Direct and indirect effects of local habitat quality, landscape use intensity and honeybees on plant-pollinator network functioning and robustness across Europe

Overview_data_sets.xlsx:Overview table of the data set collection.Study: original study where the data comes fromStudy ID: ID of the studyCountry: country where the study was conductedHabitat: main habitat that was sampledSampling year: year when the sampling took placeTaxa recorded: recorded taxa in the plant-pollinator interactionsCrop cover provided: whether original crop cover was provided (arable, arable + perennial, no)Flower survey: whether a flower survey was done in addition to the interaction sampling (yes, no)Seed set: whether seed set of one or more wild plant species was assessed at the same sitesNo. sampling rounds: Number of sampling rounds done per sampling yearNo. sites (i.e. networks): number of sites sampleddata_sem.csv: Data used to build the piecewise structural equation model (without honeybee abundance as a driver). All variables were z-transformed within study. Honeybees (Apis mellifera) were excluded from the networks (interactions between wild bees or hoverflies and visited plants).Study_id: ID of the studySite_id: ID of the site where the plant-pollinator sampling took place. One network was constructed for each site_id (pooling the sampling rounds).mean_d.sc: mean foraging specialisation (d’) of all wild pollinator species in that network (Blüthgen et al. 2006)mean_PSI.log: mean pollination service index (PSI, log-transformed) of that network (Dormann 2011)Common.I_mean.sc: mean commonness index of all wild pollinator species occurring at that site by using “Fuzzy Quantification of Common and Rare Species in Ecological Communities” (FuzzyQ) (Balbuena et al. 2021). This method estimates the commonness of each species based on combining its abundance across and occupancy of sites via fuzzy clustering using R package FuzzyQ version 0.1.0 (Balbuena et al. 2021).wNODF_z.sc: weighted nestedness (NODF) (Nestedness metric based on Overlap and Decreasing Fill) (Almeida-Neto & Ulrich 2011) of that network, standardised using 1000 null models (using the Patefield algorithm which keeps marginal totals constant, Blüthgen & Staab 2024; Patefield 1981).fc_HL_z.sc: Functional complementarity of the wild pollinators in that network, standardised using 1000 null models (using the Patefield algorithm which keeps marginal totals constant, Blüthgen & Staab 2024; Patefield 1981).z.scores_mod.sc: Modularity of that network calculated based on weighted networks using the DIRTLPAwb+ algorithm from Beckett (2016) to calculate Newman’s modularity (Newman 2006) and standardised using 1000 null models (using the Patefield algorithm which keeps marginal totals constant, Blüthgen & Staab 2024; Patefield 1981).WP.richness.sc: Number of wild pollinator species in that network.WP.abundance.sc: Number of wild pollinator individualsF.richness.sc: Number of plant species assessed in a separate flower survey. The studies “46_Hadrava”, “53_Libran_Embid”, “60_Triquet” and “71_1_Cano_Saez” did not do a separate flower survey, thus the number of plant species from the networks was taken.Prop_crops.sc: Percentage of (mostly) arable crops in a radius of 1000 m around the sampling site. For details see the method section of the related article.Edge_den_1000_clc.sc: Edge density (m/ha) in a 1000 m radius around the sampling site considering the following land use categories from CLC+ landcover raster data (© European Union, Copernicus Land Monitoring 2018): “sealed”, “permanent herbaceous”, “periodically herbaceous”, “lichens and mosses”, “non- and sparsely vegetated”, “low-growing woody plants (bushes, shrubs)” and we combined the categories “woody – needle leaved trees”, “woody – broadleaved deciduous trees” and “woody – broadleaved evergreen trees” into one “woody” category since the distinction among woodland types is likely irrelevant for pollinators sampled in the herbaceous vegetation layer.r50_IS_5_rew_75.sc: calculated robustness of the network as the fraction of primary extinctions that cause 50% of all species in the network to go extinct (both primary removals and secondary extinctions) when sequentially removing plant species from the least to the most connected species, using R package NetworkExtinctions, version 1.0.3 (Ávila-Thieme et al. 2023). The higher these R50-values, the more robust the network. The extinction threshold was set to 50 % (i.e. interaction strength threshold: each node went extinct when it lost more than 50 % of its original interaction strength). We also included a rewiring probability between each plant and pollinator species based on two functional traits: lecticity of the pollinator species (diet breadth) and flower complexity of the plant species, and the rewiring probability threshold was set to 75% (for details see method section of the related article).data_sem_hb-driver.csv:Data used to build the piecewise structural equation model (with honeybee abundance as a driver). All variables were z-transformed within study. Honeybees (Apis mellifera) were excluded from the networks. One additional variable compared to “data_sem.csv”:HB_abundance.log.sc: Number of honeybees sampled at that site, log-transformed.seedset_data.csv:Data of the seed set of eight plant species (from two studies: Scabiosa ochroleuca, n = 9 sites, 10 flower heads from 10 individual plants, Centaurea jacea, n = 5 sites, Hypochaeris radicata, n = 6, Lathyrus linifolius, n = 5, Lathyrus pratensis, n = 3, Lotus corniculatus, n = 4, Knautia arvensis, n = 5, Ranunculus acris, n = 5, up to 9 individuals per species) at a selection of sites where also networks were collected.Study_id: ID of the studySite_id: ID of the site where the plant-pollinator sampling took place.Species_long: Species name of the plant speciesunfertilized: number of unfertilized seeds of the plantfertilized: number of fertilized seeds of the plantmean_PSI.log: mean pollination service index (PSI, log-transformed) of that network (Dormann 2011)fc_HL_z.sc: Functional complementarity of the wild pollinators in that network, standardised using 1000 null models (using the Patefield algorithm which keeps marginal totals constant, Blüthgen & Staab 2024; Patefield 1981).ext_scenario_valid.csv:Data for the validation of the different extinction scenarios using R package NetworkExtinctions, version 1.0.3 (Ávila-Thieme et al. 2023), see description of robustness calculation in data_sem.csv. The extinction scenarios differed in the probability threshold for rewiring: RP = 0.25, RP = 0.5 or RP = 0.75. To evaluate which of the three robustness scenarios was the most realistic (i.e. predicted species loss best), we assessed the difference between simulated and observed species loss from the networks. For details see method section of the related article.Study_id: ID of the studySite_id: ID of the site where the plant-pollinator sampling took place.Pair_id: ID of the site-pairsIS: interaction strength threshold, set to 0.5 (i.e. interaction strength threshold: each node went extinct when it lost more than 50 % of its original interaction strength)RewiringProb: threshold of the probability for rewiring (i.e. the species only rewired when their probability RP is higher than the threshold). Flexible scenario: RP = 0.25, Medium scenario: RP = 0.5, Constraint scenario: RP = 0.75.Scenario: 3 levels: flexible, medium constraint. See above.Remaining_sp_diff: difference in the remaining plant and pollinator species between the plant species-poor and the plant species-rich site (after extinction scenario run on the rich site) of the site pair. For details see method section of the related article.Remaining_sp_diff_abs: absolute difference in the remaining plant and pollinator species References:Almeida-Neto, M. & Ulrich, W. (2011). A straightforward computational approach for measuring nestedness using quantitative matrices. Environmental Modelling & Software, 26, 173–178.Balbuena, J.A., Monlleó‐Borrull, C., Llopis‐Belenguer, C., Blasco‐Costa, I., Sarabeev, V.L. & Morand, S. (2021). Fuzzy quantification of common and rare species in ecological communities (FuzzyQ). Methods in ecology and evolution, 12, 1070–1079.Beckett, S.J. (2016). Improved community detection in weighted bipartite networks. R Soc Open Sci, 3, 140536.Blüthgen, N., Menzel, F. & Blüthgen, N. (2006). Measuring specialization in species interaction networks. BMC Ecol, 6, 9.Blüthgen, N. & Staab, M. (2024). A critical evaluation of network approaches for studying species interactions. Annual Review of Ecology, Evolution, and Systematics, 55.Dormann, C.F. (2011). How to be a specialist? Quantifying specialisation in pollination networks. Network Biology, 1, 1–20.Newman, M.E.J. (2006). Modularity and community structure in networks. Proceedings of the National Academy of Sciences, 103, 8577–8582.Patefield, W.M. (1981). Algorithm AS 159: an efficient method of generating random R× C tables with given row and column totals. Journal of the Royal Statistical Society. Series C (Applied Statistics), 30, 91–97.