (2) individual or corporate (e.g., childcare center, community garden) participants.Recruiting agricultural producers had the dual objective of getting access to agricultural sites and increasing awareness among farmers toward insect pollinators.
During each recruitment period (February-April), anyone interested to participate in the project was able to express their interest on the Abeilles citoyennes website (abeillescitoyennes.ca) or by contacting the project coordinator directly.The only condition for participation was to have access to a sampling site (e.g., farm, backyard, large garden, meadow) in one of the areas targeted by the project.The site needed to respect certain criteria, which are presented in the research protocol (Supporting Information).Participants were recruited until the preestablished maximum number of sites (n ≈ 100 sites per year; determined based on expert identification capacity) was reached.To maximize public involvement, we did not limit the number of participants per area (grid sampling) at this early stage of the project.

| Data collection protocol and specimen identification
We focused on bees (Hymenoptera: Anthophila) and hover flies (Diptera: Syrphidae), two taxa known to be very efficient pollinators and reliable ecological indicators (Dunn et al., 2020;Goulson & Nicholls, 2016;Sommaggio, 1999).Pan trapping was chosen as a standardized sampling technique as it is well suited to minimize collector biases (Le Féon et al., 2016;Westphal et al., 2008).Moreover, pan traps are cheap, they can provide reliable data to assess the overall species richness of bees and hover flies at a study site (although they may not be optimal for capturing larger bees such as species of the Bombus and Colletes genera), and their use does not require extensive training (Roulston et al., 2007;Westphal et al., 2008).
At the beginning of each sampling season, each participant received by mail a sampling kit containing the sampling protocol, nine pan traps (12-oz Party City Canada plastic bowls; 3 Â light yellow, 3 Â royal blue, 3 Â white), five pre-identified leak-proof Whirl-Pak ® bags, an aquarium net, ethanol (70% v/v), blue Dawn dish soap, recording sheets, zipper bags, and a pre-addressed and prepaid return envelope.Following a standardized protocol (see Supporting Information for details; original version and video available at abeillescitoyennes.ca/protocoles/), participants were asked to sample insects once a month, from May to September (i.e., five samplings per year), by placing pan traps at their chosen site directly on the ground (among mowed grass or low-growing vegetation), in a single line or cross, at 1-to 3-m intervals each.Pan traps were filled with water containing dish soap and were left in place for 24 h, during sunny days without forecasted rainfall for 24 h.Flexibility in sampling dates was chosen to keep the protocol as simple as possible for the participants.Trapped insects were then collected and stored in ethanol in a freezer until being mailed back to Université Laval (Quebec City, QC, Canada) once all five samplings were completed.After each monthly sampling, participants were asked to fill out a data sheet and record information regarding the date and time of sampling, the spacing and arrangement of pan traps, and any potential disturbances that have occurred at the site in the past 3 weeks (e.g., pesticide application, tilling).
In the laboratory, all insects from the received samples were sorted and all bee and hover fly specimens were prepared for identification (rinsed, air dried, fluffed-for bees only-, pinned, and labeled) and identified to species using taxonomic keys of species known to the study area (e.g., Gibbs (2011) 2019)) by expert taxonomists (including professionals and trained graduate students in Valérie Fournier's lab at the Université Laval).Bycatch specimens have been kept in ethanol in the laboratory but have not been quantified nor identified.

| Web tools
Before the project launch, a website (abeillescitoyennes. ca) was developed to allow participants to enroll in the project and access the research protocol and additional information (e.g., Q&A page, video tutorials).Further improvements were made to the website, which now includes a highly interactive platform with publicly available data on collected specimens and sites, and a password-protected user profile page where participants can enter and modify data on their socioeconomic profile and site location and characteristics.Participants are also invited to share data on ground cover, crops, and other local variables to help characterize the sampling sites and their surroundings, within a radius of 100 m (Table 1), and to upload photos of their sites, which can be viewed and analyzed by the research team.
Importantly, taxonomic data can be uploaded to the web platform by the research team and shared with participants after specimens have been identified.Therefore, participants have private access to the list and abundance of bee and hover fly species collected at their site, directly from their user profile, once data are available (Figure 1a).Data can be filtered by group (bee or hover fly), genus, or species, and can be visualized using pie charts (Figure 1b).In addition, an interactive map displaying the sampling sites and associated lists of species collected is also publicly available on the website (abeillescitoyennes.ca).
In addition to the website, Facebook and Twitter pages were also created and a newsletter was sent regularly to keep in touch with the participants, remind them to continue their monthly sampling and provide information on wild bee biology and conservation and upcoming outreach events.

| Participant demographics and site description
In total, 131 volunteers participated in the project between 2019 and 2021 and collected insects at 161 sites (21 in 2019, 132 in 2020, and 104 in 2021; some participants sampled at many sites) in 16 of the 17 administrative regions of Quebec (all but Côte-Nord; Figure 2).Despite similar recruitment efforts, more volunteers participated in the project in 2020 (n = 103) compared to 2021 (n = 85) (Table 2), which could have resulted from Quebecers having more time on their hands during the 2020 Covid-19 lockdown (S anchez-Clavijo et al., 2021).It should be noted that due to the voluntary nature of the project, site distribution was rather heterogeneous and did not allow a systematic sampling across the territory.Most sites were located in the St. Lawrence Lowlands, the most populated and agriculturally productive ecoregion of the province (Figure 2).
The demographic profile of the participants varied greatly in age, occupation, and geographic location (Figure 3).The vast majority of participants had a university (70%) or college (19%) education, as is often the case in CS projects (Dibner et al., 2018;Goldsmith et al., 2019).Among the participants, 22% were agricultural producers (horticultural or other crop growers) or livestock producers, 67% were individual participants, and 11% were corporate participants, including childcare centers, national parks, and community gardens.Although a majority of participants T A B L E 1 Environmental variables associated with sampling sites shared by participants through an online survey.

Local-scale variables (<25 m)
Substrate type (presence/absence): sand, soft bare ground, hard bare ground Vegetation structure and cover (we asked for photos) Nesting material availability (presence/absence): clusters of hollow stems, cavities, dead wood, branches

Regional-scale variables
Types of surrounding crops (<100 m) Honey bee hive density within a 1 km radius were not agricultural producers themselves, 59% of the sites were located on or adjacent to agricultural lands.Of all the participants who enrolled in the project and received at least one sampling kit, 77.8% (21/27), 72% (103/144), and 73% (85/117) of the participants returned specimens at the end of the 2019, 2020, and 2021 sampling season, respectively, and 70.4% (19/27), 66% (95/144), and 55% (64/117) completed all monthly samplings.Considering the resources and efforts associated with preparing and mailing sampling kits to participants, it is also interesting to look at these data in terms of sampling kit utilization.From all the sampling kits that were sent in 2020 and 2021 (n = 312), 242 (78%) were effectively used by participants who returned at least one insect sample at the end of the sampling season, 172 (55%) of which were used to complete all monthly samplings.Participant retention between 2020 and 2021 was 59% (61/103), corresponding to 61% (80/132) of the sites, and was higher for agricultural producers (77%; 21/27) than for individual and corporate participants (53%, 40/76).Of the 21 participants of the 2019 (pilot year) cohort, 62% (13/21) remained involved until 2021.

| Communities of bees and hover flies
Data collection by community scientists between 2019 and 2021 yielded 16,044 specimens, including 13,558 bees and 2,486 hover flies.These specimens represented ≥252 species from 33 genera of bees and 86 species from 39 genera of hover flies.To date, approximately 1,000 bee specimens from the Andrena, Nomada, and Sphecodes genera have yet to be identified to species due to specimens from these genera being more difficult to identify, meaning that the total number of bee species collected is likely higher than reported here.On average, three genera (0-13) and twelve species (0-36) of bees and/or hover flies have been collected per site in 2020 and 2021.Although presenting a detailed characterization of the insect communities sampled is beyond the scope of this article, the complete list of species collected by site is publicly available at participant.abeillescitoyennes.ca/fr/map.Data will also be deposited on GBIF (gbif.org/)once all specimens have been identified.

| Outreach
Public outreach being an important aspect of the Abeilles citoyennes program, multiple outreach events were organized for both the project participants and the general public throughout the project.Among others, webinars were offered to participants annually, encompassing  In 2019 (pilot year), pan traps were deployed for 48 h instead of 24 h in 2020 and 2021 and three samplings were requested (July-September) instead of five (May-September).
themes such as the importance of pollinating insects, the bee fauna of Quebec, and actions to help pollinators thrive.Preliminary results were also presented during these webinars, as well as a tutorial on the use of the web platform to access data shared with participants on their user page.In addition, field workshops (n = 8 from 2019 to 2022) were offered during the summer seasons, during which participants engaged in insect pollinator sampling and identification, using a simple identification key (to the family or genus level) and a stereo microscope.These workshops took place in different places and regions, including participating farms, national parks, and educational institutions.

| Abeilles citoyennes as a monitoring program
The Abeilles citoyennes project has shown that combining the help of CS for data collection and expert taxonomists for specimen identification can be a practical approach for monitoring insect pollinators across large areas, and one that we advocate for.An impressive number of specimens (>16,000) have been collected by community scientists during the project, permitting the identification of ≥252 species of bees (out of a total of 298 species recorded for Quebec; The Royal Saskatchewan Museum ( 2023)) and 86 species of hover flies (out of 413 species known in Eastern North America; Skevington et al. ( 2019)).In comparison, a recent review of the bee fauna associated with agriculture in North America reported a total number of 185 bee species for the Province of Quebec from 11 monitoring studies conducted between 1967 and 2021 (Rondeau et al., 2022) while 200 species were reported for the cities of Montreal and Quebec (Normandin et al., 2017).For hover flies, previous studies have reported a total number of 28 species in lowbush blueberry (Moisan-DeSerres et al., 2015), 33 species in cranberry crops (Gervais et al., 2018), and 48 species in urban green spaces (McCune et al., 2023).This shows the efficiency of CS projects to provide a snapshot of bee diversity across a large territory in a short period of time, which would be very difficult, if not impossible, to achieve using traditional monitoring methods.The pilot study conducted in 2019 proved to be essential to help ensure training materials were understood and that data were collected efficiently and accurately.Following the pilot sampling season, feedback from the participants was collected using an online questionnaire and informal one-on-one interviews.Changes made to address participants' feedback after the pilot year improved the Abeilles citoyennes protocol and tools tremendously.Among others, we switched from using a mobile datagathering platform to using a simple paper form to record data, in response to differences in digital literacy among participants and to improve inclusiveness.Changes were made to the protocol (e.g., reduced deployment time of pan traps from 48 h in 2019 to 24 h in 2020-2021 due to feedback about the difficulty of finding a 48-h sampling window without forecasted rainfall) and additional information and photos were also added to make the protocol more intelligible.Prototyping CS projects is a critical step to ensure long-term success and meet participant needs and expectations (Bonney et al., 2009).Unfortunately, it is often omitted by CS organizers and may impact project success and long-term participant retention.
Although pan traps have been suggested as a method of choice for insect sampling in CS projects (Le Féon et al., 2016), relying on passive lethal sampling to monitor insect pollinators also has disadvantages and should be considered carefully to minimize unnecessary lethal captures (Montero-Castaño et al., 2022).Indeed, although our sampling protocol (nine pan traps, one sampling per month Â five months) is unlikely to significantly impact bee communities (Gezon et al., 2015), the effect of such a sampling effort is presently unclear.As such, it may be prudent to assume that long periods of lethal sampling may have negative effects on populations, especially on uncommon species (Tepedino & Portman, 2021).Running CS data collection at intervals of once every few years (i.e., intermittent surveillance) or leaving gap years between periods of sampling may be a way to alleviate the possible impact of intense lethal sampling on populations (Tepedino & Portman, 2021).However, doing so may as well decrease participant retention if the program is not running every year and providing alternative ways for volunteers to engage during gap years may be necessary to retain participants over years.Lethal sampling can also be a source of concern for certain participants, who volunteer to 'help' insect pollinators and yet are asked to kill some.Even though we explained the rationale and the low incidence of pan-trap sampling on pollinator populations, a small number of potential volunteers decided not to participate after reading the protocol and what was expected of them.
While we advise that future monitoring programs include adequate numbers of sites across the studied area (and among regions, if applicable) to provide enough statistical power (Lebuhn et al., 2013) without surpassing identification capacity, limiting participation to a predefined number of sites may bring perceived issues in terms of inclusiveness, which should be considered carefully when designing new projects (Paleco et al., 2021).Perhaps, a way to work around this challenge is to provide alternative means of participation that are accessible to all without creating identification bottlenecks in the laboratory.One example would be encouraging participants to share observations of insect pollinators on established reporting platforms based on photographic collections (e.g., iNaturalist [inaturalist.ca],Bumble Bee Watch [bumblebeewatch.org],Bee Watch [abdn.ac.uk/ research/beewatch]) and provide protocols on how to take quality photos for identification purposes.Indeed, except for specific groups of bees like bumblebees, identification of most bee species cannot easily be achieved from photographs (especially those excluding lower abdomen details), but providing guidance regarding specific anatomical features to include on photographs could be useful to maximize success (Stafford et al., 2010).

| Benefits and challenges of using CS for monitoring insect pollinators
There are many benefits of using CS for monitoring insect pollinator communities, both for the participants and for those running monitoring programs (MacPhail & Colla, 2020; Figure 4).Among others, data collection by volunteers has increased the spatial scope of the present study way beyond what we, as researchers, could have accomplished alone, permitting monitoring across an area of roughly 150,000 km 2 (Figure 2).We estimate that the Abeilles citoyennes participants contributed 2,056 h (8 h per sampled site per year, including familiarizing themselves with the protocol, sampling, and data entry) of in-kind donation of time from 2019 to 2021, valued at CA $51,421 (CA $25.01/h; ca.talent.com/en/salary).Considering total project expenses of CA $160,000 (excluding in-kind volunteer labor amount) over 3 years, the participants' in-kind contribution represents 32.1% of the total project costs.CS can also bring enormous value to participants by increasing awareness and scientific literacy, as well as opportunities for participants to interact with nature (MacPhail & Colla, 2020;Mason & Arathi, 2019).In addition to engaging in data collection, webinars and field activities were organized for the Abeilles citoyennes participants from 2019 to 2022, providing opportunities for social interactions between participants and scientific staff.Considering that social interaction is often reported as one of the main reasons for volunteering in environmental projects (Larson et al., 2020;Rondeau et al., 2020;Ryan et al., 2001), these social opportunities may also benefit participant recruitment and retention over time.Importantly, we can attest that, from participating in the project, many volunteers have in turn become important players in raising awareness.In our opinion, and from our experience with Abeilles citoyennes, these benefits outweigh the cost (time and effort invested) of developing and managing the CS project.
Recruiting agricultural producer participants has been a major challenge and from those who first enrolled in the project, many did not follow up or never returned samples.The most common reason provided for dropping out was work and mental overload during the production season, which overlaps with the Abeilles citoyennes sampling period.Working with agricultural advisory clubs may help increase the number of collection sites on farmlands.Indeed, up to 10 agricultural sites in the Eastern Townships region of Quebec were sampled through the help of an agri-environmental advisory club, which mandated the monthly sampling to one of their employees.Offering a program where available volunteers would be matched with agricultural producers to collect data on their farm could be another way to increase participation among producers.Recruiting individual and corporate participants was easier, the general public being sympathetic to bee issues.However, the timing of data collection was also a barrier to the participation of other groups.For instance, school teachers might have been interested in contributing to the project, but the school calendar is not compatible with the active life period of adult bees and hover flies in eastern Canada.One possibility to involve schools in monitoring programs would be to offer one-off inventory projects to interested teachers during the spring (May, June) and fall (September) months.
Participant retention is extremely important for projects aiming at collecting long-term data.Creating valuable tools for participants, like the ability to view the list of species associated with their own site (and those of the other participants) via an online portal (Figure 1) may help to increase participant retention.However, even when such a tool is available, processing and identifying specimens can take months, if not years.As any delays in identification can have a negative effect on participant motivation, expected timelines to make results available should be disclosed to participants ahead of time.Over the course of the project, emails were sent regularly to update the participants of our progress in identifying specimens (and developing web tools to share data) and the expected timeline to access species list.
The accuracy of data collected by volunteers in CS programs is often a concern among scientists (MacPhail & Colla, 2020).Although specimens sampled by the Abeilles citoyennes participants were not identified by the volunteers themselves, it is possible that the reliability of other data, such as information regarding site characteristics (e.g., % of flowering plants, trees, or bare ground; Table 1), which needed to be estimated by the participants as percentages of land cover, was very variable among participants.Differences in sampling efforts may also have introduced biases.Indeed, while participants were asked to leave traps out for 24 h once a month and to keep the same sampling location across years, these parameters somewhat varied in practice due to logistical constraints or oversights, or because enthusiastic participants wanted to contribute more data (also reported in Le Féon et al. ( 2016)).Fortunately, unequal sampling efforts, when known, can be accounted for in data analysis by using appropriate statistical tools (Johnston et al., 2023).

| Benefits and challenges of using taxonomist expertise as part of a monitoring program
The main benefit of having specimens identified by professional taxonomists is data precision and accuracy.Indeed, identification to species can provide important data on the distribution, relative abundance, and longterm changes in insect pollinator communities or populations of individual species, but identifying bees to species is time-consuming and can only be performed by a limited number of highly trained specialists (Lebuhn et al., 2013;Tepedino & Portman, 2021).There can be considerable repercussions for species misidentifications in CS monitoring programs (e.g., incorrect assessments of population status and distribution, based on CS data; MacPhail and Colla (2020)), but using the assistance of experts can help to significantly reduce the risk of identification errors (Le Féon et al., 2016).However, passive sampling of native bees often results in a number of captured specimens that largely surpasses identification capacity (Tepedino & Portman, 2021).This is a major challenge to which Abeilles citoyennes has not escaped from.To avoid such an identification bottleneck, reconsidering the need to identify each collected specimen to species may be needed.One alternative would be to identify all or some groups of specimens to genus, which can usually be done quickly.Archived specimens that were first identified to genus could be identified to species at a later stage, if needed.
While CS can provide cost-effective ways to collect data, relying on expert taxonomists for specimen identification means that some of the main expenses cannot be reduced.Indeed, Lebuhn et al. (2013) estimated that about 85% of the cost of monitoring bees is associated with the handling, processing, shipping, storage, and identification of collected specimens.Here again, identification to genus could help reduce these costs but this would result in a significant loss of information, although changes in abundance could still be detected.The shortage of bee taxonomic experts in many parts of the world (Tepedino & Portman, 2021;Woodard et al., 2020) is another major challenge that any monitoring program should be prepared to face.In the case of Abeilles citoyennes, the fact that some bee specimens from difficultto-identify groups still need to be identified to species testifies to this challenge.

| Recommendations and further work
Considering the size of the Quebec territory (i.e., nearly three times the size of France or Texas), the inevitable variations in pollinator insect assemblages between years, and the need to adequately quantify the current collapse of pollinators in temperate regions, monitoring the diversity of wild bees in the province must continue.While Abeilles citoyennes is planning on carrying on with its activities for years to come, improvements could be made to optimize data collection and participants' satisfaction and better value participants' contributions.These include: • Recruiting additional agri-environmental advisory clubs to increase the number of collection sites on farmlands, promote knowledge transfer, and increase awareness among agricultural stakeholders.• Reconsidering the need for identification at the species level for all genera, every year.• Assessing the interest of "super volunteers" (MacPhail & Colla, 2020) and offering additional activities for them to engage with the project (e.g., sampling by visual observations or netting, providing protocols on how to take quality photos for identification purposes, matching available volunteers with producers who cannot commit the time to collect data but are willing to have monitoring performed on their farmlands).• Considering adapted sampling schedules for schools (May-June and September).• Developing additional targeted research questions to which the Abeilles citoyennes program may help answer, in addition to its broader surveillance scope.
In addition, it has often been suggested that involving participants at different stages of CS projects (in addition to data collection) may bring advantages to both the participants and the scientific team (Kennett et al., 2015;Pocock et al., 2018).For instance, CS programs could involve participants in the identification of objectives and methods to achieve them or in knowledge mobilization.In the case of Abeilles citoyennes and future similar CS projects, an interesting approach would be to recruit a lead volunteer in each region who would receive training at the start of the season, help with recruitment, and offer active support to participants in their region.

| Conservation implications
Insect biodiversity is collapsing at a rate and scale that can only be matched by large scale monitoring studies (Homburg et al., 2019;Raven & Wagner, 2021;S anchez-Bayo & Wyckhuys, 2019;Seibold et al., 2019), which few organizations have the capacity to execute.CS is one of, if not, the best tool to monitor terrestrial insect populations and assess long-term trends because it allows collection of biodiversity data at a very large scale (even at the global scale) in near real-time while mobilizing hundreds or even thousands of people to biodiversity issues.By providing a discussion of the strengths and limitations of the Abeilles citoyennes approach and by suggesting areas of improvement (Figure 4), this study can help inform the development of future large-scale monitoring programs for insect pollinators in different geographical locations.In turn, such programs could provide relevant results to guide conservation measures at various policy-relevant scales (e.g., provincial or national).As highlighted by Pocock et al. (2018), a CS program such as Abeilles citoyennes also contributes to educate and stimulate local 'champions' or ambassadors who may advocate for the conservation of wild pollinators in their community and promote good management practices.

AUTHOR CONTRIBUTIONS
Sabrina Rondeau and Amélie Gervais conceived the original project idea.Sabrina Rondeau, Amélie Gervais, Anne Leboeuf, and Valérie Fournier developed the research protocols and contributed to the design and implementation of the research.Anne Leboeuf coordinated the Abeilles citoyennes project.André-Philippe Drapeau Picard and Maxim Larrivée assisted with web development and provided advice on sampling protocols.Sabrina Rondeau, Amélie Gervais, Anne Leboeuf, and André-Philippe Drapeau Picard co-wrote the first draft of the manuscript and all authors provided feedback on subsequent drafts.Sabrina Rondeau, Amélie Gevais, and Valérie Fournier secured research funding.

ACKNOWLEDGMENTS
This project was funded by the Innov'Action Program of the Ministry of Agriculture, Fisheries, and Food of Quebec (Grant: IA119082) and received additional financial support from the Montréal Insectarium.We would like to thank all the project volunteers for their contribution to data collection, Xinbao Zhang for developing the web platform, Véronique Martel for advice during project development, Frédéric McCune for his help with hover fly identification, and the following students who helped preparing specimens for identification: Guillaume Blais, Andréa Duclos, Steven L'Heureux-Lepage, Myriam Moreault, and Saïda Rojas-Charrette.

F
I G U R E 1 Example of information shared with the participants on their user profile page: (a) list of insect species collected by the participant.Each species is clickable to display information on collection date and the sex of the specimen; (b) Pie charts showing the distribution of species (left) and specimen abundance (right) per genus.

F
I G U R E 4 Summary of the strengths and limitations of the Abeilles citoyennes project, as well as the opportunities for improvement and associated threats discussed in the main text.