Showing papers in "Journal of Applied Ecology in 2022"

Strip‐intercropping of wheat and oilseed rape enhances biodiversity and biological pest control in a conventionally managed farm scenario

Abstract: 1. Conventional agriculture in the global north is typically characterized by large monocultures, commonly managed with high levels of pesticide or fertilizer input and mechanization. Strip-intercropping, i.e., diversifying cropland by growing strips of different crops using conventional machinery, may be a viable strategy to promote natural predator diversity and associated biological pest control in such conventional farming systems. 2. We tested the influence of strip-intercropping of conventionally managed winter wheat with oilseed rape, using common machinery with 27-36 m broad strips, on arthropod predator diversity and biological pest control. We characterized spider and carabid beetle communities, calculated pest aphid and pollen beetle densities, and recorded parasitism rates for both crops (number of mummified aphids on wheat and number of parasitized pollen beetle larvae on oilseed rape). 3. We observed a significant reduction in the densities of wheat aphids (50% decrease) and pollen beetle larvae (20% decrease) in strip-intercropping areas compared to monocultures. Parasitism rates of wheat aphids increased significantly from 10% in monocultures to 25% in strip-intercropping areas. The number of parasitized pollen beetle larvae did not show the same pattern but was higher towards the center of the oilseed rape strip. Overall, the composition of predator communities benefited from the close neighborhood of the two crop species in the strips, as carabid beetles were more abundant in oilseed rape and spiders were more abundant in wheat fields. Overall, strip-intercropping reduced the dominance of one predator group and allowed for an equal representation of both spiders and carabid beetles in the mixture. 4. Synthesis and applications: Our study presents evidence of the benefits of adopting strip-intercropping with relatively large strips (adapted to existing machinery) for natural predator diversity and biological pest control in a large-scale conventionally managed farm scenario. Wheat-oilseed rape strip-intercropping reduced pest densities, increased parasitism of wheat aphids, and promoted equal representation of natural predator groups well beyond the areas of monoculture. Overall, by reducing the area dedicated to only one crop, the implementation of strip-intercropping adapted to mechanized agricultural scenarios can be used to increase crop heterogeneity at regional scales and enhance biodiversity and biological control, even in simplified landscapes dominated by large-scale conventional agriculture.

Bringing together approaches to reporting on within species genetic diversity

Abstract: The Convention on Biological Diversity (CBD, 1992) recognises three main levels of biodiversity: ‘diversity within species, between species and of ecosystems’. Genetic diversity within species (heritable variation) underpins their ability to react, adapt and be resilient, which is particularly crucial at this time of climate change, and biodiversity loss. Reporting is a key aspect of the CBD—all Parties must report progress approximately every 4 years. Reporting on changes over time allows policy makers to assess progress, evaluate policy effectiveness and learn from the outcomes. However, despite its importance, genetic diversity did not achieve similar levels of recognition to the other two levels of biodiversity in the 2020 Aichi targets (e.g. Hoban et al., 2020; Laikre et al., 2020) and where reported upon, reports were largely limited to species of agricultural or forestry importance (Hoban et al., 2021), which are largely unrepresentative of global biodiversity. Concerns about neglect of wild species' genetic diversity over the past three decades have led to several potential monitoring and reporting approaches being proposed. While we welcome this burgeoning interest, we are concerned that a choice of multiple reporting approaches may lead to confusion among policy makers, conservation practitioners and other stakeholders. Such confusion may lead to continued lack of reporting on genetic diversity of wild species, as the issue may be perceived to be too complex to resolve. Having different approaches also limits opportunities to make comparisons among countries, within countries and regions, and across time, and thus may mask genetic diversity loss. Given genetic diversity's vital role, we believe that a consolidated approach to reporting is essential if all countries are to maximise opportunities to protect biodiversity. This paper presents a simple framework to bring together several proposed reporting methods, and shows how they are related. In our proposals, genetic monitoring refers to ‘monitoring of genetic diversity within and between populations of species across contemporary time frames covering at least two different time points’ (Hvilsom et al., 2022). The examples below will show that such monitoring can make use of DNA data or proxies (Table 1), and results may be expressed as single indicators or grouped. Our focus is on monitoring genetic diversity within species, and does not include more general use of genetic data to study biodiversity (e.g. the use of molecular markers to track individual organisms, or the use of DNA barcoding to identify species). Indicators measure pressures on biodiversity, the state of biodiversity, conservation responses and benefits from ecosystem services (Butchart et al., 2010). The Group on Earth Observations Biodiversity Observation Network (GEO BON) Genetic Composition Working Group has used a collaborative international approach to develop genetic Essential Biodiversity Variables (EBVs; Hoban et al., 2022), designed for monitoring and understanding biodiversity change. These EBVs measure: (a) genetic diversity; (b) genetic differentiation; (c) inbreeding; and (d) effective population size (Ne). The first two require genetic sampling, but can usually be calculated from a single time point sample dataset. Furthermore, they can be calculated using different genetic markers (e.g. whole genome sequencing data, SNPs, DNA sequences, microsatellites), allowing cost to be reduced by using existing datasets (Kriesner et al., 2020), rather than requiring de novo sample analysis. Inbreeding and effective population size can be calculated using genetic data or inferred from proxies (Hoban et al., 2020, 2022). EBVs are summary measures of biodiversity rather than indicators. As with all indicators, these are imperfect measures of genetic change and careful interpretation and application of indicators is needed, including thoughtfully considering historic and recent population fragmentation (see Hoban et al., 2020, Hoban et al., 2022 for a more complete discussion). On average, although, they should provide relative assessment of genetic erosion, in an affordable manner, without requiring genetic data. Several countries are developing national programmes for monitoring genetic diversity. The Swedish Agency for Marine and Water Management (SwAM) has proposed three indicators to integrate genetic diversity into the national aquatic monitoring programme (Andersson et al., 2021). These focus on monitoring genetic diversity within and between populations, and on assessing the genetically effective population size; they are being applied to several marine and freshwater species using different types of DNA data. Furthermore, the Swedish Environmental Protection Agency (SEPA) has recently prioritised species for monitoring genetic diversity, and initiated monitoring using different DNA methods depending on target species and techniques available (Posledovich et al., 2021). SEPA is also using Swedish Red List data to apply the indicators proposed by Hoban et al., 2020, which use proxies for genetic diversity (Thurfjell et al., 2022). Switzerland is also implementing a strategy for a national monitoring of genetic diversity and currently runs a pilot study (https://gendiv.ethz.ch) for a small number of high priority species based on an earlier feasibility study (Fischer et al., 2020) and stakeholder analysis (Pärli et al., 2021). This will monitor genetic diversity, Ne, population structure, gene flow, inbreeding, hybridisation, genetic load and, if possible, adaptive potential. Switzerland uses historical DNA (hDNA) from collections to directly explore the temporal dimension of genetic diversity and to infer baselines of past genetic diversity. It uses individual whole genome resequencing and de novo genome assemblies for all species, perhaps making it the most powerful of the methods considered, although at high cost and complexity. As technologies mature, cost typically declines, possibly making this approach more widely applicable. Hollingsworth et al. (2020) developed a scorecard approach to assessing genetic diversity in wild species and published a report for Scotland. This was compiled using available data and expert knowledge across multiple disciplines including conservation, agriculture and forestry and statistics. The method was designed to be practical in all countries regardless of economic development, focuses on threats to genetic diversity, and is not dependent on prior genetic knowledge. It assesses: (a) demographic declines likely to lead to genetic diversity loss (genetic erosion—including declines in population size, loss of functional diversity and loss of divergent lineages), (b) hybridisation likely to lead to undesirable replacement of genetic diversity (note that not all hybridisation is unwanted—in some cases it is beneficial to adaptation or may be a natural process at contact zones—and genetic rescue can rely on crossing with allochthonous populations), (c) restrictions to regeneration/turnover likely to impede evolutionary change and (d) representativeness of ex situ collections, where applicable. The overall risk and mitigation are summarised into ‘green’, ‘amber’ or ‘red’ status for each species. The Scottish scorecard covered 26 terrestrial species with plans to expand to marine species. A version is being developed in Libya to test its application in a country facing severe resource constraints. Additionally, standard bibliographic methods are being developed to facilitate a basic inventory of genetic studies of wild and domestic species within any given country, which can then be reported, although this does not refer only to genetic monitoring but also genetic surveys (single time point studies). Despite recognition of the importance of wild species' genetic diversity, reporting under CBD was very limited (Hoban et al., 2021). This may be partly because broad-scale monitoring of genetic diversity is seen as difficult. For example, while effective population size can be measured for populations (Hoban et al., 2022), getting meaningful data across a whole country for tens of species is resource intensive and thus challenging for developing nations. DNA sequence data collection for dozens of species may cost hundreds of thousands to several million euros per reporting period: for example, Posledovich et al., 2021 and developing nations may need better access to training and equipment (Hvilsom et al., 2022). In contrast, where detailed data are available, it makes sense to use them. Differential access to data may restrict comparisons among nations or regions if some use DNA data while others use proxies. Comparable data are important for nations to share good practice, or to enable interpretation across a species' international range. Reporting requirements must be flexible enough to allow nations to participate using the best level of technology available in each country for their own requirements. Policy makers are a key audience who need access to clear, accurate information on the status of genetic diversity in order to make informed decisions affecting biodiversity (Hoban et al., 2013; Klütsch & Laikre, 2021; Vernesi et al., 2008). The multiplicity of methods may lead policy makers to conclude that reporting is too complex or impractical (Young et al., 2014). The healthy debate integral to scientific development may lead to mixed messages, even if most specialists agree on the key issues (Spierenburg, 2012). Furthermore, monitoring and reporting should be valuable to practitioners. The disconnect between conservation geneticists and conservation practitioners (‘conservation genetics gap’; Hoban et al., 2013) remains a problem (Klütsch & Laikre, 2021). These issues must be resolved for genetic diversity to be properly considered nationally and internationally, and for global species conservation plans. Standardised tools will allow practitioners to integrate genetic diversity into conservation efforts across the in situ and ex situ continuum, ensuring that this essential facet of biodiversity receives adequate attention and reporting to CBD can be achieved. All the approaches to measure and report genetic change detailed above have strengths and are already being implemented, demonstrating that they are well aligned with policy-makers' needs. Hvilsom et al. (2022) have found that they have much in common, both in terms of policy goals and selection criteria (Figure 1). As well as being essential for measuring change and informing policy, these approaches are relevant to species and habitat conservation. Existing and proposed genetic diversity conservation measures, such as gene conservation units (GCUs; Koskela et al., 2013; Minter et al., 2021) rely on monitoring to assess their efficacy. GCUs are designed to protect genetic diversity and evolutionary processes in situ, aiding adaptation to environmental change, and complementing existing approaches to species and habitat conservation. Effective population size is frequently used as an assessment criterion for GCUs. The various teams developed ideas separately but are now in frequent communication, particularly through the Coalition for Conservation Genetics (Kershaw et al., 2022), providing an opportunity to collaborate on international standards. We also recognise the benefits of engaging with initiatives such as the Earth BioGenome Project, Africa BioGenome Project, International Barcode of Life and others around the world. We should embrace pragmatism. In order to serve all nations, we believe that we should cooperate to develop a practical and flexible approach that can encompass genetic (including genomic) data as well as proxies or expert opinion. Given common themes within the various approaches to genetic diversity reporting, we propose bringing them together, using the categories outlined in the Scottish scorecard (Hollingsworth et al., 2020) as headlines, and nesting all approaches within this framework (Table 1). This categorisation would allow users (e.g. Parties to the CBD and other agreements) to select and report on those metrics most suitable for their needs and resources (using as many as possible), considering expertise, time and data availability. Equally importantly, it would provide an overview of potential steps for increasingly comprehensive reporting. The framework approach enables consideration of all main threats to genetic diversity. It can also monitor interventions, in situ and ex situ, which may incentivise active conservation of genetic diversity. To support genetic diversity reporting, we propose creation of a centrally held database to hold both monitoring and underlying data. This would allow transparency and encourage contributions from nations that may not have the resources to set up local mechanisms. The database could be established and maintained by an intergovernmental organisation such as GEO BON or IUCN, potentially linking into and informing the Red List process. Embedding genetic diversity metrics into the Red List would give appropriate weight to this crucial aspect of biodiversity (Garner et al., 2020; Willoughby et al., 2015). Experience with EUFORGEN, which has a much broader scope albeit over a smaller geographic area, suggests that a collaboratively funded coordination mechanism need not be expensive (EUFORGEN's annual budget ≈€350,000; member states contribute €2500–€35,000 each; EUFORGEN, 2019). We suggest that funding is required for at least the whole CBD reporting cycle (i.e. to 2030) to ensure its benefits are realised. Our proposal would provide all nations, regardless of economic status, with the ability to report on the pressures, state, conservation interventions and ecosystem services provided by genetic diversity. D.O. conceived the paper; D.O., L.L., S.H. and A.J.M. managed the editing process and coordinated the paper's production; All authors contributed to the design of the component approaches, contributed critically to the drafts and gave final approval for publication. R.E.S. designed the graphical abstract. The authors are grateful for the contribution of COST Action G-BiKE, CA 18134, supported by COST (European Cooperation in Science and Technology), www.cost.eu, which formed the backbone of this work. We thank the following funding agencies for support to separate co-authors: the Swedish Research Council Formas (grant 2020/012990; LL), the Swedish Research Council (grant 2019-05503; LL). The authors declare that they have no conflict of interest. This article does not contain data.

Wildlife roadkill and COVID‐19: a biologically significant, but heterogeneous, reduction

Abstract: Collisions with vehicles are a major cause of wildlife mortality. During the COVID-19 pandemic, many countries enforced lockdowns that reduced vehicular traffic and consequently wildlife-vehicle collisions. However, no study has yet explored how traffic-related mortality declined across multiple species of wildlife, leaving doubts about the species-specific impact of COVID-19 on wildlife ecology and management. We modelled how two lockdowns (in spring and autumn 2020) influenced wildlife-vehicle collisions throughout Slovenia, in central Europe, by comparing weekly roadkill in 2020 with 2010?2019 time series for European roe deer (n = 53,259), red fox (n = 9,889), Eurasian badger (n = 5,170), brown hare (n = 5,050), stone marten (n = 4,267), wild boar (n = 1,188), and red deer (n = 1,088). During the spring lockdown (16 March ? 30 April 2020), we observed far fewer collisions than expected for roe deer and badgers. During the autumn lockdown (20 October ? 31 December 2020), we observed significantly fewer collisions for roe deer and wild boar, but we noted an excess of collisions with badgers. Traffic reduction had a major influence on roe deer, whose roadkill decreased by between 156?321 individuals. Heterogeneous changes in road mortality across the seven studied species indicate that reductions in human mobility can trigger complex species-specific dynamics in wildlife assemblages, which may generate compensatory effects beyond lockdowns. For some species, such as roe deer, local reductions in the number of roadkill attained a significant fraction of the overall mortality. This could affect local population dynamics in cases where lockdowns are repeated over a number of years. Policy implications. Management aimed at reducing vehicular traffic, and therefore human disturbance and roadkill, can be evaluated using time-series analysis of data on multiple species. During times of restricted human movement, local-scale reductions should be estimated and accounted for in adaptive management, such as for planning culling quotas, to minimize their ecological and socio-economic impacts while optimizing outcomes of science-based population management.

The disproportionate value of ‘weeds’ to pollinators and biodiversity

Abstract: Agricultural intensification has been implicated in global biodiversity declines. In the European Union, agri-environmental schemes are designed to address this. For pollinating insects, funding has been provided to sow wildflower mixes. However, previous research indicates that a suite of agricultural weeds are also of great importance to pollinators. Here, we compare the biodiversity associated with the species which are considered harmful to agricultural production and legally deemed as ‘injurious’ by the United Kingdom 1959 Weeds Act (common ragwort Jacobaea vulgaris, creeping thistle Cirsium arvense, spear thistle C. vulgare, curled dock Rumex crispus and broadleaved dock R. obtusifolius), with plant species recommended for pollinator-targeted agri-environmental options. In our field study, the abundance and diversity of pollinators visiting the weed species averaged twice that of the recommended plants and included the main insect orders (Coleoptera, Diptera, Hymenoptera and Lepidoptera). This relationship was also seen in a meta-analysis of literature data, which indicates that fourfold more flower-visitor species and fivefold more conservation-listed species are associated with the weeds. Additionally, the literature shows that twice the number of herbivorous insect species are associated with these plants. We suggest that several factors are responsible for this pattern. Injurious weed species are widely distributed, their flower morphology allows access to a wide variety of pollinator species, and they produce, on average, four times more nectar sugar than the recommended plant species. Freedom of information requests to public bodies such as local councils, Natural England and Highways England indicate that c. £10 million per year is spent controlling injurious weeds. Meanwhile, the cost of the four pollinator-targeted agri-environmental options in the United Kingdom exceeds £40 m annually. Synthesis and applications. Our results clearly show that weeds have an underappreciated value to biodiversity. Unfortunately, current UK agricultural policy encourages neither land sparing for nor land sharing with weeds. The UK government is, however, currently committed to overhauling agricultural payments to encourage more wildlife- and climate-friendly practices. Thus, the challenge of reconciling the conflicts between agricultural production and these native and biodiverse species should be a renewed priority to land managers, researchers and policymakers.

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

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

Primary forest loss and degradation reduces biodiversity and ecosystem functioning: A global meta‐analysis using dung beetles as an indicator taxon

Abstract: Because of continuing degradation or deforestation in areas of undisturbed primary forest, there is a need to study the relative merit of strategies that mitigate their impacts on biodiversity and associated ecological functionality. Here, we provide a global synthesis of forest degradation or deforestation using 48 studies published in peer-reviewed journals that use dung beetles as indicators given their sensitivity to anthropogenic disturbance and their relevance in performing essential ecological functions in terrestrial ecosystems. We evaluated forest cover associated with undisturbed primary forest degradation (i.e. degraded primary forest) and undisturbed primary forest deforestation (i.e. secondary forest, forestry plantations and forestry restoration implementation) on species richness, total abundance, biomass, functional groups' presence and ecological functions provided by dung beetles. Additionally, we determined whether if dung beetle responses to forest disturbances were geographically dependent. We found lower diversity and a decrease in ecological functions associated with all classes of disturbance in primary forest. However, the effects were less severe in the case of forest degradation compared to complete deforestation with natural regeneration of secondary forest, development of forest plantations or active forest restoration by planting indigenous trees. The Neotropical and Oriental regions are particularly vulnerable, given the elevated rates of undisturbed primary forest deforestation and its negative impact on their assemblages' diversity and ecological functions. Synthesis and applications. Our results show that efforts for the conservation of remaining undisturbed primary forests need to be prioritized, especially in tropical latitudes. However, in regions where primary forest conservation is not feasible, logging management programs in degraded primary forest may have a potential role in reducing negative impacts for dung beetle diversity and ecological functions. Moreover, we conclude that despite the negative effect of primary forest deforestation and implementation of secondary forest, forestry plantation and forestry restoration, they can be useful for partial recovery of diversity and ecological functions performed by dung beetles in areas lacking any primary forest (undisturbed or degraded) vegetation cover.

Flower plantings support wild bee reproduction and may also mitigate pesticide exposure effects

Abstract: Sustainable agriculture relies on pollinators, and wild bees benefit yield of multiple crops. However, the combined exposure to pesticides and loss of flower resources, driven by agricultural intensification, contribute to declining diversity and abundance of many bee taxa. Flower plantings along the margins of agricultural fields offer diverse food resources not directly treated with pesticides. To investigate the potential of flower plantings to mitigate bee pesticide exposure effects and support bee reproduction, we selected replicated sites in intensively farmed landscapes where half contained flower plantings. We assessed solitary bee Osmia lignaria and bumble bee Bombus vosnesenskii nesting and reproduction throughout the season in these landscapes. We also quantified local and landscape flower resources and used bee-collected pollen to determine forage resource use and pesticide exposure and risk. Flower plantings, and their local flower resources, increased O. lignaria nesting probability. Bombus vosnesenskii reproduction was more strongly related to landscape than local flower resources. Bees at sites with and without flower plantings experienced similar pesticide risk, and the local flowers, alongside flowers in the landscape, were sources of pesticide exposure particularly for O. lignaria. However, local flower resources mitigated negative pesticide effects on B. vosnesenskii reproduction. Synthesis and applications. Bees in agricultural landscapes are threatened by pesticide exposure and loss of flower resources through agricultural intensification. Therefore, finding solutions to mitigate negative effects of pesticide use and flower deficiency is urgent. Our findings point towards flower plantings as a potential solution to support bee populations by mitigating pesticide exposure effects and providing key forage. Further investigation of the balance between forage benefits and added pesticide risk is needed to reveal contexts where net benefits occur.

Emerging stability of forest productivity by mixing two species buffers temperature destabilizing effect

Abstract: Forest ecosystems provide humankind with a wealth of ecosystem services. The necessary transition to a bioeconomy as a pathway towards the UN 2030 Sustainable Development Goals will require joint efforts in the near future to achieve a stable, sustainable flow of forest goods and services through nature-based solutions, the urgency of which is exacerbated by the pressure of accelerating climate change. In this context, the temporal stability of primary productivity is essential, not only for the current functioning of ecosystems (McCann, 2000) but also for securing the long-term provision of ecosystem services associated with productivity (Baeten et al., 2019). There is compelling evidence that tree species diversity may enhance forest primary productivity (Huang et al., 2018; Jactel et al., 2018; Liang et al., 2016) and its temporal stability (Jucker et al., 2014; Schnabel et al., 2019). The greater productivity is mainly due to tree species niche complementarity and facilitation which allow for a greater supply, capture or use of light, water and nutrient resources (Ammer, 2019). The stabilizing effect of species diversity can be explained by the insurance hypothesis, which states that species diversity insures against declines in functioning (i.e. greater stability) because some species will maintain functioning even if others fail (Yachi & Loreau, 1999). The temporal stability of productivity is usually estimated as the inverse of the coefficient of temporal variation (TS = μ/σ). Thus, species diversity may enhance temporal stability by increasing the mean productivity (μ) ‘performance-enhancing effect’ or by reducing the variance (σ) ‘buffering effect’ over time (Hector et al., 2010; Tilman et al., 1998; Yachi & Loreau, 1999). Overyielding (Figure 1) has frequently been used to evaluate the effect of species diversity on community productivity in forest ecosystems (Pretzsch & Forrester, 2017), although its effect on the temporal stability of productivity (performance enhancing effect) is unclear, with previous research providing contrasting results (del Río et al., 2017; Jucker et al., 2014; Schnabel et al., 2019). However, there is consensus that the buffering effect emerges mainly from asynchrony fluctuations in productivity among species (Loreau et al., 2021). The existing analyses of forest ecosystems also point to a major role of species asynchrony (del Río et al., 2017; Dolezal et al., 2020; Jucker et al., 2014; Morin et al., 2014; Schnabel et al., 2021; Yuan et al., 2019), although the underlying mechanisms remain still unclear. Species asynchrony in mixtures may result from asynchronous species-specific responses to environmental fluctuations and intrinsic rhythms, although other factors such as species interactions or demographical stochasticity may also play a part (Loreau et al., 2021; Loreau & De Mazancourt, 2013). The two first mechanisms could theoretically be inferred from species-specific behaviours in monospecific stands, so between-species asynchrony in the respective monocultures may indirectly explain part of the species diversity stabilizing effect (Figure 1). Beyond the underlying mechanisms, for forest management, it is important to infer to what extend mixing species can increase temporal stability of productivity. For this aim, it can be useful to estimate the average of species temporal stability in monocultures weighted by the species abundance in the mixture (Figure 1), referred to henceforth as ‘additive effect’ (Forrester & Pretzsch, 2015; Jourdan et al., 2021), since it reflects the reduction in variance that can result from averaging out variations in species productivities when species are added and behave as they would do in monocultures. Climate change can modify forest productivity (Pretzsch et al., 2014) and increase its variability due to the higher vulnerability to extreme climatic events (Anderegg et al., 2015; Choat et al., 2012). Several recent studies have revealed the impact of climate conditions on the temporal stability of ecosystem productivity and, more generally, on diversity–stability relationships (Garcia-Palacios et al., 2018; Hallett et al., 2014; Ma et al., 2017). Only a few studies have focused on this aspect for forest ecosystems (del Río et al., 2017; Jing et al., 2022; Jourdan et al., 2021; Ouyang et al., 2021), which still do not allow, a comprehensive understanding of climate as a driver of the temporal stability of forest productivity. Although it is accepted that productivity increases with the number of species following an asymptotic pattern (Liang et al., 2016; Tilman et al., 1997), it is becoming increasingly patent that the underlying mechanisms start as soon as two species are mixed in forest ecosystems (Pretzsch & Forrester, 2017). To what extent this pattern can be extended to temporal stability of productivity is of great relevance to forest ecosystems management given that (i) there are already large areas of mixed forests composed of two species; (ii) from a management perspective, transition from monospecific to two-species mixed forests is often a more realistic option than direct conversion to multispecies forests. In Europe, approximately 33% of the forested area is composed of monocultures (Van Brusselen et al., 2020). However, in relation to climate change, there are concerns about the ability of these forests to maintain their productivity over the long term. Recent observations suggest that monocultures are more vulnerable to biotic and abiotic hazards than mixed-species forests (Jactel et al., 2017; Knoke et al., 2008). In this study, we examine the temporal stability of stand growth (as an indicator of productivity) and its main drivers in monocultures and two-species mixtures along a climate gradient in Europe. We use a unique dataset of 261 plots including beech (Fagus sylvatica L.), oak (Quercus robur L. and Q. petraea [Matt.] Liebl.) or spruce (Picea abies [L.] H. Karst.) in combination with pine (Pinus sylvestris L.). All these tree species are native, productive and economically valuable for the European forestry sector, with the greatest values of growing stocks (Freudenschuss et al., 2020). In a previous study, we used 93 beech and pine plots to study temporal stability of productivity in mixed and monospecific stands at different organizational levels (tree, population, community; del Río et al., 2017). Here, we focus on the community level and specifically aim to: (i) assess the effect of two-species mixing on stand growth performance (overyielding) and temporal stability; (ii) evaluate direct and indirect effects of climate conditions on temporal stability; and (iii) determine which driver is the most relevant to explain temporal stability in mixtures. We hypothesize that: (H1) mixing two species results in overyielding and greater temporal stability of productivity; (H2) climate, overyielding and species asynchrony are the main direct drivers of temporal stability; (H3) the additive effect explains an important part of the temporal stability in two-species mixtures. The data used come from 87 triplets distributed in three transects with different tree species compositions across Europe (Figure 2). Each triplet was composed of a two-species mixed plot and two related monospecific plots of their component species growing under similar site conditions (261 plots). The studied species compositions were beech–pine (32 triplets), oak–pine (35 triplets) and spruce–pine (20 triplets). The triplets covered a wide range of climatic conditions, with mean annual temperatures ranging from 3.2 to 11.1°C and mean annual total precipitation from 502 mm to 1336 mm (more details in Table S1 in Supporting Information). Plots were established in mature, mostly monolayered, fully stocked stands without signs of recent thinning interventions. Plot sizes ranged from 0.02 to 1.55 ha, where the diameter of all trees was measured, and two increment cores per tree were taken at a 1.3 m stem height in a sample of approximately 20 trees per species and plot. Annual ring widths were measured and cross-dated using standardized dendrochronological techniques. See Pretzsch et al. (2015, 2020) and Ruiz-Peinado et al. (2021) for more details on field measurements and main stand characteristic calculations (Table S2). Fieldwork permits were obtained from the respective forest owners when required. Ethical approval was not required. The annual stand basal area increment (BAI, m2 ha−1 year−1) was used as an indicator of community productivity (del Río et al., 2017; Dolezal et al., 2020), as this variable can be precisely estimated from field measurements and in forest stands is highly correlated with net above-ground productivity. The studied period was 2000–2013 for the beech–pine transect and 2004–2017 for the oak–pine and spruce–pine transects, the last year corresponding to triplet establishment. The series of annual stand basal area increments for the studied period (14 years) were estimated based on measured tree diameters and tree ring width series. Using data from cored trees, tree diameter increment–diameter models were fitted by year, species and plot to estimate annual diameter increments of noncored trees. Annual climate data were obtained from meteorological weather stations located in the proximity of each triplet (50 triplets). When local station data were not available, national digital climatic atlas data (24 triplets) or more general gridded data (13 triplets) were used (see Table S1). For analyses, we considered the average mean annual temperature (T), total annual precipitation (P) and the Martonne aridity index (M = P/[T + 10]) (Martonne, 1926) during the studied period, as they describe the large variability of climates in the study area (Figure 2), and are related to productivity variation at large scales in a simple way (Huang & Xia, 2019). In a preliminary analysis, we also explored the effects of the standard deviation and coefficient of variation of climate variables (Craven et al., 2018), but finally we discarded them due to their high correlations to the mean values (Figure S1). The asynchrony in a mixed plot (Asynmixed) was estimated following the synchrony metric proposed by Gross et al. (2014), that is, the average across species of correlation coefficients, which for two species mixtures results in the correlation coefficient between the two species growth series during the 14-year period, Asynmixed = 1 − corr (BAIsp1_mixed, BAIsp2_mixed). Additionally, we estimated in a similar way the asynchrony between the two species growing in monospecific stands (Figure 1) (Asynmono = 1 − corr (BAIsp1_mono, BAIsp2_mono)). We also tested the metric community-wide asynchrony (Loreau & de Mazancourt, 2008), but the results were very similar as the two asynchrony metrics were highly correlated, r = 0.93. To test whether OY was significantly greater than 1, we used a one-sided Student's t-tests. We tested mean OY of all triplets and by transect, as well as the relative productivity by species and by transect. The effect of drivers other than climate on TS in mixed stands (TSmixed) was explored by simple linear regression using only data of mixtures. As potential drivers, we tested Asynmixed and OY. We further explored the relationships between TSmixed and additive effect (AE). To unravel the direct and indirect effects of climate and other drivers on TSmixed, we applied structural equation modelling (SEM) (Shipley, 2016), using data of mixtures. As direct drivers of TSmixed, we included climate variables, OY and Asynmixed (Hector et al., 2010; Loreau & De Mazancourt, 2013; Figure 1). We also considered AE as direct factor, which may explain the effect of averaging species-specific variability. We included a path from Asynmixed to OY, since a positive relationship has been previously reported (Allan et al., 2011), which could represent an indirect effect of Asynmixed on TS. We also assumed that Asynmono reflects the asynchrony in species-specific responses to environmental fluctuations and intrinsic rhythms in a given site, so it may explain a large part of Asynmixed and AE. Accordingly, as Asynmixed and the additive effect could covary, we included the covariance between them in the model. Moreover, we expected that climate could influence TS indirectly through changes in overyielding (Jactel et al., 2018) and in Asynmixed (Ma et al., 2017). The preliminary analyses of the effect of different drivers on TS indicated that both temperature and precipitation modulated TS and that overyielding did not have any effect. Thus, overyielding was removed in the final fitted SEM to reduce paths due to the limited number of data from mixed stands (n = 87; Figure S2). All endogenous variables were log-transformed to achieve normality. We fitted the SEM based on a maximum likelihood method and used the χ2 test, the comparative fit index (CFI) and standardized root mean square residual (SRMR) to evaluate the fit of the model. SEM fitting was performed using the R package lavaan (Rosseel, 2012). We found that the overall mean overyielding across mixture types was 1.062 (ln(OY) > 0 p-value = 0.0446), that is, growth was 6.2% greater in mixtures than expected by the growth in monocultures. Overyielding (in %) was 9.5% (ln(OY) > 0 p-value = 0.0511) in the beech–pine mixture, 5.6% (OY >1 p-value = 0.0329) in the oak–pine mixture and 2% (OY >1 p value = 0.2195) in the spruce–pine mixture, with high variability within each mixture (Figure 3, Table S3, Figure S3). Beech, oak and spruce significantly benefited from the mixture by increasing their growth, while pine showed similar (beech–pine and oak–pine) or slightly lower (spruce-pine) growth than did the monocultures (Table S3). The temporal stability of stand growth was 12.3% greater in mixed (TSmixed = 6.12) than in monospecific (TSmono = 5.45) stands (p value = 0.0016, Table S4a, Figure S4). There were no significant differences between the three species mixtures (Table S4e). TS was significantly improved for both beech and pine in the beech–pine mixture, for only pine in oak–pine mixtures and only spruce in spruce–pine mixtures (Figure 3, Table S4b–d). Increasing the mean temperature had an overall negative effect on TS (p < 0.0022) in both monospecific and mixed stands, indicating a greater variability in stand growth at warmer sites (Figure 4a, Table S5a,b). In contrast, the effect of annual precipitation was only significant when the identity of species composition was considered in the analysis (Table S5b). The TS of beech–pine mixtures slightly increased with higher precipitation, whereas that of spruce–pine decreased (Figure 4). The stability of the oak–pine mixture growth was not affected by precipitation. The stabilizing effect of mixing species (i.e. difference between mixed and monospecific stands) was stronger under higher precipitation for the beech–pine mixture (Figure 4b) and weaker for spruce–pine (Figure 4d). When comparing the three mixtures, there were only slight differences among the precipitation effect on their TS (Table S5c; Figure S5). The effect of climate variables on overyielding was also tested, but we did not find any significant effect. Simple linear regressions (Table S6) indicated that among the different drivers tested, the additive effect had the strongest relationship (positive) with TSmixed (R2 = 0.34, p < 0.0001). Asynmixed was also positively correlated with TSmixed (R2 = 0.21, p < 0.0001), while overyielding did not show any relation. Interestingly, Asynmixed had a significant effect on overyielding. The results of the SEM confirmed the direct negative effect of temperature (−0.21) and reflected the relevance of species asynchrony on TSmixed (Figure 5; Table S7). Precipitation had an indirect weak influence (0.06 = 0.25 × 0.24) on TSmixed through its positive effect (0.25) on Asynmixed. Asynmono indirectly explained TSmixed by the effect on Asynmixed and by AE (Figure 5, Table S7). It is noteworthy that Asynmixed increased TSmixed beyond the AE. This reflects the effect of species interactions, which may modify the species fluctuations in mixed stands in comparison to monospecific stands, emerging in more stable forest stands. However, the AE effect was greater (0.50) than the direct effect of Asynmixed (0.24). Based on observational data of three relevant mixtures with Scots pine and another tree species across Europe, we demonstrate that adding one species to monocultures yields important benefits in terms of the level and stability of community productivity. We further reveal the negative effect of temperature on TS and the relative importance of different factors acting on the stability gain. The significant mean OY and greater stability in mixtures confirm results from previous overarching analysis that include two-species forest mixtures (Jactel et al., 2018; Jucker et al., 2014; Pretzsch & Forrester, 2017), as well as a previous analysis on beech–pine transect (del Río et al., 2017). It highlights general complementarity and buffering effects in mixtures in terms of stand growth. However, the magnitude of mixtures’ benefits can vary with species composition (Figure 3), which suggests the importance of species traits on diversity effects (Baeten et al., 2019; Craven et al., 2018; Schnabel et al., 2021) and the need to assess specific species compositions. Previous results from experimental studies point to a high relevance of climate conditions for TS (Craven et al., 2018; Schnabel et al., 2021), but their results cannot be easily generalized to other sites. Our observational approach allowed us to address for the first time the effect of climate on TS for specific forest mixtures along their main distribution range. We identify the destabilizing effect of temperature on stand productivity for all the studied forest types, as found for other plant communities (Ma et al., 2017; Valencia et al., 2020). Greater temperatures may be linked to sites where the species show greater climate sensitivity, which might increase growth variability. Nonetheless, the large variability observed (Figure 4a) points to the need for a deeper analysis considering monthly climate variables and local growing seasons to clarify the reasons for the observed temperature effect. Although the temperature destabilizing effect was common for monocultures and mixtures, the positive effect of mixing species on TS may counterbalance the negative effect of temperature. The greater TS in mixed stands was, on average, equivalent to the TS of corresponding monocultures under ≈ 2°C lower temperature (Figure S6), although uncertainty is large. In accordance with other studies (Ouyang et al., 2021; Valencia et al., 2020), TS was also modulated by annual precipitation, but we found distinct effects depending on species identities in monospecific and mixed stands (Figure 4) (Jourdan et al., 2021). We did not find any climate influence on OY, as found in previous analyses of beech–pine and spruce–pine transects for a shorter period (Pretzsch et al., 2015; Ruiz-Peinado et al., 2021), but against those of the oak–pine transect (Pretzsch et al., 2020). Nevertheless, precipitation may have an indirect positive influence on OY, as found by Jactel et al. (2018), through its effect on Asynmixed (Table S6). Asynchrony in species productivities is often reported as the main driver of the greater TS with increasing species diversity (Blüthgen et al., 2016; Schnabel et al., 2021; Yuan et al., 2019). By exploring Asynmixed and Asynmono, we demonstrated the relevant effect of asynchrony through two complementary mechanisms (Figure 5), that is, the difference between intrinsic species-specific fluctuations (responses to climatic variations and intrinsic rhythms) and species interactions, which results in the stabilizing effect of mixing species as predicted by the insurance hypothesis (Yachi & Loreau, 1999). Accordingly, our results reveal the substantial influence of AE on TS (Jourdan et al., 2021), which suggests the potential stabilizing effect by mixing any tree species (van der Plas, 2019), even when mixed by patches (i.e. spatial stability, (Loreau et al., 2021)). Asynmixed also had a positive effect on OY, which indicates the presence of temporal niche complementarity and points asynchrony in species growth as a key driver of forest ecosystem functioning (van der Plas, 2019). The lack of influence of OY on TS reveals that TS increased by the variance buffering effect (Schnabel et al., 2021) and that OY and TS were independent effects (Cardinale et al., 2013; Jing et al., 2022). However, caution is needed for causal interpretation of our SEM results given the observational, not experimental, approach. We found that on average, in two-species stands, growth was 6% higher than expected and that temporal stability was 12% higher than in monospecific stands. Although the level and stability of productivity may increase with the number of species (Liang et al., 2016; Schnabel et al., 2021; Vilà et al., 2013), we demonstrate that adding one additional species to monocultures has already a strong effect. Monocultures of conifers, such as pine and spruce stands, are efficient systems for timber supply because of their high growth rates and simplified management. However, admixing just one species in these stands could stabilize the provision of wood and other ecosystem services linked with the level and stability of growth, such as nutrient and water cycling, carbon sequestration and storage or protective functions (Knoke et al., 2008), beyond the reduction and distribution of risks under higher climate uncertainty (Jactel et al., 2017). These findings underline that promoting two-species mixtures can be a realistic and effective nature-based climate solution, supporting the sustainability of forest productivity and contributing further to climate change mitigation (Mori et al., 2021). Miren del Río, Hans Pretzsch and Andrés Bravo-Oviedo conceived the ideas and designed methodology; Hans Pretzsch, Ricardo Ruiz-Peinado, Magnus Löf, Jorge Aldea, Mathias Steckel and Michael Heym compiled and elaborated the data; Miren del Río, Ricardo Ruiz-Peinado and Charlotte Poeydebat performed the analyses; Miren del Río led the writing with inputs from Hans Pretzsch, Ricardo Ruiz-Peinado, Hervé Jactel, Luís Coll, Magnus Löf, Sonia Condés, Andrés Bravo-Oviedo; Miren del Río, Hans Pretzsch, Ricardo Ruiz-Peinado, Lluís Coll, Magnus Löf, Jorge Aldea, Christian Ammer, Admir Avdagić, Ignacio Barbeito, Kamil Bielak, Felipe Bravo, Gediminas Brazaitis, Jakub Cerný, Catherine Collet, Lars Drössler, Marek Fabrika, Michael Heym, Stig-Olof Holm, Gro Hylen, Aris Jansons, Viktor Kurylyak, Fabio Lombardi, Bratislav Matović, Marek Metslaid, Renzo Motta, Thomas Nord-Larsen, Arne Nothdurft, Jan den Ouden, Maciej Pach, Marta Pardos, Quentin Ponette, Tomas Pérot, Ditlev Otto Juel Reventlow, Roman Sitko, Vit Sramek, Mathias Steckel, Miroslav Svoboda, Kris Verheyen, Sonja Vospernik, Barbara Wolff, Tzvetan Zlatanov contributed data. All authors contributed critically to the drafts and gave final approval for publication. This study was supported by the projects REFORM (ERA-Net SUMFOREST, PCIN2017-026/-027, MICIN, Spain), CARE4C (Marie Skłodowska-Curie No 778322, HORIZON2020) and CLU-2019-01 - iuFOR-UVa and VA183P20-SMART. J.C. was supported by the National Agency of Agricultural Research (Project No. QK21020307); K.B. by the Polish Government MNiSW 2018–2021 Matching Fund No. 117/H2020/2018; M.M. by Estonian Research Council grant (PRG1586), and EMÜ Projects P180024MIME, P200029MIME; R.S. by the Slovak Research and Development Agency, project No. APVV-18-0390. The authors declare no conflicts of interests. Data available via the Dryad Digital Repository https://doi.org/10.5061/dryad.0rxwdbs3r (del Río et al., 2022). Data S1 Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

A review of existing and potential blue carbon contributions to climate change mitigation in the Anthropocene

Abstract: The atmosphere concentration of CO2 is steadily increasing and causing climate change. To achieve the Paris 1.5 or 2 oC target, negative emissions technologies must be deployed in addition to reducing carbon emissions. The ocean is a large carbon sink but the potential of marine primary producers to contribute to carbon neutrality remains unclear. Here we review the alterations to carbon capture and sequestration of marine primary producers (including traditional ‘blue carbon’ plants, microalgae, and macroalgae) in the Anthropocene, and, for the first time, assess and compare the potential of various marine primary producers to carbon neutrality and climate change mitigation via biogeoengineering approaches. The contributions of marine primary producers to carbon sequestration have been decreasing in the Anthropocene due to the decrease in biomass driven by direct anthropogenic activities and climate change. The potential of blue carbon plants (mangroves, saltmarshes, and seagrasses) is limited by the available areas for their revegetation. Microalgae appear to have a large potential due to their ubiquity but how to enhance their carbon sequestration efficiency is very complex and uncertain. On the other hand, macroalgae can play an essential role in mitigating climate change through extensive offshore cultivation due to higher carbon sequestration capacity and substantial available areas. This approach seems both technically and economically feasible due to the development of offshore aquaculture and a well-established market for macroalgal products. Synthesis and applications. This paper provides new insights and suggests promising directions for utilizing marine primary producers to achieve the Paris temperature target. We propose that macroalgae cultivation can play an essential role in attaining carbon neutrality and climate change mitigation, although its ecological impacts need to be assessed further.

Seasonal progression and differences in major floral resource use by bees and hoverflies in a diverse horticultural and agricultural landscape revealed by DNA metabarcoding

Abstract: The decline in pollinating insects is well documented globally, leading to potentially severe impacts on floristic biodiversity and human health due to the loss of pollination ecosystem services (Klein et al., 2007; Lundgren et al., 2016; Smith et al., 2015). Pollinator declines have occurred due to a combination of habitat loss, climate change, pests and diseases and the use of pesticides (Potts et al., 2010). As the availability of floral resources limits pollinators (Goulson et al., 2015), understanding foraging preferences is a key knowledge need for their effective conservation. Gardens are important, heterogeneous habitats, covering significant areas in urban landscapes (Loram et al., 2007). Gardens can provide pollinators with pollen, nectar and nesting sites (Osborne et al., 2008), supporting pollinators in agricultural (Timberlake et al., 2020) and urban (Potter et al., 2019) settings while increasing habitat connectivity within the landscape (Goddard et al., 2009). The limited number of studies in the United Kingdom (Wignall et al., 2019) and elsewhere in Northern Europe (Schonfelder & Bogner, 2017) on the public perception of pollinators suggests that attitudes towards their conservation is very positive. However, while there is a wealth of information available on the best plants for pollinators, only a small number of recommendation lists are based on empirical evidence (Garbuzov & Ratnieks, 2014), with most plants sold in UK garden centres relatively unattractive to flower-visiting insects (Garbuzov et al., 2017). Moreover, these lists broadly target pollinators, leading to generalisation across a wide range of functional groups and species. Consequently, there is a clear need to provide scientific evidence for effective floral use in gardens to support pollinators. Although foraging can vary between pollinator groups (Bänsch et al., 2020), most studies in gardens focus on a single group (de Vere et al., 2017). Honeybees and bumblebees are the most frequently studied, however, non-corbiculate bees and hoverflies have important roles in pollination and ecosystem function (Klein et al., 2007). Additionally, seasonality and annual variation can influence forage choice (Petanidou et al., 2014), highlighting the need to provide information on floral use throughout the year. There are conflicting perspectives as to whether native or non-native plants are preferred by pollinators, but it is imperative to understand this for effective conservation. When surveying pollinator visits to a variety of plants, Salisbury et al. (2015) found a greater abundance of pollinators on native and near-native taxa than those defined as exotic. Additionally, introduced plant species have been shown to attract fewer species of flower visitors than natives and those closely related to natives (Memmott & Waser, 2002). DNA metabarcoding has been used to identify pollen within honey (de Vere et al., 2017; Jones, Brennan, et al., 2021), from the bodies of insects (Lucas et al., 2018a; Richardson et al., 2021), and from brood provision in nests (Vaudo et al., 2020). The advantages of pollen metabarcoding approaches include increased taxonomic resolution (Brennan et al., 2019) and the elimination of the taxonomic expertise required for pollen microscopy (Hawkins et al., 2015). DNA metabarcoding overcomes the limitations of observational methods by revealing interactions previously unseen due to spatial and temporal limitations (Arstingstall et al., 2021), however, it must be accompanied by a comprehensive reference library to ensure accurate identification. In the United Kingdom, the Barcode UK project provides 98% coverage of all native flowering plants and conifers using three plant DNA barcode markers, rbcL, matK and ITS2, allowing reliable identification at the species and genus level for the majority of plants (de Vere et al., 2012; Jones, Twyford, et al., 2021). Bees and hoverflies were sampled monthly from March to October during 2018 and 2019 at the National Botanic Garden of Wales, UK (51°50′33.4″N 4°08′49.2″W). The site is a diverse landscape (230 ha) set within a predominately semi-improved (based on the extent of agricultural improvement) landscape and consists of formal garden and organic farmland, designated as a National Nature Reserve,Waun Las NNR (Figure 1). The Botanic Garden contains over 5,000 plant taxa from throughout the world, including many horticultural plants grown throughout Western Europe. Eight areas were selected for pollinator sampling covering broadleaved woodland and hedgerows, horticultural and grassland habitat. The most abundant plants across both years per transect area per season are provided in Appendix S1. Within each sampling area, a 210 m × 2 m transect was established and divided into 3 × 70 m sections, walked independently of each other. Transect walks were preferentially undertaken between 11:00 and 15:00 when the temperature was over 10°C. When this was not possible, transects were walked on dry days with little wind. All bees and hoverflies seen on the transect were caught individually and stored at −20°C prior to pollen removal. Further information on field sampling is provided in Appendix S2. Permission for field work and ethical approval was granted by the National Botanic Garden of Wales. Pollen was washed from insects following a modified version of the protocol described by Lucas et al. (2018b). Insects were first transferred to a sterile 1.5 ml collection tube using sterile forceps and cleaned with 70% ethanol between each insect. The tube used to catch insects was washed with 1 ml of 1% sodium dodecyl sulphate (SDS) and 2% polyvinylpyrrolidone (PVP) solution, ensuring any pollen residue on the sides was collected and transferred to the tube containing the insect. Samples were shaken using a TissueLyser II (Qiagen) for 1 min at 8.5 Hz, stood at room temperature for 5 min, then shaken again for 20 s at 8.5 Hz. Each insect was removed using sterile forceps and placed into a 1.5 ml microcentrifuge tube containing 70% ethanol, prior to species identification (see Taxonomic assignment of insects, Appendix S2). The tube containing the detergent and pollen pellet was centrifuged at 16,200 g for 5 min and the supernatant removed. The pollen pellet was resuspended in 400 μl buffer, made up of 400 μl AP1 from the DNeasy 96 Plant Kit (Qiagen) and 80 μl (1 mg/ml) of Proteinase K (Qiagen). A modified version of the DNeasy 96 Plant Kit was used for DNA extraction. Samples were incubated in a water bath at 65°C for 1 hr and 1 μl RNase (Qiagen) added before disruption using a TissueLyser II for 4 min at 30 Hz with 3 mm tungsten carbide beads. The remaining steps were carried out according to the manufacturer's protocol, excluding the use of the QIAshredder and the second wash stage. A negative control was included within each extraction. Two barcode regions, rbcL and ITS2 were amplified via a two-step PCR protocol (Table S1, Appendix S3). The initial PCR used a final volume of 20 μl: 2 μl template DNA, 10 μl of 2× Phusion Hot Start II High-Fidelity Mastermix (New England Biolabs UK), 0.4 μl (2.5 μM) forward and reverse primers, and 7.2 μl of PCR grade water. Each PCR was repeated twice more and pooled before purification using the Illumina 16S metabarcoding protocol, with a 1:0.6 ratio of product to Agencourt AMPure XP beads (Beckman Coulter). The purified product was amplified further to anneal custom unique and matched i5 and i7 indices to each sample (Ultramer, Integrated DNA Technologies). This second stage PCR used a final volume of 25 μl: 5 μl of purified first-round PCR product, 12.5 μl of 2x Phusion Hot Start II High-Fidelity Mastermix (New England Biolabs UK), 1 μl of i5 and i7 Index Primer and 6.5 μl of PCR grade water. All thermal cycling conditions are available in Appendix S2. Tag addition was confirmed with visualisation on a 1% agarose gel. A second Illumina clean-up stage was followed with a 1:0.8 ratio of product to beads. Products were quantified using a Qubit 4.0 (Thermo Fisher Scientific) and pooled at equal concentrations. The negative extraction and PCR controls from each plate were sequenced with the pollen samples on an Illumina MiSeq (2 × 300 bp) at Liverpool University's Centre for Genomic Research (Liverpool, UK). Laboratory contamination controls can be found in Appendix S2. Sequence reads were processed following Ford and Jones (2020). Initially, raw sequences were trimmed to remove low-quality regions, paired and merged. Only sequences greater than 450 bp (rbcL) and 350 bp (ITS2) were used in downstream analysis. Identical reads were dereplicated within each sample and clustered at 100% identity across all samples with singletons (sequence reads occurring once across all samples) removed. Sequences were compared to a custom reference library containing 5887 plant species (Jones, Brennan, et al., 2021), comprising native plants of the United Kingdom (Stace, 2019), naturalised and alien species (Preston et al., 2002) and horticultural species from the IRIS BG database at the National Botanic Garden of Wales. Sequences were compared against the reference library using blastn, summarising the top 20 BLAST hits and combining all sequences with identical BLAST results across all 20 hits. Sequences with bit scores below the first percentile were excluded. Sequences were assigned so that if the top bitscore matched a plant species, the sequence was assigned to that species. If the top bitscore matched different species within the same genus, the sequence was assigned to that genus. If the top bitscore belonged to multiple genera of the same family, then a family designation was made for that sequence. Sequences returning top bitscores of multiple families within different orders were removed, assuming that these were poor quality sequences. The botanical veracity of the plants identified across all insect samples was assessed by considering whether those plants were present within the study site and wider landscape. Taxonomic assignment of sequences was compared between markers on a sample-by-sample basis for further verification. Once the identifications were complete, a consensus identification was reached to combine the taxa identified by both markers at differing taxonomic resolution using a rule-based, objective, and conservative decision process (see Using rbcL and ITS2 markers, Appendix S2). The number of rbcL and ITS2 sequences for each consensus taxon within a sample were then summed to combine the results of each marker. Sequences assigned to taxa identified using one marker alone were retained. Plants identified to genus and species were assigned to a status category following Stace (2019). The category ‘native and near native’ comprised native species and also genera that include native species and horticultural varieties which are functionally similar. Naturalised plants were those which have been introduced and become widespread and self-perpetuating in the wild. All remaining non-native plants were classified as horticultural. The DNA metabarcoding data were treated as semi-quantitative with relative read abundance used for all analyses (Deagle et al., 2019), either using the proportion of taxa as a percentage or, for the models, the number of sequences, controlling for sequencing depth by setting the total number of sequences per sample as an offset, comparable to proportion (Jones, Brennan, et al., 2021; Appendix S2). Using the package mvabund (Wang et al., 2012), a multivariate generalised linear model with a negative binomial distribution was used to understand how pollen load composition changed through time. The data best fit a negative binomial distribution due to the strong mean–variance relationship (Figure S1, Appendix S3), likely from distributions of rare taxa where mean abundance is low, a common observation in multivariate abundance data. To understand the effect of time and pollinator type on plant composition, the effect of season (coded as 1–3, starting with spring), year and pollinator group/order was included as predictor variables, with the number of sequence reads for each plant taxon set as the multivariate response variable. The number of reads per sample was included as an offset to control for differences in sampling depth (Deagle et al., 2019; Jones, Brennan, et al., 2021). Seasonal changes in the composition of pollen loads were visualised using non-metric multidimensional scaling (NMDS) ordination of Bray–Curtis dissimilarity indices (based on the proportion of reads returned for each plant taxa), using the vegan package (Dixon, 2003). A Chi-square contingency test was used to investigate differences in major taxa (constituting over 5% of sequences) between pollinator orders (based on the relative read abundance overall), with Holm correction for multiple testing. Each pollinator group was split into categories based on a unique ecological functional trait (see Functional diversity analysis, Appendix S2) and Chi-square contingency tests were used to investigate differences in taxa constituting over 1% of sequences between functional categories within broader groups. To investigate the change in use of native plants over time, the plant taxa were grouped by their status categories. A multivariate generalised linear model was run, with season and year included as predictor variables and the response variable being the number of reads, retaining the use of the offset. All statistical analyses were carried out in R v 4.0.2 using the consensus identification. Analysis of rbcL and ITS2 was also carried out separately to support the use of combining markers (Appendix S4). Throughout the study, 382 insects were caught with successful sequencing of pollen from 369 individuals (Table 1). No insects were caught in October despite surveys being carried out. Pollinators were grouped into hoverflies (Syrphidae, n = 195), bumblebees (Bombus spp., n = 108), honeybees (Apis mellifera, n = 44) and all other non-corbiculate bees (n = 22; Table S2, Appendix S3). A total of 40,800,709 reads were returned with 22,510,682 remaining after stringent quality control (11,305,697 rbcL and 11,204,985 ITS2). Using the rbcL and ITS2 regions combined, 191 plant taxa were identified with the majority of taxa identified at genus level (Appendix S1). Six taxa were found on over 50% of insects sampled (Figure 2): bramble (Rubus spp.), thistles, knapweeds and cat's ear (Cirsium/Centaurea/Hypochaeris spp.), buttercups and lesser celandine (Ranunculus/Ficaria spp.), angelica and hogweed (Angelica/Heracleum spp.), daisy family (Asteraceae) and meadowsweet (Filipendula ulmaria). An average of 17 (SD = 9.76) plant taxa were found on each individual insect with an average of 4 (SD = 2.55) taxa contributing >1% of reads (Table 1). Overall, we found little variation in foraging habits between the four pollinator groups. Neither pollinator group nor pollinator order predicted pollen composition when all plant taxa were included in the model (pollinator group: LR1,363 = 1753.8, p = 0.999, order: LR1,365 = 953.9, p = 1.000). The ability of the model to predict pollen composition was greatest when characterising pollinators by their taxonomic order (Diptera, Hymenoptera) rather than group (bumblebees, honeybees, non-corbiculate bees and hoverflies; Table S3, Appendix S3). There was, however, a significant difference in the composition of plant taxa constituting over 5% of sequences carried by Diptera and Hymenoptera (x2 = 46.26, df = 5, p < 0.001; Figure 3). A large proportion of pollen sequences from hoverflies (Diptera) belonged to Angelica/Heracleum spp., but these were not found to be as valuable for bees (Hymenoptera). Cirsium/Centaurea/Hypochaeris spp. contributed a large proportion of sequences for bees but made up a lower proportion of sequences for hoverflies, while hoverflies used Ranunculus/Ficaria spp. more abundantly than bees. Within pollinator groups, differences in foraging were found between ecological functional categories (Figure 4). A significant difference was found in the composition of plant taxa represented by over 1% of sequences from short- and long-tongued bumblebees (x2 = 50.179, df = 20, p < 0.001). A large proportion of pollen was attributed to Ranunculus/Ficaria spp. across short-tongued species (Bombus hypnorum, B. lapidarius, B. pratorum, B. lucorum/terrestris agg.) while long-tongued species (B. hortorum, B. pascuorum), utilised more Trifolium pratense and Rubus spp. (Figure 4). Honeybees' foraging habits were broadly similar to bumblebees but utilised a greater proportion of Impatiens glandulifera than any other group (Figure 4). Within the non-corbiculate bees, the total proportion of pollen collected was significantly different between body size groups (x2 = 433.01, df = 52, p < 0.001), with extra small bees carrying mostly Heuchera spp., small-sized carrying mostly Rudbeckia/Helenium spp. and medium-sized carrying mostly Taraxacum officinale. Pollen composition from hoverfly species differed between various larval requirements (x2 = 235.4, df = 48, p < 0.001), with carnivorous and detritivorous species utilising a greater diversity of plant taxa than herbivorous species (Figure 4). Season was a good predictor of pollen composition (LR2,367 = 2632.8, p < 0.001), regardless of year of sampling (LR2,366 = 816.2, p = 0.828; Figures S2–S4, Appendix S3). There were 147 taxa found in 2018 and 170 in 2019, and of these 71 were identified in both years. NMDS ordination scaling shows that pollen samples collected in the same season are most similar to each other (Figure 5). Seasonal progression is visible for each pollinator group when assessing the most abundantly foraged plants throughout the year (Table 2) using the consensus data and rbcL and ITS2 separately (Figures S5–S7, Appendix S3). Ranunculus/Ficaria spp. Rubus spp. Cirsium/Centaurea/Hypochaeris spp. Rubus spp. Aster spp. Clematis spp. Rubus spp. Ranunculus/Ficaria spp. Taraxacum officinale Cirsium/Centaurea/Hypochaeris spp. Impatiens glandulifera Rubus spp. Actaea spp. Heuchera spp. Impatiens glandulifera Geum spp. Ranunculus/Ficaria spp. Taraxacum officinale Bidens/Coreopsis spp. Heuchera spp. Rudbeckia/Helenium spp. Hoverfly Ranunculus/Ficaria spp. Angelica/Heracleum spp. Cirsium/Centaurea/Hypochaeris spp. Rubus spp. Angelica/Heracleum spp. Bidens/Coreopsis spp. Rudbeckia/Helenium spp. The largest proportion of DNA reads returned from pollinators were attributed to native and near-native plants (Figure 6). Native and near-native plants were predominately used in the spring and the use of naturalised and horticultural plants increased during the summer and autumn (LR1,367 = 58.10, p = 0.001) (Figure 6), regardless of year of sampling (LR1,366 = 3.14, p = 0.369). In transects with more horticultural plants, we see that pollinators use a diverse array of plants with no dominant taxa identified (Figure S9, Appendix S3), compared to predominately native areas (Figures S8, S10–S12, Appendix S3). Using DNA metabarcoding, we reveal the most frequently visited plants by key pollinator groups, across a broad taxonomic range covering bumblebees, honeybees, non-corbiculate bees and hoverflies. We show that while common resources are shared across all groups, differences are seen in the major taxa visited by hoverflies (Diptera) and bees (Hymenoptera) and between ecological functional categories within. This choice in foraging is strongly influenced by season, with clear changes in floral use through the year. Pollinators were shown to predominately utilise native and near-native plants, with increased use of horticultural and naturalised plants towards the end of the season. Floral resources were shared overall among pollinator groups (hoverflies, bumblebees, honeybees and non-corbiculate bees), but clear differences were seen between the taxa used most abundantly by Diptera (hoverflies) and Hymenoptera (bees). In comparison to hoverflies, bees utilised thistles more and umbelliferous plants less. A possible explanation for the preference differences between the major plants of Diptera and Hymenoptera is that the accessibility of nectar may be limited by the morphology of the plants, influencing which plants are visited by pollinators. The hoverflies recorded here generally have shorter tongues than bees (King, 2012), and may have difficulty fully removing nectar from the long corollas found in the genera Cirsium, Centaurea and Hypochaeris. While hoverflies are evidently able to utilise this resource, the issue of accessibility may be a reason for hoverflies prioritising the shorter, open flowers of Angelica/Heracleum spp. We demonstrate that within broad pollinator groups, resources may be partitioned further based on ecological functional traits shared by species. When studying the diversity and abundance of pollen collected by insects, Cullen et al. (2021) found that traits had a greater impact than local floral diversity itself, highlighting the importance of understanding this relationship. Tongue length is widely known to affect forage choice in bumblebees and is thought to influence species' vulnerability to extinction as long-tongued species tend to specialise more on species with long corollae (Goulson et al., 2005). While we did find differences in forage relating to bee size in non-corbiculate bees, the small sample size and long sampling period mean these results must be interpreted with caution and further work is required. Non-corbiculate bees comprise a cosmopolitan suite of ecologically distinct taxa in the United Kingdom. However, this study was limited to bees within Halictidae and Andrena, along with the kleptoparasitic Nomada which are all relatively small (thoracic width < 3 mm) making comparisons within this group difficult. As body size limits the foraging distance of bees (Greenleaf et al., 2007), the floral resources used by these species may have been predicted by the species immediately available to them in the area sampled. The relationship with floral resources is more complex in hoverflies, as larval requirements influence the habitats which species occupy (Schirmel et al., 2018) although the link between these requirements and floral resources used is little studied. While we identified differences in floral resource use between these functional guilds, we also highlight that hoverflies use plants for mate seeking, therefore additional work is required to fully understand which plants are being used for food, breeding sites or oviposition in phytophagous species (Rotheray & Gilbert, 2011). Season of collection was found to be the biggest predictor of plant use, with pollinators relying on key plants within each season (Appendix S5). The phenological patterns of plants result in shifting of floral availability, temporally limiting the foraging habits of insects. These shifts in available resources require pollinators to alter their use of resources throughout the season to survive, with those with long flight periods utilising a greater diversity of plant taxa than those with short flight periods (Ogilvie & Forrest, 2017). Pollinators use native and near-native plants more often than non-native plants, however, the non-native plants play a key role at the end of the flowering season. These findings are supported by Salisbury et al. (2015) who showed that native and near-native plants attracted a greater number of pollinators than non-native plants in a garden, however, the non-native plants extended the flowering period. The greater use of naturalised plants in summer and autumn can be attributed to the high use of Impatiens glandulifera by honeybees, highlighting the importance of this species for nectar provision. However, such an observation comes with a broader conservation caveat since I. glandulifera is a highly invasive, non-native plant and so it must not be grown due to its ability to displace other plant species (Chittka & Schürkens, 2001). While a lower proportion of non-native plants were used compared to native and near-native plants, they may contribute by increasing the diversity of pollinator diets. For example, Taraxacum officinale is used abundantly in the spring, however, it must be supplemented with additional resources as it lacks essential amino acids needed for pollinator health (Génissel et al., 2002). The multi-locus metabarcoding approach used here allows the relationship between plants and pollinators to be studied on a fine scale, improving both the number of plant taxa that can be detected and the level of discrimination achievable with the use of one marker alone (Jones, Twyford, et al., 2021) or alternative methods (Brennan et al., 2019). We highlight the ability of DNA metabarcoding to not only provide a greater depth of information, but also to support knowledge provided by traditional techniques, for example here the frequent use of taxa with large open inflorescences by hoverflies (Branquart & Hemptinne, 2000). Due to potential biases in sampling, along with extraction, amplification and sequencing of DNA (Bell et al., 2016), the data should be treated as semi-quantitative, with the abundance of DNA reads treated as estimates of relative abundance (see Analysing DNA metabarcoding data using semi-quantitative approaches, Appendix S2). Frequent taxa may be over represented, and rare taxa more difficult to detect, however, this is also the case using pollen microscopy (Hawkins et al., 2015). Recent developments suggest that in some cases metabarcoding data may be quantitative (Richardson et al., 2021), particularly regarding the most abundant taxa in a sample (Bänsch et al., 2020), however, further work is needed to fully understand this relationship (Piñol et al., 2019). Furthermore, species-level discrimination in plants using DNA metabarcoding is challenging due to no single marker meeting the requirements for an ideal barcode (CBOL Plant Working Group, 2009). While genus-level designations have limitations in understanding fine-scale plant–pollinator interactions, these provide a conservative approach to identification, using the most universal and discriminative plant DNA markers available, to provide accurate taxonomic information across a wide study scale (Jones, Twyford, et al., 2021). Our conclusions therefore focus on the plants most abundantly used by pollinators, and how we can provide these in gardens and wider landscapes. The results of this study allow us to provide an evidence-based plant recommendation list to support a range of pollinators throughout the season including native and horticultural plants across a range of growth forms (Table S4, Appendix S3). We improve on previous lists by providing foraging information from the perspective of the insect, increasing both the temporal and spatial scope possible compared to using observations of plants (Arstingstall et al., 2021). This recommendation list is based on taxa found within the United Kingdom, with relevance to Northern Europe and can be used by gardeners, land managers, plant producers and policy makers to inform decisions on planting within gardens and urban greenspace to ensure pollinators are appropriately supported. N.d.V., L.J. and A.L. have received funding through the Welsh Government Rural Communities – Rural Development Programme 2014–2020, which is funded by the European Agricultural Fund for Rural Development and the Welsh Government. A.L. was supported by a Knowledge Economy Skills Scholarship (KESS2), part-funded by the Welsh Government's European Social Fund (ESF). Pollinator icons contained in the plant recommendation list were created by Thomas McBride. We acknowledge the support of the Supercomputing Wales project, which is part-funded by the European Regional Development Fund (ERDF) via Welsh Government. None of the authors have a conflict of interest. The study was conceived by A.L., N.d.V. and S.C. Data collection and laboratory work was carried out by A.L.; The data were compiled by A.L. and analysed by A.L. and L.J. with suggestions from N.d.V., S.C. and G.B.; The manuscript was written by A.L. with contributions from all the authors. All authors gave final approval for publication. Raw sequence data are available on the Sequence Read Archive at PRJNA763761. Data available via the Dryad Digital Repository https://doi.org/10.5061/dryad.rjdfn2z9s (Lowe et al., 2022). All code is available at https://github.com/colford/nbgw-plant-illumina-pipeline. Appendix S1: Supporting Data Appendix S2: Supporting Methods Appendix S3: Supporting Tables and Figures Appendix S4: Supporting Results Appendix S5: Supporting Discussion Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Hotspots in the grid: Avian sensitivity and vulnerability to collision risk from energy infrastructure interactions in Europe and North Africa

Abstract: sion birds for which sufficient GPS tracking data at high spatial resolution were available. We also map vulnerability hotspots where mitigation at existing EI should be prioritised to reduce collision risks. As tracking data availability improves our method could be applied to more species and areas to help reduce bird- EI conflicts.

Leaf functional traits predict shade tolerant tree performance in cloud forest restoration plantings

Abstract: Data available via the Dryad Digital repository https://doi.org/10.5061/dryad.q2bvq83mj (Toledo-Aceves et al., 2022). Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Activity of forest specialist bats decreases towards wind turbines at forest sites

Abstract: Global carbon dioxide emission is the main driver of climate change (Solomon et al., 2009), threatening biodiversity and human economies worldwide (Bellard et al., 2012; Walther et al., 2002). To mitigate this threat, many countries are promoting wind energy production as a sustainable form of energy from renewable sources (Gielen et al., 2019). However, a growing body of literature indicates that the construction and operation of wind turbines may lead to habitat loss and an increased mortality risk for wildlife (Grünkorn et al., 2017; Kuvlesky et al., 2007; Saidur et al., 2011). For instance, past studies documented high fatality rates of bats and birds at wind turbine rotors (Arnett et al., 2016; Thaxter et al., 2017). Indeed, it was suggested that wind turbines may be the most significant anthropogenic factor causing multiple mortality events in bats (O'Shea et al., 2016). Consistent with this notion, past studies estimated that annual losses of bats at wind turbines may reach several hundred thousand in countries of the temperate zone (Hayes, 2013; Voigt et al., 2015; Zimmerling & Francis, 2016). This is mirrored in observed and modelled population declines of high collision risk species in North America and Europe (Frick et al., 2017; Friedenberg & Frick, 2021; Printz et al., 2021). Our current understanding of the wind energy–bat conflict is based almost exclusively on studies conducted at wind turbines operating in open landscapes. However, over recent years, turbines have been increasingly built at forest sites throughout Europe, particularly in Central and Northern Europe (Gaultier et al., 2020), despite guidelines recommending the contrary when alternative sites are available (Rodrigues et al., 2014). For instance, in Germany, more than 2,000 wind turbines (7.5% of all onshore turbines) operate currently at forest sites (Mackensen, 2019; Quentin & Tucci, 2021). To reduce further greenhouse gas emissions, recent pledges aimed at doubling the share of renewable energy production by increasing the area assigned for wind energy development from 0.8% (as of 2021) to 2.0% of the total surface area until 2030 (BMWK, 2022). Since land use pressure on open landscapes is already high and critical distances between wind turbines and settlements need to be maintained, several German federal states expand wind energy production in forests. Although non-primary forests of the temperate zone are usually managed for timber production, they offer valuable habitats for many species (Götmark, 2013; Hilmers et al., 2018; Spiecker, 2003). Forests constitute important hunting grounds for forest specialist bats and provide shelter for many more bat species (Dietz & Kiefer, 2014; Müller et al., 2013; Plank et al., 2012). Thus far, it is largely unknown how wind turbines in forests affect forest-associated bats. Although not at high risk of colliding with turbine rotors, forest specialist bats foraging below the canopy may be impacted by indirect wind turbine effects (Hurst et al., 2020). For instance, studies in open landscapes documented a reduced bat activity close to wind turbines compared to control sites without turbines, suggesting an avoidance behaviour and an indirect habitat loss for several species (Millon et al., 2015). Another study documented decreased bat activity along transects towards turbines (Barré et al., 2018), an observation that was confirmed for small wind turbines (Minderman et al., 2017). The underlying cause for this avoidance remains unclear, but bats may respond to turbine-generated noise (Allen et al., 2021; Finch et al., 2020) or potentially to artificial light (Bennett & Hale, 2014). Turbine construction in forests is further accompanied by fragmentation and degradation (Lesiński et al., 2007), while the creation of clearings and aisles is leading to a loss of foraging habitats and daytime roosts in trees (Hurst et al., 2020). However, forest fragmentation may also lead to increased activity of those bats which are more adapted to open and edge habitats and to an increased collision risk for these species at forest wind turbines (Kirkpatrick et al., 2017). In temperate forests, diverse vegetation structure and vertical stratification facilitate the cohabitation of three foraging guilds: open-space foragers which hunt insects above the canopy and in clearings, edge-space foragers which hunt along structures like forest edges or within gaps, and narrow-space foragers which hunt in dense vegetation and are especially adapted to life in forests (Denzinger & Schnitzler, 2013). The effect of habitat changes related to turbine construction and operation on bats may be guild specific due to different ecological requirements. The activity of open- and edge-space bats could even increase towards wind turbines caused by their attraction to clearings and forest edges (Kirkpatrick et al., 2017). Conversely, narrow-space foragers might respond negatively or not at all to the turbine-related habitat changes as they do not profit from open or semi-open habitats. In addition, a structure-rich forest vegetation could influence how far turbine effects on bats may extend into the surrounding forest, as dense vegetation may block visual signals and mitigate noise pollution. Lastly, turbine effects on bats may depend on the season, since bat activity varies throughout the year (Heim et al., 2016). For instance, most fatalities at turbines have been reported in late summer, coinciding with the post-weaning period of juveniles and the migration season (Kruszynski et al., 2022). Here, we asked how wind energy production affects bat assemblages in non-primary forests of Central Europe. This is a critical question since all bat species are protected by national and international legislation. Knowledge of factors that impact forest-associated bats is key to formulate adequate mitigation and compensation measures to protect bats when expanding wind energy production in forests. In our study, we used call activity as a proxy for the abundance of bats and thus conducted acoustic surveys along distance gradients towards wind turbines in 24 forests. Compared to earlier distance-gradient studies on bat activity at wind turbines, our focus on forest sites is novel and offers new insights about the consequences of wind turbine integration in forests accounting for vegetation structure. We predicted (I) that bat activity decreases with increasing proximity to the nearest turbine and that this effect will be stronger at larger wind turbines, where sensory pollution is presumably stronger. Moreover, we expected (II) that bat responses differ across functional guilds with strongest impacts for the activity of forest specialists, that is, narrow-space foragers and (III) that bat responses may vary across seasons and with vertical vegetation heterogeneity as a measure of forest structure. Our study aims to contribute to a sustainable wind energy development in forests from the perspective of bat conservation. Ultimately, this will help to reconcile the two important environmental goals of mitigating climate change and protecting biodiversity. We conducted our study in Hesse, a federal state in Central Germany characterized by temperate low range mountains and a forest cover of 42% (316–545 m a.s.l., 50°81′N, 8°81′W, Figure 1). We selected 24 forests ranging from coniferous monocultures to mixed and deciduous stands. Forest patch size varied between 184 and 6,337 ha (1,798 ha ± 1,745 ha; mean ± standard deviation, hereafter). Wind turbines in our study sites had been erected between 2006 and 2017 (6 ± 3 years). Tower height ranged between 145 and 212 m (194 m ± 16 m; N = 24) while rotor diameter ranged between 82 and 126 m (mean: 112 m ± 11 m). Studied turbines were located individually in cleared forest patches that ranged in size between 0.16 and 11.77 ha (median: 1.75 ha). To minimize the confounding effects of other anthropogenic disturbances and edge effects, we excluded study sites adjacent to highways and factories and established all transect points at a distance of more than 473 m (median) to the forest edge (91–1,884 m). Fieldwork permits were obtained from the respective forest owners. Ethical approval was not required. At each forest site, we used a distance-gradient study design with sampling points at 80, 130, 250 and 450 m distance to our focal turbine at the edge of the wind farm. In one study site each, one 80, 130 and 250 m point had to be skipped because of smaller clearings. For acoustic monitoring, we used automated bat recorders (BATLOGGER A+; Elekon). At each sampling point, we installed one recorder per forest stratum: near-ground in the clutter-free understorey (approx. 2.5 m height) and a second recorder in the lower canopy, where height varied according to forest succession stage (range: 4 m–22 m; 13 m ± 4 m). Recordings were conducted in 45 nights between mid-May and mid-September 2020, from 9 pm to 5 am. Per night, we recorded simultaneously at two geographically close transects and at each sampling point in the two designated forest strata. At every recording point, we recorded bat calls once per sampling period (1: May 17–June 5; 2: June 8–July 7: 3: July 13–August 15; 4: August 18–September 17) with intervals of 17–58 days (33.29 ± 11.26 days) in between. Some exceptions were caused by technical failures and unforeseeable logging activities (four recording nights at 156 recording points, three at 15 points, two at 1 point and one at 14 points). We employed BATLOGGER default settings with a trigger frequency between 15 and 155 kHz, thus covering the call frequency range of species expected in the local bat assemblage. We set a pre-trigger time of 500 ms, a post-trigger time of 1,000 ms and a recording intersection time of 20s. We used the CrestAdvanced trigger algorithm to enhance the recording probability of quiet calls and minimize sensitivity towards disturbing noise (Elekon AG, 2020). At each sampling point, we assessed four environmental variables that were assumed to influence bat activity: As a proxy for habitat heterogeneity, we estimated vegetation cover at heights of 0, 0.5, 1, 2, 4, 8, 16 and 32 m to the nearest 5% within a 10 m radius around distance points. We then calculated the diversity of the layers at each distance point using the Shannon–Weaver index to obtain vertical vegetation heterogeneity (Bibby et al., 2000). Furthermore, as a proxy for age structure, we measured the average tree canopy height in the immediate surrounding of sampling points with the help of a laser rangefinder (Forestry 550; Nikon) and used aerial photographs (Google Ireland Limited) to measure the distance between sampling points and the nearest outer forest edge. Finally, we calculated the proportion of deciduous and coniferous trees based on the Copernicus land cover map (ESA, 2018) within a 200 m radius around distance points, hereafter called tree composition. To capture differences in turbine characteristics, we retrieved the rotor diameter of each turbine from the publicly accessible database of Hessian environmental agency (HLNUG, 2019). We used the software BatExplorer (version 2.1.7.0; Elekon) to manually assign echolocation calls to bat species, only relying on the automatic call identification for Pipistrellus pipistrellus. We identified bat species based on echolocation call characteristics such as peak frequencies and call shapes from the literature (Barataud, 2020; LFU Bayern, 2020; Skiba, 2009). We subsequently grouped all call sequences into one of three ecological guilds (Denzinger & Schnitzler, 2013): open-space foragers (consisting of the genera Eptesicus, Vespertilio and Nyctalus), edge-space foragers (Pipistrellus ssp. and Barbastella barbastellus) and narrow-space foragers (genera Myotis and Plecotus). Sequences that could not be identified because of poor recording quality were discarded (0.4%). To obtain a proxy for the local bat abundance and prevent overestimation of single bats, we calculated the number of bat activity minutes for each of the three ecological guilds per night, sampling point and stratum. We divided recordings of all nights into 60-s intervals and counted minutes with at least one echolocation call, hereafter called activity minutes (Heim et al., 2016). If calls of more than one bat species appeared in one interval, they were considered as two separate activity minutes. Recordings with only social calls were discarded to avoid a bias towards species with higher detection and identification probability for social calls. In the following, we use the amount of activity minutes as a metric measure to describe bat activity. We conducted all statistical analyses with the software R (version 4.0.3; R Core Team, 2021). First, we split the dataset into three subsets, one for each foraging guild, because recorded activities were quantitatively too different between guilds to be fitted in the same model. For each guild, we tested if bat activity (response variable) decreases with increasing proximity to wind turbines. Due to the nested structure of our data, we used generalized linear mixed models (glmmTMB package; Brooks et al., 2017) with sampling points nested in study site as random effects. We used a negative binomial distribution to account for overdispersion (nbinom1 for open- and edge-space foragers, nbinom2 for narrow-space foragers) and, apart from that, applied the same model structure for all guilds. Models included turbine distance, vertical vegetation heterogeneity, canopy height, tree composition, rotor size, forest stratum and sampling period as fixed effects. Moreover, we added forest edge distance as fixed factor to correct for its potential influence on the distribution of bats in the studied forests, as well as the interactions of turbine distance with sampling period and rotor size. We checked the variance-inflation factor (VIF) of the regression, which assesses for each coefficient whether a correlation with other predictors may lead to an increased variance. VIF was below 2 for all predictors and we thus excluded multicollinearity (car package; Akinwande et al., 2015; Fox & Weisberg, 2019). All numerical predictors were standardized to allow direct comparison of estimates (Schielzeth, 2010). We worked with full models (Tredennick et al., 2021) and ensured their goodness-of-fit with the DHARMa package for residual diagnostics (Hartig, 2020). We checked that all models were informative looking at the difference in AIC value compared to null models and marginal R2 values (Table S1). Rotor diameters were not randomly distributed across forest sites and small rotors were biased towards deciduous forests. To exclude misinterpretations, we repeated above described analyses with only the data obtained from deciduous forests, thereby obtaining a balanced representation of rotor sizes. Additionally, we tested for potential confounding edge effects of the turbine clearing on bat activity by applying our model to a subset including only data sampled at 250 and 450 m distance to the wind turbine. Results did not qualitatively change in the additional analyses compared to models based on the complete data set (Tables S2 and S3). Accordingly, we considered our original results to be robust. During 5 months of data sampling, we obtained 678 recordings of complete nights, out of which 17 did not contain any bat calls. In total, we recorded 61,988 activity minutes of which 83% belonged to edge-space foragers, 12% to narrow-space foragers and 5% to open-space foragers (Table 1). The activity of narrow-space foragers was almost halved at the distance points closest to wind turbines (80 m) compared to 450 m distance points (Figure 2, Figure S1). This distance effect showed temporal variation, as it was apparent for the first three sampling periods (mid-May to mid-August) and absent for the last sampling period (mid-August to mid-September, Figure 3). Furthermore, the activity decrease was only observed towards turbines with rotors larger than 93 m diameter (Table 2, Figure 4). Activity increased with vertical vegetation heterogeneity, but no difference was observed between recordings made at the canopy and ground level. Bats were most active between mid-July and mid-September (Table 2, Figures S4–S8). The activity of edge-space foragers did not vary with turbine distance or rotor size (Figure 2, Figure S2). However, activity was higher at the canopy level than at ground level and increased with vertical vegetation heterogeneity and with tree height. Edge-space foragers were most active between mid-July and mid-August (Table 2, Figures S4–S8). The overall activity of open-space foragers did neither change with the distance to the wind turbine (Figure 2, Figure S3) nor with rotor size. Yet, in the last sampling period (mid-August to mid-September), we observed an increase in activity minutes close to turbines (Figure 5). Activity of open-space foragers was higher at canopy than ground level and increased with the proportion of coniferous trees in the forest. Bats were most active between mid-July and mid-August (Table 2, Figures S4–S8). We studied bat activity at wind turbines in 24 temperate forests in Central Germany and discovered a relationship with turbine distance, season and turbine size, but different patterns depending on bat foraging guild. Strikingly, activity of narrow-space foragers decreased with increasing proximity to turbines. This effect was notable over distances of several hundred metres. Our findings highlight that forest-dwelling bats, being at low risk of colliding at turbines, might still be affected by wind turbines in forests. This complements research from open landscapes, where narrow-space foraging bats showed a similar negative response towards wind turbines (Barré et al., 2018; Millon et al., 2015). However, our study is the first to confirm this pattern for forests, a highly important habitat from the perspective of bat conservation. We found that the activity of narrow-space foragers, mainly Myotis bats in our study area, decreased significantly towards turbines. This is in line with earlier studies on Myotis activity in open landscapes (Barré et al., 2018), even when focussing on small wind turbines (Minderman et al., 2012), highlighting the sensitivity of narrow-space foragers to wind turbines both in forests and open landscapes. Furthermore, we found that the activity decline of narrow-space foragers towards wind turbines was weaker in late summer, which confirmed the results of another open landscape study comparing wind turbine sites to control sites (Millon et al., 2015). In our study, we observed a distance effect particularly at turbines with large rotors. This suggests that avoidance might be caused by turbine-generated noise, which is presumably related to turbine size and diminishes over distance (Katinas et al., 2016). An adverse effect of noise on Myotis activity is also implied by a study on small wind turbines, where bats were particularly repelled by operating turbines (Minderman et al., 2012). Many narrow-space foragers locate their prey passively by detecting acoustic cues (Denzinger & Schnitzler, 2013). Therefore, these bats tend to avoid noisy environments, suggesting either a masking of prey sounds by anthropogenic sound emissions (Schaub et al., 2009) or a startling effect (Luo et al., 2015). In conclusion, we found a hitherto unknown avoidance behaviour of narrow-space foragers towards wind turbines in forests, indicating an indirect habitat loss for bats of this functional guild, possibly caused by noise. For edge-space foragers, which were mostly P. pipistrellus in our study, we neither found support for avoidance of, nor attraction towards wind turbines in forests. In contrast, recent open landscape studies observed a strong decrease in the activity of P. pipistrellus at hedgerows with decreasing distances to turbines on the one hand (Barré et al., 2018), and an increased activity at wind turbine sites in comparison to control sites on the other hand (Richardson et al., 2021). Possibly, the discrepancy between findings may be explained by different habitat matrices. Specifically, the erection of wind turbines in forests creates clearings and a network of edge structures which is an ideal foraging habitat for edge-space foragers. Indeed, it was observed that members of the edge- and open-space foraging guild were more active in spruce plantation after clear-cuttings (Kirkpatrick et al., 2017). In conclusion, clear-cutting for turbine construction probably poses a spatially restricted benefit for edge-space foragers. Activity of open-space foragers did not change in relation to turbine distance except for an activity increase with increasing turbine proximity in late summer. Our overall findings contrast with a previous open-landscape study that showed decreased activity for N. leisleri, but not for N. noctula and E. serotinus close to turbines (Barré et al., 2018), suggesting that open-space foragers might not be coherent in their responses to wind turbines. Different responses may even be related to intraspecies variation across bat individuals, as was suggested by GPS tracking studies on N. noctula around wind turbines (Reusch et al., 2022; Roeleke et al., 2016). In contrast, our finding of open-space foragers being attracted to wind turbines in late summer aligns with numerous previous studies suggesting an attraction effect of wind turbines on open-space foragers, hypothesizing various, yet untested causes (Guest et al., 2022). Given the seasonality of the attraction, open-space foragers possibly confuse forest turbines with tall trees, when searching for orientation points or stop-over roosts during fall migration (Cryan et al., 2014; Jameson & Willis, 2014). However, a recent study from Northern Germany shows an avoidance behaviour of N. noctula in late summer towards wind turbines, which argues against a general attraction of open-space foragers towards turbines in this season (Reusch et al., 2022). In conclusion, we could not confirm avoidance behaviour towards turbines for the entire guild. Yet, our findings of a seasonal attraction to turbines in forests are of high relevance in context of collision risks for open-space foragers. High activity of edge- and narrow-space foragers coincided with heterogeneous vertical vegetation structure. Similar positive effects of different measures of vegetation structure on forest-associated bats have been shown before and can be explained by a higher availability of microhabitats (Adams et al., 2009; Langridge et al., 2019; Müller et al., 2013). Furthermore, activity of edge-space foragers increased with tree height, suggesting a preference for more mature forest stands, probably due to their dependency on semi-open foraging habitats which rarely occur in early succession stages. In contrast, activity of open-space foragers was not affected by vertical vegetation structure or tree height, indicating that forest vegetation parameters are less important for aerial hawkers. For most bats, we observed a higher activity in the canopy than near-ground, confirming that the forest canopy is an important bat habitat (Adams et al., 2009; Erasmy et al., 2021; Müller et al., 2013; Plank et al., 2012). Lastly, we found a similar activity of most bats in mixed and coniferous forests which is consistent with a recent study suggesting that bats can find suitable roosts even in monocultural forest plantations (Buchholz et al., 2021). In conclusion, our findings indicate that forests with diverse vegetation structure present valuable habitats for a variety of bats, while forest type alone seems to be less important. The high activity of open-space foragers in conifer-dominated forests is likely related to high proportions of standing deadwood and clearances in these forests, leading to reduced attenuation of echolocation calls and an increased recording probability (Lawrence & Simmons, 1982). Our study highlights that the activity of forest-associated bats declines towards wind turbines at forest sites. Narrow-space foragers such as Plecotus spp. and Myotis spp. seemingly avoid wind turbines in forests and show reduced activity by about 50% from 450 to 80 m turbine distance. This avoidance is possibly caused by habitat degradation triggered by turbine-generated noise, since it was strongest towards turbines with large rotors. Consequently, legally protected forest bat specialists lose large habitat areas when wind turbines are erected at forest sites. Hence, we argue that this habitat loss should be compensated by taking nearby old forest stand out of forestry use, thus creating refugia for forest specialist bats. We also plead for a general caution when siting wind turbines in forests, since the response of bats was independent of vegetation structure and tree composition. We do not necessarily argue for a complete ban of wind energy production in forests, because in some countries there is little other option for renewable energies. Where absolutely necessary, turbines should only be built in managed forests with low vertical vegetation heterogeneity, as bat activity is expected to be low in these forests. This approach would most likely also account for birds and insects, which have been reported to die in considerable numbers through wind turbines (Thaxter et al., 2017; Voigt, 2021). However, as forest-related studies on birds and insects are still lacking, we urge to fill these research gaps to provide a basis for comprehensive recommendations on wind energy development in forests. This research was funded by Deutsche Bundesstiftung Umwelt (DBU). We are grateful for study permissions and organizational support by Hessian nature conservation agencies and forest owners. We thank Laura de Vries, Lea Vetter, Marcel Becker and Nina Wallmann for supporting fieldwork, Katharina Rehnig and PGNU for helping with acoustic analyses and Finn Rehling for advises during data analysis and visualization. Open Access funding enabled and organized by Projekt DEAL. N.F., F.P. and C.C.V. conceived the ideas and designed the methodology; J.S.E. and A.D. collected the data; J.S.E. analysed the data and wrote the manuscript. All authors contributed critically to the drafts and gave final approval for publication. Our study relied on local authors and research assistants. We frequently presented our research progress to stakeholders and sought their feedback. None of the authors have conflicting interests. Data available via the Dryad Digital Repository https://doi.org/10.5061/dryad.m0cfxpp66 (Ellerbrok et al., 2022). Appendix S1 Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Turnover in floral composition explains species diversity and temporal stability in the nectar supply of urban residential gardens

Abstract: 1. Residential gardens are a valuable habitat for insect pollinators worldwide, but differences in individual gardening practices substantially affect their floral composition. It is important to understand how the floral resource supply of gardens varies in both space and time so we can develop evidence- based management recommendations to support pollinator conservation in towns and cities. 2. We surveyed 59 residential gardens in the city of Bristol, UK, at monthly in-tervals from March to October. For each of 472 garden surveys, we combined floral abundances with nectar sugar data to quantify the nectar production of each garden, investigating the magnitude, temporal stability, and diversity and composition of garden nectar supplies. 3. We found that individual gardens differ markedly in the quantity of nectar

Leveraging functional traits of cover crops to coordinate crop productivity and soil health

Abstract: The authors declare no competing interest. Data available via the Dryad Digital Repository https://doi.org/10.5061/dryad.bg79cnpdm (Zhang, Xue, et al., 2022). Table S1 Table S2 Figure S1 Figure S2 Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Temporal and spatial heterogeneity of semi‐natural habitat, but not crop diversity, is correlated with landscape pollinator richness

Abstract: Agricultural expansion and loss of (semi-)natural habitats are major drivers of pollinator declines (Potts et al., 2010), with associated threats to the pollination services these species provide to wild plant populations (Clough et al., 2014; Martins et al., 2015; Pauw & Bond, 2011) and crop yields (Fijen et al., 2018; Sritongchuay et al., 2020; Webber et al., 2020). However, reversing these trends by converting agricultural fields to semi-natural habitat comes with high opportunity costs (smaller surface productive land) that may not outweigh the benefits of increased productivity (Kleijn et al., 2019). Increasing the diversity of mass-flowering crops is often raised as a promising strategy to complement resources to semi-natural habitats, which ultimately could benefit pollinator biodiversity levels in agricultural landscapes (Fahrig et al., 2011). Whether this approach could work likely depends on the capacity of different crops to sustain complementary diverse pollinator communities, or to supplement the characteristics of semi-natural habitats. However, we know surprisingly little about the potential of mass-flowering crop diversity to support rich pollinator communities (but see Raderschall et al., 2021; Sirami et al., 2019) and about the characteristics in semi-natural habitats that make them so relevant for pollinators. For instance, is semi-natural habitat the main determinant of the diversity of pollinator communities or can (diversity in) flowering crops boost pollinator diversity by adding new and abundant resources? Does this vary over space and time? Do different crop types complement each other and partially provide for different species? Ecological theory suggests that niche diversity or habitat heterogeneity is a key driver of species coexistence and therefore species diversity (Benton et al., 2003; Chesson, 2000; Reverté et al., 2019). Compared to crops, semi-natural habitats are generally much more heterogeneous, both within and between landscapes, and vary, for example, in the composition of flowering plants and availability of nesting substrates (Williams & Kremen, 2007). In addition to this spatial heterogeneity, the same semi-natural habitats also vary markedly in floral composition across the growing season as early flowering species senesce and are replaced by later flowering species (CaraDonna et al., 2017). This may not only imply that these habitats provide resources for species with different host plant preferences and phenologies (Mallinger et al., 2016), but also results in continuity of resources over time for generalist species that accept a wide range of host plants (Schellhorn et al., 2015). The relative role of temporal and spatial heterogeneity in resource supply in semi-natural habitats is virtually unexplored. A higher within and between habitat heterogeneity could explain why species diversity in semi-natural habitats is generally higher than in crops (Fijen et al., 2019). Nevertheless, even though the spatial and temporal heterogeneity of floral resources is inherently very limited in crops, each new insect-pollinated crop that is introduced into an agricultural landscape provides a potential new niche. This could mean that agricultural landscapes with insect-pollinated crops support richer pollinator communities than similar agricultural landscapes without insect-pollinated crops. Hence, promoting the cultivation of different mass-flowering crop types in agricultural landscapes might represent a strategy to boost insect pollinator communities in agricultural landscapes. Indeed, growing a single mass-flowering crop has been found to increase pollinator abundance and richness both locally (Diekötter et al., 2014; Holzschuh et al., 2013) and at the landscape scale (Beyer et al., 2020; Westphal et al., 2003). More recent studies have examined how pollinator abundance and diversity in one insect-pollinated crop depends on presence of other crops with responses differing between pollinator species groups (Aguilera et al., 2020; Martins et al., 2018). Whether flowering crop diversity increases landscape-level pollinator diversity and how this compares to the contribution of semi-natural habitat remains untested. Here, we address these questions using a dataset of pollinator communities occurring in South-Italian agricultural landscapes in two types of semi-natural habitats and 15 different crops across the entire growing season. In this study, we evaluate the relative importance of crop diversity and sources of heterogeneity in semi-natural habitats to promote insect pollinator richness in agricultural landscapes. To this aim, we sampled pollinator communities occurring in South-Italian agricultural landscapes in two types of semi-natural habitats and 15 different crops across the entire growing season (i.e. 4 months). We monitored pollinator communities in 26 agricultural landscapes along a gradient of increasing semi-natural habitat cover. We specifically asked (a) Does diversity of insect-pollinated crops contribute to pollinator diversity in agricultural landscapes? (b) How important is temporal heterogeneity in resources compared to spatial heterogeneity in driving pollinator species richness in agricultural landscapes? and (c) Does semi-natural habitat cover moderate pollinator richness similarly in semi-natural habitats and adjacent crops? We addressed these questions using a resampling approach, in which we analysed whether different sampling scenarios (e.g. sampling pollinators in one crop type vs. sampling different crop types; sampling pollinators in different landscapes at the same time vs. sampling pollinators at different times in the same landscape) would result in significant differences in the cumulative species numbers. We mainly focused on cumulative species numbers rather than species densities as this more accurately reflects the landscape-level species pool and enables a better understanding of the contribution of each site or habitat type. The study was conducted in a Mediterranean agricultural landscape located in southern Italy, in the same general region as the study of Fijen et al. (2018). The study area covered approximately 1,400 km2 and is dominated by wheat cultivation, but many other crops are cultivated, such as olive, faba bean and chickpea for food and feed, and onion and leek for seed production. In total, 26 study landscapes (750 m radius) were selected within the region based on a wide gradient of insect-pollinated crop types (0–8 per landscape) and semi-natural habitat cover (from c. 0.2%–72% semi-natural area; Table S1). Landscapes were separated from each other by 19 ± 18 km (mean ± 1 SD), except for one landscape pair where the borders of the landscape slightly overlapped. However, the radii of the landscapes were above the mean maximum foraging range of most solitary bees, c. 200–300 m c.f. (Zurbuchen et al., 2010). Therefore, we decided to keep that landscape pair in the analyses. The centre of 22 landscapes was a mass-flowering focal crop (18 leek hybrid seed production fields and four onion [hybrid] seed production fields). Four landscapes contained no mass-flowering crops. The semi-natural habitat cover for each landscape was quantified via Google earth aerial imagery and initial classifications were validated through field visits. Grasslands, woodlands, fallow arable fields and road verges (i.e. estimated as 1 m wide area along each side of roads) were considered as semi-natural habitats. The sampled herbaceous semi-natural habitats were mostly located at road verges with annual ruderal plants, while woody semi-natural habitats were usually forest edges or hedgerows. The most abundantly flowering species in herbaceous transects consisted of ruderal species of Brassicaceae (e.g. Diplotaxis erucoides, Sinapis arvensis, S. alba), Malvaceae (e.g. Malva sylvestris), Papaveraceae (e.g. Papaver rhoeas, Fumaria officinalis), Asteraceae (e.g. Anthemis tinctoria, A. arvensis) and Boraginacea (e.g. Echium plantagineum, Borago officinalis). Flowering woody plants consisted mostly of Prunus spinosa, Crataegus monogyna and several species of wild roses (Rosa sp.). Access to fields was kindly granted by the company Nunhems Netherlands BV (BASF). In each landscape, two important pollinator groups (bee and syrphid pollinators) were surveyed every 2 weeks in semi-natural habitats and each blooming mass-flowering crop during the main growing season (end of March to end of July 2018), resulting in eight rounds per landscape. Standardized transects of 150 m long and 1 m wide (150 m2) that were subdivided in three 50 m2 subareas were used, to ensure an even time distribution across the whole transect. Transects were monitored for 15 min pure observation time (i.e. excluding handling time). During surveys, we visually recorded all the observed species interacting with flowers. The pollinator species that could not be recognized in situ, were caught using butterfly nets and identified to species or morphospecies level in the laboratory. For this study, we discarded all the individuals that had not been identified to (morpho) species level (c. 6% of individuals). We ascertained that these unidentified individuals did not influence the results because they were few and evenly distributed across habitat types and landscapes (Figure S1). Surveys were conducted with temperatures above 18 degrees Celsius, on sunny and calm days (<5 bft wind), and roughly between 8 a.m. and 5 p.m. (c.f. Fijen & Kleijn, 2017). Days and times of the surveys were randomized across landscapes. This study did not require ethical approval for sampling pollinators. Within each landscape, semi-natural areas were sampled in each round with two transects in flowering herbaceous vegetation (including pioneer vegetation, grasslands and ruderal vegetation), and with a varying number of transects in woody vegetation (including shrubs and trees). Herbaceous transects were always located in focal areas containing flowers whenever these were locally present. This was achieved by slightly shifting, from one round to the next, the exact location of the transects up to 50 m to the right or the left to avoid sampling sites without flowers and therefore pollinators. The number of woody transects depended on the availability of flowering woody vegetation, and therefore varied between 0 and 2 transects per landscape per round. In total, nine landscapes had no woody vegetation and hence no transects in woody vegetation. During each round, all flowering insect-pollinated crops were surveyed when at least 10% of the flowers were open as even a small percentage of crop flowers represent large numbers of flowers. If there were multiple fields of the same crop in a landscape, we only sampled only one of the fields. For each flowering crop type, we selected one crop field close to the centre of the landscape and located one fixed 150 m2 transect starting at least 20 m from the edge of the field. Transects in crops were not moved, as variation in flowering stage within crop fields was negligible. Overall, this sampling provided data from 416 transects in herbaceous semi-natural habitats, 179 transects in woody semi-natural habitats and 122 transects in the 15 following insect-pollinated crop types (see also Table S2). The sampled crop types were as follows: basil Ocimum basilicum, broccoli Brassica oleracea var. italica, cauliflower Brassica oleracea, chickpea Cicer arietinum, dill Anethum graveolens, faba bean Vicia faba, leek Allium porrum, flax Linum usitatissimum, lucerne Medicago sativa, onion Allium cepa, rucola Eruca vesicaria, sulla Hedysarum coronarium, sunflower Helianthus annuus, clover Trifolium sp. and vetch Vicia sp. All these crops were flowering because they were either used for vegetable/herb seed production (e.g. leek and onion), oil-seed production (e.g. sunflower and flax), for food or feed (e.g. faba bean and chickpea), or because they had not yet been harvested for animal feed (e.g. sulla, clover and vetch). To analyse the data, we used a resampling approach. This was necessary because all our variables of interest had been sampled with different intensities and replication. Not correcting for these differences could result in the variable with the largest sample size being most strongly related to species richness, only because of the wider environmental gradient sampled. In contrast, using a method with, for example, sample size as an offset would not account for the fact that with an increasing number of samples the probability of finding new species decreases, which would lead to underestimated species richness in more sampled sites. Our robust resampling with replacement approach allowed us to correct for differences in sample size but still use all the data in our extensive dataset and estimate confidence intervals from which infer significance. In this study, we used four sets of analyses. To compare the diversity of pollinators supported by crops and semi-natural habitats, we compared the total number of pollinator species observed in herbaceous and woody semi-natural habitats with (a) a mix of 12 different crop types, (b) faba bean, the most frequently occurring early flowering crop type and (c) leek, the most frequently occurring late-flowering crop type. Mean estimates of accumulated species richness and 95% confidence intervals were obtained by randomly resampling 250 combinations of 12 transects in each habitat type or crop (mixture). Twelve was the highest number of transects we considered acceptable for estimating the cumulative number of species by means of resampling because the maximum number of crops was restricted at 15. The cumulative species numbers in the mixture of crops were estimated by selecting one transect each in 12 different crops in any landscape and round, thus maximizing the potential effects of crop diversity. This was compared with similarly obtained cumulative species numbers in herbaceous and woody semi-natural habitats from all landscapes and rounds. Because faba beans only flowered in rounds two/three and leek only flowered in rounds seven/eight, we compared cumulative pollinator species numbers of these crops with estimates from herbaceous and woody semi-natural habitats that were also based on resample analyses from these rounds only. To subsequently test for significant differences, we used linear models, with resampled cumulative species richness estimates as the response variables and habitat type as explanatory variable. We present the results as rarefaction curves to visualize differences in total richness, and different rates of species accumulation as the number of transects increases. To test whether landscape-level pollinator diversity was influenced by the diversity of insect-pollinated crops, we also analysed whether the cumulative species richness in herbaceous semi-natural habitats across the eight sampling rounds was related to the number of crops grown in each landscape across the season (n = 26), using simple regression analysis. For this analysis, we focused on herbaceous semi-natural habitats only for several reasons. Sampling effort in herbaceous semi-natural habitats was completely balanced across landscapes, and 97% of all pollinator species were found in these habitats. Furthermore, including crop transects would lead to bias in this analysis, because landscapes with higher crop diversity had inherently more crop transects. To better understand the relationship between crop diversity and pollinator diversity, we analysed for each pollinator species in how many different crops they had been observed and whether they had additionally been observed in herbaceous and/or woody semi-natural habitats. To examine the importance of temporally stable habitats, we analysed the capacity of individual herbaceous semi-natural habitats to accumulate species richness across the season and compared it to the richness accumulated in multiple herbaceous semi-natural habitats from different landscapes in a single round (i.e. same site different times vs. same time different sites). We refer to these two different drivers of species richness as temporal heterogeneity and spatial heterogeneity in resources, respectively. Although other stressors such as competition or diseases might contribute to species spatial/temporal turnover, resource heterogeneity is expected to be the main limiting factor, and the main differentiating characteristic of semi-natural habitats compared to coexisting crops. We first pooled each pair of transects in herbaceous semi-natural habitats within each landscape and round to increase sampling effort/precision. Then, we calculated species richness accumulated in herbaceous semi-natural habitats across the eight rounds within each landscape (26 data points; one per landscape). We also calculated species richness accumulated in herbaceous semi-natural habitats from eight different landscapes within the same round. We resampled eight random landscapes 10 times and averaged the results to provide stable estimates that are representative for all spatial samples per round. This was done 26 times within each round to run balanced models, as we have 26 replicates in the temporal heterogeneity dataset. To test for significant differences, we fitted two separate linear models, in which accumulated species richness was the response variable, and the source of heterogeneity (temporal/spatial) was the explanatory variable. In the first model, we tested for general differences in accumulated species richness due to spatial and temporal heterogeneity. For this analysis, we selected three random samples in each of the eight rounds from the spatial heterogeneity data pool (n = 24) to more or less balance the temporal heterogeneity data pool (n = 26). For the second model, we compared the accumulated species richness due to temporal heterogeneity to that of spatial heterogeneity in each round (i.e. temporal heterogeneity compared to spatial heterogeneity in each round). Groups were compared using post-hoc Tukey tests. To study whether landscape semi-natural cover moderates the distribution of pollinator species over semi-natural habitats and crops, we analysed how the number and percentage of shared species between semi-natural habitats and nearby crops were influenced by semi-natural cover in the landscape. We first calculated the number of species in herbaceous semi-natural habitats in each landscape (throughout the whole season). Then, we calculated the number and percentage of these species that were also found in a single transect of each crop sampled in that same landscape. We fitted two general linear models with number and percentage of shared species as response variables, and crop type and percentage of semi-natural habitat in the landscape as explanatory variables. The Gaussian error structure of the models was chosen based on model fit and performance of the residuals. All models were checked for outliers and for normal distribution of the residuals. We conducted all the analyses in R (R Core Team, 2019). We used the package stats (base R) to run the linear models, dplyr to manage data (Wickham et al., 2021), ggplot2 to create the graphs (Wickham, 2016), vegan for species accumulation curves (Oksanen et al., 2020) and spadeR to count the number of shared species (Chao et al., 2000). Code and data are freely available (see corresponding section). In total, 26,123 individuals belonging to 49 genera and 372 different species or species complexes were found in the 717 surveyed transects (see Table S3 for a species list). The most frequently observed species were Sphaerophoria scripta complex (35% of transects), Apis mellifera (32%), Andrena flavipes (25%), Lasioglossum villosulum/medinai (22%), Syritta pipiens (21%) and Eristalis tenax (20%). Herbaceous semi-natural habitats consistently hosted more diverse pollinator communities than the two most frequently cultivated crops or, perhaps more interesting, any combination of 12 different crops (Table 1; Figure 1). Herbaceous semi-natural habitats furthermore supported more species-rich pollinator communities than woody semi-natural habitats. The cumulative number of species observed on leek crops alone was higher than the cumulative number of species observed on a sample combining 12 different crops (non-overlapping confidence intervals; Figure 1 and Table 1). Flowering leek fields were particularly attractive and hosted a rich community of pollinators (mean 42.40 ± 0.37 SE species per 12 crop fields). In contrast, faba bean fields were visited by a relatively small number of species (mean 19.97 ± 0.36 SE). Crop diversity was not related to pollinator richness in herbaceous semi-natural habitats (Figure 2A; t24 = −0.092, p = 0.927). A total of 236 pollinator species were exclusively encountered in semi-natural habitats, and 13 only in crops. Crops were visited by 136 species, of which 90% were also observed in semi-natural habitats. Most crop visiting species visited only a few crops (1–3) but 20 species were observed in more than three crops (Figure 2B). The honeybee was the most ubiquitous species in crops, visiting 13 of the 15 crop types (Figure 2B). Because this species is managed by farmers in these landscapes, it is a poor representative of how most pollinators use crop resources. Temporal heterogeneity was at least as important as spatial heterogeneity for pollinator species richness (t48 = 0.946, p = 0.349; Figure 3A). The species richness due to spatial heterogeneity varied strongly between sampling periods, being highest in mid-season (May) and lower in early/late season (Figure 3B). The number of shared species between semi-natural habitats and nearby mass-flowering crops increased roughly from five to nine along a gradient of 0%–72% semi-natural cover scale (t105 = 3.725, p < 0.001, β = 0.059; Figure 4A). This was in line with a general increase in species density with increasing cover of semi-natural habitats at the landscape scale (Figure S2). However, the percentage of shared species (8.3% ± 4.9; mean ± 1 SD) remained stable along this landscape gradient (t105 = 0.705, p = 0.482, β = 0.015). These differences were fairly stable across crop types (Tables S4 and S5). Understanding what habitat and landscape characteristics determine the size of the pollinator species pool in agricultural landscapes is important for the development of productive agroecosystems that support high diversity of pollinators. In this study, we compared the relative importance of different habitat types for determining pollinator species richness. We found that the capacity of croplands to support pollinators seems to depend more on crop identity than on crop diversity, as a single crop hosted more pollinator species than a mixture of 12 different crops. In addition, pollinator richness in semi-natural habitats was not influenced by the number of flowering crop types in that landscape. We also found that semi-natural habitats generally hosted a richer pollinator community than individual crops or mixtures of crops. Furthermore, only few species visited more than two different crops, probably because crops offer a limited set of resources and provide suboptimal habitat conditions for most pollinators (e.g. disturbed ground or use of pesticides). The temporal continuity and heterogeneity in resources provided by semi-natural habitats was at least as important as spatial heterogeneity for determining pollinator species richness, highlighting the value of habitats that provide resources throughout the season. Lastly, our results suggest that landscape-moderated increases in crop pollinator diversity are driven by more complex habitats supporting larger pollinator species pools because the percentage of shared species between crops and semi-natural habitats was stable across a landscape complexity gradient. Pollinator richness in landscapes was not related to crop diversity. This indicates that other landscape characteristics, such as landscape composition (i.e. what specific habitat types compose a landscape) or edge density (i.e. amount of very small semi-natural habitats) might be more important for pollinators (Hass et al., 2018). The most likely explanation for this is that mass-flowering crops are only temporally available habitats, and therefore do not provide a permanent niche for pollinators (Schellhorn et al., 2015). Most pollinator species probably cannot complete their life cycle in crop habitats (e.g. they need above ground cavities) and visit crops only to forage. Furthermore, only a subset of pollinator species makes use of crops (Kleijn et al., 2015; Senapathi et al., 2015), and these are usually generalist species that can readily exploit abundant resources when they become available (Fijen et al., 2019). We show that most crop pollinators visit only one or two crops, with the exception of honeybee that was abundant in most crops. This suggests that even these generalist species have preferences or are constrained in the floral resources they can use and thus remain partially dependent on semi-natural habitats for foraging, not only for reproduction and shelter. Whether pollinators make use of crops probably depends on flower morphology and crop phenology (i.e. crop identity). For example, the highly attractive leek crop has open flowers and blooms in June when many pollinator species are active but nectar is scarce (Timberlake et al., 2019). Leek fields hosted twice as many species as the early flowering faba bean that has complex flowers that are not accessible to many insect pollinators. Crop diversity can, therefore, probably at best increase pollinator abundance, as demonstrated for faba bean pollinators (Raderschall et al., 2021), and, indeed, some studies have shown that cultivation of a single mass-flowering crop such as faba bean, red clover or oilseed rape can also (temporarily) increase abundances (but see Riggi et al., 2021 for bumblebee diversity) of crop pollinators (Beyer et al., 2020, Westphal et al., 2003, Westphal et al., 2003) by providing temporally abundant floral resources (Holzschuh et al., 2013). Ninety-seven percent of all observed pollinator species were found in semi-natural habitats. This highlights that semi-natural habitats, particularly herbaceous semi-natural habitats, are the main source of pollinators and effectively determine how many species can exist in agricultural landscapes. Crops generally supported high densities of less species than semi-natural habitats, but the right crop can temporarily attract a large proportion of the local species pool, as illustrated by leek in our study, since it hosted a similar number of species as woody semi-natural habitats in the time when leek bloomed. In our study, we found surprisingly few crop pollinator species that were only found in woody habitats, suggesting that woody habitat mainly provides resources that are complementary to herbaceous semi-natural habitat but offer few unique niches for pollinators (Eeraerts et al., 2021; Rivers-Moore et al., 2020). These complementary resources most likely include nesting sites for cavity nesting species (Rivers-Moore et al., 2020), that are not available in herbaceous habitats or crops. Semi-natural habitats contain many different niches and the resources that pollinators need are therefore temporally and spatially spread out (Schellhorn et al., 2015). Our study assessed the relative importance of the commonly studied spatial heterogeneity (number of bee species supported by multiple habitats in a specific period during the growing season) with the importance of the rarely studied temporal heterogeneity (one habitat repeatedly sampled throughout the growing season) and found that temporal heterogeneity was just as important for species richness as spatial heterogeneity. This suggests that most work on the contribution of semi-natural habitats to crop pollination underestimate the total number of species that rely on these semi-natural habitats because they are generally done during a small period around crop flowering (see Kleijn et al., 2015). Importantly, the richness accumulated due to temporal turnover might be even stronger than shown here, because the richness accumulated through different landscapes (spatial heterogeneity) also includes some temporal heterogeneity due to topographic and phenological mismatches between these landscapes (Olliff-Yang & Ackerly, 2020). In conclusion, the combination of temporal and spatial heterogeneity in the resources (e.g. flowers and nesting sites) that semi-natural habitats provide is likely the main reason why they can support richer pollinator communities than a combination of many different flowering crops. Pollinator richness increased in landscapes with more semi-natural habitat cover, which supports the widely accepted view that landscape simplification jeopardizes pollinator communities (Kennedy et al., 2013). This has well-known pernicious effects for the provision of pollination services at the landscape scale. Interestingly, our results provide a key nuance, since small semi-natural habitats in very simplified landscapes hosted a decent density of different pollinator species (c. 11 species vs. 15 species in patches of the same size in highly naturalized landscapes). Hence, these (stable) small semi-natural patches in highly intensified landscapes might be very important for pollinator conservation, acting as refuges where pollinator species concentrate (Boetzl et al., 2021; Li et al., 2020; Martínez-Núñez et al., 2020; Redhead et al., 2020). Maintaining even small patches of semi-natural habitat can therefore help conserving pollinator populations in simplified landscapes. The number of shared species between semi-natural habitats and nearby crops increased with landscape complexity, yet the proportion of shared species remained constant across landscapes. This suggests that the proportion of pollinators that can use crops and non-crop pollinators is rather similar in simple and complex landscapes. This apparently goes against ecological theory that suggests that specialist bee species have higher extinction rates in highly fragmented landscapes (Harrison et al., 2017; Redhead et al., 2018), but the equal proportion across the landscape complexity gradient may at least partly be explained by the fact that many pollinator species only opportunistically make use of crop resources (Fijen et al., 2019; Senapathi et al., 2015). Whether pollinator species visit crops is strongly limited by the species’ tolerance to crops, or, conversely, their preference for semi-natural habitats, and our results show that this tolerance/preference is proportionally constant along a landscape complexity gradient. This strongly supports the concept of ecological intensification, because it shows that by increasing the cover of semi-natural habitats at the landscape level it simultaneously and proportionally increases the crop and non-crop pollinator species pools (Bommarco et al., 2013; Kleijn et al., 2019), with subsequent benefits for ecosystem services and crop pollination (Morandin & Kremen, 2013). We show that increasing crop diversity cannot be used as a strategy to maintain species-rich pollinator communities in agricultural landscapes. Instead, promoting spatially and temporally heterogeneous habitats is key to increase the availability of niches and support a high number of pollinator species throughout the season. Conservation of semi-natural habitat, and the restoration or promotion of set-aside small patches of semi-natural habitat can contribute to maintaining relatively rich pollinator communities, and helps to keep the proportion crop and non-crop pollinator species fairly stable across wide landscape complexity gradients. Nunhems Netherlands BV (BASF) kindly provided access to field sites and assisted in the logistics of this study. We thank two anonymous reviewers for their constructive comments. None of the authors have a conflict of interest. T.P.M.F. and D.K. designed the sampling and led the project; T.P.M.F., D.K. and C.M.-N. conceived the main ideas of the manuscript; C.M.-N. analysed the data and wrote the first draft of the manuscript with inputs from T.P.M.F. and D.K.; C.G., D.H., W.V. and T.P.M.F. conducted the fieldwork; I.R. identified species in the laboratory. All authors contributed significantly to the final version of the manuscript. Data and code available via the Figshare https://doi.org/10.6084/m9.figshare.19153214.v1 (Martinez-Nuñez et al., 2022). Supinfo Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Experimental comparison of fish mortality and injuries at innovative and conventional small hydropower plants

Abstract: Resolving the controversy about hydropower is only possible based on reliable data on its ecological effects, particularly fish welfare. Herein, we propose a comprehensive assessment of conventional and innovative hydropower using a dataset of 52,250 fish. The effects of hydropower on fish were most harmful at sites with Kaplan turbines, showing ≤83% mortality. Innovative hydropower, often termed ‘fish-friendly’, caused ≤64% mortality. Our findings suggested that the runner peripheral speed, number of turbine blades and turbulence at turbine outlets were the most important factors. Synthesis and applications. To reduce the impact of hydropower on fish, site-specific characteristics such as head drop, bypass options and river-specific species composition need to be more intensively considered in optimal turbine technologies and operation modes.

Even short‐term revegetation complicates soil food webs and strengthens their links with ecosystem functions

Abstract: 1. 干旱生态系统退化是气候变化和人类活动造成的世界性问题。各国政府采取了一系列恢复措施，例如植被重建，以恢复这些退化的干旱生态系统。但是目前为止，有关植被重建如何影响土壤食物网及其生态系统功能的实验论证相对匮乏。 2. 本研究利用中国内蒙古自治区锡林郭勒盟乌拉盖管理区不同退化强度草地上建立的大型植被重建控制实验，研究了植被重建如何影响土壤食物网中关键生物组分(植物、微生物和线虫)以及生态系统功能(土壤碳和氮矿化)。 3. 植被重建四年后，我们发现植物、土壤细菌和真菌的生物量得到了明显恢复，但土壤线虫多度恢复较差。植被重建增加了植物和细菌多样性以及土壤碳和氮矿化速率，也改变了植物和土壤微生物群落的结构，但不影响土壤真菌或线虫的多样性。研究表明退化草地土壤食物网多样性的恢复滞后于土壤食物网生物量/多度的恢复。 4. 相较于高退化强度，低退化强度上的植被重建对植物、土壤细菌、真菌和线虫的影响更强，这表明退化草地的恢复措施应考虑不同的草地退化程度。此外，植被重建还增加了植物、土壤食物网和生态系统功能之间的相互作用。研究表明在植被重建过程中，土壤食物网复杂性的增强有助于退化草地土壤养分和植被生产力的恢复。. 5. 总结与应用:我们的研究结果表明，植被构建对植物(初级生产者)和土壤微生物(中间营养级)的影响比对土壤线虫(较高营养级)的影响更大，即使是短期植被恢复也会增强土壤食物网复杂性和生态系统功能的恢复。这些发现强调了草地管理者采取积极人工恢复措施的必要性，即便是短期的合理恢复措施也有助于干旱区土壤食物网和生态系统功能的恢复和改善。 Data available via the Dryad Digital Repository https://doi.org/10.5061/dryad.pnvx0k6q0 (Wang et al., 2022). Appendix S1 Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Semi‐natural habitat surrounding farms promotes multifunctionality in avian ecosystem services

Abstract: 1. Farmland birds can suppress insect pests, but may also consume beneficial insects, damage crops and potentially carry foodborne pathogens. As bird communities shift in response to farming practices, so too do the benefits (services) and costs (disservices) from birds. Understanding how and why ecosystem services and disservices

The widespread trade in stingless beehives may introduce them into novel places and could threaten species

Abstract: Data are available via Zenodo https://doi.org/10.5281/zenodo.5295749 (dos Santos et al., 2021). Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

A methodological guide to observe local‐scale geodiversity for biodiversity research and management

Abstract: Current global environmental change calls for comprehensive and complementing approaches for biodiversity conservation. According to recent research, consideration of the diversity of Earth's abiotic features (i.e. geodiversity) could provide new insights and applications into the investigation and management of biodiversity. However, methods to map and quantify geodiversity at local scale have not been developed although this scale is important for conservation planning. Here, we introduce a field methodology for observing plot-scale geodiversity, pilot the method in an Arctic–alpine tundra environment, provide empirical evidence on the plot-scale biodiversity–geodiversity relationship and give guidance for practitioners on the implementation of the method. The field method is based on observation of geofeatures, that is, elements of geology, geomorphology and hydrology, from a given area surrounding a location of species observations. As a result, the method provides novel information on the variation of abiotic nature for biodiversity research and management. The method was piloted in northern Norway and Finland by observing geofeatures from 76 sites at three scales (5, 10 and 25 m radii). To explore the relationship between measures of biodiversity and geodiversity, the occurrence of vascular plant species was recorded from 2 m × 2 m plots at the same sites. According to the results, vascular plant species richness was positively correlated with the richness of geofeatures (Rs = 0.18–0.59). The connection was strongest in habitats characterized by deciduous shrubs. The method has a high potential for observing geofeatures without extensive geological or geomorphological training or field survey experience and could be applied by conservation practitioners. Synthesis and applications. Consideration of geodiversity in understanding, analysing and conserving biodiversity could facilitate environmental management and ensure the long-term sustainability of ecosystem functions. With the developed method, it is possible to cost-efficiently observe the elements of geodiversity that are useful in ecology and biodiversity conservation. Our approach can be adapted in different ecosystems and biodiversity investigations. The method can be adjusted depending on the abiotic conditions, expertise of the observer(s) and the equipment available.

Nonlinear response of the soil seed bank and its role in plant community regeneration with increased grazing disturbance

Abstract: The soil seed bank represents valuable rebuilding capital that may rescue an ecosystem from state transition once vegetation has crossed an apparent threshold from the desired to degraded state. However, almost no research has explored the response of transient and persistent seed banks and their role in plant community regeneration along a gradient of wetlands from intact to a seriously degraded state due to increased grazing disturbance. Seven grazing disturbance levels from nondisturbed to highly degraded alpine marsh ecosystems were selected on the eastern Tibetan Plateau. Akaike information criterion (AIC) was used to select the best-fit model to predict the response of the plant community, soil seed bank and Bray–Curtis dissimilarity index to increased grazing disturbance. Both the plant community and seed bank showed a nonlinear change with increasing grazing disturbance. Species richness and seed density of the transient seed bank first decreased and then increased with increased disturbance, but the persistent seed bank showed a reverse trend, with an obvious threshold. Species composition of the persistent and transient seed banks exhibited little change compared to the plant community as disturbance increased. Similarities between both the persistent and transient seed banks and plant community also showed a nonlinear change with increased disturbance, while the persistent seed bank had a higher similarity with the plant community than the transient seed bank. Synthesis: At high grazing disturbance, persistent seed banks are more important than transient seed banks in plant community regeneration. Alpine wetland ecosystems have intrinsic resilience because the persistent seed bank has a pool of species above the threshold. However, ecosystem resilience declines if the species pool of the persistent seed bank is depleted below the threshold. The restoration potential of the seed bank has limits, and it will gradually be exhausted when species losses due to increased grazing intensity exceed the threshold of state transition.

Evaluating CAP wildflower strips: High‐quality seed mixtures significantly improve plant diversity and related pollen and nectar resources

Abstract: In recent decades, intensification has led to a massive biodiversity loss in agricultural landscapes throughout Europe (Seibold et al., 2019) due to effective weed control, field enlargements, short crop rotation as well as the ploughing of grassland (Baessler & Klotz, 2006; Meyer et al., 2013). The remaining plant communities are often species poor, dominated by highly competitive grasses and nitrophilous ruderals and are often isolated. The reduced availability of flowering plants as resources of pollen and nectar in agricultural landscapes has led to a significant decline in pollinator diversity (Biesmeijer et al., 2006) affecting important ecosystem services such as pollinator performance (Clough et al., 2014). The integration of agri-environment schemes (AESs) in the common agricultural policy (CAP) is an attempt to counteract this negative trend (Scheper et al., 2013). Flower strips are among the most commonly applied AES in Germany and aim to enhance biodiversity by providing shelter as well as food resources for different animal taxa, for example pollinators (Ouvrard et al., 2018). Perennial flower strips are sown on a part of the arable field once at the beginning of the funding period of 5 years. Farmers receive subsidies to compensate for yield loss and maintenance costs. Although vegetation is fundamental for many animal species groups, it was often not or only sporadically evaluated in mainly faunistically focused studies on flower strips (Zingg et al., 2019, but see Schmidt et al., 2022). To sustainably improve flower strip performance, the design of AESs and thus their ecological effectiveness, it is necessary to evaluate their quality over several years to understand what determines their failure or success (Albrecht et al., 2021). Existing studies that focus exclusively on vegetation often refer to mixtures containing mainly cultivar species or to species-poor (<15 species) wild plant mixtures (Delphia et al., 2019; Piqueray et al., 2019; Uyttenbroeck et al., 2015). Since many insect species depend on a specific plant species or genus (Potts et al., 2010), perennial wildflower strips (WFSs), sown with species-rich native seed mixtures from regional seed propagations, are expected to better provide the desired diversity of nectar and pollen (Wood et al., 2017). However, to our knowledge, floristic evaluations of such WFSs and their development over several years are scarce (Pywell et al., 2011). A previous study showed that high diversity seed mixtures with regionally typical native plants ensured high species diversity and cover of native perennial wildflowers within the AES funding period of 5 years and beyond (Schmidt et al., 2020). However, that study was conducted only on small experimental plots and it is unclear if such an approach will provide the desired functional diversity in agricultural practice (Lepŝ et al., 2007). To date, factors influencing vegetation composition of WFSs over time have rarely been investigated (but see Piqueray et al., 2019). Spontaneously establishing species from the soil seed bank or immigrating from adjacent structures can affect species assemblage and diversity in two ways. Spontaneous forbs may significantly increase plant species richness and the flowering aspect, and expand the food supply. Dominance of spontaneous grasses, however, may jeopardise success. With increasing soil fertility, vegetation stands often become taller and denser with increased competition (Piqueray et al., 2021), especially on shaded sites. Contrary to general recommendations, farmers often place flower strips on heavily shaded, north-exposed forest edges or in areas with already existing high weed pressure because the yield would be low there anyway. Whether measures designed for animal taxa adapted to predominantly open and sunny agricultural landscape can develop as desired under these conditions is questionable. In addition, the evaluation of conservation actions should not only consider local field conditions, but also the landscape context (Kleijn et al., 2011), as other studies verified, for example, the importance of habitat connectivity (Brudvig et al., 2009). A high amount of semi-natural habitats of open landscapes like grasslands or a generally high habitat diversity in the vicinity may increase overall plant species richness by providing a higher local species pool (Moser et al., 2002), or they decrease the establishment success due to a higher immigration of competitive weeds. Moreover, these effects could vary depending on local site conditions, for example soil fertility. Since the participation of farmers in AES programmes and thus the spatial distribution of WFSs varies greatly at the landscape level, similar landscape effects could possibly appear by considering the area under flower strip management. As landscape effects can change with increasing distance, multiple scales have to be considered. The study was conducted in the federal state of Saxony-Anhalt, Germany. The climate is rather dry with an annual mean temperature of 9.3 degree Celsius and annual precipitation of 579 mm (long-term mean 1981–2010, German Meteorological Service, 2020). Sandier soils such as brown earth dominate in the north and east of the federal state, interrupted by loamy gleys and fen soils along rivers and in lowlands. A loess belt with fertile chernozem soils can be found in the central part of the state. More than 60% of Saxony-Anhalt is used as arable land, mainly for cereal, rape and maize cultivation. The average field size in the study areas was 18 ± 22.3 ha (mean ± SD). Study sites with WFS (n = 40) were randomly chosen from 272 WFSs with a minimum length of 200 m that were established in 2015 or 2016 by farmers in agricultural practice on arable fields under the Saxony-Anhalt AES directive for the 5-year funding period (Fenchel et al., 2015; Figure 1). Selected WFS sites varied along a gradient regarding landscape heterogeneity and area under WFS management in the surroundings. As controls, we selected cereal crop fields lacking WFSs (n = 20). Controls were stratified-randomly selected on cereal crop fields from the same landscape units at a minimum distance of 1,000 m. WFSs and controls were always selected in a 2:1 ratio per landscape unit (Figure 1; Figure S1). The uneven spatial distribution of the study plots corresponds to a regionally varying participation of farmers in the AES. At the time of the first data collection in 2017, WFSs were in their second/third year, and at the end of the observation period in 2019 in their fourth/fifth year. Seed mixtures, obligatorily used for AESs within the CAP funding period 2014–2020, contained 30 native forbs from a certified regional seed propagation (parent seeds taken from the wild with proof of provenance, see Mainz & Wieden, 2019) and cost about 500 € per hectare, with a sowing rate of 0.4–0.5 g/m2. The investigated WFSs were sown with two different seed mixtures due to varying soil conditions (loess-loam, n = 14 or sand-loam, n = 26), containing 18 identical and 12 different species. Management by farmers was on a voluntary basis; fertiliser and pesticide application was forbidden (Fenchel et al., 2015). For further information about the seed mixtures and WFS regulations, see Appendix S1. On each of the selected WFSs and control cereal fields, the presence of all vascular plant species was recorded along a 5 m × 200 m transect at the field edge and within 2 m to the adjacent landscape structure (see Figure S2). Species per cent cover was estimated in four permanently marked 2 m × 2 m plots per site within the 1,000 m2 transect. To avoid edge effects, each quadrat was systematically placed in the centre of a 5 m × 50 m section. Permanent plots were recorded each year from 2017 to 2019 once between mid-May and the end of June. Nomenclature follows Jäger (2017). Vegetation surveys were granted by the Ministry of Environment, Agriculture and Energy Saxony-Anhalt and did not require ethical approval. For each year and plot, we calculated total plant species richness, the number and cumulative cover of sown and spontaneous forbs and grasses. Values regarding species richness were derived from the 1,000 m2 transect. To calculate the cumulative cover of sown/spontaneous forbs and grasses, the cover of the respective species per permanent plot was summed and then averaged per WFS site to avoid pseudo-replication. The calculation of shading was based on the assumption that the sun rises exactly in the east and sets in the west (Table 1). In our study, 50% of WFSs received sunlight almost all day, while 50% were shaded. Since the proportion of area under WFS management and of non-intensively used open habitats varied greatly in the vicinity of the study sites, we included both as metric landscape variables in our analyses. Habitat types were mapped within a 1,000 m radius around each vegetation transect using an adjusted standard habitat mapping key of Saxony-Anhalt (Schmidt et al., 2022 on the basis of Peterson & Langner, 1992). The mapped data were digitalised, and habitat proportions (WFS, open habitats) and habitat diversity were calculated with four radii (250, 500, 750 and 1,000 m) around the centre of the vegetation transect using ESRI ArcGIS 10.4.1. In order to distinguish between the effects of environmental variables per year and per sown and spontaneous forbs, the categorical factors ‘year’ and ‘status’ were included (Table 1). Statistical analyses were performed in R version 4.0.2 (R Core Team, 2020) and figures created using ggplot2 3.3.2 (Wickham, 2016). Mann–Whitney U tests (stats 4.0.3) were used to analyse statistical differences in total species richness and PFI between WFSs and controls. Generalized linear mixed models (GLMM) or linear mixed models (LMM) were fitted to evaluate whether forb species richness (sown and spontaneous forbs, count data, GLMM with negative binomial error distribution to account for overdispersion), forb cover (sown and spontaneous forbs, LMM) or cover of grasses (LMM) on WFSs were affected by environmental variables or years (Table 1), using lme4 1.1-21 (Bates et al., 2015) and MuMIn 1.43.17 for multi-model selection and averaging (Bartoń, 2020). Since percentage cover data are strictly bounded but not binomial, we logit-transformed the cover data prior to statistical analysis (Warton & Hui, 2011) to achieve normally distributed residuals and avoid heteroscedasticity. Assumptions were checked graphically as recommended by Smith et al. (2009). We modelled the effects of six environmental variables at plot and landscape level (Table 1), separately evaluating the landscape effect at four spatial scales (250, 500, 750 and 1,000 m). By selecting those environmental variables, we avoided strong inter-correlations among the predictors (|r| > 0.6, Pearson's correlation analysis, Appendix S2). Moreover, we included all two-way interactions between plot-level and plot-level variables and between plot-level and landscape-level variables (except for cover of grasses, which was analysed as a separate response variable), and the interactions between the proportion of WFS and the other landscape-level variables. Possible dependence in the data due to spatially close locations, repeated measurements in the same permanent WFS plots and usage of different seed mixtures was controlled by incorporating landscape units, WFS site and mixture as random variables in the models. Multi-model selection was based on Akaike information criterion (AIC) and relative importance values of all predictors were calculated using AIC weights from all analysed models. For model averaging, we selected all models with ΔAIC < 4 compared to the best model according to AIC. Temporal beta diversity was analysed to assess the variation in forb communities through the observation period. Specifically, we calculated total dissimilarity of sown and spontaneous forbs based on the Sørensen index and its nestedness and turnover components, using betapart 1.5.4 (Baselga & Orme, 2012). Mann–Whitney U tests (stats 4.0.3) were used to analyse statistical differences in beta diversity between sown and spontaneous forbs. In each of the three study years, plant species richness and cover of non-crop species were considerably higher on WFSs than on controls (p < 0.001, Mann–Whitney U test). On controls, 8.3 ± 1.5 (mean ± SE) spontaneous species (forbs and grasses) both with a very low cover occurred in addition to the crop (Figure 2), while total plant species richness on WFSs was 61.5 ± 1.0 (mean ± SE) per 1,000 m². On WFSs, most plant species were forbs (Figure 2a; Table S1). The species richness of sown and spontaneously established forbs was similar in the first study year. Contrary to the constant number of sown forb species, the species richness of spontaneous forbs increased significantly over the study years (Figures 2a and 3; Appendix S3). Of the 30 forb species sown, 22 species per 1,000 m² were found continuously in each of the survey years, resulting in an establishment rate of 73%. Achillea millefolium, Centaurea jacea, Daucus carota, Lotus corniculatus and Silene vulgaris had the highest frequency as they appeared on most WFSs in all years. Each of the sown forbs established on at least two WFSs and only three (sand-loam mixture) and five forbs (loess-loam mixture) were not detected on at least 50% of the WFSs in any of the 3 years. Mean temporal beta diversity of sown forb communities was <0.3 in both annual comparisons (Figure 4). Species turnover and nestedness contributed nearly balanced to dissimilarity. In addition to sown forbs, 28.4 ± 0.8 (mean ± SD) forb species established spontaneously with Fallopia convolvulus and Tripleurospermum inodorum being the most frequent species. Temporal beta diversity of spontaneous forb communities decreased over time, but was significantly higher than values for sown forbs (Mann–Whitney U test, p < 0.001). On average, turnover represented approximately 70% of spontaneous forb temporal beta diversity between 2017 and 2018, and was lower between 2018 and 2019, but still higher than nestedness. At all scale levels, forb species richness decreased with increased shading, with sown forbs being more affected, as shown by the significant interaction (Figure 5a–c). Soil fertility was important in most models, but only significant as interaction with status at two scale levels. Landscape-level factors did not show significant effects in the averaged model estimates and were inconsistent in their importance across scale levels and years (Figure S3). Forbs had the largest share of total plant cover on WFSs in all study years on most sites. The cover of sown forbs was significantly higher than the cover of spontaneous forbs, but decreased over the study period (Figure 2b; Appendix S3). Forb cover was negatively affected by the cover of grasses (Figure 3). While the cover of the sown forbs decreased strongly with more shading (Figure 5e), the cover of the spontaneous forbs remained largely stable (Figure S4), as indicated by the significant interaction. Soil fertility, the proportions of open habitats and WFS and the interaction status × cover of grasses had a higher relative importance (Figure S5), but did not show significant effects in the full averaged models. The 13 species with the highest cover of all forbs over all study sites and years were sown species, with A. millefolium, Galium album, D. carota, Leucanthemum vulgare and C. jacea being the most abundant. In the first study year, grass cover was similar to the cover of spontaneous forbs, but lower than the cover of sown forbs (Figure 2b). Grass cover increased and peaked in 2018, being higher on shadier and more productive sites (Figure 3; Appendix S3; Figure S6). The proportion of open habitats and habitat diversity had a higher predictor importance in models explaining grass cover at all scale levels, with a significant interaction year × habitat diversity at the 250 m scale level (Figure S7). Especially WFSs with a high cover of grasses in the first year of the study also showed high grass cover in the following years (2017 ~ 2018: rPearson = 0.44; 2018 ~ 2019: rPearson = 0.68, see Figure S8). The grass species Holcus lanatus, Bromus sterilis, Elymus repens and Poa trivialis were by far the most common species causing undesired monodominance (Table S2). According to the calculated PFI, the availability of feeding resources was significantly higher on WFSs than on controls, where nearly no pollen and nectar-producing forbs occurred (p < 0.001, Mann–Whitney U test, Figure 6). On WFSs, pollen and nectar were mainly supplied by sown forbs, which provided more than twice as much from June to October than spontaneously established forbs (Figure 7). From November to May, spontaneous forbs provided most of the pollen and nectar, albeit at mainly low values. The spontaneously established Capsella bursa-pastoris, Taraxacum sect. Ruderalia and Veronica persica had the highest PFI factors of all detected species. The sown L. vulgare, D. carota, A. millefolium, C. jacea and Trifolium pratense were most important for pollinators according to the PFI as they all have a generally high nectar and pollen production, and were found in high cover on the WFSs (see Table S1 for total forb species list and corresponding PFI factors and mean PFI values). The PFI decreased from 2017 to 2018/2019, with sown forbs declining less than spontaneous forbs. In the last 3 years of the 5-year AES funding period, total plant species richness on WFSs was more than six times higher than on control cereal fields. This confirms the effectiveness of perennial WFSs to enhance plant diversity in agricultural landscapes (Balzan et al., 2014). Although the permanent plots were located at field edges, which are usually more species-rich than the interior (Bellanger et al., 2012), the majority of controls had less than 10 species per 1,000 m², validating the negative effects of intensive agriculture found by Meyer et al. (2013) and Baessler and Klotz (2006). Within the field, plant diversity is probably even lower, highlighting the indispensable need for biodiversity-enhancing measures in agricultural landscapes. With an average 73%, the mean establishment rate of sown wildflowers in WFSs implemented by farmers was similar to experimental conditions in a previous study (Schmidt et al., 2020), and remained stable over all study years. The very low beta diversity further indicated a high level of similarity and persistence for sown forb communities on most sites from the second/third to the fourth/fifth year of establishment. None of the sown species of the two investigated mixtures completely failed to establish. Which sown species were actually present on a WFS varied greatly between sites. The reasons for such spatial variability are difficult to determine and are not always only linked to site conditions (Lepŝ et al., 2007). Inter- and intraspecific competition can also play a role (Wassmuth et al., 2009). However, in terms of risk diversification (‘insurance effect’), our species-rich mixtures were suitable for guaranteeing a good performance of the WFSs over the AES funding period. Altogether, forb cover on WFSs reached over 60%, which is very high (see Appendix S4). Contrary to Lepŝ et al. (2007), we found at least as many spontaneously emerging forb species as sown species on WFSs, both contributing to the high plant diversity. Thus, sowing a balanced ratio between competitive/weak and high/low growing forbs at a low sowing rate of 0.4–0.5 g/m2 left enough gaps for desired spontaneous forbs, also reducing the cost of the seed mixture considerably. The higher dissimilarity of spontaneous forb communities was mainly caused by species turnover, with beta diversity and turnover component decreasing with time, indicating the change from annual (e.g. Filago arvensis) to perennial (e.g. Euphorbia cyparissias) spontaneous forb communities. As a result of the extreme drought in 2018 and 2019, however, some annuals reappeared in vegetation gaps in 2019. Overall, the cover of spontaneous forb species was comparatively low. Weedy forbs became dominant only on very few WFSs (Figure S4). In contrast to grass-dominated sites, however, these WFSs still provided abundant nectar and pollen. The cover of spontaneous forbs remained relatively stable with regard to all investigated environmental factors. Altogether, shading negatively affected sown forbs most. On heavily shaded areas, only a few sown species, such as A. millefolium and D. carota, established successfully in our study. However, as many plant species in the mixtures are light-demanding species of grassland or mesophilic/thermophilic fringe communities, they are rather weak competitors to species which dominate shaded communities of nutrient-rich sites in agricultural landscapes, such as B. sterilis or E. repens. The cover of grasses increased parallel to shading, significantly reducing the cover of the sown forbs. Consequently, WFSs should not be established on north-exposed, heavily shaded field edges. For these sites, mixtures better adapted to shady conditions could be developed and applied. However, flower strips are an AES that aims to promote species of the agricultural landscape and thus of open and sunny habitats. Therefore, heavily shaded sites would only marginally benefit AES target species, if at all (Schmidt et al., 2022). The dominance of competitive grasses can considerably reduce the cover of sown forbs (Haaland et al., 2011). Grasses probably emerged from the soil seedbank, migrated from neighbouring vegetation or may already have been present due to insufficient seedbed preparation. They have an ecological function, for example, as host plants for some pre-imaginal butterfly stages, but grass seeds are usually still sufficiently present in the remaining semi-natural landscape structures. Thus, although it is common practice in other countries (Piqueray et al., 2019), grasses should not be included in WFS seed mixtures. When grass cover was high in the first year of the study, it was also high in the following years, confirming the findings of Weidlich et al. (2018), and indicating the importance of early established sown forbs to counteract grass dominance. Measures to promote biodiversity are particularly necessary in highly productive landscapes (Haenke et al., 2009), where semi-natural habitats are rare due to intensive agricultural use. In our study, sites with a higher soil fertility had a lower, but still high species richness of sown forbs. Weed pressure is known to be higher on fertile soils (Piqueray et al., 2019), but our study shows that by using site-adapted native seed mixtures, WFSs can be established successfully on poor as well as on highly productive soils. Overall, landscape effects played only a minor role compared to plot-level factors explaining plant composition at all scale levels, and relationships were often not consistent neither between years nor along a gradient of landscape levels (but see Appendix S3 for conditional averaged model estimates and Figures S3, S5 and S7). We expected that a high habitat diversity would provide a higher plant species pool and that a higher spatial connectivity to open habitats and other WFSs would presumably lead to an enhanced exchange of diaspores by wildlife, agricultural machinery or wind (Zonneveld, 1995). Although not confirmed by the full averaged model estimates, we found a trend that open habitats had differing effects on forb cover, depending on shading at the 250 m scale level or proportion of WFS at the 500 m scale level and soil fertility at the 1,000 m scale level. Furthermore, the cover of grasses increased less on WFSs with a high habitat diversity at higher scale levels over time. Possibly, the WFS lifetime of 5 years is too short to observe an approximation to the plant species pool at landscape level. However, we studied only plants as sessile organisms and local seed banks may play a more important role in the first colonisation period. For more mobile animal species groups, studies have found effects of landscape structure on the ecological effectiveness of WFSs (Haenke et al., 2009; Hellwig et al., 2022; Schmidt et al., 2022). In our study, grass cover and shading had a decisive effect on the performance of perennial WFSs. Dispersion in the data, however, indicates other unverifiable processes, for example farmers' disregard of recommended practices in terms of seedbed preparation, seed application or management. As the WFS selection in our study was random and anonymous, documentation of implementation and management was not possible, and these effects could not be included in our statistical analyses. According to our newly developed PFI, sown native species with potentially high pollen and/or nectar production contributed most to the food supply for pollinating insects on WFSs. Large amount of pollen and nectar, especially during the main flying time in summer, would lead to a clear preference of pollinators visiting WFSs, as flower cover is positively related to the absolute number of insect species (Ouvrard & Jacquemart, 2018; Warzecha et al., 2018). Plant species with high PFI values, like C. jacea, T. pratense and Knautia arvensis, were also found to be of high relevance for insects in other field studies (Haaland & Gyllin, 2010; Wood et al., 2017). Nevertheless, species with lower PFI values, like Campanula rotundifolia, can provide important pollen resources for oligolectic insects. Hence, a high plant and functional group diversity is associated with a higher availability of floral resources over time and different morphological adaptions of the fauna (Balzan et al., 2014; Wix et al., 2019). In our study, spontaneously established forbs accounted for one quarter to one third of the potential total pollen and nectar food provision, particularly species flowering in early spring like Draba verna or Veronica spp., and thus, probably also supported the local pollinator community, as shown by Warzecha et al. (2018). Spontaneous forbs can comprise a large amount of the wild bee pollen diet over the whole vegetation period (Ouvrard et al., 2018; Wood et al., 2017) meaning that their additional contribution to the ecological performance of the flower strip should not be underestimated (Di Pasquale et al., 2013). Thus, the evaluation of WFS performance should include the diversity of both sown and spontaneously established forbs. The decline of the PFI from 2017 to 2018/2019 is most likely due to the general decrease in total plant cover as a result of low precipitation in 2018 and 2019 (100–200 mm below the long-term mean, German Meterological Service, 2020). However, the proportion provided by sown and spontaneously established species remained largely stable over the years, regardless of weather conditions, making the results reproducible. The PFI was intended to improve the assessment of pollinator food supply, by weighting vegetation survey (relevé) data with plant species traits (nectar and pollen production, flowering period). A bias can arise from including plant cover (and not flower cover) recorded in a snapshot, but at the time of highest detectable species diversity. Nevertheless, statistical analysis showed that the PFI correlated strongly with the species richness and abundance of wild bees on WFSs (Schubert et al., 2021), indicating the success of this AES. Species-rich perennial native WFSs provided a diverse, forb-rich vegetation and related feeding resources for pollinators over the 5-year AES funding period, shaped by the good performance of the sown wildflower mixtures and a high diversity of spontaneously established forbs, which colonised gaps caused by the low sowing rate of 0.4–0.5 g/m2. The overall high plant species richness on WFSs is suitable to support a high number of pollinator species (Wix et al., 2019), which probably provide important ecosystem services both for agriculture and nature conservation (Balzan et al., 2014). WFSs could be established successfully largely independent of the landscape context. WFS implementation, however, should be avoided in heavily shaded sites, where grasses often become dominant and AES target species are unlikely to benefit. If grasses or other weedy species occur in dense and high stands, an appropriate and especially timely management, for example mowing, is necessary as also recommended in other studies (Kirmer et al., 2018; Wix et al., 2019). To avoid failures in implementation and management of WFSs and to achieve the maximum cost–benefit ratio, advisory services for farmers are necessary, as already practised in some EU countries (Leventon et al., 2017). This work was supported by the Ministry of Environment, Agriculture and Energy Saxony-Anhalt (Grant numbers: A03/2016, A02/2019) and by the Graduate Scholarship Programme of Saxony-Anhalt. Niels Hellwig acknowledges support by the German Federal Ministry of Food and Agriculture (MonViA; www.agrarmonitoring-monvia.de/en). The authors thank their colleagues and students Thomas Stahl, Eike Christoph, Heiner Hensen, Lenka Šebelíková, Anna Müllerová and Kamila Vítovcová for assisting with the data collection. They also thank the reviewers and the editor for their useful comments, which have significantly improved this manuscript. Open Access funding enabled and organized by Projekt DEAL. The authors declare that there is no conflict of interest. A.S. contributed to data collection, conducted trait measurements and led the writing of the manuscript; A.S. and N.H. analysed the data; A.S., S.T. and A.K. designed the experiment; Significant inputs from S.T., A.K., K.K. and N.H. helped to interpret and critically discuss the results. All authors contributed critically to the drafts and gave final approval for publication. Data available via the Dryad Digital Repository https://doi.org/10.5061/dryad.qnk98sfj9 (Schmidt et al., 2021). Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Effects of Spartina invasion on the soil organic carbon content in salt marsh and mangrove ecosystems in China

Abstract: en

Restoration promotes recovery of woodland birds in agricultural environments: A comparison of ‘revegetation’ and ‘remnant’ landscapes

Abstract: Ecological restoration in rural environments is a global challenge for the 21st century. Restoration measures—such as agri-environment activities, woodlots, natural regeneration and conservation plantings—collectively alter landscape structure with the aim of restoring conservation values that are characteristic of natural ecosystems. We tested the landscape-scale benefits of restoration for woodland birds, species of conservation concern in southern Australia, by assessing the richness and composition of avian communities in rural landscapes along a gradient in habitat restoration, benchmarked against landscapes with comparable extent of native vegetation. We selected 43 landscapes (each 8 km2) in Victoria, Australia, representing: (a) a trajectory of decline in the extent of remnant native wooded vegetation (‘remnant’ landscapes), (b) a trajectory of gain in planted vegetation (‘revegetation’ landscapes) and (c) a similar gradient comprising a mix of remnants and planted vegetation (‘mixed’ landscapes). In each landscape, repeat surveys of birds were undertaken at 12 sites, stratified in relation to land cover. Species richness of all terrestrial and woodland birds showed similar positive responses to total wooded cover in each landscape type, but woodland birds had reduced richness in ‘revegetation’ relative to ‘remnant’ and ‘mixed’ landscapes. Across all landscapes, key factors influencing richness were the extent of wooded cover and proportion comprised of plantings, scattered trees in farmland and mean annual rainfall. The composition of woodland bird assemblages differed between ‘remnant’ and ‘revegetation’ landscapes with predictable differences associated with foraging traits. Synthesis and applications. Restoration plantings stimulate recolonisation of otherwise-depleted landscapes, effectively reversing a decline in woodland birds. Key insights include: (a) benchmarking ‘revegetation’ against ‘remnant’ landscapes provides a valuable means to quantify restoration outcomes at the landscape scale; (b) time-lags in vegetation maturation contribute to a trajectory of recovery that differs from a trajectory of decline, in both richness and composition of the avifauna; (c) scattered trees have a critical role for avifaunal conservation in farm landscapes; (d) restoration plantings are most effective in ‘mixed’ landscapes, where complementary resources from remnant and planted vegetation are juxtaposed; and (e) restoration plantings on individual farms contribute to landscape-scale biodiversity gains while also having socio-ecological and production benefits.

Plant biodiversity declines with increasing coffee yield in Ethiopia’s coffee agroforests

Abstract: This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2022 The Authors. Journal of Applied Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society. 1Department of Ecology, Environment and Plant Sciences, Stockholm University, Stockholm, Sweden 2Department of Plant Biology and Biodiversity Management, Addis Ababa University, College of Natural and Computational Sciences, Addis Ababa, Ethiopia

Epidemiological differences between sexes affect management efficacy in simulated chronic wasting disease systems

Abstract: Sex-based differences in physiology, behaviour and demography commonly result in differences in disease prevalence. However, sex differences in prevalence may reflect exposure rather than transmission, which could affect disease control programmes. One potential example is chronic wasting disease (CWD), which has been observed at greater prevalence among male than female deer. We used an age- and sex-structured simulation model to explore harvest-based management of CWD under three different transmission scenarios that all generate higher male prevalence: (1) increased male susceptibility, (2) high male-to-male transmission or (3) high female-to-male transmission. Both female and male harvests were required to limit CWD epidemics across all transmission scenarios (approximated by R0), though invasion was more likely under high female-to-male transmission. In simulations, heavily male-biased harvests controlled CWD epidemics and maintained large host populations under high male-to-male transmission and increased male susceptibility scenarios. However, male-biased harvests were ineffective under high female-to-male transmission. Instead, female-biased harvests were able to limit disease transmission under high female-to-male transmission but incurred a trade-off with smaller population sizes. Synthesis and applications. Higher disease prevalence in a sex or age group may be due to higher exposure or susceptibility but does not necessarily indicate if that group is responsible for more disease transmission. We showed that multiple processes can result in the pattern of higher male prevalence, but that population-level management interventions must focus on the sex responsible for disease transmission, not just those that are most exposed.

City climate and landscape structure shape pollinators, nectar and transported pollen along a gradient of urbanization

Abstract: Urbanization gradients influence both landscape and climate and provide opportunity for understanding how plants and pollinators respond to artificially driven environmental transitions, a relevant aspect for the ecosystem service of pollination. Here, we investigated several aspects of pollination along an urbanization gradient in landscape and climate. We quantified wild hoverfly and bee abundances with trapping, standing crop of nectar with spectrophotometer, and the pollen transported by flower visitors with DNA metabarcoding, in 40 independent sites from semi-natural to built-up areas in Northern Italy. Direct and indirect effects were fitted considering landscape and climate variables. Linear and nonlinear relationships were detected along the urbanization gradient. Pollinator abundances increased quadratically and peaked at 22% of impervious cover with an 81% growth, and they decreased with green-patch distance by 37% and urban park largeness by 60%. This indicates that pollinators are more abundant at intermediate levels of urbanization. Climatically, pollinators diminished by up to 46% in areas with low spring–summer temperature seasonality: urban areas likely posing thermic stress. Furthermore, the sugar mass available in nectar increased by 61% with impervious cover and by 79% with precipitations, indicating that city nectars were less consumed or flowers more productive. Furthermore, the species richness of pollen decreased by 32% in highly urbanized areas, and contained a high incidence of exotic plants, hinting for anthropized, simplified plant communities. Synthesis and applications. Urbanization influences pollinator abundances, nectar resources and transported pollen in direct and indirect ways. Pollinators are negatively affected by a thermally harsh climate in highly urbanized areas with isolated green areas and large parks. Suburban landscapes demonstrated the highest pollinator presence. In the city core, flowers contained more nectary sugar in association with more precipitations, while pollinators collected pollen from a small number of plants, mainly exotic. These findings highlight the strong influence of urban landscape and climate on pollinators and plants, showing that cities are heterogenous realities. Patterns from this study will serve as basis for pollinator-friendly planning, mitigation and management of urban landscapes.

Herbaceous production lost to tree encroachment in United States rangelands

Abstract: Rangelands of the United States provide ecosystem services that benefit society and rural economies. Native tree encroachment is often overlooked as a primary threat to rangelands due to the slow pace of tree cover expansion and the positive public perception of trees. Still, tree encroachment fragments these landscapes and reduces herbaceous production, thereby threatening habitat quality for grassland wildlife and the economic sustainability of animal agriculture. Recent innovations in satellite remote sensing permit the tracking of tree encroachment and the corresponding impact on herbaceous production. We analysed tree cover change and herbaceous production across the western United States from 1990 to 2019. We show that tree encroachment is widespread in US rangelands; absolute tree cover has increased by 50% (77,323 km2) over 30 years, with more than 25% (684,852 km2) of US rangeland area experiencing tree cover expansion. Since 1990, 302 ± 30 Tg of herbaceous biomass have been lost. Accounting for variability in livestock biomass utilization and forage value reveals that this lost production is valued at between $4.1–$5.6 billion US dollars. Synthesis and applications. The magnitude of impact of tree encroachment on rangeland loss is similar to conversion to cropland, another well-known and primary mechanism of rangeland loss in the US Prioritizing conservation efforts to prevent tree encroachment can bolster ecosystem and economic sustainability, particularly among privately-owned lands threatened by land-use conversion.