Selection of sampling sites

The selection of sites aimed to ensure representativeness across habitats to best estimate the diversity of wild bees. A total of 32 sites were chosen, based on their expected diversity of plants and bees. These sites encompass a wide range of habitats within the SVNP (Fig. 2), most of which are classified using EUNIS codes. Amongst the habitats particularly favourable to bees, mesic grasslands (EUNIS E2) are mown twice a year. These grasslands are characterised by light to moderate fertilisation and generally support a diverse range of plant species adapted to moderately fertile soils. Additionally, due to the presence of the Semois River, numerous wet habitats are found within the SVNP. Seasonally wet and wet grasslands (E3) are ecosystems where water is present seasonally or even permanently. These areas are typically mown once or twice a year and develop on mesotrophic soils, which are relatively rich in minerals and nitrogen. Tall-herb communities of humid meadows (E5.42) are distinguished by spongy, moist soils resulting from the accumulation of organic matter. These habitats often emerge due to the abandonment of traditional agricultural practices, such as grazing and mowing, allowing for dense growth of large-leaved plants. As a result, they represent a successional stage that wet meadows may progress towards. Valley mires, poor fens and transition mires (D2) form in depressions, gentle slopes and areas waterlogged by rising water tables. Another wet habitat, riverine and fen scrubs (F9), represents near-climax formations, which are more stable in particularly wet sites with a high water table near the surface. In contrast to these wet habitats, a few drier habitats remain in the SVNP, though they are geographically restricted. Amongst these, dry grasslands (E1) are typically low-growing and result from past agro-pastoral practices. Today, only a few parcels in the SVNP have not been impacted by either reforestation or intensive agriculture. Dry heaths (F4.2) are characterised by siliceous soils, with open areas covered by mosses or lichens and plant species from the Ericaceae family. The region's soil was also exploited for its mineral content, with schist quarries historically used to extract materials for road construction, as well as the aforementioned slate quarries. These last two types of habitats are not associated with any EUNIS code.

Figure 2.  

Distribution of sampling sites. The 32 sites are located in the Semois Valley National Park, Belgium. The colour code represents the habitat type at each sampling site.

In total, eight sites of mesic grassland, eight sites of wet grassland and eight sites of tall-herb communities in humid meadows were selected. Additionally, five slate quarries, one schist quarry, one site combining mires, dry heaths and riverine scrubs, and one site combining dry grasslands and mires were selected. To maintain consistency in our sampling protocol and, due to the insufficient presence of the latter habitats in the National Park to include eight sites per habitat type, these last eight sites were grouped under the category 'other habitats' for the purposes of presenting the results in figures. Although ecologically heterogeneous, these eight additional sites aim to provide a comprehensive overview of bee diversity within the SVNP.

All sites were spaced at least one kilometre apart to minimise spatial autocorrelation during the inventories, ensuring that individuals collected from different sites were unlikely to be present in multiple locations, thus avoiding double sampling effort (Gathmann and Tscharntke 2002). The sampled sites were identified by the SVNP project leaders and employees. Private landowners of the selected sites were contacted and informed in advance. Additionally, forest rangers from the Département Nature et Forêt (DNF), responsible for safeguarding state and public land, were consulted. For each sampled site, an official collection permit has been delivered by the Service Public de Wallonie (SPW).

Sampling protocol

Throughout the study, each of the 32 sites was sampled monthly from April to August, resulting in a total of five sampling sessions per site across the entire inventory period. These months coincide with the majority of the active season for wild bees in Belgium (Duchenne et al. 2020). To ensure data consistency, a standardised sampling protocol was followed. Each sampling session involved 40 minutes of effective net collection time, which refers to the time spent actively searching for individuals within the site. The size of the sites ranged from 0.15 ha to 12.39 ha (mean = 1.97 ha, median = 1.14 ha). When a bee was captured, the stopwatch was paused to place the individual in a vial and to record the relevant collection information as outlined below. The effective collection time resumed once the bee was secured in the vial.

Bees were exclusively captured using nets, with the exception of honeybees (Apis mellifera), which were not collected. Research has shown that Apidae and Megachilidae species are more frequently caught with nets than with coloured traps, whereas the reverse is true for Halictidae, which are more commonly found in traps than nets (Vandaudenard 2023). Furthermore, net collection tends to favour larger, slower species, such as bumblebees (Bombus spp.) (Prendergast et al. 2020). While combining net collection with pan traps may seem like a viable alternative, pan traps were not used in this study due to several limitations. When left unattended between monthly sampling sessions, traps are often stolen or overturned, either by humans or large herbivores present in the Park. Additionally, smaller specimens collected in these traps can be challenging to identify due to prolonged exposure to water, which hampers the use of identification criteria, particularly those related to the general aspect and colour of body hair. It is also important to note that net sampling provides valuable information about the plant species visited by the bees.

The captures were carried out along a variable transect, where the collector moved freely across the site, primarily guided by the presence of flowers or nesting sites, rather than following a linear transect, where the collector remained on a straight line across the site. The variable transect approach enhances the capture of bee diversity by allowing the collector to focus on key plants and areas with abundant flowering (Westphal et al. 2008). Collected individuals were placed in vials containing ethyl acetate-soaked paper for euthanisation. During collection, data such as altitude, longitude, latitude, site reference, bee behaviour (whether in flight, on the ground or on flowers) and, when applicable, the plant species visited were recorded. The sampling protocol also required specific weather conditions. Thus, all captures were conducted between 9 AM and 5 PM, with temperatures above 15°C, no wind or rain and low cloud cover.

Specimen curation and analyses

One day after euthanasia, the specimens were pinned dorsally through the mesosoma. For males, genitalia were extracted from the metasoma using pins, as this organ is often essential for species identification (e.g. Rasmont et al. (2021), Wood (2023)). All bees were identified to species level using identification keys under a binocular microscope and each identification was subsequently validated by a taxonomist expert of the group. Specifically, Apidae identifications were confirmed by Frédéric Carion, Guillaume Ghisbain and Achik Dorchin, Megachilidae by Clément Tourbez, Halictidae by Thomas Brau and Simone Flaminio, Andrenidae by Thomas Wood, Colletidae by Romain Le Divelec and Melittidae by Maxence Gérard.

After all specimens were identified, an accumulation curve was generated using the iNEXT package (Hsieh et al. 2016) to illustrate the rate of species discovery relative to sampling effort. This method allows for extrapolation of the effort required to detect additional species. Finally, we estimated the total number of species in the region by following the methodology proposed by Chao (1984), Chao (1987).

All records collected for the present work were published through GBIF (Gérard 2025).

Diversity and abundance of species

Our standardised survey resulted in the collection of 1,119 specimens, representing 120 of the 419 bee species recorded in Belgium, i.e. 28.6% of the national bee fauna. By comparison, over a geographical area with a similar size, three years of sampling in the industrial region of the Hainaut Province yielded 9,410 specimens, encompassing 180 species (Fiordaliso et al. 2022). This discrepancy suggests that many species are yet to be discovered in the Semois National Park, as indicated by our accumulation curve (Fig. 15). Indeed, using the method proposed by Chao (1984, 1987), we estimate the maximum number of species in the province to be 161, with a confidence interval between 137 and 217 species (Fig. 15). Consequently, a single year of sampling has likely led to the identification of 55-88% of the species present in the National Park. The large standard error highlights that more sampling effort is needed to refine this estimation. As the species accumulation curve did not reach a plateau, direct comparisons of observed diversity with other protected areas remain challenging. However, when compared to a bee survey conducted in a European national park with similar forest-dominated habitats and a comparable position on the accumulation curve, the results appear relatively similar. For instance, in Wielkopolska National Park (Poland), 110 bee species were recorded over a six-month sampling period (Banaszak-Cibicka et al. 2018). It is important to note that Wielkopolska National Park is approximately half the size of SVNP, suggesting again that additional sampling efforts in SVNP would be necessary to enable robust conclusions.

Figure 15.  

Accumulation curve and expected number of species collected in National Park of the Semois Valley (Belgium). The sampling effort is represented by the number of specimens collected (x axis). The dotted line represents the predicted number of species (y axis), based on the number of specimens collected. The total predicted species richness and the associated confidence interval (in red) are calculated using the Chao method.

The genus Bombus (family Apidae) was the most abundant in our study, with more than one-third of the individuals collected (38%, n = 424) belonging to this genus. The two most frequently collected taxa were Bombus pascuorum (14.92%, n = 167) and species belonging to the subgenus Bombus sensu stricto (5.72%, n = 64), a subgenus represented by species in which the females of some species can only be reliably distinguished through genetic or semio-chemical analyses (Rasmont et al. 2021). The third most prevalent species was another bumblebee, B. hortorum (5.09%, n = 57). Although widespread and abundant across Europe (Ghisbain et al. 2021b), this species is classified as Near Threatened in Belgium (NT), with populations having recently declined (Drossart 2019). Its abundance in the National Park is likely due to the well-preserved grasslands, which host a rich diversity of Fabaceae plants. Moreover, this is one of the few bumblebee species partially associated with woodland habitats (Folschweiller 2020). While the prominence of some bumblebee species suggests that a restricted set of certain common species within this group are resilient to global environmental changes (Rasmont et al. 2015), it may also be an artefact of our sampling method, as the use of nets tends to overestimate the abundance of larger, social species (Prendergast et al. 2020). Amongst the solitary bees, Trachusa byssina (4.2%, n = 47, Megachilidae) was the most abundant. This finding is particularly noteworthy, as the species is protected in Belgium and is rarely recorded in other studies (e.g. Fiordaliso et al. (2022), who found only five specimens amongst 9,410 bees collected). Trachusa byssina is a summer species largely associated with Fabaceae plants and it was primarily collected from Lotus, which was present in all the habitat types sampled. Additionally, Trachusa byssina constructs its nests in the ground using tree leaves, which it seals with conifer resin. It is, thus, one of the few solitary bee species that benefit from mixed and coniferous forests. The second most abundant solitary species was Lasioglossum calceatum (3.57%, n = 40, Halictidae), one of the most common spring species in Belgium, primarily visiting Taraxacum spp. Colletes daviesanus (2.86%, n = 32, Andrenidae), an oligolectic species on the Asteraceae family, with a summer phenology, was the third most common solitary bee in our sampling. This species is widespread and abundant in Belgium.

Of the 120 bee species collected in this survey, 25 (around 21%) are either brood parasites (e.g. Nomada, Sphecodes) or inquilines (bumblebees of the subgenus Psithyrus). Parasitic taxa constitute a diverse guild of wild bees (Michener 2007) with a great potential as indicator taxa for assessing the health of bee communities (Sheffield et al. 2013). Due to their higher trophic level, their presence depends on the presence of their hosts and the resources used by the latter. Based on Sheffield et al. 2013, a proportion exceeding 20% of parasitic taxa amongst the total bee species sampled is considered high. Further detecting the presence of these parasites, that are most generally undersampled, overlooked and understudied due to their relatively lower abundance (Ghisbain 2023), will, therefore, be critical for properly characterising the communities of the National Park.

The overall sampling includes 34 singletons, representing 28.3% of the species, a proportion notably higher than those reported for wild bee communities in literature (e.g.Williams et al. (2001), Fiordaliso et al. (2022)). This elevated number of species collected as single specimens supports the findings of the species accumulation curve, suggesting that additional sampling effort is necessary to fully capture the bee diversity of the SVNP. Additionally, some species may naturally occur at low population densities. The SVNP offers a wide range of heterogeneous microhabitats, some of which are highly fragmented among forest patches, potentially limiting the abundance of certain species by restricting access to essential resources. Furthermore, since some species have very short activity periods and each site was sampled only once per month, the low probability of catching them might explain why they are not represented in our sampling.

Threatened and protected species

Of the 120 species identified, 22 are classified as threatened at the national level, based on the Red List of Belgian bees (CR: Critically Endangered (n = 4), EN: Endangered (n = 8), VU: Vulnerable (n = 10); Drossart (2019)), accounting for 18.3% of the total number of species collected (Fig. 16). Amongst the most remarkable species, Bombus humilis, Bombus sylvarum, Lasioglossum costulatum and Nomada sexfasciata were the four species Red-listed as CR. All species collected and their Red List status in Belgium are listed in Table 1.

Figure 16.  

Species distribution within the four types of habitats, according to their Red List status at the Belgian scale, following Drossart (2019). CR: Critically Endangered, EN: Endangered, VU: Vulnerable, NT: Near Threatened, LC: Least Concern, DD: Data Deficient, NE: Not Evaluated.

We recorded 12 species legally protected in Wallonia within the National Park. The most frequently encountered were Trachusa byssina (n = 47) and Macropis europaea (n = 15). The diet of Macropis europaea is highly specialised, as it forages exclusively on Lysimachia species - plants that thrive in particularly wet habitats - to collect oil. Its phenology is largely confined to the summer months.

Eighty-one bee species were recorded in mesic grasslands, 21 of which were exclusive to this habitat type in our sampling. Mesic grasslands also had the highest number of bee species with threatened conservation statuses, comprising 15 species: three categorised as Critically Endangered (CR), four as Endangered (EN) and eight as Vulnerable (VU). This habitat was particularly crucial for long-tongued bees from the Apidae family, including all six threatened bumblebee species in our dataset and the two Belgian species of Eucera, likely due to the abundance of flowering plants of the Fabaceae family. Similarly, the ‘other habitats’ category supported a substantial number of species, with 13 species exclusively found in these environments. Eleven of these were associated with slate and schist quarries, which provide suitable habitats for species typical of forested, thermophilic and/or rocky environments, such as Osmia parietina and Megachile ligniseca (Schmidt et al. 2017, Falk and Lewington 2019). The quarries which we sampled remain relatively open within large forest patches, offering nesting and floral resources that might be absent as the forest fully regenerates. Thus, maintaining this type of habitat is vital for maintaining a high bee diversity within the National Park. The wet meadow had the fewest species exclusive to it, yet harboured nearly as many threatened species (n = 11) as the tall-herb communities of humid meadows and the ‘other habitats’ (seven and nine, respectively). Nearly half of the Trachusa byssina population, a legally protected species in Wallonia, was also found in wet meadows, where they mainly foraged on big trefoil (Lotus pedunculatus). Additionally, almost all Melitta nigricans specimens were located in wet grasslands and tall-herb communities, where their host plant, the purple loosestrife (Lythrum salicaria), was abundant. Species richness across habitat types was then assessed using a generalised linear model (GLM) with a generalised Poisson distribution to address overdispersion. Statistical analyses were conducted using the R packages glmmTMB (Brooks et al. 2017) and emmeans (Lenth 2017). No significant differences in species richness were detected between habitat categories (p > 0.05). The lack of significant results is likely due to the limited number of sites per habitat category (n = 8 per category), which restricts the statistical power to detect differences amongst groups. Nevertheless, this finding underscores the importance of each habitat type, as they support distinct species.

Eighteen species were found across all four habitat categories, suggesting a broader ecological niche. Amongst them, Bombus soroeensis is Red-listed as Vulnerable (VU). This species is often misidentified in Belgium due to its close resemblance to the far more common Bombus lapidarius. Bombus soroeensis is typically associated with forest and boreal climates (Folschweiller 2020) and climate warming may pose a threat to its Belgian populations.

Finally, we identified two species that were not evaluated (NE) in the Belgian Red List (Drossart et al. 2019). The first, Nomada braunsiana, is a cleptoparasitic bee previously documented by a single record in Belgium (Vertommen et al. 2024). This species is widespread across Europe, with occurrences reported in most neighbouring countries (Smit 2018). However, its host species, Andrena curvungula and Andrena pandellei, are rare and were not detected in our sampling. Recent findings, including its discovery in the Netherlands in 2020 (Fernhout and Rhebergen 2020), suggest that N. braunsiana may be expanding its range. The second species with a NE Red-list status, Andrena afzeliella, belongs to the Taeniandrena subgenus and was recently confirmed as part of the Belgian fauna (Wood 2023). Advances in resolving the taxonomic challenges within this subgenus facilitated its identification. A review of historical collections revealed that A. afzeliella has been present in Belgium for a considerable period, albeit previously unrecognised (Wood 2023). Andrena afzeliella exhibits a dietary preference for flowers of the Fabaceae family and is bivoltine, with generations occurring in late spring and mid-summer (Praz et al. 2022).

We recorded 120 bee species during a five-month sampling period in the Semois Valley National Park (SVNP), Belgium. Of these, 22 are listed as VU, EN or CR on the Belgian Red List, representing a notably high proportion of threatened species. We identified four ecological categories of particular importance for wild bee conservation:

Forest and Clearing Species: A significant number of threatened species are associated with forests and clearings. Conservation measures should focus on the promotion of habitat heterogeneity by creating and maintaining early successional stages and preventing the recolonisation or abandonment of existing openings (Taki et al. 2013, Odanaka and Rehan 2020). These disturbed areas should be promoted along more preserved parcels of trees, as canopy gaps and standing dead trees in old-growth forests may support distinct bee communities (Eckerter et al. 2022).

Fabaceae-Associated Species: Many bee species depend on Fabaceae plants naturally found in mesic grasslands, though some Fabaceae, like Lotus pedunculatus, are also found in marshy areas. In degraded habitats targeted for restoration, it is essential to ensure the presence of Fabaceae species. Integrating these plants into agricultural practices, as a natural alternative to industrial fertilisers, may further support these species. Legume crop seed mixes can be optimised to include both early- and late-flowering species, enhancing support for bumblebee populations that typically exhibit extended periods of activity (Pywell et al. 2005, Cole et al. 2022).

Thermophilic Species: Numerous threatened species in our sample are adapted to warm, open environments, highlighting the critical role of regional quarry sites in supporting bee diversity. Effective management of these areas should focus on maintaining early-successional stages rich in floral resources, while preventing forest colonisation of exposed nesting sites (Kettermann et al. 2022). We advocate granting protection status to several slate quarries, accompanied by targeted management practices to conserve their unique ecological characteristics. Additionally, further consideration should be given to the potential role of nearby watercourses in providing similar nesting habitats.

Wetland Community: Wetland habitats host a unique, albeit less diverse, community including species specialised on Lysimachia, such as Macropis europaea and Macropis fulvipes, as well as their associated parasite Epeoloides coecutiens. Protecting these habitats is, therefore, essential for maintaining this distinctive assemblage.

Repeated collection efforts and further observations of wild bees in the National Park of the Semois Valley are expected to reveal species that have been overlooked in the present study. Further work should ultimately allow us to compare the diversity of the Park to other protected areas in Belgium and abroad and contribute to anticipate the potential threats that these bee communities species will face in a context of global change. These results will guide the SVNP and other protected areas in Belgium in focusing conservation efforts on sites that harbour both the highest species diversity and the greatest number of threatened species.