Different types of semi‐natural habitat are required to sustain diverse wild bee communities across agricultural landscapes

Wild bees provide essential pollination services to crops and wild plants (Potts et al., 2016; Wei 2021), but are jeopardized by habitat loss intensive agriculture (Goulson 2015). To counteract declines in agricultural landscapes, management measures conserve these semi-natural habitats (SNH) would ensure floral nesting resources for throughout the season. Hence, landscapes with higher amounts of SNH generally associated an increased abundance richness pollinators (e.g. Holzschuh 2010) enhanced (Garibaldi 2011). The type, structure composition may be critical drivers different taxa (Bartual 2019), our understanding relative contribution types their diverse pollinator meta-communities at landscape scale is scarce. Some recent studies suggest that bee communities Central Europe more abundant flower-rich grasslands than woody such as hedgerows forest edges Bartual 2019; Rivers-Moore 2020). Additionally, sown flower strips locally contributing sustaining populations generalist species (Ganser In addition, role can vary across season, example, due distinct flowering phenologies dominant plant (Cole 2017; Eeraerts 2021). bumblebee species, have been shown track season 2017), shifting main pollen source from mainly spring herbaceous still abundantly summer (Bertrand 2019). Thus, conservation should consider how promote resource continuity (Schellhorn Since target groups, rare or crop pollinating bees, rely on key (Sutter they distinctively benefit over could complementary spatial temporal niches, combining them support communities, partly β-diversity (Rivers-Moore Therefore, information about importance underlying local allow developing tailored landscapes. A promising tool evaluate factors species-habitat networks (Marini This approach applies bipartite interactions framework considering whole a unit meta-community. analysis contribute valuable relevant roles entire community reveal strongly linked certain (i.e. specialists). allows example assessing uniqueness terms its meta-community landscape. Besides drivers, quantity quality provided Sutter landscape-level percentage arable cover) configuration edge density) important (Martin contrast positive relationships between amount diversity, findings effects inconsistent (Hass 2018; 2010). potential reason effect depend (Martínez-Núñez Maurer it crucial properties interactive influence aspects within this study, we integrated network seasonal analyses study five major supporting varying configuration. We analysed data standardized transect surveys extensively conventionally managed meadows, strips, 25 Switzerland address following questions: (i) What meta-communities? (ii) Does (iii) bees? (iv) Are flower-habitat good predictors richness? (v) How do drive within—and among—different SNH? Data here were collected two 2014 (n = 17) 2020 8) northern Swiss lowlands (Figure 1a). Agricultural 1 km radius selected along gradient (17%–88% cover 51–157 m/ha density), ensuring least 3 centres (except landscapes). They considered independent, since average foraging ranges typically <1 (Greenleaf 2007). small-scaled mosaic (few orchards vineyards) permanent intensity, dominated each landscape, sampled transects (hereafter habitats): meadows (intensive meadows), (‘biodiversity promoting area’: no fertilizer application; first cut after 15th June), (inclusive border) 1b). survey, one patch per type (wherever possible, see Appendix S2, Table S1) 100 m (2 width; 2019 details). survey conducted 2020, were, analogous 2 wide same types. However, was subdivided into sections proportional (similar Cappellari & Marini, These randomly placed patches corresponding (including crops, not here). both surveys, three sampling rounds April, May/June July 9 am 6 pm during dry warm weather conditions (min. 14°C) low wind. Transects fixed allowed rounds. When present once round (see S1 overview). During walks, min used recording visiting section, pausing clock catching processing samples. Back laboratory, samples stored 70% ethanol −80°C until insect identification. 2014, experts determined morphologically, while barcoding cytochrome oxidase subunit I gene region company Microsynth Ecogenics GmbH (Balgach, Switzerland). Identified classified groups: when listed most available Red List ‘vulnerable’, ‘endangered’ ‘critically endangered’ (Amiet, 1994; interpreted adequate caution age); Kleijn al. (2015). excluded honeybee Apis mellifera L. all presence attributed beekeeping surrounding. did require permission fieldwork, nor ethical approval bees. assessed described (2019). Similarly, 10 plots x 0.5 m) (10 m; horizontally vegetation; vertically vegetation edges). m2 estimated vascular number single flowers multiplied area. Flower area calculated circle, radii (or inflorescences case Asteraceae Plantago sp.) retrieved trait databases: Casanelles-Abella (2021), Info Flora (https://www.infoflora.ch/de/), PlantNET (https://plantnet.rbgsyd.nsw.gov.au/) Naturegate (https://luontoportti.com/). section. Using geographic system (ArcGIS Pro version 10.7, ESRI), (1 radius) categories: Arable crop, orchard, vineyard, hedgerow, forest, meadow, urban green (>25% areas) space (<25% areas). Based resulting raster maps (pixel size: × m), total density (m/ha) using r package landscapemetrics (Hesselbarth, metrics inform simplification widely proxies Albrecht 2020; Hass 2018). built bee-habitat round, nodes links framework, found meta-community, where likely through dispersal (Leibold 2004). assess what extent depends specific specialists), strength (Collado Marini sum dependencies (fraction appearances) particular (Bascompte 2006). Compared traditional richness, which treats equally, provides information, complete picture perspective. complementarity (pooled rounds), further functional branch length dendrogram based qualitative differences assemblages (Devoto 2012). investigate whether differs types, modularity (Cappellari modular networks, share others thereby form modules (Olesen If mostly habitats, correspond habitats. Then, species' identified calculating z (standardized module, within-module degree) c values (level other modules, among-module connectivity; Olesen Here, high value show strong preference habitat, generalists. Strength, related (Appendix S2). All performed (Dormann 2008). First, completeness years Results showed coverage satisfactory differ respectively Figure S1). together compared without bias. examine rounds, round. fitted generalized linear mixed models negative binomial error distribution (square-root transformed) Gaussian distribution. Habitat interaction explanatory variables nested ID random effects. applied Tukey post-hoc tests test significant among (within rounds). groups (rare, pollinator, other), response group, two-way group (using above). variation unique contributions uniqueness), pooled step, beta diversity (LCBD) ‘beta.div’ function adespatial (Dray 2021; examined LCBD (square root model variable factor. second disentangled components according method proposed Legendre (2014) quantitative (abundance-weighted) Jaccard index, ‘beta.div.comp’ package. quantifies turnover nestedness tested (turnover nestedness) Wilcoxon rank test. third hypothesis supports relatively sets hence expected chance, Modularity overall DIRTLPAwb+ algorithm (Beckett, 2016). observed obtained 1000 null representing visits any controlling (Patefield algorithm; Dormann 2014). degree connectivity determine thresholds (c 0.62; 2.6) (2007). context abundance, strength, models. (in averaged rounds) entered variables. positively correlated (coefficient |r| 0.64) lower AIC thus better fit analyses. models, (log-transformed) fourth fifth model, (z-scores) (sampling pooled) model. way above, separately, z-scores Moreover, explored well alternative descriptors like complementarity, explained separate S3). describe availability only richness. included year additional factor account possible years. statistical software 4.1.1 (R Core Team, Models lme4 (Bates 2015) assumptions checked inspection residual DHARMa (Hartig, 2022). continuous improve convergence algorithms. together, recorded 2072 104 (530 flower-visiting 61 744 60 798 62 July). Of those, 24 21 pollinators. See S4 list species. Overall, supported highest changed (significant richness; 1; 2a,b). similar significantly extensive edges. May/June, July, became (Table (forest hedgerows) lower. qualitatively identical very repeated S5, Consistently, patterns shifts capacity specialists; 2c). also robust reduced dataset S2; S6). Relative differed studied (abundance pollinators): additionally (LCBD: F 0.99, p 0.42; marginal R2 0.03, conditional 0.32). (70% ± 5%, mean SE), rather (30% 5%; one-tail Rank Sum Test: W 103, < 0.001), indicating harboured large extent. fact, 17% exclusively (unique species; S7). Among these, (5 S8). (Q) 0.2, (one-tail Z test: 0.001; S4). detected four (1) (2) (3) (4) 3), corroborating Eighteen exceeded and/or S9, S5). Floral had 2). Bee (linear relationship) (when outlier excluded; hump-shaped relationship, affected cover, well-connected (high less connected (low moderate levels cover; >50%; 2, 4). contrast, influenced methodological approaches shed light habitat-level shaping demonstrate sustained consistently growing particularly many specialists At time, gradually April set highlighting niches results emphasize need take perspective dynamics drove While meadows—and lesser meadows—sustained gained late Extensively continuously early suite (Albrecht 2007), especially uniquely Ekroos (2020). often offer few active crucial, colony building bumblebees (Williams times scarcity – mown—in (Ouvrard Our indicate enhance pollinators, Schubert assuring periods long (Rundlöf 2014), region. line, social solitary (von Königslöw Interestingly, corroborates drives scale, although exact mechanism study. meadow extensification schemes, addition establishing agroecosystems (Ekroos Ganser Despite (Pfiffner implies evidence ecosystems (Penado 2022) conserving meta-communities. For oligolectic longhorn Eucera nigrescens module consisting intensively c). preferred forage region, Vicia sepium (Westrich, borders nutrient rich meadows. Although Bombus pascuorum sweat Lasioglossum malachurum association respectively, specialized apparently regularly Simultaneously, series use Even though cannot make direct inference determining evaluates if several (and which). useful develop targeted 2021).We note, however, almost inevitably under-sampling lead overestimation uniqueness, therefore needs caution. highly replicated inherently abundances. resources, α- (Hendrickx previous 2010; Lami except abundance. landscape: remaining field elements. Because central-place foragers restricted specialization result findings. dampen (Beduschi Consequently, structurally enhances facilitate might increase resilience disturbance, better-connected re-colonized faster (Tscharntke level und offers increases species-rich promoted resilient communities. illustrates combined analyses, tools baseline informed recommendations. besides providing insights scale. Combining (α- β-diversity), sustain Especially phenologies, complementarily Locally, specialists, maintaining ideally increasing highlight schemes play safeguarding specialist level, simple smaller sizes dense infrastructure areas under agri-environment schemes) seem exchange possibly disturbances. Actions recommendations help agroecosystems, Corina Matthias conceived ideas; Maurer, Louis designed methodology data; Carlos Martínez-Núñez led writing manuscript; Loïc Pellissier ideas visualization results. authors contributed critically drafts gave final publication. brings countries includes scientists country carried out. thank Stefanie Bossart, Lea Bona, Bettina Schär, Barryette Oberholzer Nadine Ahorn Christoph Grünig Jeanette Kast leading work identification Laura Bosco her calculations. Lorenzo anonymous reviewers comments earlier manuscript. Further, grateful farmers giving fields. acknowledge Biodiversa project VOODOO (Viral eco-evolutionary domestic global change www.voodoo-project.eu) funder Switzerland: SNSF 31BD30_186532/1. Open access funding Agroscope. declare conflict interest. via Dryad Digital Repository https://doi.org/10.5061/dryad.9cnp5hqn3 (Maurer Please note: publisher responsible content functionality supplied authors. Any queries (other missing content) directed author article.