Skip to main content
NCBI home page
UKPMC Funders Author Manuscripts logoLink to UKPMC Funders Author Manuscripts
. Author manuscript; available in PMC: 2016 Oct 14.
Published in final edited form as: J Appl Ecol. 2015 Oct 14;53(2):530–540. doi: 10.1111/1365-2664.12540

Small beetle, large-scale drivers: how regional and landscape factors affect outbreaks of the European spruce bark beetle

Rupert Seidl 1,*, Jörg Müller 2,3, Torsten Hothorn 4, Claus Bässler 2, Marco Heurich 2, Markus Kautz 5
PMCID: PMC4816203  EMSID: EMS67412  PMID: 27041769

Summary

1. Unprecedented bark beetle outbreaks have been observed for a variety of forest ecosystems recently, and damage is expected to further intensify as a consequence of climate change. In Central Europe, the response of ecosystem management to increasing infestation risk has hitherto focused largely on the stand level, while the contingency of outbreak dynamics on large-scale drivers remains poorly understood.

2. To investigate how factors beyond the local scale contribute to the infestation risk from Ips typographus (Col., Scol.), we analysed drivers across seven orders of magnitude in scale (from 103 to 1010 m2) over a 23-year period, focusing on the Bavarian Forest National Park. Time-discrete hazard modelling was used to account for local factors and temporal dependencies. Subsequently, beta regression was applied to determine the influence of regional and landscape factors, the latter characterized by means of graph theory.

3. We found that in addition to stand variables, large-scale drivers also strongly influenced bark beetle infestation risk. Outbreak waves were closely related to landscape-scale connectedness of both host and beetle populations as well as to regional bark beetle infestation levels. Furthermore, regional summer drought was identified as an important trigger for infestation pulses. Large-scale synchrony and connectivity are thus key drivers of the recently observed bark beetle outbreak in the area.

4. Synthesis and applications. Our multiscale analysis provides evidence that the risk for biotic disturbances is highly dependent on drivers beyond the control of traditional stand-scale management. This finding highlights the importance of fostering the ability to cope with and recover from disturbance. It furthermore suggests that a stronger consideration of landscape and regional processes is needed to address changing disturbance regimes in ecosystem management.

Keywords: bark beetle infestation risk, Bavarian Forest National Park, forest disturbance dynamics, graph theory, Ips typographus, landscape connectivity, large-scale control, multiscale analysis, Picea abies, spatial synchrony

Introduction

Bark beetles are the most important biotic forest disturbance agent in temperate and boreal forest ecosystems. The host–insect system is highly sensitive to climatic changes, as both beetle development and host tree defences are dependent on climate variables (Netherer & Schopf 2010). As a result, bark beetle disturbances around the globe have already increased in response to the ongoing climatic changes (Bentz et al. 2010; Seidl, Schelhaas & Lexer 2011; Weed, Ayres & Hicke 2013). Unprecedented outbreaks are currently observed, for instance, in North America and Central Europe, where native bark beetle species (i.e. the mountain pine beetle Dendroctonus ponderosae Hopkins and the European spruce bark beetle Ips typographus L.) have killed trees in the order of tens of millions of hectares (Raffa et al. 2008; Meddens, Hicke & Ferguson 2012; Lausch, Heurich & Fahse 2013). As climate change continues, a further intensification of bark beetle disturbances is expected for the coming decades (Hicke et al. 2006; Seidl et al. 2009, 2014; Jönsson et al. 2011).

Here, we focus on I. typographus, which is the most important bark beetle species in Europe in terms of tree mortality. Comprehensive summaries of the state of the knowledge on I. typographus biology are given by Wermelinger (2004) and Kausrud et al. (2012). Recently, empirical approaches have been applied to determine the most relevant drivers of I. typographus damage at different spatial and temporal scales (e.g. Marini et al. 2012; Overbeck & Schmidt 2012; Thom et al. 2013; Pasztor et al. 2014). However, these approaches have almost exclusively addressed a single focal scale, studying either the stand (e.g. Pasztor et al. 2014), landscape (e.g. Lausch, Heurich & Fahse 2013), country (e.g. Thom et al. 2013) or continental (e.g. Seidl, Schelhaas & Lexer 2011) scale. This focus often necessitates from prevailing data limitations, as the grain of information frequently increases with extent. This, however, can impair our inferential potential with regard to the drivers of disturbance, as either local processes are increasingly neglected in large-scale studies, or the effect of large-scale drivers such as climatic variation and landscape structure is erroneously attributed (in part) to the local level and the error term.

In Central Europe, the scale at which forest managers are currently tackling the impacts of global change on disturbance regimes is primarily the tree to stand scale (Jactel et al. 2009), while large-scale drivers pertaining to the scales of regions or landscapes (e.g. spatial connectivity and contagion) are frequently neglected in considerations of management. This, however, could result in an inflated perception of the efficacy of stand management measures and foster an unwarranted sense of control regarding the risks from large-scale disturbance. It is thus highly relevant to determine whether and to what degree the probability of disturbance can be modified through local-scale actions and how much they remain beyond the influence of stand management [and thus need to be embraced as residual uncertainty in ecosystem stewardship (Seidl 2014)]. Addressing this issue calls for a multiscaled perspective in the analysis of forest disturbances (Simard et al. 2012; Seidl et al. 2013). The utility and need for such a perspective is aptly illustrated by Raffa et al. (2008): for D. ponderosae they showed that the sensitivity of bark beetles to climate variation is strongly contingent on the status of the beetle population, differing between epidemic and non-epidemic conditions. This suggests that dependencies and interactions across scales require consideration in order to understand (and subsequently predict) the complex outbreak dynamics of bark beetles in forest ecosystems.

We here present the – to our knowledge – first quantitative multiscale study of the most important forest insect pest in Eurasia, I. typographus. Addressing drivers across seven orders of magnitude in scale (from 103 to 1010 m2), our aim was to determine the influence of large-scale drivers on bark beetle outbreak dynamics at the Bavarian Forest National Park (Germany). Analysing a 23-year time series of bark beetle dynamics, our particular objective was to assess whether and how strongly regional and landscape drivers affected the temporal variation in bark beetle infestation probability. We hypothesized that – in addition to stand-scale variables – the variation in infestation observed between 1990 and 2012 was significantly driven by variables beyond the reach of stand-level management, in particular by climate variability and large-scale bark beetle pressure (i.e. regional variables), as well as by the spatial structure of the beetle and host populations in the ecosystem (i.e. landscape-scale variables).

Materials and methods

STUDY AREA

The Bavarian Forest National Park (BFNP) is a 24 222 ha forest landscape located in the state of Bavaria in south-eastern Germany. It is characterized by mountainous terrain and dominated by Norway spruce Picea abies (L.) Karst. Together with the adjacent Šumava National Park (Czech Republic) it represents one of the most extensive areas of protected temperate forest ecosystems in Europe. The BFNP was established in 1970 and was significantly extended in 1997. In order to consistently analyse the longest possible time series of bark beetle dynamics at BFNP, we here focus solely on the initial portion of the park (i.e. an area of 13 319 ha). An outbreak of I. typographus started in the 1990s at BFNP. The outbreak dynamics were characterized by two distinct epidemics (1996–2000 and 2005–2009), with a cumulative area of approximately 6500 ha of spruce forest being infested by 2012 (Lausch, Heurich & Fahse 2013) (Fig. 1). Consistent with the ‘benign neglect’ strategy of the park, the outbreak unfolded largely without human interference (Müller et al. 2010). Only a limited area in the vicinity of the border of the park was salvaged (5·7% of the total area) to dampen the spread of the epidemic into adjacent areas. Due to this unprecedented bark beetle activity (43·9% of the landscape disturbed in 23 years) and the fact that processes such as forest disturbance and succession were allowed to proceed with minimal human intervention (which is rare in the intensely managed landscapes of Central Europe), the BFNP offers an unique opportunity to glean insights into bark beetle dynamics. Previous studies at BFNP focused on spatio-temporal outbreak patterns (Kautz et al. 2011; Lausch, Heurich & Fahse 2013), the role of antagonists in outbreak dynamics (Fahse & Heurich 2011) and ecological impacts of the I. typographus outbreak (Müller et al. 2010). However, the role of large-scale drivers and regional synchrony in the outbreak at BFNP remains an unresolved question. Here, we analysed factors affecting bark beetle infestation probability at BFNP at three different spatial scales, that is the stand (103 m2), landscape (108 m2) and regional (1010 m2) scale (Fig. 1).

Fig. 1.

Fig. 1

Open in a new tab

The study area and the three spatial scales considered in the analyses.

POTENTIAL DRIVERS OF BARK BEETLE DISTURBANCE

Stand scale

Data on bark beetle infestation were derived from annual aerial surveys conducted in late summer or early autumn (Heurich et al. 2010; Kautz et al. 2011). The thus determined binary infestation status recorded for the 23-year period from 1990 to 2012 at a spatial grain of 30 × 30 m grid cells (n = 114 235) was used as the response variable for the stand-level analysis. We defined the Moore neighbourhood around each focal cell as the stand level in our analysis (i.e. an area of 8100 m2). This definition reflects the small-scale structure prevailing in Central Europe’s forests; the use of different stand definitions did not substantially change the results of our analyses (see Appendix S1, Supporting Information). Climatic predisposition was characterized by the thermal limitations for beetle development (represented here by long-term mean growing season temperature and radiation) as well as by tree water availability (using precipitation sum and soil wetness index) as proxy for tree defence capacity (Wermelinger 2004; Kausrud et al. 2012; Netherer et al. 2015). Data were derived from local weather stations and spatial climate modelling (Beudert et al. 2015). Bark beetle damage in the previous year was used as a proxy for the effects of local population density, and the current availability of host trees (i.e. mature individuals of P. abies) was used to describe stand predisposition (Thom et al. 2013; Marini et al. 2013) (see Table 1 and Appendix S2 for details). Also wind damage in the previous 2 years was considered as a potential explanatory variable, but was omitted from the analysis since no damage from wind was recorded for the BFNP between 1990 and 2012.

Table 1.

The explanatory variables considered (see also Appendix S2)

Scale Variable groups and description Mean ± SD
Stand 8·1 × 103 m2 Site predisposition
 Mean growing season temperature (°C) 11·2 ± 0·7
 Mean growing season precipitation sum (mm) 1105 ± 93
 Mean growing season radiation (kWh m−2) 1488 ± 73
 Soil wetness index (Beudert et al. 2015) (ordinal) –
 Stand predisposition and population dynamics
 Previous year bark beetle infestation (% of cells in the stand) 0·54 ± 3·78
 Host tree availability (% of cells in the stand that is mature Picea abies) 83·7 ± 21·6
Landscape 1·3 × 108 m2 Spatial structure of the bark beetle population
 Average path length between all pairs of
  nodes in the networks of bark beetle infested cells (dim.)		 5·3 ± 3·6
 Standardized graph diameter, that is the
  maximum shortest path joining any two nodes,
  divided by the number of nodes in the component (dim.)		 0·37 ± 0·18
 Mean degree, that is the number of edges adjoining a node (dim.) 4·4 ± 2·1
 Graph density, that is the ratio of the number of
  actual connections to the number of potential connections (dim.)		 0·012 ± 0·008
 Integral index of connectivity (Pascual-Hortal & Saura 2006) (unit scale) 3·9 × 10−6 ± 8·1 × 10−6
Spatial structure of the potential host tree population
 Standardized graph diameter, that is the maximum
  shortest path joining any two nodes divided by the
  number of nodes in the component (dim.)		 0·26 ± 0·13
 Mean degree, that is the number of edges adjoining a node (dim.) 6·5 ± 1·2
 Graph density, that is the ratio of the number of actual
  connections to the number of potential connections (dim.)		 8·6 × 10−3 ± 3·6 × 10−3
 Landscape coincidence probability, that is the probability
  that two random nodes belong to the same component
  (Pascual-Hortal & Saura 2006) (%)		 29·4 ± 21·4
 Integral index of connectivity (Pascual-Hortal & Saura 2006) (unit scale) 4·2 × 10−5 ± 3·6 × 10−5
Region 9·0 × 1010 m2 Climate variation
 Spring (MAM) temperature anomaly (Δ°C) 0·00 ± 1·00
 Summer (JJA) temperature anomaly (Δ°C) 0·00 ± 0·74
 Autumn (SON) temperature anomaly (Δ°C) 0·00 ± 1·02
 Winter (DJF) temperature anomaly (Δ°C) 0·00 ± 1·80
 Spring (MAM) precipitation anomaly (%) 0·0 ± 36·0
 Summer (JJA) precipitation anomaly (%) 0·0 ± 28·1
 Autumn (SON) precipitation anomaly (%) 0·0 ± 38·1
 Winter (DJF) precipitation anomaly (%) 0·0 ± 28·7
Bark beetle population pressure
 Outbreak stage in the previous 2 years (ordinal) –
 Outbreak stage in the previous year (ordinal) –
 Outbreak stage in the current year (ordinal) –
 Outbreak stage in the next year (ordinal) –
 Outbreak stage in the next 2 years (ordinal) –
Open in a new tab

dim., dimensionless; MAM, March, April and May; JJA, June, July and August; SON, September, October and November; DJF, December, January and February.

Landscape scale

At the landscape scale, we used graph theory (Urban & Keitt 2001; Dale & Fortin 2010) to characterize the spatial structure and connectivity of both the host and beetle populations. In this approach, the landscape is analysed via the nodes and edges of a network. The network topology is an emergent property of the underlying dynamics of the system and contains valuable information on the spread (of information, agents, etc.) and stability of a system. Here, we used the aerial surveys described above as the underlying data for spatial graph analysis. Graph nodes represent potential habitat patches or local beetle populations (with currently infested cells being used as a surrogate for the bark beetle population), while links between nodes indicate functional connections among populations and patches (i.e. via the dispersal of bark beetles). With regard to the bark beetle population, a contiguous area of cells classified as infested was defined as a node of the bark beetle population network. Salvaged cells were included here, as only trees with visible signs of bark beetle attack were cut in the limited salvage operations conducted at BFNP. For the host, contiguous areas of cells containing mature P. abies were considered as nodes of the host tree network.

Whether nodes were connected was determined based on an in-depth analysis of the spatial dependency of I. typographus attacks, showing that 95% of new infestations occur within distances of ≤253 m of previous attacks (Kautz et al. 2011). Using this value as a threshold for edges, we determined the respective connections of the host and beetle networks for every year of the study period. A sensitivity analysis of the selected threshold distance for establishing connections is presented in Appendix S3. Consequently, we used the current network of the potential host tree population as well as that of the previous year’s bark beetle population as potential landscape-scale drivers, hypothesizing a positive relationship between infestation probability at BFNP and host and beetle connectivity, respectively. We applied scale-free graph indicators in our analysis (see Table 1 for the complete list, and Minor & Urban (2008) for details on the calculation of these metrics).

Regional scale

The regional scale is here represented by the joint area of Bavaria, Austria and the Czech Republic, that is the administrative entities surrounding the BFNP, representing a total forest area of 9.02 × 106 ha (Fig. 1). At this scale, we in particular investigated the effects of temporal variation in climate parameters as well as the role of regional-scale bark beetle population pressure (Table 1). Climate anomalies relative to the period mean were extracted from the NCEP/NCAR climate reanalysis data set for 1990–2012 (Kalnay et al. 1996) and averaged over all grid cells falling into our study region. Three-month aggregates were used in order to account for the multiweek developmental period of I. typographus. In addition to climate variability, regional-scale bark beetle damage (in m3 timber per year) was used as a proxy for the large-scale bark beetle pressure in the region surrounding the BFNP. While bark beetle population dynamics emerge from local-scale processes (e.g. beetles successfully finding, colonizing and reproducing in individual trees), we here analysed cross-scale feedbacks (cf. Raffa et al. 2008) by testing the lagged influence of regional-scale beetle damage on the outbreak dynamics at BFNP. As the BFNP was frequently suspected to be the source of bark beetle outbreaks in the region (Müller 2011), we tested whether the temporal variation in bark beetle infestation probability at BFNP did influence or was influenced by the dynamics at the roughly three orders of magnitude larger regional scale. Large-scale disturbance data were extracted from Seidl et al. (2014) and categorized into epidemic, gradation and non-epidemic conditions using the 33rd and 66th percentiles of annual damage levels as cut-offs. It has to be noted that the regional and BFNP infestation data are strictly independent of each other (i.e. the latter is not contained in the former), as the regional estimate is based on salvage statistics for managed forests, while the BFNP data were determined from aerial surveys for the largely unsalvaged national park landscape. Both interannual variations in climate and bark beetle population pressure were found to be highly synchronized throughout the region (Appendix S4), supporting their subsumption as regional variables. For the latter, this synchrony is in parts related to large wind-throw events occurring in the region.

ANALYSES

To determine the role of large-scale drivers on bark beetle infestation probability in a multiscale analysis framework, we first controlled for the effect of stand-scale variables. To this end, we fitted a time-discrete hazard model (Singer & Willett 1993), using bark beetle infestation as the response variable and local variables as predictors (binomial link function). The time-discrete hazard model accounts for the fact that once an event occurs (i.e. a cell is infested), it cannot occur again at a later point in time. In other words, the occurrence of an event is conditional on the event not to have happened previously. In addition to local predictors (Table 1), this model also contained individual dummy variables for the 23 years of the study period (i.e. year-specific intercepts). These latter values represent the annual variation in bark beetle infestation probability at BFNP that is not explained by local-scale predictors. These year-specific intercepts were subsequently used as the response variable for analysing the effect of regional- and landscape-scale variables. The stand-level model was not further analysed as its sole purpose in this contribution was to determine the variation in infestation probability not explained at the local scale.

The association of large-scale variables with the interannual variation in infestation probability not explained at the stand level (n = 23) was first scrutinized by means of independence tests, a powerful permutation-based approach to test the null hypothesis that two variables (measured on arbitrary scales) are independent of each other (Hothorn et al. 2006). Subsequently, to determine the joint influence and significance of regional as well as landscape variables on bark beetle infestation probability, a regression modelling approach was used. Since the response variable was scaled to the unit interval (0, 1), we used beta regression to address the resulting issue of heteroscedasticity (Ferrari & Cribari-Neto 2004; Cribari-Neto & Zeileis 2010). The parameters of the final regression model were determined in a two-stage approach. First, we selected the most influential variable per group (i.e. regional climate variation, regional bark beetle population pressure, landscape-scale spatial structure of the bark beetle population, and landscape-scale spatial structure of the host tree population, cf. Table 1) by means of independence tests. This a priori selection was necessary as the candidate variables within the groups were found to be significantly correlated. Secondly, we used an information-theoretic approach based on Akaike’s information criterion (AIC) to determine which model formulations were most strongly supported by the data. The best fit was determined by the model with the lowest AIC value (AICmin), and models with a ΔAIC = AIC − AICmin of <2 are reported as plausible. The model weight wi (i.e. the weight of evidence in favour of model i being the actual best model) was determined by dividing the likelihood of the ith model (i.e. exp (−0.5 × ΔAICi)) by the sum of all likelihoods (Burnham & Anderson 2002). To evaluate the relative importance of each predictor, we summed wi across all the models in the full set in which the respective predictor occurred (∑wi). Since we also hypothesized interaction effects between the connectivity of beetle and host populations as well as between the effects of regional climate and population pressure, we also included the respective interaction terms in the analysed model formulations. The temporal autocorrelation structure in the explanatory variables was explicitly accounted for by means of sandwich estimators of the covariance (Zeileis 2004), and the intercept and precision parameter of the beta regression model were omitted for determining the multiplicity-adjusted significance levels of parameters. The R Language and Environment for Statistical Computing was used for all calculations (R Development Core Team, 2013, version 2.13.1), in particular employing the packages igraph (Csardi & Nepusz 2006; version 0.5.5-4), betareg (Cribari-Neto & Zeileis 2010; version 2.4-1), sandwich (Zeileis 2004; version 2.2-9) and coin (Hothorn et al. 2006, version 1.0-20).

Results

ANNUAL BARK BEETLE INFESTATION PROBABILITY

The annual infestation probability at BFNP showed two distinct outbreak waves over the 23-year study period (Fig. 2), varying between 0·05% in 1992 and 10·62% in 1999 (time-series mean: 3·53%). Besides the six stand-level explanatory variables, the annual intercepts of the time-discrete hazard model were also significant, indicating that a considerable part of the temporal variation in infestation probability was not explained by the stand-scale drivers. In particular, while a substantial portion of the first outbreak wave could be attributed to local factors (cf. the grey shaded area in Fig. 2), the second wave remained largely unexplained by stand-level variables. This remaining temporal variation in infestation probability was subsequently analysed for the influence of landscape and regional drivers.

Fig. 2.

Fig. 2

Open in a new tab

Temporal variation in infestation probability at the Bavarian Forest National Park. The black line indicates the observed value (infested cells by total number of potential host cells), while the red line gives the yearly intercepts of the time-discrete hazard model fitted with local explanatory variables. The latter thus illustrates the residual temporal variation in infestation probability that is not explained by variables at the stand scale.

LANDSCAPE-SCALE DRIVERS

The structure and connectedness of the bark beetle population on the landscape showed considerable variation over time. The number of distinct subnetworks of bark beetle outbreaks on the landscape (i.e. a measure of metapopulation structure), for instance, varied between 37 and 261 (mean: 151), and the average path length within the subnetworks ranged from 1·52 to 12·7 (mean: 5·52) over the 23-year analysis period (Fig. 3). This variation in the structure and connectedness of the bark beetle population was found to be significantly correlated with the bark beetle infestation risk in the subsequent year (Appendix S5). A higher spatial connectedness and higher order of networks generally increased the infestation probability on the landscape in the following year. However, this effect of the spatial structure of the bark beetle population was strongly conditional on the connectedness of the host tree population, with high connectivity between potential host patches amplifying the effect of a highly connected bark beetle population (Fig. 4). In general, the host tree network showed decreasing graph density (i.e. the ratio of the actual to the potential connections of a network) over time, mainly as an effect of the beetle depleting its host resource. The standardized graph diameter (i.e. an indicator of potential movement speed through the host network) was the host-related indicator most strongly associated with infestation probability.

Fig. 3.

Fig. 3

Open in a new tab

Graph analysis of the bark beetle population in the Bavarian Forest National Park. The two selected years are indicative of non-epidemic conditions (2001, top) and epidemic conditions (2007, bottom), respectively. Node size is scaled with the square root of the size of the respective outbreak patch represented by the node. Connections are determined from an analysis of spatial infestation spread (Kautz et al. 2011). For clarity, a subset of the landscape (dashed box) is magnified.

Fig. 4.

Fig. 4

Open in a new tab

The influence of the spatial structure of the bark beetle population on subsequent infestation probability at the Bavarian Forest National Park is conditional on the connectivity of the host tree population. Number symbols indicate the years of the 23-year study period, and a linear relationship (and confidence interval) is included to aid visual interpretation. The indicator used to describe the spatial structure of the bark beetle population in the preceding year is the mean path length of the bark beetle network, whereas host tree connectivity was evaluated by means of standardized graph diameter (with the median value used to discriminate between the panels).

REGIONAL-SCALE DRIVERS

First, by analysing different temporal lags from t − 2 to t + 2 year, we tested for an upward feedback of bark beetle outbreaks through the scales (i.e. from the landscape to the region, in other words: did the BFNP act as a trigger of regional bark beetle outbreaks), or whether the regional outbreak development influenced the local dynamics at the BFNP. Here, our data strongly supported the notion that the regional-scale outbreak dynamics affected the disturbance pattern at BFNP and not vice versa (see also Appendix S5). Epidemic conditions at the regional scale in the preceding 2 years were positively associated with the bark beetle infestation probability at BFNP. This effect was, however, modulated by the variation in summer precipitation levels: in drier than average summers, no significant effect of regional-scale bark beetle population pressure was detected, with high infestation probabilities occurring in all outbreak stages (Fig. 5). Our data thus suggest that under low regional population pressure, drought can trigger bark beetle outbreaks. For years with above average summer precipitation, however, a high regional bark beetle population pressure in previous years was found to be strongly related to infestation probabilities, amplifying bark beetle damage at the BFNP. Other climatic factors, such as the variation in temperature, were found to have only little effect on the bark beetle infestation probability at BFNP (Appendix S5).

Fig. 5.

Fig. 5

Open in a new tab

The influence of regional bark beetle population pressure (previous 2 years) on infestation probability at the Bavarian Forest National Park (BFNP) is contingent on the regional variation in summer precipitation. Local infestation probability closely tracks regional developments under ample water supply. In dry summers, however, high local infestation probability is possible regardless of regional outbreak stage. Summer precipitation is the precipitation in the months June, July and August, and for illustration purposes, the data set was split into two groups of higher (left) and lower (right) than average precipitation.

MULTISCALE DRIVERS OF BARK BEETLE OUTBREAK DYNAMICS

Analysing the effects of landscape and regional drivers jointly by means of beta regression modelling generally confirmed the effects and relationships of the individual analyses at their respective scales. The strongest support was found for a direct positive effect of a highly connected bark beetle population on the landscape, with the average path length of the bark beetle network in the previous year highly significant in all models with ΔAIC < 2 and a predictor weight ∑wi of close to one (Table 2). Furthermore, while a direct effect of the host tree connectivity on the landscape was only moderately supported by the data, the variable was found to be highly influential as an interaction term with the spatial structure of the bark beetle population (Fig. 4). Also at the regional scale, the interaction effect between regional climate variability and bark beetle outbreak stage was found to be of high importance in the regression analysis. However, a main negative effect of summer precipitation was also retained in two of the five most plausible model formulations, indicating that decreasing levels of summer precipitation generally increase the probability of bark beetle infestations. A more detailed analysis showed that landscape-scale factors of connectivity strongly drove the first outbreak wave in the late 1990s, while the second outbreak wave in the late 2000s was largely determined by regional drivers (Appendix S6). Overall, the beta regression models with ΔAIC < 2 were all well able to describe the temporal variation in bark beetle infestation probability at the BFNP, with pseudo-R2 of >0.70 for all five models.

Table 2.

Results of the information theory-based selection of the beta regression models best supported by the data (n = 23) (see also Appendix S6)

Landscape variables
 Regional variables
Spatial structure of bark beetle population: average path length (previous year) Spatial structure of the host tree population: standardized graph diameter Interaction beetle–host connectivity and structure Climate variation: summer precipitation Bark beetle population pressure: outbreak stage in previous 2 years Interaction climate variation: outbreak stage
# ΔAIC wi Pseudo-R2 ∑wi = 0·999 ∑wi = 0·364 ∑wi = 0·935 ∑wi = 0·421 ∑wi = 0·264 ∑wi = 0·640
1 0·00 0·166 0·743 0·468 (<0·001) – −1·362 (<0·001) −0·020 (0·085) – 0·026 (0·040)
2 0·00 0·166 0·743 0·468 (<0·001) – −1·362 (<0·001) – – −0·006 (0·421)
3 1·00 0·100 0·719 0·418 (<0·001) – −1·323 (<0·001) – – –
4 1·58 0·075 0·765 0·408 (<0·001) −1·428 (0·968) −1·074 (0·107) – – 0·006 (0·383)
5 1·58 0·075 0·765 0·408 (<0·001) −1·428 (0·947) −1·074 (0·095) −0·018 (0·134) – 0·025 (0·055)
Open in a new tab

Discussion

We tested the hypotheses that large-scale drivers have a significant influence on local bark beetle outbreak dynamics, and that complex interactions exist between these large-scale drivers. Based on a 23-year time series of the P. abies – I. typographus system analysed across seven orders of magnitude in scale we found support for both hypotheses, documenting that bark beetle outbreaks are significantly affected by regional and landscape factors. Our analysis underlines that graph-based indicators are an intuitive yet powerful means to condense the complex spatial information of beetle and host populations into simple indices of high explanatory value for natural disturbance dynamics. Considering that the high spatial resolution data required to conduct such analyses are becoming increasingly available through advances in remote sensing, such approaches could increasingly be used in disturbance risk assessment in the future. In this context, it has to be noted, however, that the 30-m grain used here still aggregates over several trees and precludes a more detailed analysis of tree-level infestation status, which might be relevant for detecting transitions from non-outbreak to outbreak dynamics. Furthermore, automatic detection and attribution of bark beetle damages from remote sensing products still remain challenging (Kautz 2014). With regard to time-series analysis a particular uncertainty remains the stringent attribution of bark beetle damage to a given year, as trees infested by a second generation of beetles in late summer might only become apparent as beetle kill in the following spring. However, since these problems pertain equally to bark beetle data at the stand, landscape and regional scale, we deem them to have only a minor impact on the main findings of our analysis.

One particularly interesting finding is that regional bark beetle population pressure was strongly related to the temporal outbreak pattern at the BFNP. While the BFNP and its no intervention policy towards disturbances has been frequently alluded to be responsible for regional bark beetle outbreaks (Müller 2011), our analysis indicates that the temporal outbreak dynamics observed at the BFNP are not only consistent with the regional-scale development but even influenced by it. High regional infestation levels also lead to a high local risk, a relationship that is not even buffered or broken by years with ample water supply. Our study thus suggests important cross-scale interactions, yet the mechanisms underlying this finding need to be investigated further. Besides increased population pressure due to a higher number of reproducing beetles in the region one relevant mechanism could be the increased connectivity between bark beetle metapopulations under regional outbreak conditions. Yet, also local-scale interactions, such as a negative feedback on population dynamics through density dependence of reproductive success and intraspecific competition need to be considered here (cf. Marini et al. 2013). However, this would require the analysis of bark beetle population data (e.g. from beetle traps), which are unfortunately not available with landscape-scale coverage.

In the context of ecosystem management, our findings highlight the existence of important drivers at scales beyond the influence of stand-scale management. Consequently, addressing bark beetle risk needs to better incorporate landscape and regional processes in order to be effective. Our analyses suggest that a focus on the connectivity between host and beetle populations could be a key element of such a landscape-scale risk management approach. In the short term, outbreak probabilities could be reduced by specifically targeting spatial connectivity in outbreak spots (e.g. after wind-throw) with management actions aimed to reduce bark beetle densities (e.g. pheromone traps or the removal of infested trees prior to the dispersal of the offspring). Since such measures might also have negative consequences on biodiversity (Thorn et al. 2014), our findings on spatial connectivity might be used in the future to distinguish areas where removing infested trees is critical from others where the habitat benefits of retaining beetle-killed trees outweigh the risk for further infestation. Over longer time-scales, stand structure and composition could also be managed for increased heterogeneity in order to reduce large, well-connected patches of mature Norway spruce, for example by reintroducing naturally occurring admixed species such as European beech Fagus sylvatica L. and silver fir Abies alba Mill., and promoting a diversity in stand ages and structures.

While landscape drivers can be influenced by (spatially coordinated) management measures, regional-scale drivers such as climatic variability and large-scale bark beetle pressure are beyond the influence of individual ecosystem managers. From the manager’s perspective, an important response strategy besides anticipatory risk mitigation is thus to strengthen the ability to cope with and recover from disturbance events (Seidl 2014). The fact that such considerations of resilience are becoming increasingly important is underlined by our finding of the high sensitivity of bark beetle infestations to summer drought (see also Hart et al. 2014; Netherer et al. 2015). As such conditions are expected to become more frequent in the future (IPCC 2012), our data suggest that infestation probability might increase under climate change. Particularly noteworthy in this context is the finding of a drought-mediated high infestation probability even when regional bark beetle population pressure is low (Fig. 5). This supports the hypothesis that variation in climate can act as a trigger for bark beetle outbreaks (see, e.g. Marini et al. 2012) and facilitate a regionally synchronized transition from low to high population levels. The broader finding that variation in water availability is more important than variation in temperature is also in line with previous findings for Central Europe (Seidl, Schelhaas & Lexer 2011; Thom et al. 2013). However, these sensitivities might differ at the leading edge of the bark beetle distribution (e.g. in high latitudes or altitudes), where insect development is more strongly limited by temperature (Jönsson et al. 2011).

More broadly, our results highlight the importance of contingencies and interacting drivers for understanding bark beetle outbreak dynamics. We, for instance, found strong support for the hypothesis that the effects of climate vary with the stages of population dynamics in the P. abies – I. typographus system. This result is consistent with findings for bark beetles and defoliators in North America (Raffa et al. 2008; Bouchard & Auger 2014). Furthermore, also an important role of the interplay between the spatial structure of the beetle and host populations was strongly supported by our data. This underlines that neither a beetle- nor host-centric perspective is sufficient to understand bark beetle outbreak dynamics, but that an integrative view of the host tree–bark beetle system is needed.

CONCLUSION

In conclusion, our analysis showed that a multiscale perspective is important to understand the outbreak dynamics of I. typographus in Central Europe. Besides local site and stand variables, regional as well as landscape-scale processes are prominent drivers of bark beetle outbreaks. This first quantitative description of multiscale controls for the P. abies – I. typographus system is well in line with findings for other disturbance regimes (e.g. Raffa et al. 2008; Simard et al. 2012; Nash et al. 2014). On the one hand, the importance of large-scale controls highlights the limitations of stand-level management to address disturbances in a ‘command and control’ manner (as was the historic default in Central Europe) as our results underline that the bark beetle infestation risk is strongly dependent on variables that are beyond the influence of forest managers. On the other hand, however, the finding of strong large-scale controls of bark beetle outbreaks also opens possibilities to explore the utility of large-scale drivers and regional synchrony as early warning indicators (cf. Scheffer et al. 2009) of increasing local disturbance risk. This calls for regional, cross-border disturbance monitoring systems as well as for a strengthened role of landscape analysis in ecosystem management.

Supplementary Material

Appendix S1-S6

Acknowledgements

R.S. acknowledges support from the Austrian Science Fund (FWF) under the grant P 25503-B16, as well as from a European Commission’s FP7 Marie Curie Career Integration Grant (PCIG12-GA-2012-334104). We thank three anonymous reviewers for their helpful comments on an earlier version of the manuscript.

Footnotes

Data accessibility

The data used in this study are archived in the Dryad Digital Repository doi: 10.5061/dryad.c5g9s (Seidl et al. 2015).

Supporting Information

Additional Supporting Information may be found in the online version of this article.

References

  1. Bentz BJ, Régnière J, Fettig CJ, Hansen EM, Hayes JL, Hicke JA, Kelsey RG, Negrón JF, Seybold SJ. Climate change and bark beetles of the western United States and Canada: direct and Indirect effects. BioScience. 2010;60:602–613. [Google Scholar]
  2. Beudert B, Bässler C, Thorn S, Noss R, Schröder B, Dieffenbach-Fries H, Foullois N, Müller J. Bark beetles increase biodiversity while maintaining drinking water quality. Conservation Letters. 2015;8:272–281. [Google Scholar]
  3. Bouchard M, Auger I. Influence of environmental factors and spatio-temporal covariates during the initial development of a spruce budworm outbreak. Landscape Ecology. 2014;29:111–126. [Google Scholar]
  4. Burnham KP, Anderson DR. Model Selection and Inference: A Practical Information-Theoretic Approach. Springer; Berlin: 2002. [Google Scholar]
  5. Cribari-Neto F, Zeileis A. Beta Regression in R. Journal Of Statistical Software. 2010;34:1–24. [Google Scholar]
  6. Csardi G, Nepusz T. The igraph software package for complex network research. InterJournal Complex Systems. 2006;1695 [Google Scholar]
  7. Dale MRT, Fortin M-J. From graphs to spatial graphs. Annual Review of Ecology, Evolution, and Systematics. 2010;41:21–38. [Google Scholar]
  8. Fahse L, Heurich M. Simulation and analysis of outbreaks of bark beetle infestations and their management at the stand level. Ecological Modelling. 2011;222:1833–1846. [Google Scholar]
  9. Ferrari S, Cribari-Neto F. Beta regression for modelling rates and proportions. Journal of Applied Statistics. 2004;31:799–815. [Google Scholar]
  10. Hart SJ, Veblen TT, Eisenhart KS, Jarvis D, Kulakowski D. Drought induces spruce beetle (Dendroctonus rufipennis) outbreaks across northwestern Colorado. Ecology. 2014;95:930–939. doi: 10.1890/13-0230.1. [DOI] [PubMed] [Google Scholar]
  11. Heurich M, Ochs T, Andresen T, Schneider T. Object-orientated image analysis for the semi-automatic detection of dead trees following a spruce bark beetle (Ips typographus) outbreak. European Journal of Forest Research. 2010;129:313–324. [Google Scholar]
  12. Hicke JA, Logan JA, Powell J, Ojima DS. Changing temperatures influence suitability for modeled mountain pine beetle (Dendroctonus ponderosae) outbreaks in the western United States. Journal of Geophysical Research. 2006;111:G02019. [Google Scholar]
  13. Hothorn T, Hornik K, van de Wiel MA, Zeileis A. A Lego system for conditional inference. The American Statistician. 2006;60:257–263. [Google Scholar]
  14. IPCC . Managing the risks of extreme events and disasters to advance climate change adaptation. A special report of working groups I and II of the intergovernmental panel on climate change. In: Field CB, Barros V, Stocker TF, Qin D, Dokken DJ, Ebi KL, Mastrandrea MD, Mach KJ, Plattner GK, Allen SK, Tignor M, Midgley PM, editors. Intergovernmental Panel on Climate Change. Cambridge University Press; Cambridge, UK: 2012. p. 582. [Google Scholar]
  15. Jactel H, Nicoll BC, Branco M, Gonzalez-Olabarria JR, Grodzki W, Längström B, et al. The influences of forest stand management on biotic and abiotic risks of damage. Annals of Forest Science. 2009;66:701. [Google Scholar]
  16. Jönsson AM, Harding S, Krokene P, Lange H, Lindelöw Å, Økland B, Ravn HP, Schroeder LM. Modelling the potential impact of global warming on Ips typographus voltinism and reproductive diapause. Climatic Change. 2011;109:695–718. [Google Scholar]
  17. Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, et al. The NCEP/NCAR 40-year reanalysis project. Bulletin of the American Meteorological Society. 1996;77:437–471. [Google Scholar]
  18. Kausrud K, Økland B, Skarpaas O, Grégoire J-C, Erbilgin N, Stenseth NC. Population dynamics in changing environments: the case of an eruptive forest pest species. Biological Reviews. 2012;87:34–51. doi: 10.1111/j.1469-185X.2011.00183.x. [DOI] [PubMed] [Google Scholar]
  19. Kautz M. On correcting the time-lag bias in aerial-surveyed bark beetle infestation data. Forest Ecology and Management. 2014;326:157–162. [Google Scholar]
  20. Kautz M, Dworschak K, Gruppe A, Schopf R. Quantifying spatio-temporal dispersion of bark beetle infestations in epidemic and non-epidemic conditions. Forest Ecology and Management. 2011;262:598–608. [Google Scholar]
  21. Lausch A, Heurich M, Fahse L. Spatio-temporal infestation patterns of Ips typographus (L.) in the Bavarian Forest National Park. Germany. Ecological Indicators. 2013;31:73–81. [Google Scholar]
  22. Marini L, Ayres MP, Battisti A, Faccoli M. Climate affects severity and altitudinal distribution of outbreaks in an eruptive bark beetle. Climatic Change. 2012;115:327–341. [Google Scholar]
  23. Marini L, Lindelöw Å, Jönsson AM, Wulff S, Schroeder LM. Population dynamics of the spruce bark beetle: a long-term study. Oikos. 2013;122:1768–1776. [Google Scholar]
  24. Meddens AJH, Hicke JA, Ferguson CA. Spatiotemporal patterns of observed bark beetle-caused tree mortality in British Columbia and the western United States. Ecological Applications. 2012;22:1876–1891. doi: 10.1890/11-1785.1. [DOI] [PubMed] [Google Scholar]
  25. Minor ES, Urban DL. A graph-theory framework for evaluating landscape connectivity and conservation planning. Conservation Biology. 2008;22:297–307. doi: 10.1111/j.1523-1739.2007.00871.x. [DOI] [PubMed] [Google Scholar]
  26. Müller M. How natural disturbance triggers political conflict: bark beetles and the meaning of landscape in the Bavarian Forest. Global Environmental Change. 2011;21:935–946. [Google Scholar]
  27. Müller J, Noss RF, Bussler H, Brandl R. Learning from a ‘benign neglect strategy’ in a national park: response of saproxylic beetles to dead wood accumulation. Biological Conservation. 2010;143:2559–2569. [Google Scholar]
  28. Nash KL, Allen CR, Angeler DG, Barichievy C, Eason T, Garmestani AS, et al. Discontinuities, cross-scale patterns, and the organization of ecosystems. Ecology. 2014;95:654–667. doi: 10.1890/13-1315.1. [DOI] [PubMed] [Google Scholar]
  29. Netherer S, Schopf A. Potential effects of climate change on insect herbivores in European forests—General aspects and the pine processionary moth as specific example. Forest Ecology and Management. 2010;259:831–838. [Google Scholar]
  30. Netherer S, Matthews B, Katzensteiner K, Blackwell E, Henschke P, Hietz P, et al. Do water-limiting conditions predispose Norway spruce to bark beetle attack? New Phytologist. 2015;205:1128–1141. doi: 10.1111/nph.13166. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Overbeck M, Schmidt M. Modelling infestation risk of Norway spruce by Ips typographus (L.) in the Lower Saxon Harz Mountains (Germany) Forest Ecology and Management. 2012;266:115–125. [Google Scholar]
  32. Pascual-Hortal L, Saura S. Comparison and development of new graph-based landscape connectivity indices: towards the priorization of habitat patches and corridors for conservation. Landscape Ecology. 2006;21:959–967. [Google Scholar]
  33. Pasztor F, Matulla C, Rammer W, Lexer MJ. Drivers of the bark beetle disturbance regime in Alpine forests in Austria. Forest Ecology and Management. 2014;318:349–358. [Google Scholar]
  34. R Development Core Team . R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing; Vienna, Austria: 2013. [Google Scholar]
  35. Raffa K, Aukema B, Bentz B, Carroll A, Hicke JA, Turner MG, Romme WH. Cross-scale drivers of natural disturbances prone to anthropogenic amplification: the dynamics of bark beetle eruptions. BioScience. 2008;58:501–518. [Google Scholar]
  36. Scheffer M, Bascompte J, Brock WA, Brovkin V, Carpenter SR, Dakos V, et al. Early-warning signals for critical transitions. Nature. 2009;461:53–59. doi: 10.1038/nature08227. [DOI] [PubMed] [Google Scholar]
  37. Seidl R. The shape of ecosystem management to come: anticipating risks and fostering resilience. BioScience. 2014;64:1159–1169. doi: 10.1093/biosci/biu172. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Seidl R, Schelhaas M-J, Lexer MJ. Unraveling the drivers of intensifying forest disturbance regimes in Europe. Global Change Biology. 2011;17:2842–2852. [Google Scholar]
  39. Seidl R, Schelhaas M-J, Lindner M, Lexer MJ. Modelling bark beetle disturbances in a large scale forest scenario model to assess climate change impacts and evaluate adaptive management strategies. Regional Environmental Change. 2009;9:101–119. [Google Scholar]
  40. Seidl R, Eastaugh CS, Kramer K, Maroschek M, Reyer C, Socha J, Vacchiano G, Zlatanov T, Hasenauer H. Scaling issues in forest ecosystem management and how to address them with models. European Journal of Forest Research. 2013;132:653–666. [Google Scholar]
  41. Seidl R, Schelhaas M-J, Rammer W, Verkerk PJ. Increasing forest disturbances in Europe and their impact on carbon storage. Nature Climate Change. 2014;4:806–810. doi: 10.1038/nclimate2318. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Seidl R, Müller J, Hothorn T, Bässler C, Heurich M, Kautz M. Data from: small beetle, large-scale drivers: how regional and landscape factors affect outbreaks of the European spruce bark beetle. Dryad Digital Repository. 2015 doi: 10.1111/1365-2664.12540. http://dx.doi.org/10.5061/dryad.c5g9s. [DOI] [PMC free article] [PubMed]
  43. Simard M, Powell EN, Raffa KF, Turner MG. What explains landscape patterns of tree mortality caused by bark beetle outbreaks in Greater Yellowstone? Global Ecology and Biogeography. 2012;21:556–567. [Google Scholar]
  44. Singer JD, Willett JB. It’s about time: using discrete-time survival analysis to study duration and timing of events. Journal of Educational Statistics. 1993;18:155–195. [Google Scholar]
  45. Thom D, Seidl R, Steyrer G, Krehan H, Formayer H. Slow and fast drivers of the natural disturbance regime in Central European forest ecosystems. Forest Ecology and Management. 2013;307:293–302. [Google Scholar]
  46. Thorn S, Bässler C, Gottschalk T, Hothorn T, Bussler H, Raffa K, Müller J. New insights into the consequences of post-wind-throw salvage logging revealed by functional structure of saproxylic beetles assemblages. PLoS One. 2014;9:e101757. doi: 10.1371/journal.pone.0101757. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Urban D, Keitt T. Landscape connectivity : a graph-theoretic perspective. Ecology. 2001;82:1205–1218. [Google Scholar]
  48. Weed AS, Ayres MP, Hicke JA. Consequences of climate change for biotic disturbances in North American forests. Ecological Monographs. 2013;83:441–470. [Google Scholar]
  49. Wermelinger B. Ecology and management of the spruce bark beetle Ips typographus – a review of recent research. Forest Ecology and Management. 2004;202:67–82. [Google Scholar]
  50. Zeileis A. Econometric computing with HC and HAC covariance matrix estimators. Journal of Statistical Software. 2004;11:1–17. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Appendix S1-S6
NIHMS67412-supplement-1.docx (662.8KB, docx)

RESOURCES