Evaluating the effectiveness of retention forestry to enhance biodiversity in production forests of Central Europe using an interdisciplinary, multi‐scale approach

Abstract Retention forestry, which retains a portion of the original stand at the time of harvesting to maintain continuity of structural and compositional diversity, has been originally developed to mitigate the impacts of clear‐cutting. Retention of habitat trees and deadwood has since become common practice also in continuous‐cover forests of Central Europe. While the use of retention in these forests is plausible, the evidence base for its application is lacking, trade‐offs have not been quantified, it is not clear what support it receives from forest owners and other stakeholders and how it is best integrated into forest management practices. The Research Training Group ConFoBi (Conservation of Forest Biodiversity in Multiple‐use Landscapes of Central Europe) focusses on the effectiveness of retention forestry, combining ecological studies on forest biodiversity with social and economic studies of biodiversity conservation across multiple spatial scales. The aim of ConFoBi is to assess whether and how structural retention measures are appropriate for the conservation of forest biodiversity in uneven‐aged and selectively harvested continuous‐cover forests of temperate Europe. The study design is based on a pool of 135 plots (1 ha) distributed along gradients of forest connectivity and structure. The main objectives are (a) to investigate the effects of structural elements and landscape context on multiple taxa, including different trophic and functional groups, to evaluate the effectiveness of retention practices for biodiversity conservation; (b) to analyze how forest biodiversity conservation is perceived and practiced, and what costs and benefits it creates; and (c) to identify how biodiversity conservation can be effectively integrated in multi‐functional forest management. ConFoBi will quantify retention levels required across the landscape, as well as the socio‐economic prerequisites for their implementation by forest owners and managers. ConFoBi's research results will provide an evidence base for integrating biodiversity conservation into forest management in temperate forests.


| Forest biodiversity and the need for integrated forestry
Since the UN Convention on Biological Diversity (CBD) was signed in 1992, the conservation of biodiversity has become a global commitment. Yet, species extinction rates continue to accelerate, as recently documented by the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES, 2019).
In Europe, forests are the dominant natural vegetation form, and present the primary evolutionary background of today's plant and animal species and communities. Forests provide habitat for numerous species and therefore play a key role in the conservation of biodiversity (European Environment Agency, 2016;Forests Europe, 2015).
Forests are also valued for safeguarding ecosystem services, such as production of wood, protection against natural disasters, and provision of recreational opportunities. Further, forests and forestry traditionally play an important role in European culture, provide employment, and contribute to local economies. In view of these manifold human demands for wood and other ecosystem goods and services, the vast majority of forests in Europe are managed to serve economic, social, and environmental functions at the same time (Forests Europe, 2015).
These multiple functions are not always easily reconciled (Niemelä et al., 2005;Verkerk et al., 2014). For example, temperate forests managed for high economic benefit are unlikely to be rich in biodiversity, because trees and stands are typically harvested at economic maturity, which can be reached, depending on the tree species, after 60-200 years. This is relatively early in the life span of forests whose trees may live for several hundred years, and long before the development of old and structurally rich forest successional stages with their high level of biodiversity and many unique species (Gustafsson et al., 2012;Hilmers et al., 2018;Scherzinger, 1996).
Thus production forests, here understood as "forests available for wood supply" (FAWS) (UNECE/FAO, 2000) lack in many cases the structural heterogeneity of natural forest ecosystems. Specifically, they lack late successional stages and structural elements (e.g., snags, coarse woody debris, and canopy gaps) created by decay processes or natural disturbances. This is of major hindrance for various groups, such as epiphytic lichens and plants, saproxylic fungi and invertebrates, and cavity-nesting mammals and birds, which depend on the specific resources available in old forests (Goldberg, Kirby, Hall, & Latham, 2007;Kraus & Krumm, 2013;Siitonen, Martikainen, Punttila, & Rauh, 2000). Forests rich in late successional stages and structural elements are extremely rare in Europe. Indeed, across Europe, only marginal proportions (1.5%) of forest are permanently taken out of production ("no active intervention," Forests Europe, 2015, p. 160) and set aside in protected areas such as National Parks and other strictly protected forests. Reserves without human intervention are valuable reference areas for research on natural forest ecosystems, and important cornerstones for the conservation of forest biodiversity. However, a reserve-based segregative approach alone cannot ensure the conservation of forest biodiversity because small reserve sizes and insufficient connectivity limit the effectiveness of these remaining islands of natural forest for biodiversity conservation (Bollmann & Braunisch, 2013).
Thus, integrative conservation measures in production forests are a crucial complement in safeguarding forest biodiversity not only at the local scale of protected areas, but in particular across entire landscapes. Increasingly, retention-forestry approaches (Bauhus, Puettmann, & Messier, 2009) are applied "to provision for continuity in structural, functional, and compositional elements from the preharvest to the postharvest forest" (Gustafsson et al., 2012). Retention forestry, that is, retaining small patches or structural elements within a production forest matrix, is implemented particularly in public forests, and often related to requirements originating from EU legislation. Integration of biodiversity conservation into forest management is a major policy goal throughout Europe (cf. the EU Forest Strategy, European Commission, 2013;and the EU Biodiversity Strategy, European Commission, 2011; the German National Strategy on Biological Diversity, BMU (2007); and the German Forest Strategy 2020, BMELV, 2011). In the federal system of Germany, the responsibility for programme implemention is focused on state level (ForstBW, 2016).

| State of knowledge and research gaps
Retention forestry has been practiced and extensively studied in clear-cutting systems for 30 years (Fedrowitz et al., 2014;Gustafsson et al., 2012;Mönkkönnen, Ylisirniö, & Hämäläinen, 2009). However, there is still a need to extend research on the effectiveness of retaining structural elements for biodiversity conservation from clearcutting to the continuous-cover forestry systems (Gustafsson et al., 2019), which dominate major parts of the temperate regions in Europe.
Few studies have examined retention approaches in selectively logged forests in the temperate zone (Müller, 2005); thus, retention guidelines, as they are currently implemented across central Europe, are based on plausibility and expert knowledge rather than on scientific evidence (Vítková, Bače, Kjučukov, & Svoboda, 2018).
For example, in the state forests of Bavaria, Germany, an average of ten habitat trees per hectare has to be retained in near-natural stands (BaySF, 2009), whereas in the neighboring state of Baden-Württemberg, one group of habitat trees consisting of about 15 trees is to be secured per 3 ha (ForstBW, 2016). Across Europe, all countries use their own, variable prescriptions (Sotirov, 2017;Winter et al., 2014). This also holds for FSC and PEFC certification standards in Europe: In national certification standards, quantitative retention targets are either not specified, or they vary greatly (e.g., 1-10 living habitat trees per ha; 1-20 dead trees per ha) among European countries (Gustafsson et al., 2019).
Trade-offs between conservation and production objectives may explain different thresholds set in conservation programmes in production forests. Yet, there is obviously also uncertainty about the thresholds required for retention amounts to achieve the desired effects on biodiversity. Recommendations to date, that is, for the required amount of deadwood and habitat trees, have been based mostly on single taxonomic groups such as forest birds, saproxylic insects, lichens, or fungi (Müller & Bütler, 2010;Sandström et al., 2019).
Studies investigating a wide spectrum of species including multiple taxa and trophic groups in the same study system are rare (Franklin, Macdonald, & Nielsen, 2019;Müller & Bütler, 2010;Paillet et al., 2018;Ranius & Fahrig, 2006;Vítková et al., 2018). Further, relationships between forest structure and biodiversity have not been studied comprehensively in a landscape context (Mori, Tatsumi, & Gustafsson, 2017). Landscape ecology and community ecology theory (Leibold et al., 2004;Lindenmayer & Franklin, 2002;Tscharntke et al., 2012) for instance would predict that a network of stands rich in deadwood may contribute more to biodiversity than uniformly distributed retention of moderate amounts of deadwood across the landscape (Müller & Bütler, 2010). In other words, the amounts and spatial distribution of structural elements, such as deadwood, required at the landscape scale to allow for functional connectivity of taxa dependent on old forests have not received sufficient attention (Müller & Bütler, 2010;Percel, Laroche, & Bouget, 2019;Ranius & Fahrig, 2006). Increasing the knowledge of the minimum landscape-scale requirements for biodiversity conservation through retention forestry would help to improve existing retention schemes (Mori et al., 2017).
Evidence-based prescriptions for the amount and distribution of retention elements, however, will not guarantee the implementation of recommended measures into forestry practice. Successful biodiversity conservation in multi-functional forests necessitates that conservation objectives are compatible with the aspirations of landowners and other goals of forest management (BMELV, 2011), and that potential trade-offs are known and regulated by respective conservation policies. Conservation measures often represent a loss in income, cause issues of work-safety, and result in higher management costs for forest owners (Rosenkranz, Seintsch, Wippel, & Dieter, 2014). Hence, the success of biodiversity conservation approaches such as retention forestry is greatly affected by the socio-economic context, including local knowledge, traditions, motivations, and practices; yet, little is known about these human dimensions underlying the conservation of forest biodiversity in Central Europe and elsewhere (Bennett et al., 2017;Gorenflo & Brandon, 2006;Maier & Winkel, 2017;Rutte, 2011).
The inter-and transdisciplinary approach and the multi-scaled design of the Research Training Group ConFoBi (Conservation of Forest Biodiversity in Multiple-use Landscapes of Central Europe) explicitly address these two major gaps in forest biodiversity research, namely the influences of the landscape context and the relevance of the socio-economic context for the effectiveness of retention measures to maintain biodiversity in multi-functional forests of temperate Europe.
To the best of our knowledge, such a broad integrative analysis of the ecological and social preconditions for forest biodiversity conservation has not yet been attempted in Central Europe.
In the following, we first present the rationale and lead questions underlying the ConFoBi research programme, then outline how these ideas were translated into a research design, describe how this design was implemented in the Black Forest, south-west Germany, characterize the pool of 135 ConFoBi study plots and place the study system in a European context. We then briefly present specific projects, their methods and linkages. Finally, we discuss perspectives of ConFoBi for research and conservation, and encourage scientists from multiple disciplines to join the ConFoBi Research Training Group.

| Guiding principles
The ultimate goal of ConFoBi is to provide an evidence-based framework for socio-political decision-making, which rests on a comprehensive analysis of retention forestry and its consequences for and linkages to ecological and societal systems in the multiple-use landscapes of temperate Europe. Key principles guiding ConFoBi's research toward this goal are (a) the application of a transdisciplinary approach, (b) the development of a consistently interdisciplinary research programme, and (c) the implementation of a real-world representative study system spanning across multiple spatial scales, from plot (1 ha) to landscape.

| Transdisciplinarity
Biodiversity conservation often fails because of a lack of communication and understanding between researchers and practitioners (Mehring, Bernard, Hummel, Liehr, & Lux, 2017;Pregernig, 2014;Tinch et al., 2018). To overcome this deficiency, ConFoBi chose to apply a transdisciplinary approach to maintain a regular dialogue be-

| Interdisciplinarity
Successful biodiversity conservation in multi-functional forests requires a solid ecological basis, but lastly depends on the compatibility of conservation measures with the human dimensions of forest management. In response to these requirements, ConFoBi adopted also an explicitly interdisciplinary approach, integrating multi-scale ecological studies of forest biodiversity with social and economic studies of the preconditions and consequences of biodiversity conservation. To maximize interdisciplinary synergies, ConFoBi researchers of numerous disciplines work in the same set of study plots and surrounding landscapes and hence also in the same socio-economic environment.

| A real-world study system
Experiments have played an important role in research on the effectiveness of retention forestry (see Gustafsson et al., 2012 for an overview). Experimental results will generate valuable knowledge on the potential outcomes of retention approaches. However, whether these results are transferable to the conditions of real landscapes under different types of multi-functional forest management remains questionable (Mori et al., 2017). In order to study retention effects in a setting representative of regular forestry practices, and to generate results directly relevant to forest owners and practitioners (see above, Section 2.1.1), ConFoBi is using plots in state-owned forests available for wood supply, as a real-world study system for its research programme. ConFoBi concentrates its research in a specific region, the Black Forest in southern Germany, with its specific forestry and biodiversity conservation structures and practices, to allow for transdisciplinarity. While the study system is regional, the findings are interpreted within a larger framework to achieve relevance for temperate forests with continuous-cover forestry across Europe.

| Research programme and lead questions
ConFoBi assesses whether and how structural retention measures contribute to the conservation of forest biodiversity and analyses the potential of retention forestry to be adopted by landowners and be supported by stakeholders, using montane forests of the Black Forest, Germany, as a model system. (1 ha) forest structure, that is, abundance, quality, heterogeneity, and spatial distribution of structural elements within the forest and (b) to the landscape context, that is, abundance, quality, heterogeneity, and spatial distribution of forests in the surrounding landscape. Among structural elements, the focus is on habitat trees (large trees with hollows, cracks, crevices, crown deadwood, and other habitat features sensu, Bütler, Lachat, Larrieu, and Paillet, 2013) and deadwood, whereas site conditions and tree species compositions will be considered as covariates. Module C uses a subset of the ConFoBi plots selected along social gradients (e.g., ownership, protection status) to assess how forest practitioners in different settings perceive and practice biodiversity conservation, and will assess costs and benefits to model and optimize the economic efficiency of retention measures. Using a translational approach (Musacchio, 2009;Schlesinger, 2010), module D focuses on the interface between science and practice to assess how knowledge is generated and evidence is translated into practice, and to provide integration and communication between ConFoBi and forest and conservation managers and policy makers. In summary, ConFoBi: • investigates the effects of forest structure including structural elements such as habitat trees and deadwood, and landscape context on multiple taxa including different trophic levels and functional groups (Modules A, B); • analyses how forest biodiversity conservation is perceived and practiced, and what costs and benefits it creates (Module C); and • identifies how biodiversity conservation can be effectively and efficiently integrated in multi-functional forest management through a translational approach focusing on epistemologies (how knowledge is generated and transferred) and by developing evidence-based guidelines for practitioners (Module D).

| The Black Forest as a model system
ConFoBi has implemented its research programme in the southern Black Forest, Germany, as a model system for temperate forests.
The Black Forest is a forest-dominated low mountain range within a multiple-use landscape typical of central Europe. It extends over an area of about 5,000 km 2 and has a forest cover of 75% (365,000 ha). In the western and southern parts, the geology is dominated by granite and gneiss, whereas in the eastern and northern parts, sandstone prevails. The macroclimate is strongly influenced by the elevational gradient, ranging from 120 to 1,493 m a.s.l., with an average annual temperature between 4°C at the higher elevations and 10.4°C in the lowlands (Gauer & Aldinger, 2005). Norway spruce (Picea abies (L.) H. Karst.) is the most important tree species (42.8%), in particular in the northern and eastern Black Forest, whereas silver fir (Abies alba Mill.) (18.5%) and beech (Fagus sylvatica L.) (15.3%) maintain a higher share of the tree species composition in southern and western parts. In recent decades, the forests in the Black Forest have become older, more mixed and more structurally diverse (Kändler & Cullmann, 2016).
In 2012, 71.5% of the forested area was covered by mixed stands, 29.5% was stocked with trees older than 100 years, two-layered and multi-storeyed forests made up 55% and 26%, respectively (Kändler & Cullmann, 2016). The average amount of deadwood (>10 cm diameter, incl. stumps >20 cm diameter) is 33.4 m 3 /ha, and there are on average five habitat trees per ha; the latter comprise 83% deciduous trees and more than two-thirds of these trees have a diameter (dbh) >50 cm (Kändler & Cullmann, 2016).
The forested landscape is further characterized by a fine-scale mosaic of ownerships, with some regional differences in the relative share of ownership types. In the Black Forest 39% of forests are privately owned, 27% are state owned, and the remaining 34% are owned by municipalities and public corporations (Kändler & Cullmann, 2016).
Forests in the region are managed using a variety of silvicultural systems. Owing to strict forest legislation, a prevailing paradigm of close-to-nature forest management, and the wide coverage of certifications systems (PEFC and FSC), clear-cutting is basically not practiced in the Black Forest. Instead, regeneration methods such as shelterwood, group shelterwood ("Femelschlag"), strip cutting, and single-tree selection ("Plenter" forest) are employed to create and maintain structurally diverse and species-rich continuous-cover forests (Bauhus & Pyttel, 2015), although the selection systems tend to disadvantage light demanding species such as pine and oak (Bauhus, Puettmann, & Kuehne, 2013).
In the State of Baden-Württemberg, including the study area, a retention programme ("AuT-Konzept" [translated as: old and deadwood programme]; ForstBW, 2016) stipulates the retention of one group of habitat trees ("Habitatbaumgruppe") consisting of about 15 trees per 3 ha throughout state forests. Moreover, efforts have been taken to assist forest biodiversity on private property through contractual nature conservation, extension services, or subsidies.

| Selection and establishment of study plots
The study design of ConFoBi is based on an "all-measurements-on-allplots" approach with a common pool of 135 quadratic study plots (1 ha) distributed along two environmental gradients: (a) landscape-scale forest connectivity (measured by the proportion of forest in the 25 km 2 surrounding plot centers) and (b) retention-related forest structure at the plot scale, that is, richness in habitat trees and deadwood per ha.
For logistic and authorization reasons, all plots were selected in state forests and outside of wildlife protection areas with restricted access.
Plots were preselected on the basis of a set of criteria (Table 1) to reduce variation in confounding factors. Major criteria were topography (<35° slope, >500 m a.s.l.), stand age (>60 years), and the absence of waterbodies and human infrastructures. The remaining F I G U R E 1 Illustration of the ConFoBi research concept. The study system (top) is represented (gray arrows) in a research system (bottom) consisting of four modules areas were classified into three forest-connectivity classes (<50%, 50%-75%, >75% forest) based on the amount of forest in the surrounding landscape, as calculated within a circular moving window of 25 km 2 based on a raster map of 25 × 25 m resolution.
To determine the structural gradient, using designated habitat trees of the AuT-Programme was not an option. At the time of plot selection, in 2016, AuT was still in its initial implementation stage, habitat trees had been selected for only minor parts of the study, and designated habitat trees were not (yet) older than the surrounding stands. Instead other proxies for forest structures that are expected to be enhanced by retention forestry, that is, old or dead trees rich in microhabitats, were used for plot selection. We screened the study area for 1 ha-plots with low (0), medium (1-9), and high (≥10) numbers of standing dead trees. Trees were visually assessed using stereo aerial color-infrared imagery, as provided by the state agency of spatial information and rural development of Baden-Württemberg (LGL), together with a stereo viewer. Fifteen plots from each category (low, medium, and high structure) were then randomly selected for each of the forest-connectivity classes, ensuring that the centers of two neighboring plots were >750 m apart, resulting in 135 plots arranged in a stratified design with 3 × 3 = 9 categories.
These 135 candidate plots were cross-checked with the local forest managers for the possibility to largely exclude forestry operations (e.g., harvesting and road construction) on these plots until the end of the anticipated maximum ConFoBi funding period (2016)(2017)(2018)(2019)(2020)(2021)(2022)(2023)(2024)(2025). Individual plots, for which operations were already scheduled, were randomly replaced by plots of the same category. Plots were established in the field by marking the center with a metal rod and a magnet at ground level. Afterward the borders were identified with a differential GPS and marked. A schematic overview of sampling design on the plots is shown in Figure 4. All plots were inventoried by measuring the DBH of every tree (above 7 cm DBH) and the amounts of lying and standing deadwood.
Because habitat trees are an important structural element in retention approaches, for each study plot we mapped the 15 largest trees (measured by crown area from aerial images) and quantified in great detail their existing microhabitats (such as hollows, cracks, and crevices; Larrieu et al., 2017). Abiotic measurements in the plots include air temperature at 150 cm height at 1-hr intervals in the center of every plot and quantification of light levels at the understorey, as well as soil nutrients. Additional abiotic data (e.g., precipitation) are derived from existing sources on a landscape scale (atlas data). Landscape patterns have been described using a range of metrics (measured around plot centers), including amount of forest cover, land cover composition, and edge density (Appendix 1).

| The study system in a European context
The ConFoBi study system is embedded in a typical multiple-use landscape of Central Europe; its study plots are part of a forest landscape with a variety of silvicultural systems and ownership types. Yet, the selection of study plots ensured that a wide range of the conditions found in European montane forests are represented; this in particular holds for the two design gradients, TA B L E 1 Criteria used to identify potentially suitable plots as well as geodata sources used for plot selection. After preselection based on the general criteria, potential plots were classified according the two design gradients, forest structure, and landscape pattern The distribution of the 135 study plots along these two design gradients is shown in Figures 5 and 6. The landscape gradient ranges from a forest matrix with high connectivity, to highly fragmented forests with an open matrix ( Figure 5). Mean volumes of standing and lying deadwood amount for 14.0 and 43.6 m 3 /ha, respectively ( Figure 6).
The plots richest in deadwood with close to 500 m 3 /ha are located in strict forest reserves. According to studies in forest reserves of temperate Europe, natural amounts of >200 m 3 /ha of standing and downed deadwood can be expected in montane beech-fir forests (Bujoczek, Szewczyk, & Bujoczek, 2018

F I G U R E 3 Location of the 135
ConFoBi study plots in the Black Forest (main map; light gray: open land; dark gray: forest), and in Germany and the State of Baden-Württemberg (insert). All plots are 1 ha in size and >750 m apart. Point size indicates three levels of landscape-scale forest connectivity (small <50%, medium 50%-75%, large >75% forest cover) in the 25 km 2 surrounding a plot. Point color indicates three levels of plot-scale forest structure (red: 0, yellow 1-9, black ≥20 standing dead trees per ha). Plots richest in structure (≥20 standing dead trees) include stands >200 years of age and plots in strict forest reserves, where harvesting has been excluded F I G U R E 4 Schematic overview of the sampling design on the ConFoBi study plots. All plots are 1 ha in size and north south aligned. Wooden poles mark the plot center as well as all four corners and all four sides. The center point is permanently marked with a white plastic reference point and a strong magnet on ground level. Trees along the borders and in the corners of the plot are also marked with long lasting light blue color. The following measurements are collected on all 135 study plots, except for a few additional measurements, for example, for pilot studies, that are taken on subsets of plots (n plots in brackets) only: Flights with unmanned aerial vehicles covering the whole plot, within the whole plot full inventory of all trees with a DBH above 7 cm as well as full presence list of all herbaceous plants; at the center point temperature, bird counts, automatic acoustic recorders for soundscapes, light traps for moths (n = 28); automatic camera traps for large mammals at the center point and the locations of the Flight Interception Trap (FIT); a V-transect is aligned from the north-western corner via the central point of the southern border to the north-eastern corner, data for the ForMIn (Forest management Index) within a 4m wide strip and data on standing deadwood within a 10 m wide strip are collected; terrestrial laser scans (TLS) on five locations (plotcenter, two bat recording, two insect collecting sites); light measurements (photosynthetic active radiation = PAR) along a transect subdivided into twelve subplots of 40 cm × 40 cm plus one soil sample from the middle of the transect, the transect was placed north south in the grid cell of 10 m × 10 m with the highest variability of crown height of each study plot; tree microhabitats (TreMs) on the fifteen trees with the largest crown identified from aerial images; epiphytes on the five trees with the largest crown identified from aerial images and on five trees of the most common species and of average DBH of each plot; ticks are collected with a 1 m × 1 m flag along a 100 m transect aligned northwest to south-east through the center point in four 25 m steps (n = 34); six floral subplots of 5 m × 5 m which detail species list plus cover, in addition one soil sample and one hemispherical photograph were taken at the center of each subplot; FIT in the north-western and southeastern area of the plot; automatic acoustic bat recorders placed in one area of the plot with high structure and one area with low structure; sifting leaf litter for weevils, centipedes, and millipedes along deadwood next to beech trees (n = 43)

| Complementing other research efforts
To our knowledge, ConFoBi is the first research programme that investigates how retention of forest structures affects forest biodiversity in uneven-aged and selectively harvested continuous-cover forests of temperate Europe (Gustafsson et al., 2019). With its explicit focus on retention practices and by integrating the landscape and socio-economic contexts of forest biodiversity, ConFoBi complements other research efforts on forest biodiversity. Synergies mayarise with the "Biodiversity Exploratories" (Fischer et al., 2010), "FunDivEUROPE" (Baeten et al., 2013) and its follow-up process "SoilForEUROPE" (http://websie.cefe.cnrs.fr/soilf oreur ope/), as well as small FOREST (Valdés et al., 2015). Other related projects in-

| Knowledge and expertise for integrated forest management
ConFoBi's research results will lead to quantitative target values describing the optimal amount, quality, and distribution of retention sites for different species and taxonomic groups as well as to recommendations how to efficiently integrate these targets into multi-functional forest management. ConFoBi will provide an extensive case example for both reseach and management of forest biodiversity across the multiple-use landscapes of the temperate zone.
In its first stage, ConFoBi focusses on describing patterns of biodiversity in relation to forest structures at spatial scales from plot to landscape. In later stages, the focus will shift from biodiversity patterns to functional relationships (e.g., between forest structures and organisms; among taxa and across trophic levels) and processes, which shape the observed patterns. To understand the socio-economic context of biodiversity conservation in managed forests, ConFoBi will move from describing general knowledge structures to actions and impacts, to assess how practitioners integrate conservation in forest management. Finally, in an interdisciplinary synthesis, ConFoBi will quantify retention targets, F I G U R E 5 Gradient of forest cover at the landscape scale (connectivity). Boxplot and frequency distribution of the 135 study plots by forest cover in 25 km 2 surrounding plotcenter n and in particular the amounts and distribution of habitat trees and deadwood, required across the landscape for effective biodiversity conservation, identify the trade-offs between forestry and biodiversity conservation, and will elaborate the socio-economic prerequisites for their implementation by forest owners and managers. Wider applicability of the results will be ensured in cooperation with research groups elsewhere, for example, in the newly established COST Action BOTTOMS-UP (https ://www.cost.eu/ actio ns/CA182 07/#tabs|Name:overview).
ConFoBi emphasizes the ecological foundations and socio-economic frameworks of retention approaches in continuous-cover forestry (Gustafsson et al., 2019). Preliminary results indicate positive overall effects of retention of structural elements for forest biodiversity but also high variability among taxa and landscape F I G U R E 6 Gradients of forest structure at the plot scale. Boxplots and frequency distributions of the amount of lying (top) and standing (bottom) deadwood on the 135 study plots (1 ha)

A1 Remote sensing based methods for the assessment of forest structures
The project develops remote-sensing methods to assess abundance, heterogeneity, and spatial distribution of structural elements, as predictors of biodiversity across multiple temporal and spatial scales. Advanced remote-sensing techniques, such as LiDAR and digital stereophotogrammetry, in conjunction with platforms such as Unmanned Aerial Vehicles and terrestrial laser are used and new algorithms for extraction and unification of data are developed to provide structural information at scales appropriate for other ConFoBi projects.

A2 Retention of structural elements in selectively used forests
Live or dead standing trees that provide tree-related microhabitats (TreMs) such as cavities, large dead branches, loose bark, epiphytes, bracket fungi, cracks, or trunk rot are defined as habitat trees. In natural forests habitat trees and TreMs may be highly clumped and variable, which may affect structure and viability of communities and meta-communities of species reliant on microhabitats. A2 analyses and predicts TreMs on potential habitat trees, and assesses factors relevant for habitat tree selection, such as distribution, quality and longevity of habitat trees and TreMs, forest management intensity, and landscape context.

B1 Epiphyte and microhabitat diversity and function on habitat trees
The project quantifies the role of habitat trees for conserving forest epiphytes, and assesses how habitat trees differ in their epiphytic lichen and bryophyte diversity from trunk up to the crown and in variation of the species identity of the habitat tree, the main tree species of the stand, the diameter and microhabitat characteristics of the habitat tree, and landscape-scale forest connectivity.

B2 Mechanisms of vegetation change and diversity in retention forestry
The project addresses the role of habitat trees and deadwood, and their landscape context, for plant diversity by disentangling various effects of abiotic changes on plant performance. B2 determines the composition and spatial distribution of understorey higher plants and ground-dwelling mosses, to assess how heterogeneity in forest structure influences resource heterogeneity and availability of light for understorey vegetation, and hence species diversity and composition, and to quantify the role of forest structure and resource heterogeneity on trait distributions and functional diversity of the understorey.

B3 Diversity and functions of plant-insect interactions along a retention gradient
The project addresses the overall hypothesis that stand-scale retention measures influence the diversity and trophic interactions of hymenoptera and other arthropods, which are mediated by forest composition and configuration. B3 analyses the relationship between components of insect diversity and environmental variables related to retention (e.g., deadwood, microhabitats). To quantify trophic interactions and food webs plant galls with their gall-forming communities and cavity-nesting Hymenoptera are studied.

B4 Functional connectivity among saproxylic beetles in dead-wood patches
The project investigates species and genetic diversity of saproxylic beetles and hymenoptera, and assesses gene flow of dead-wood specialists with different dispersal ability as a function of abundance, distribution and isolation of deadwood across the landscape.
Species are sampled at dead-wood patches of different size and distance by means of trapping and metabarcoding. Genetic diversity is quantified using metabarcoding for total diversity and species-specific markers (RADSeq) for genetic distance. Finally, B4 infers thresholds for functional connectivity among dead-wood patches.

B5 Landscape-moderated use of forest structures by bats
The project assesses linkages between multi-scale forest structures and bat activity, richness, and diversity, as well as the functional importance of structures, specifically for foraging and commuting. Automated acoustic recorders are deployed to detect bat species(-groups) and to quantify their activity. Presence, richness, diversity, and activity of bats are then related to measures of forest structure, including LiDAR-information capturing the 3D-characteristics of subcanopy space, and landscape-scale connectivity metrics. Availability of food will be assessed (with B3, B4, and B6) to address inter-trophic relationships that may explain the observed results.

B6 Multi-scale assessment of bird-forest relationships
The project quantifies occurrence and abundance of bird species based on repeated aural-visual point counts in order to assess multiscale structural predictors of avian diversity. B6 first focusses on linkages between retention elements, landscape patterns, and birds; thereafter, on this basis, B6 assesses inter-trophic relationships of birds that may explain the patterns observed. Assessment of food availability (with B3-B5) will allow elucidating functional relationships.

B7 Soundscapes
This project examines how acoustic diversity can be quantified to reflect avian diversity, and how these metrics can be combined with other variables to describe a forests potential for biodiversity. To test the hypothesis that acoustic diversity of avian vocalizations varies in response to multi-scale forest structures B7 collects audio files and analyses acoustic indices to identify and track phenological changes in avian communities. These indices are validated against observed species richness and abundance (B6), then related to additional taxonomic groups (B3-B5) and metrics of forest structure (A1, A2, B2) at the plot and landscape level to model optimal combinations of indices and quantitative benchmarks to predict biodiversity.

B8 Ungulate-forest relationships
The project focusses on the relationship between roe deer Capreolus capreolus and forest structure, addressing both bottom-up and top-down effects. Local abundance and habitat selection of roe deer, quantified through camera trapping and pellet counts, are related to biomass and composition of understorey vegetation, and to forest structure and landscape pattern. Effects of deer on plant community composition and understory species richness are addressed through browsing surveys and germination experiments based on deer feces. Finally, co-occurrence and direct and indirect biotic interactions among roe deer and other taxa (with B1-6) are assessed.

B9 Effects of forest structures on fungi
The project quantifies the impact of multi-scale forest structure and fragmentation on the diversity of fungi with focus on woodinhabiting species and interactions between fungi and other organismic groups. It combines standard monitoring techniques with genetic analyses of selected species to determine genetic differentiation among local populations along gradients of forest fragmentation. Taxonomic, functional and genetic diversity measures of fungi are related to various forest structural and landscape metrics, as well as to diversity and composition of plants and animals (B1-8).

Box 1 (Continued)
settings; further, our findings suggest high potential for optimizing the integration of retention into management practice. The first ConFoBi publications described technical and methodological advances for quantifying forest structures; for example, we found a high potential for optimizing the quality in reconstruction of 3D forest models from aerial images based on drones (Frey, Kovach, Stemmler, & Koch, 2018), showed how the abundance and diversity of tree-related microhabitats can be predicted with readily available forest attributes (Asbeck, Pyttel, Frey, & Bauhus, 2019), recorded a lichen species new for Germany (Wirth, Tønsberg, Reif, & Stevenson, 2018), optimized the trap design to capture flying arthropods (Knuff, Winiger, Klein, Segelbacher, & Staab, 2019), and showed that the occurrence of specialist herbivore communities might be best explained by plant species composition rather than the abiotic environment C Human dimensions

C1 Economic valuation of biodiversity-oriented forest management strategies
The project hypothesizes that orienting forest management toward biodiversity has economic implications that can be quantified on different spatial scales. First, the value of biodiversity is expressed as an opportunity cost within a forward-looking simulationoptimization approach; thereafter, C1 quantifies social benefits of biodiversity, investigating optimal retention levels and spatial allocation of retention under multiple uncertainties including climate change. Finally, C1 aims to propose cost-effective measures to promote biodiversity. For this, C1 collaborates with the A and B projects for structural and biodiversity data, and with C2 and D1 to find out which role economic parameters play in decision-making related to biodiversity.

C2 Local biodiversity knowledge and forest conservation practices
This project connects three distinct fields of research. It conceptually links socio-psychological findings of research on forest owners, their attitudes, practices and institutional integration and the work on traditional ecological knowledge and practices in conservation to the literature on forest and biodiversity conservation discourses and professional forest management paradigms among practitioners and policy makers. These questions have never systematically been related to each other in an empirical study. After a systematic review, exploratory interviews, and document analysis, qualitative interviews are combined with participatory observation. Finally, a representative mail survey will quantify the derived patterns.

D1 Professional epistemologies and integration of biodiversity-related knowledge into socio-political decision-making
The project assesses under which conditions specific stocks of biodiversity-related knowledge are taken up in different decisionmaking contexts. D1 is built on the hypothesis that both within biodiversity science and biodiversity policy and management there are distinguishable professional epistemologies that have an impact on problem definition, agenda setting as well as the formulation and implementation of problem-solving strategies. The project uses a qualitative-interpretative approach to investigate the specific "thought styles" of ConFoBi-relevant scientific disciplines as well as those found in practical decision-making contexts. Major approaches are document analysis, expert interviews, and participatory observations.

D2 Evidence-based biodiversity management of forests
This project explores the scientific basis of management principles for conservation in forests. One challenge is the plethora of studies of different design and quality, sometimes yielding equivocal conclusions about causal links between management and biodiversity. Experts integrate the information available to them into a belief, which may or may not be a reasonable summary of the state of knowledge. D2 aims to critically evaluate the foundation of such models, both mathematical and mental, and juxtapose them with scientific evidence in the literature. D2 compares three representations of understanding: general causal knowledge with high levels of published evidence; specific causal assumptions represented in mathematical and computer models with biodiversity as a state variable; and intuitive causal belief of biodiversity-generating processes in scientists and forest managers.
Box 1 (Continued) (Knuff, Staab, Frey, Helbach, & Klein 2019). Based on meta-analyses, ConFoBi researchers confirmed that crown-damaged trees improve nesting opportunities for cavity-nesting birds (Gutzat & Dormann, 2018) and that woodpeckers select cavitiesby relative rather than absolute tree size (Basile, Mikusinski, & Storch, 2020), but found that bird guilds are affected differently by forestry measures including retention, according to their life history, biome, and forest type . A joint study by social and remote-sensing scientists of ConFoBi found expert ratings of forest structure, despite large individual bias, were on average significantly related to technical structural complexity indices based on terrestrial laser scanning (Frey, Joa, Schraml, & Koch, 2019), and a review concluded that local ecological knowledge holds significant promise for integrating conservation objectives into forest management under changing environmental conditions (Joa, Winkel, & Primmer, 2018;Joa & Schraml, 2018). Analyses of the opportunity costs arising from retention forestry suggest that conservation practices, such as habitat networks of deadwood islands, will only marginally impact profitability when conservation and production goals are balanced through suitable planning tools (Augustynczik, Yousefpour, Rodriguez, & Hanewinkel, 2018). Interdisciplinary modeling suggested that integration of uncertainty into conservation planning may reduce the trade-off between production and conservation objectives in forest landscapes (Augustynczik et al., 2017;, and a diversification of forest management regimes is recommended for securing various model taxa, including saproxylic beetles , as well as tree microhabitats and birds under climate change  ConFoBi's results will be seminal for integrating biodiversity conservation into forest management by providing an interdisciplinary evidence base for optimizing the effectiveness of retention approaches.

| Complementing ConFoBi: an invitation
While ConFoBi is primarily a Research Training Group for doctoral students, we have also set it up to provide a research platform that can be extended in scope, spatial, and temporal scales, as well as interdisciplinarity. Whereas the first set of studies primarily addresses disciplinary questions, later studies will focus on synergies across projects and disciplines.

CO N FLI C T O F I NTE R E S T
Authors declare no conflict of interest. Examples of measures of topography, vegetation, and landscape pattern for the 135 ConFoBi study plots. Topography and forest characteristics are mean values and total counts, respectively, referring to 1 ha; landscape metrics were calculated for various scales (see Schindler, Wehrden, Poirazidis, Wrbka, & Kati, 2013; respective moving window sizes are given in brackets). Definitions of landscape metrics according to McGarigal (2015). Note that Landsat forest classes (conifer, broad-leaved, and mixed) were merged into class "forest," remaining classes to "nonforest."