Contrasting vulnerability of monospecific and species‐diverse forests to wind and bark beetle disturbance: The role of management

Abstract Wind and bark beetle disturbances have increased in recent decades, affecting Europe's coniferous forests with particular severity. Management fostering forest diversity and resilience is deemed to effectively mitigate disturbance impacts, yet its efficiency and interaction with other disturbance management measures remain unclear. We focused on Central Europe, which has become one of the hotspots of recent disturbance changes. We used the iLand ecosystem model to understand the interplay between species composition of the forest, forest disturbance dynamics affected by climate change, and disturbance management. The tested measures included (a) active transformation of tree species composition toward site‐matching species; (b) intensive removal of windfelled trees, which can support the buildup of bark beetle populations; and (c) reduction of mature and vulnerable trees on the landscape via modified harvesting regimes. We found that management systems aiming to sustain the dominance of Norway spruce in the forest are failing under climate change, and none of the measures applied could mitigate the disturbance impacts. Conversely, management systems fostering forest diversity substantially reduced the level of disturbance. Significant disturbance reduction has been achieved even without salvaging and rotation length reduction, which is beneficial for ecosystem recovery, carbon, and biodiversity. Synthesis and applications: We conclude that climate change amplifies the contrast in vulnerability of monospecific and species‐diverse forests to wind and bark beetle disturbance. Whereas forests dominated by Norway spruce are not likely to be sustained in Central Europe under climate change, different management strategies can be applied in species‐diverse forests to reach the desired control over the disturbance dynamic. Our findings justify some unrealistic expectations about the options to control disturbance dynamics under climate change and highlight the importance of management that fosters forest diversity.


| INTRODUC TI ON
Forest disturbances are an integral part of forest dynamics, contributing to ecosystem functioning, creating heterogenous landscapes, and fostering biodiversity (Beudert et al., 2015;Turner et al., 2004). In production forests, however, disturbances place social management objectives at risk and compromise the provision of valued ecosystem services (Lindroth et al., 2009;Seidl & Blennow, 2012). Research suggests that all types of ecosystem services are negatively affected (Thom & Seidl, 2016), and these impacts will continue to increase in the future (Seidl, Schelhaas, Rammer, & Verkerk, 2014). Efforts to prevent or mitigate disturbance impacts have therefore become an integral part of forest management in Europe. The applied measures include, for example, improvement of tree vigor and morphology, modification of stand structure and composition, or reduction of fuel loads and breeding substrate for insects (Berryman, 1988;Gardiner & Quine, 2000;Jactel et al., 2009;Wermelinger, 2004). Research has also highlighted some controversies related to active disturbance management. These particularly include an effort to replace complex ecosystem regulation dynamics by oversimplified technological processes, which often erode ecosystem resilience (Cox, 2016) and produce collateral effects interfering with local management objectives . For example, long-term outbreak prevention via salvage logging can increase forest vulnerability to future disturbances via creation of vulnerable complexes of mature stands with high growing stock . Intensive disturbance management can also affect the quality of ecosystem services and modify natural ecological interactions in the forests .
Forest disturbance management has received increased attention in response to the recently intensified disturbance regimes and the increased rate of social and ecological impacts (Müller, 2011;. Moreover, model projections indicated that disturbance intensification will continue to increase in the future, which highlights the need to revise current management strategies (e.g., Dobor et al., 2019Dobor et al., , 2020Hlásny et al., 2019;Honkaniemi et al., 2020;Kausrud et al., 2012;Seidl et al., 2018). This requires a comprehensive understanding of the interactions between vegetation and disturbance dynamics affected by climate change, and management, which strives to interact with this overly complex and potentially unstable system. Quantifying the outcomes of disturbance management in ecosystems affected by climate change is therefore beyond our current understanding, which was mostly developed under past and more stable conditions. Among different disturbance agents, bark beetles (Coleoptera: Curculionidae, Scolytinae) have shown remarkable climatic sensitivity (Cudmore et al., 2010;. Recent intensification of bark beetle disturbance in Europe has been greater than that of any other disturbance type, including wind and wildfires (Seidl, Schelhaas, et al., 2014). While total canopy mortality has doubled in Europe over the last three decades , impact from bark beetles has increased by 600% (Seidl, Schelhaas, et al., 2014). These outbreaks highlight the prominent role of climate change as the driver of bark beetle disturbance (Bentz et al., 2019;Jönsson et al., 2009). In Europe's Picea abies-Ips typographus system, climate change increases the number of bark beetle generations, reduces winter mortality, and compromises fitness of host trees (Huang et al., 2019). Climate change also synchronizes the outbreaks over areas large several hundreds of kilometers via the large-scale impacts of regional climate extremes . Outbreaks of I. typographus amplified by climate change thus represent one of the major threats to forestry economies and the environment in Europe (Grégoire et al., 2015;Komonen et al., 2011).
In Europe's production forests, management has traditionally strived to control bark beetle populations to prevent or mitigate their impacts (Berryman, 1988;Wermelinger, 2004). These measures either aim to directly control populations of bark beetles or to modify forest structure and composition to create environment that is less conducive to the outbreaks (Wermelinger, 2004). Direct control mainly endeavors to reduce the amount of breeding substrate for beetles by removing trees affected by wind, snow, and rime; remove infested trees from the forest prior to beetles' emergence; and reduce beetle populations using insecticides or various trapping devices (Faccoli & Stergulc, 2008;Stadelmann et al., 2013). Conversely, indirect control includes silviculture practices, which, for example, aim to reduce tree competition for resources using thinning, reduce the concentration of host trees via change in tree species composition, or modify harvesting regimes to reduce the share of mature and vulnerable trees (Björkman et al., 2015;Jactel et al., 2009;Zimová et al., 2020). Indirect control can also aim to modify the forest configuration on the landscape to reduce the connectedness of local bark beetle populations and complexes of host trees (Honkaniemi et al., 2020;Seidl et al., 2018).
Efficiency of outbreak management measures in reducing the level of tree mortality is generally not sufficiently understood to inform management decisions Kausrud et al., 2012).
Rare examples of quantitative assessments include the studies of Faccoli and Stergulc (2008) (Guyot et al., 2016) and have higher survival rates (Griess et al., 2012;Neuner et al., 2015). Adaptive change in tree species composition can dilute the host trees in the forest, increase semiochemical diversity, and strengthen resilience mechanisms (Seidl,   .
Here, we investigate how management of functionally linked wind and bark beetle disturbances performs in differently managed forests, and how this performance can be affected by climate change.
In particular, we investigated how adaptive change in tree species composition affects the vulnerability of forests dominated by P. abies to natural disturbances and how the transformation of species composition interacts with other disturbance management measures. We focused on Central Europe, which has become one of the hotspots of recent disturbance change, and where the revision of current disturbance management strategies is urgently required. Because this analysis requires considering landscape-scale climate-sensitive disturbance dynamics, disturbance interactions, and dynamic feedback from vegetation, we addressed these questions using the forest landscape and disturbance model iLand .

| Study region
The study area is in the Western Carpathians (the Low Tatras mountain range) in Slovakia (Lon 20.088-20.275, Lat 48.920-49.061), covering an area of 16,050 ha ( Figure 1). The landscape has 70% forest cover and elevation range of 620 to 1550 m a. s. l. The forests are chiefly managed for softwood timber production, though recreation, game management, and nature conservation also occur.
Intense elevation and climate gradients, and the temperate continental climate (Kottek et al., 2006) resulted in the presence of multiple zonal forest communities in the natural species composition, with a dominance of broadleaved species (Rizman et al., 2005).
Due to the intense production-oriented management applied over the last 200 years, however, the forests are currently dominated by Norway spruce (Picea abies), which is found (often in monocultures) across the majority of site types on the landscape (Figure 1)

| Simulation model
We investigated the interactions between the climate, management, disturbances, and vegetation dynamics using the process-based model iLand  (http://iland.boku.ac.at). iLand is an ecosystem model that simulates forest landscape dynamics, including growth and regeneration, disturbance dynamics, and management in a spatially explicit manner. The main entity in the model is a tree, for which the demographic processes are simulated. Processes at the stand and landscape scale constrain the dynamics of individual trees and thus allow for the scaling of tree-scale processes to large areas . The model explicitly simulates tree competition for resources such as light, water, and nutrients. A light use efficiency approach (Landsberg & Waring, 1997) is used to simulate the production physiology. Carbon starvation is used as a processoriented indicator of tree stress, which can result from competition for resources as well as suboptimal environmental conditions for tree growth (e.g., drought).
iLand's mechanistic representation of forest disturbances and vegetation dynamics, as well as the climatic sensitivity of these processes, makes it well suited for the research of disturbance dynamics under climate change (e.g., Dobor et al., 2018;Seidl, Rammer, & Blennow, 2014). Wind disturbances are initiated by the wind speed of severe wind events provided as an external input to the simulation. The wind impact is simulated iteratively, with the forest structure (including the appearance of new edges) being updated after each iteration in the event of breakage or windthrow ; Supplement A). The implementation of bark beetle disturbances considers bark beetle phenology and development, spatially explicit dispersal of beetles, colonization, and tree defense, as well as temperature-related overwintering success . Outbreaks are typically triggered by wind disturbance; salvage removal of windfelled trees can therefore be applied to reduce the outbreaks . A detailed description of the implementation of wind and bark beetle disturbance in iLand is provided in Supplement A.
Flexible implementation of management operations, which include planting after harvests or natural disturbances, thinning, harvesting, and postdisturbance salvaging, allows for testing the effects of various disturbance management strategies (Dobor et al., 2019Honkaniemi et al., 2020). iLand integrates an agent-based model of forest management , in which general stand treatment programs (i.e., a sequence of management interventions applied over the course of stand development) are dynamically adapted to the forest state emerging from the simulation.
These features allow for testing the efficiency of measures taken in response to the simulated disturbance, considering the dynamically changing vegetation structure. The tested management interventions are implemented in the model as follows: • Planting is applied after harvests and stand-replacing disturbances that affected a prescribed level of growing stock (75% in this study). Planting is based on prescriptions defining details of tree species, seedling dimensions, and spacing between the seedlings. Planting prescriptions can differ between stands, depending on site conditions or management objectives. The already established regeneration can be kept or removed.
• Thinning and harvesting are applied based on prescribed timing and intensity of removal. Different criteria on tree removal can be defined to implement practices such as clearcutting, shelterwood, or selection cut. Each stand has a stand treatment program assigned that defines the sequence of interventions.
• Salvaging is applied to harvest timber affected by disturbances.
Different intensities of salvage removal can be prescribed, affecting forest carbon stocks, dynamics of secondary disturbances, and the deadwood patterns. The incidence of disturbances and subsequent salvage logging supersede regular management operations, resetting the default stand treatment program.
The model was extensively tested across a range of ecosystems in Europe and North America in previous studies (Honkaniemi et al., 2020;Silva Pedro et al., 2015;Thom et al., 2017). A detailed evaluation of simulated productivity, natural mortality, and regeneration patterns for the landscape studied here was conducted by Dobor et al. (2018). All testing exercises conducted for Central Europe proved good ability of the model to simulate ecosystem dynamics in this environment.

| Basic simulation setup
Prior to scenario simulations, an 800-year spin-up run was performed to estimate the initial litter, dead wood, and soil C pools. The spin-up run was also used to initialize stand structures in a manner consistent with the internal logic of the model. The procedure used assimilates information on the current composition and structure of forest stands (here based on forest management plan records; see Supplement B) to ensure that the resulting initial forest state for simulation is consistent with the model internal logic and represents the current structure and composition of the forest (see Thom et al., 2018 for details).
The scenario simulations were run for 100 years starting from the end-conditions of the preceding spin-up run. Each simulation was driven by five climatic scenarios (reference climate and four projected climates, see Supplement C). Each simulated forest development was exposed to five prescribed wind events, with parameters sampled from the distribution parameterized based on past meteorological observations for the area. The intensity of events was set to reach the average annual amount of disturbed trees recorded in the national forest disturbance statistics for the period 1990-2010, which range between 0.9 and 2.2 m 3 ha −1 year −1 Konôpka et al., 2016). Each scenario run was replicated 10 times to account for the stochasticity in the simulations. The value of the so-called background infestation probability (a parameter related to bark beetle disturbance  was varied between the replicates in a range of 0.0005-0.0025 (Honkaniemi et al., 2020).
The implemented baseline management included tending, thinning, and harvesting, with timing and intensity of operations modeled after the management practice currently applied in the region (Halaj & Petráš, 1998). Depending on the site, 3-4 thinning operations were applied and rotation periods ranged from 90 to 140 years.
In spruce monocultures, clearcut management was applied, whereas shelterwood management was simulated in mixed stands.
The simulations were run under the conditions defined by

| Disturbance management experiment
The previously described baseline management was modified to ac- We organized these measures around two management narratives that are being intensively discussed in the Central European production forestry; (a) the industry demand-driven effort to maintain high proportions of Norway spruce in the forest, and (b) efforts to adapt the tree species composition to climate change and intensified disturbances via recovering natural species composition, which has been markedly altered over the last two centuries in many production forests in Central Europe (Klimo et al., 2000;Spiecker et al., 2004).
To address these two objectives, we implemented two seedling planting schemes on cleared areas. The first one promoted Norway spruce in species composition; depending on site, its share ranged from 50% to 70% (natural regeneration was, however, acting concurrently  practice. We simulated the reduction of the rotation length by 40% relative to the currently applied rotation (100 years for spruce stands, and 115 years for broadleaved species, on average). The rotation length was, however, not allowed to be <60 years. This level of reduction still conforms with the criteria for the production of softwood timber and can be expected to dampen the disturbance dynamics to a certain extent.

| Disturbance management performance
Under the reference climate, the total level of growing stock disturbed was substantially reduced by different combinations of management measures (Figure 3a,b). In management systems promoting spruce in species composition, however, the mitigation effect highly varied within (i.e., the interreplicate variation) and between treatments. Still, the average reduction effect over the simulation period was 8% for salvaging, 3% for rotation length reduction, and 13% for the combination of latter two measures (separate effects on wind and bark beetle disturbance are provided in Supplement E).
Management systems promoting adaptive changes in tree species composition more efficiently reduced the disturbance than the previous systems, though lead times were long (from ca. 2060) ( Figure 2b). The simultaneous application of different treatments amplified the disturbance reduction effect. Systems containing salvaging (C2 and C4) were most efficient, reducing the disturbance by 19%-25% relative to the reference treatment A (Figure 3c).

| Underlying changes in forest structure
The tested management interventions affected forest susceptibility to disturbance mainly via changes in forest age structure and spe-

| D ISCUSS I ON
Previous studies indicated that Norway spruce forests may not be sustained in many regions of Europe because of intensifying outbreaks of bark beetles, genetic maladaptation to future climates, and sensitivity to climatic stress (Frank et al., 2017;Marini et al., 2012;Seidl, Schelhaas, et al., 2014;Zang et al., 2014). The role of active management of forests disturbances, however, has not been included in these investigations although it is an integral aspect of European forest management (Berryman, 1988;Wermelinger, 2004). We showed that contrast in the vulnerability of monospecific forests and forests managed for diversity is considerably amplified by climate change.
We also found that management measures that were successfully applied in the past are becoming inefficient under warmer climate amplifying the disturbance dynamics, which particularly applies to the forests dominated by Norway spruce.

| Implications for ecosystem management
We found that the studied ecosystem was relatively stable under past climate and the level of disturbance was low. Such dynamics agree with the national forest damage statistics (e.g., Gubka  Consistent with previous studies, we found that the change in tree species composition toward a higher proportion of less vulnerable and site-adapted species has paramount importance in managing forests under climate change (Jandl et al., 2019). Diverse ecosystems generally show lower disturbance rates compared with monospecific forests (Griess et al., 2012;Neuner et al., 2015) and are also superior in the provisioning of ecosystem services (Mori, 2017). Still, some previous studies suggested that the pest control effect may depend more on species composition of the forest than on diversity (Koricheva et al., 2006). Accordingly, the here presented disturbance mitigation effect needs to be considered as a function of both replacement of vulnerable Norway spruce by other tree species and the increase in stand-and landscape-scale diversity, which, for example, dilutes the host trees and prevents the large-scale spread of bark beetles (Honkaniemi et al., 2020;Silva Pedro et al., 2015).
We found that the same disturbance reduction effect can be  of ecosystem services in different ways (Leverkus, Benayas, et al., 2018;Roberge et al., 2016

| Methodological aspects and limitations
Here, we used a highly complex simulation model to investigate the interactions between disturbance dynamics, management, vegetation feedbacks, and climate change. Although such an approach allowed identifying and attributing the effects of different management actions, the uncertainty related to the representation of individual processes and model assumptions needs to be carefully considered (Huber et al., 2020). Although the model's use is supported by numerous testing exercises that particularly addressed forest productivity, regeneration, and natural mortality (e.g., Dobor et al., 2018), reproducing complex disturbance regimes in models remains challenging (Seidl et al., 2011). High stochasticity of disturbance events complicates testing the simulation outputs against the observed impacts (but see for example. .
We here prescribed the intensity of wind impacts to match the long-term observations, whereas the simulated windthrow pattern and the interaction with bark beetle dynamics were simulated as emergent properties of the used simulation framework. More comprehensive testing of the simulated disturbance patterns against observation would provide useful support to the presented findings.
Given the high complexity of our experimental design, we only investigated one level or salvaging intensity and rotation length reduction, though management practice may require more complex information (see e.g., Dobor et al., 2019Dobor et al., , 2020Zimová et al., 2020).
The tested intensities were, however, near to the logistic limits of the current forest management and can thus be interpreted as the reachable maximum under the operational management conditions.
The complexity of our design can be further increased by including other management variants, which stress, for example, adaptive changes in species composition, including altitudinal shift of zonal trees species (Moser et al., 2010) and introduction of species that do not participate in the actual species composition. In the Central European forestry, these species may include, for example, native oak species (Quercus sp.) which are expanding their ranges under climate change (e.g., Mette et al., 2013) as well as introduced species of which the Douglas fir (Pseudotsuga menziensii) has received considerable attention (Spiecker et al., 2019). Although such changes would not directly affect the disturbance dynamics in the current modeling framework (their vulnerability is similar to other species on the study landscape), they could indirectly affect the disturbance dynamics via different rates of establishment on disturbed sites.

| CON CLUS IONS
Management of wind and bark beetle disturbances constitutes an integral part of European forestry; yet, many approaches are based on plausible or intuitive narratives rather than on tested and data-driven concepts. We presented a new perspective on disturbance management in the Central European production forests and particularly on the interactions of adaptive changes in species composition with other management measures. Consistent with previous studies, we found a contrasting sensitivity of monospecific and species-diverse forests to disturbance impacts. However, we showed that climate change further amplifies this contrast and favors management fostering forest diversity, which can exert better control over disturbance dynamics even without pervasive measures compromising the biodiversity and resilience of the forest. These findings can justify some misconceptions about disturbance management under climate change and can support the formulation of improved management strategies.

ACK N OWLED G M ENTS
This study was supported by the grant "EVA4.