Towards a real-time tracking of an expanding alien bee species in Southeast Europe through citizen science and floral host monitoring

Citizen science, a practice of public participation in scientific projects, is popular in Western countries, however, it is still a relatively novel approach in Southeast Europe. In this region, citizen science can be a useful tool for increasing the understanding of alien species. One such species is the sculptured resin bee, Megachile sculpturalis, a putatively invasive alien pollinator native to East Asia. It was introduced to France in 2008, from where it quickly spread across West and Central Europe. However, our knowledge of its eastern distribution is scarce since it is based mostly on isolated findings. We combined citizen science and data extraction from online sources (e.g., naturalist’s databases and social media) covering 6 years, and 3 years of targeted floral resource monitoring in the search for M. sculpturalis across regions of southeastern Europe. We collected presence data and information on M. sculpturalis abundances across an urban-rural gradient from eight countries: Hungary, Slovenia, Croatia, Serbia, Bosnia and Herzegovina, Montenegro, Romania, Bulgaria, and the region of the Crimean Peninsula. We present the first country records for Romania, Bulgaria, and Montenegro, identify the dynamic expansion front in southern Serbia and provide new southernmost occurrences in Southeast Europe. We also collected data on species ecology (e.g., phenology, pollen/nectar sources, nest characteristics) and gathered evidence of reproducing populations of this species across the studied region. Citizen science data provided a five times larger spatial coverage, including recordings from remote locations, than the data collected by expert field surveys and provided critical additional data about the species biology, thanks to exceptionally engaged participants. We emphasize the importance of close collaboration between regional scientist teams and citizen participants and the benefits of this approach for monitoring a species with a continent-wide spread potential.


Introduction
Invasive alien species are recognized as one of the main drivers of global biodiversity decline [1][2][3][4], with numbers on the rise over the last 60 years [5]. Detections of early phases of incipient invasions are rare [6], likely due to initially low population densities and limited geographical extent [7][8][9], but are crucial for understanding invasion processes and successful management [10,11]. In most cases, newly introduced species are only detected once the populations have become abundant; thus, reconstructing the invasion history becomes difficult [9]. Moreover, the feasibility of management depends on the particular stage of the invasion process [12], with eradication programs being more successful in the early stages [13,14]. Consequently, there is a We predicted that citizen science data would reveal a much broader distribution than the floral host monitoring by a team of researchers.

Study area
The study region encompassed the eastern and southeastern expansion fronts of M. sculpturalis in Europe. This included most of the countries from the Balkan Peninsula (Croatia, Bosnia and Herzegovina, Serbia, Montenegro, Albania, North Macedonia, Romania, Bulgaria, and Greece) and two adjacent Central European countries (Slovenia and Hungary). We referred to the study area as Southeast Europe for the simplicity of reporting while acknowledging the inconsistency in this term (topographical versus political versus biogeographical). We also report data from the Crimean Peninsula to cover the entire eastern extent of the M. sculpturalis distribution in Europe (supplementary figure S1 (available online at stacks.iop.org/ERC/4/085001/ mmedia)).

Data collection
We collected data using two main approaches: 1) systematic/targeted fieldwork by researchers, and 2) opportunistic, mostly citizen science data (figure 1). Systematic/targeted fieldwork followed a protocol designed specifically for M. sculpturalis in a previous project [42]. Opportunistic data was collected within citizen science projects and did not follow standardized field protocols [73,74]. Additionally, this data set included searching online data sources for M. sculpturalis records.

Opportunistic recording
Citizen science data was gathered via contributory citizen science approaches (after Miller-Rushing et al [75] and Shirk et al [76]), within two European citizen science projects established for this purpose (Central Europe: http://www.beeradar.info since 2018; Serbia: https://srbee.bio.bg.ac.rs/azijska-pcela-smolarica/azijskapcela-projekat-ucesce since 2020). Two research teams published calls for reports on their homepages in late spring and summer before and during the known activity period of the species for three consecutive years (2019-2021). Calls were advertised in print, online web pages, on local and national TV and radio stations (data collection 'CSP interaction (via conventional media)', and on social media (data collection 'CSP interaction (via social media)', e.g., Facebook, Instagram, Twitter, various blog posts; figure 2; [77,78]). Report calls were written in the language of the countries where they were published, but with the most important information available in English to enable participation beyond the initially targeted countries (e.g., Slovenia, Greece, Hungary).
Participants reported observations of the target species via social media, contact forms on project homepages, email, or mobile phones. Reports included a mandatory photo or video of the specimen, observation date, and location (coordinates, ZIP code). We validated all photos by assigning species identifications to observations and provided feedback directly to the participants (figure 1). In addition to species record data, which was our primary focus [79], we asked participants to specify their observations to obtain additional information on biotic interactions and ecological preferences termed 'secondary data' (figure 1), following Callaghan et al [80] and Wang Wei et al [81]. We used submitted images to determine host plants, bee nesting activity, the number and sex of the individuals, and direct or indirect competing behavior. We also asked participants where they had heard about the project to improve our outreach ability in the future.
Web data extraction: We manually searched four public naturalists' platforms (data collection 'retrieval from naturalists' databases', see supplementary table S1c) for records of the sculptured resin bee, as well as various local social media groups (data collection 'extraction from social media') using keyword searches 'sculpturalis', 'megachile sculpturalis', 'giant resin bee', 'sculptured resin bee', and available common names for this species in local languages, 'azijska pčela smolarica' (used so far in Serbia, Bosnia & Herzegovina and Croatia) and 'óriás művészméh' (in Hungary); for a complete list of searched Facebook groups see supplementary table S1c) [74,77]. We also searched for internet-based publications on web pages and blog posts for records (data collection 'web data extraction -miscellaneous').

Systematic/targeted research
Data was gathered by repeated (3-10) observations of blooming essential floral resources (supplementary table S1d) during the periods of M. sculpturalis activity (July and August), and under optimal weather conditions (sunny to partly cloudy days, with no rain or strong wind). The localities of the plants of interest were either recorded during previous fieldwork or by additional targeted fieldwork searches, often owing to information from citizen scientists. Floral monitoring was mostly conducted by experienced researchers (authors), except in one location in Greece, where citizen scientists were also engaged. If the study species was found, we recorded the number of individuals (timed counts in combination with snapshot counts -methodology described in [43]), collected specimens if possible, and documented the abundance and blooming status of plants following the protocol described by Bila Dubaić et al [43]. This protocol also allowed us to gather data on 'not detected' records for study sites (figure 3): when necessary conditions were met (optimal weather conditions and the sufficient abundance of blooming key floral resources), but no records of M. sculpturalis were obtained [79,[82][83][84][85].

Data exploration
Reported sites less than 250 m apart were merged into a single location. To characterize the recording locations for the urban-rural-natural gradient, we used the degree of urbanization approach based on the human population distribution grid 'JRC-GEOSTAT 2018' [86] and two additional grids from DG AGRI and DG REGIO [87][88][89]. Each location is assigned to one of the three main categories based on 1×1 km resolution grid cells and their clustering pattern [89]: (a) 'high-density clusters' (contiguous grid cells with 1,500 inhabitants each and a total population of 50,000, corresponding to urban centers), (b) 'urban clusters' (contiguous grid cells with 300 inhabitants each, and a total population of 5,000), or (c) 'rural grid cells' (cells outside urban clusters: cells with less than 300 inhabitants each, or cells with higher density but not contiguous with urban clusters). This analytical scale is suitable for capturing both the variability of effective human presence (and respective activity) and the foraging activity pattern of a large, highly mobile bee species. To enable a more meaningful resolution among various landscape types within the broad category 'rural grid cells', we divided it into convenient sub-categories: 'rural/moderate density' (cells having 25 inhabitants each), and 'rural / low density' (cells having <25 inhabitants each). For a straightforward reference to the simplified landscape ecology context, hereafter, we use the terms: urban/high density, urban/intermediate density, rural/moderate density, and rural/low density landscapes.
To process georeferenced data and produce maps we used ArcGIS Pro 2.8.3 [90]. We used the Minimum Bounding Geometry tool in ArcGIS to calculate minimum convex polygons; and in RStudio [91] alpha hulls with relaxed assumptions of connectivity of data points allowing for disconnected hulls and convex angles of the hulls using the packages 'alphahull' [92], 'sf' and 'sp' [93] with alpha=2 according to IUCN recommendations [94] to compare the hull sizes, separately for two main data sources -opportunistic recordings and systematic/ targeted research ( figure 3; supplementary figure 4). However, we excluded isolated records from Bulgaria and Romania from the hulls since they might represent long-distance dispersal events (cf [43]) similar to those found on the Crimean Peninsula, i.e., not yet a continual distribution.

Results
In total, we compiled 270 records of the sculptured resin bee from 178 locations in seven countries of the Southeast European region and the Crimean Peninsula (figure 3, table 1; supplementary table S1a). Opportunistic recording provided 133 records from 114 locations, two of which are the first country records: Romania (2019) and Bulgaria (2021) (records #074 and #008; see supplementary table S1a for respective data sources). Our systematic/targeted field research resulted in 137 records from 64 locations, including the first country record for Montenegro in 2021 (published herewith). We also tentatively documented the species' absence from Greece and parts of Montenegro based on a targeted field search at three localities (figure 3; supplementary table S1b). The study area of documented bee presence based on opportunistic recordings was estimated as five times bigger than the area of our systematic/targeted research coverage (minimum convex

Opportunistic data recording
In total, 71 records of Megachile sculpturalis (26% of total records) originated from our citizen science projects: 46 from interaction via social media, seven from interaction via conventional media, and 18 from miscellaneous types of interactions (documented and reported by colleagues, by self-initiated internet research, or by family members of the authors). Intensive searches on social media platforms yielded 38 records, including the first record for Romania (data collection 'extraction from social media'), while 'retrieval from naturalists' databases' delivered 17 records. Further, seven records were from miscellaneous source types. Opportunistic observations displayed 49 bee-plant interactions, including 17 plant taxa (supplementary table S1d).
Data sources were unevenly distributed across the urban-rural gradient, as citizen science provided mostly data from urbanized areas (figure 4). The greatest share of recording locations are in the category of urban/high density − 96 (55%); 38 are in urban/intermediate density (22%); 44 are in rural/moderate density (16%), and 16 in rural/low density landscape types (9%) (supplementary figure 3). The altitudes of the recorded locations ranged from 0 to 1,060 m (supplementary table S1a). There was one exceedingly late record in Hungary, a live female observed on 1st October.
Out of all reported records, 41% were verified as our target species, and the remaining observations were misidentifications, most commonly of Bombus pascuorum (23%) and Apis mellifera (20%). Few other insects were also reported (1-6 times): other Anthophila (Amegilla sp., Andrena sp., Bombus sp., Halictus sp., Lithurgus    diameter. Four specimens emerged from the incomplete nest ( figure 5(a): the rounded cage on the left side) during the same period, with two females, one male, and one specimen of unknown sex (supplementary table S2).

Systematic/targeted data recording
The floral resource monitoring approach conducted in 2019, 2020, and 2021 resulted in 136 presence records from 63 locations: 54 in Serbia (2019-2021), five in Bosnia and Herzegovina (2020), and four in Montenegro (2021) representing the first record of M. sculpturalis for this country ( figure 3). In Montenegro, observed population densities ranged from five individuals to aggregations containing more than 30 bees. In total, 129 bee-plant interactions were recorded at ten plant taxa (supplementary table S1d).
During seven independent 15-minute observations (five by an engaged citizen scientist, two by researchers) of two floral resources (Styphnolobium japonicum sp and Vitex agnus-castus) in Greece across five locations (supplementary table 1b), no individuals of M. sculpturalis were spotted.
Recordings gathered by systematic/targeted research are mostly located in highly urbanized areas -urban/ high density (51%), and urban/intermediate density (30%). Rural locations are much less represented in this data set (11% for both sub-categories) ( figure 4); most of these rural locations were first reported by citizen scientists, and then re-visited by experts.

Discussion
We report the range expansion of M. sculpturalis, across countries of Southeast Europe. Citizen science data provided about five times larger geographical coverage than fieldwork by researchers, as well as a longer time period of records (dating back to 2017). Before extended focused research was initiated after 2019 (cf [42,43]), only scattered evidence was formally published in the scientific literature [40,41,44,49] or made available via well-known international platforms [95,96] for this region. Web data extraction from social media and local naturalists' sources provided essential information for the reconstruction of the early colonization history in the region. In contrast, repeated visitation of the same localities and use of standardized sampling provided data about population dynamics, especially useful in the case of (potentially) invasive species. To obtain observations of the target species from least populated areas (rural/moderate density and rural/low density), it is advised to use different methods. We show that different collecting approaches complement each other and provide comprehensive information on the distribution and ecology of the target species.

Range expansion in southeast / east europe
Records presented in this study confirm the successful spread of Megachile sculpturalis in Southeast European countries (figure 3; supplementary figures S1-S2). Previously published findings for most countries were generally sporadic [40][41][42][43]49], except for Serbia and Hungary [42,43]. We compiled numerous further findings, particularly from 2021, with the most remarkable growth of recording incidence in Croatia. Our data on reproducing populations over several years confirm that M. sculpturalis is now naturalized and widespread in the region, presumably fully established several seasons ago. The two southeasternmost records from Serbia (Niš and Vlasotince) and one easternmost (Negotin) are interesting as they represent the 'apparent' range expansion of about 140-150 km in the period 2020-2021 (supplementary figure S6). Remarkably, the intensive field search during the season of 2020 in Niš [43] yielded no records, indicating a recent colonization of this area. However, even the intensive and detailed field search for such a large and distinctive bee may not result in successful detection during the early phases of local colonization due to initially low population abundance [43]. Accordingly, we may regard the area of southeastern to eastern Serbia as a tentative but very widely defined southeastward expansion front. This does not imply that the bee could spread effectively and continually with such a speed (150 km/year). Rather, our focused and tedious floral resource monitoring, combined with citizen engagement, may closely approach the ideal of real-time expansion tracking by reducing the usual time lag between the spreading dynamics and detection within/along well-defined fronts and trajectories.
We documented the presence of M. sculpturalis in several coastal locations in Montenegro, representing the new country record and the southernmost occurrence in Southeast Europe known so far. The Montenegro population also represents the most southeastern location on the Adriatic coast, most likely resulting from the continuous eastward spread along the coastline of Croatia (Supplementary figure S2).
In contrast with the surrounding countries (Croatia and Serbia), where we confirmed the widespread presence of M. sculpturalis, there are still very few findings in Bosnia and Herzegovina, as well as most of Montenegro and SW-Serbia. The Dinaric Mountains range stretching across this area probably represents a dispersal barrier for coastal populations. On the other hand, Bosnia's mostly lowland peri-Pannonian zone was presumably already colonized from the north-more widely than currently documented [42].
Our data (supplementary table S1) indicates the existence of two main expansion trajectories in Southeast Europe: from the E-Pannonian core (2015) southward-southwestward and from N-Italy (2014-2015) through Slovenia (2016-2018) eastward-southeastward, along the Adriatic coast of the western Balkans (Supplementary figure S2). The spreading trajectory along the Adriatic coast was probably not yet merged with the Pannoniancentral-Balkans expansion axis through Serbia. However, the two sections of the eastern range were most likely merged within Croatia and/or Slovenia before 2021. We have no indication if it has reached the southern Balkans (North Macedonia, Albania, and Greece), but this is expected in the near future. A similar spatial pattern of spreading and detection within Southeast Europe was documented for a few introduced species of apoid Hymenoptera [27,97,98]. The ongoing studies on the genetic structure of the Balkan populations should provide a clearer image of colonization routes and dynamics in this region.
The first records of M. sculpturalis for Romania (from 2019) and Bulgaria (from 2021) are both remote from other Southeast European records. The Romanian record is positioned about 450 km linear distance from the closest contemporary record in Serbia (reduced to 290 km in 2021) and 630 km from the records in the Crimean Peninsula. The Bulgarian record is positioned 300 km away from the closest records in southern Serbia, 740 km from the Crimean Peninsula, and 230 km to the Romanian record. These two records in the eastern Balkans might indicate further cases of long-distance dispersal events, similar to the ones in Hungary [28,40] and the Crimean Peninsula ( figure 3; [49]). The lack of further records probably indicates insufficient investigations, but the species establishment needs yet to be confirmed in both countries, Romania and Bulgaria. The addition of two more and repeated records over four years in the Crimean Peninsula (now spanning about 65 km) and first findings of nesting females outside the Peninsula in Odessa, Ukraine [99] in 2022 clearly confirm that M. sculpturalis is fully established in that region.

Functional biotic interactions revealed by participants
Aside from occurrence records, citizen scientists documented important aspects of the ecology of M. sculpturalis, such as interactions with host plants. Two formerly unknown host plant species for M. sculpturalis were documented (Dipsacus sp. and Jacobaea maritima), which contribute to still inadequate understanding of the diet breadth of this species. Although the targeted searches on floral resources counted higher numbers of discrete bee-plant interactions, the opportunistic recordings documented more plant taxa, including several nectar sources. However, both approaches resulted in frequent documentation of S. japonicum (supplementary table S1a, d), supporting previous hypotheses on the importance of this pollen provisioner for the successful establishment of this pollinator [33,43]. In the future, the foraging behavior of the bee and various other aspects of functional biotic interactions should be examined (e.g., interactions with other pollinators) for a better understanding of the invasiveness potential of this alien species and to test keystone hypotheses in invasion ecology [100,101].

Limitations of citizen science in southeast europe
Data gathered by citizen science approaches can differ in quality. It can be influenced by major biases: uneven spatial coverage, recording density, sampling efforts, and species detectability across space and time [19,102]. In an attempt to overcome these issues, we integrated data from various approaches by combining structured data gathered by a floral host protocol and unstructured data derived by opportunistically obtained observations of participants. Despite our efforts to integrate data from various sources, we can not rule out that our data set, especially the systematic/targeted search, reflects a sample bias towards urban areas. Rural recordings were largely provided by citizen scientists (89%), some from very remote, inaccessible locations, thereby providing valuable information about this species' spread, dispersal ability, and behavior. Thus, complementary monitoring regimes conducted by field ecologists and citizen scientists provided reliable presence and absence data -both important for following the expansion of M. sculpturalis.

Social media and public outreach
The large size and unique appearance of the sculptured resin bee enables species recognition by a broader public. Distinctive features of M. sculpturalis can be used in public outreach to communicate more efficiently about global species homogenization and draw attention to the project [78,95,96,103]. The educational aspects of citizen science programs dealing with (invasive) alien species are especially relevant for preventing future introductions [78,104,105].
Calls for (invasive) alien species monitoring on different social platforms can reach widespread and heterogeneous audiences resulting in numerous reports [106]. Naturalist forums on social networks store rich biodiversity data sets and are too valuable not to be considered or included in monitoring efforts [107], while the role of the users and data collectors as citizen scientists should be acknowledged. Hence, social media is an efficient tool for science communication, advertising, and disseminating calls for participation, and is a source for data extraction of biological information, as we demonstrated [74,78,108,109].
Language can be a barrier when working on cross-country and multilateral cultural scales [110]. For example, data extraction on social media comes with challenges, as misspellings and misidentification are common. Hence, a simple keyword search (e.g., 'sculpturalis') is usually not enough. Also, considering that many people on social media national groups use common names in their official language for M. sculpturalis, it is critical to have a native speaker in the network.

Conclusion
We demonstrated the effectiveness of citizen science as a research approach and social media as an efficient tool in data collection for an insufficiently studied region of Southeast Europe. The citizen science approach enabled collecting data from otherwise inaccessible areas, such as private properties, sparsely populated rural areas, and a much larger geographical range than sampled through the targeted fieldwork. Our study also highlights how individual citizen scientists can show enormous dedication, self-initiative, and provide valuable information.
Furthermore, we demonstrated that combining citizen science and research-led targeted fieldwork is an effective approach for investigating the distribution and expansion fronts of highly mobile and fast-spreading potentially invasive alien species, such as M. sculpturalis. While citizen science provided broader geographic coverage, targeted field work provided new insights into the current southeastern range expansion.
We see an urgent need to broaden the search efforts and further investigate the ecology of the sculptured resin bee. We propose that citizen science projects should be launched in other European countries within the species' known and potential future range. Collaborations between research groups, researchers, and volunteers in the framework of citizen science projects and the exchange of experience among involved parties can result in mutually beneficial knowledge. Citizen science projects could be further improved by developing multidisciplinary cooperation of biologists and computer scientists (e.g., for automated data mining or/and managing highly heterogeneous data sources, machine learning approaches). We conclude that citizen science studies are an efficient tool for biodiversity surveys and public outreach, especially in low-income regions such as Southeast Europe.