Epiphyte metapopulation persistence after drastic habitat decline and low tree regeneration: time-lags and effects of conservation actions

Authors


Correspondence author. E-mail: victor.johansson@slu.se

Summary

  1. Old trees have declined in Europe due to agricultural intensification and forestry. For shade-intolerant epiphytic species occurring on old trees in semi-open landscapes, host tree numbers have further decreased because of shading by developing secondary woodland. Moreover, in this habitat, regeneration that could replace the extant old trees is low. This suggests that epiphytic species associated with old trees are declining. However, for species with low extinction rates, the decline may be slow and hard to elucidate.
  2. We investigated the persistence of five old-oak-associated epiphytic lichens with different traits by simulating metapopulation dynamics using Bayesian incidence function models for dynamic landscapes. With an oak-rich landscape as a reference, we investigated effects of (i) drastic habitat decline, (ii) conservation actions such as clearing around trees or increased regeneration rate, (iii) low tree regeneration and (iv) clearing and increased regeneration after 100 years of low regeneration.
  3. After drastic habitat decline, the number of occupied trees continued to decrease, displaying long time-lags before reaching new metapopulation equilibriums. Lichen extinction risks increased with decreasing habitat and were highest for species that only colonise very old trees or have large dispersal propagules. In landscapes with low tree densities, conservation actions had only minor effects on lichen extinction risks.
  4. Low tree regeneration rates increased lichen extinction risks, but species declines were slow. Conservation actions that increased regeneration after 100 years of low regeneration decreased the extinction risks to very low levels.
  5. Synthesis and applications. Due to low rates of local extinction, epiphytes display long time-lags to reach new equilibriums after habitat loss. Thus, we should expect ongoing declines in epiphyte metapopulations in landscapes where old trees have recently declined. Slow extinction gives an opportunity to improve persistence by conservation actions, but the success depends on species traits and the current density of old trees. In landscapes with many old but few young trees, epiphytes may persist if conservation actions quickly address the need to increase tree regeneration rates. The best conservation approach for long-term persistence of epiphytic lichens is to ensure regular tree regeneration in landscapes with a current high density of old trees.

Introduction

Habitat loss in current human-dominated landscapes is a major driver of declining biodiversity (Fahrig 2003). Old trees constitute habitat for many organisms in traditional agricultural landscapes and natural forests in Europe (e.g. Berg et al. 1994; Kirby et al. 1995; Thor, Johansson & Jönsson 2010). However, old trees have declined due to agricultural intensification and forestry, and a large proportion of the old trees is now restricted to remnant wooded grasslands and old-growth forests (e.g. Kirby et al. 1995; Nilsson 1997). In semi-open landscapes, poor management practices result in the development of secondary woodland, which not only leads to increased mortality of the old trees, but also leads to a decreased number of suitable trees for shade-intolerant epiphytic species (Kirby et al. 1995; Paltto et al. 2011). In some regions, there is a lack of natural tree regeneration (i.e. young trees that can eventually replace the old ones) due to intensive browsing (e.g. Kirby et al. 1995; Kouki, Arnold & Martikainen 2004; Plieninger & Schaar 2008) or the absence of forest fires (Lankia et al. 2012). Even if regeneration rates are soon restored to compensate for tree death, there will be significant gaps in the tree age structure (e.g. Plieninger & Schaar 2008; Fischer et al. 2009; Ranius, Niklasson & Berg 2009), such that species that depend on old trees will face enhanced extinction risk.

Epiphytic species associated with old trees have declined or can be expected to decline in the future (Berg et al. 1994). Decreasing population size decreases colonisation rates (Hanski 1999), moving species closer to the threshold where colonisations cannot compensate for local extinctions (Hanski & Ovaskainen 2002), potentially resulting in species extinction (Hanski 1999). However, species with slow colonisation–extinction dynamics can be expected to have slow responses to habitat change and may remain in the landscape long after habitat decline (Ovaskainen & Hanski 2002; Vellend et al. 2006), constituting part of an extinction debt (sensu Tilman et al. 1994). This is a major challenge in biodiversity conservation (Kuussaari et al. 2009) because for most species we do not know the duration of these time-lags after habitat decline.

Sessile organisms, for example terrestrial cryptogams and vascular plants, are often long-lived and have slow colonisation–extinction dynamics that may depend on the dynamics of their habitat patches (Verheyen et al. 2004; Snäll, Ehrlén & Rydin 2005; Johansson, Ranius & Snäll 2012). Local extinctions may either be stochastic from intact patches or deterministic as patches disappear. For some species, stochastic extinctions seem negligible (Snäll, Ehrlén & Rydin 2005; Jönsson, Edman & Jonsson 2008; Johansson, Ranius & Snäll 2012) whereby their extinction rates are set by the rate of patch destruction (Snäll, Ribeiro & Rydin 2003). Many of these species are dispersal limited as their colonisation rates depend on connectivity to surrounding occupied patches (Snäll, Ehrlén & Rydin 2005; Jönsson, Edman & Jonsson 2008; Johansson, Ranius & Snäll 2012), in accordance with metapopulation theory (Hanski 1999). The colonisation rates may depend on species traits. Species with narrow niches (few suitable patches) and large dispersal propagules (low propagule production; Jakobsson & Eriksson 2000; or short dispersal; Löbel, Snäll & Rydin 2009) are expected to have lower colonisation rates than species with wide niches and small dispersal propagules (Johansson, Ranius & Snäll 2012).

Pendunculate oak Quercus robur is a key substrate for many epiphytic species (Niklasson & Nilsson 2005) because it lives longer than most other tree species in Northern Europe (Drobyshev & Niklasson 2010). An example is epiphytic lichens, of which many depend on the characteristic coarse bark of old oaks (Thor & Arvidsson 1999). For several of these, a large proportion of the global occurrences have been recorded in south-eastern Sweden (‘Global biodiversity information facility 2012’). In Sweden, oaks declined drastically during the early 19th century (Eliasson & Nilsson 2002). Today, old oaks are mainly restricted to remnant wooded grasslands, where oak regeneration is often poor (Ranius, Niklasson & Berg 2009). Moreover, many old oaks are today surrounded by secondary woodland (approximately 15% in a county in south-eastern Sweden; Claesson & Ek 2009), resulting in lower occupancy of several oak-dependent lichens (Paltto et al. 2011). However, the oaks in these areas have the potential to become suitable patches if bushes and young trees are cleared around them, which is a common restoration action (Read 2000; Claesson & Ek 2009).

Metapopulation models can be used to increase our understanding of the colonisation–extinction dynamics of populations living in fragmented landscapes (e.g. Hanski 1999) and to investigate the effects of habitat loss and conservation actions (e.g. Morris & Doak 2002). However, a key problem when studying organisms with slow dynamics is that data on colonisations and extinctions are very time- or resource-consuming to collect. Therefore, estimating rates from data at one point in time (snapshot data) is useful. Verheyen et al. (2004) extended the incidence function model (IFM; Hanski 1994) to dynamic landscapes by adding a temporal dimension (patch age) to the model, a feature that was lacking in the original IFM. The model has been used to estimate colonisation and extinction rates of forest plants (Verheyen et al. 2004) and epiphytes (Johansson, Ranius & Snäll 2012).

The aim of this study is to investigate differences in the persistence of five oak-associated epiphytic lichens with different traits in scenarios of decreasing tree densities, low tree regeneration and conservation actions. We did this based on simulations using Bayesian incidence function models (Johansson, Ranius & Snäll 2012) and an oak-rich landscape in Sweden as our reference. First, we investigated the effect of drastic habitat decline on species persistence. These simulations also reveal the time-lags until a new equilibrium is reached. We decreased the number of trees in the study landscape to 10% or 40% of the reference level and estimated the probability that the extinction risks were higher than in the reference landscape. Second, we investigated the effect of conservation actions in landscapes where the oak density has decreased to 10% or 40% of the oak density in the reference landscape. Specifically, we estimated the probability that species persistence was improved by actions such as creating 15% more old habitat trees (by simulating a situation when oaks that are currently unsuitable because of shading become suitable again due to clearing of their neighbourhood) or increasing the tree regeneration rate such that the tree density of the reference landscape was achieved. Third, we estimated the probability that a low tree regeneration rate decreased species persistence by decreasing the rate to 10% in the reference landscape. Fourth, we estimated the probability that conservation actions after 100 years of low tree regeneration increased species persistence, that is, in a landscape with a gap in the tree age structure. We investigated effects of (i) setting the regeneration rate to that of the reference landscape, (ii) creating 15% more old oaks (simulating clearing around old trees) in combination with the regeneration rate of the reference landscape or (iii) doubling the regeneration rate during 100 years after the age gap.

Materials and methods

Study species

We investigated differences in persistence of five oak-associated epiphytic lichens, with varying niche breadth (number of suitable patches in the landscape) and propagule size (Table 1). We have shown earlier that these species traits may affect colonisation rates: species with narrow niches (i.e. colonising only old trees) or large dispersal propagules have lower colonisation rates compared to species with wide niches (i.e. colonising young and old trees) and small propagules (Table 1; Johansson, Ranius & Snäll 2012). We chose the study species because they are associated with old oaks and are of conservation concern; all species except C. phaeocephala are red-listed in Sweden (Gärdenfors 2010). Chaenotheca phaeocephala may also occur on other old deciduous trees, but apart from oaks we only found it on two old Sorbus intermedia in the study landscape (Johansson, Ranius & Snäll 2012).

Table 1. Study species characteristics. Dispersal mode from Smith et al. (2009). Lower age limit (i.e. the age at which the tree becomes suitable for colonisation) and mean colonisation and stochastic extinction rates from Johansson, Ranius & Snäll (2012). Generally the size of soredia>spores>conidia
Lichen speciesMain dispersal modeLower age limit on oaksaColonisation rateaExtinction ratea
  1. a

    95% credible intervals (Bayesian confidence intervals) for the best model in Johansson, Ranius & Snäll (2012). Colonisation and extinction rates are presented as the mean yearly rate among suitable trees in the reference landscape.

Buellia violaceofusca G.Thor and Muhr.Soredia107–145(6·4–12·8) × 10−40
Chaenotheca phaeocephala (Turner) Th. Fr.Spores92–99(2·8–4·9) × 10−2(4·9–26·0) × 10−4
Cliostomum corrugatum (Ach.:Fr.) Fr.Spores/conidia127–146(1·9–3·2) × 10−30
Lecanographa amylacea (Ehrh. ex Pers.) Egea and TorrenteSpores201–296(2·6–23·7) × 10−40
Schismatomma decolorans (Turner and Borrer ex Sm.) Clauzade and VězdaSoredia209–290(2·9–17·6) × 10−40

Model and simulated scenarios

To predict future epiphyte metapopulation dynamics, we used a Bayesian incidence function model for dynamic landscapes (Verheyen et al. 2004; Johansson, Ranius & Snäll 2012; Appendix S1, Supporting Information) that was fitted to snapshot data on the five study species on 2083 oaks (Johansson, Ranius & Snäll 2012). These species-specific models were built by comparing models with different combinations of the explanatory variables using the deviance information criterion (Spiegelhalter et al. 2002, Appendix S1, Supporting information). The data for the models were collected in a landscape in the oak-rich county of Östergötland, south-eastern Sweden, which has a long history of oak-rich wooded meadows. In the present study, we modified this landscape to simulate 11 scenarios for investigating how epiphyte population persistence is affected by drastic habitat decline, low tree regeneration and conservation actions.

First, we investigated the effect of drastic habitat decline. We compared the reference landscape (F1; Table 2) with landscapes where the oak density at year 0 had declined to 10% (F2) and 40% (F3) of the density in the reference landscape (‘10% landscape’ and ‘40% landscape’ henceforth). The oak density in these two landscapes represented the mean and upper 95% quantile, respectively, of the density in randomly selected equal-sized areas in the county of Östergötland (for details see Appendix S2, Supporting information). In all three scenarios, we use a regeneration rate of oaks that kept the age structure constant over time (‘normal rate’ henceforth; Appendix S2, Supporting information).

Table 2. Extinction risks (%) for the five epiphytic lichens in scenarios of habitat decline, low tree regeneration and conservation actions. ‘Constant’ means a regeneration rate that keeps the tree age structure constant over time. ‘Clearing’ means creating 15% more old oaks than the initial conditions, simulating a situation when unsuitable shaded oaks immediately become suitable due to clearing of bushes and young trees around them. ‘High regeneration’ means a regeneration rate that in the long run generates the same number of trees as the reference landscape. ‘Low regeneration’ is 10% of the regeneration rate that keeps the tree age structure constant. - means that the species was not included in the scenario as the extinction risk was negligible in a worse scenario
Scenario Oak numbers % Chaenotheca phaeocephala Cliostomum corrugatum Buellia violaceofusca Schismatomma decolorans Lecanographa amylacea
  1. a

    After 100 years of low (10%) oak regeneration.

Constant (Reference landscape)F110000001·0
ConstantF21000·22·415·061·8
ConstantF3400000·417·0
ClearingF41000·21·813·659·4
ClearingF540000·416·8
High regenerationF61000·21·412·059·2
High regenerationF740000·417·2
Low regenerationF810000·612·643·8
Normal regenerationaF91000003·8
Normal regeneration and clearingaF101002·6
Double regenerationaF111001·6

Second, we investigated the effect of two conservation actions – clearing around shaded oaks and increasing tree regeneration rates in the 10% landscape and in the 40% landscape. To simulate the effect of clearing around old trees that are unsuitable habitat patches due to shading (Paltto et al. 2011), we randomly (Appendix S2, Supporting information) created 15% more old trees (>100 years) than the initial number in the 10% landscape (F4; Table 2) or the 40% landscape (F5). We chose 15% as it corresponds to the proportion of old oaks in the county of Östergötland that today may be unsuitable for lichen colonisation due to shading from development of secondary woodland, but could be made suitable by clearing (see Appendix S2, Supporting information). To investigate the effect of increased tree regeneration, we increased the regeneration rate in the 10% landscape (F6) and the 40% landscape (F7) so that the number of old oaks observed in the reference landscape was obtained in the long run. This means creating equally many oaks every year as in the reference landscape, for example, by fencing parts of pastures or planting oaks. In all these scenarios, projections were made for 180 years before the conservation actions to better mimic the present situation in south-eastern Sweden where old oaks declined drastically approximately 180 years ago (Eliasson 2002). We thus illustrate effects of conservation actions in landscapes that had an oak density similar to the reference landscape before 1830, and where the oak density in 1830 declined drastically to 10% or 40%.

Third, we investigated the effect of low tree regeneration in the reference landscape (F8; Table 2). The tree regeneration rate, 10% of the rate that keeps the age structure constant over time, was based on tree age structures from five real oak-rich nature reserves, with obvious age gaps due to low tree regeneration during the last 100 years (Ranius, Niklasson & Berg 2009; Appendix S2, Supporting information).

Fourth, we investigated the effect of conservation actions (clearing around old oaks and increased tree regeneration) in a landscape with high current old oak density but poor regeneration (i.e. as in F8). In these scenarios, we assumed that the regeneration had been low during 100 years prior to the projections and that conservation actions were initiated at year 0. We thus create a 100-year gap in the tree age structure by assuming a regeneration rate of 10% of the rate that keeps the age structure constant over time. This mimics the situation in many currently oak-rich areas (Ranius, Niklasson & Berg 2009). We investigate effects of (i) setting the regeneration rate to that of the reference landscape (F9), (ii) creating 15% more old oaks (simulating clearing around old trees) in combination with the regeneration rate of the reference landscape (F10) or (iii) doubling the regeneration rate during 100 years after the age gap (F11).

In all scenarios, we start the simulations from the spatial configuration of species, trees and tree ages of the reference landscape. For creating the 10% and 40% landscapes, we randomly choose 10% and 40% of the trees in the reference landscape, respectively, to set the initial conditions. New trees were created in polygons representing the current spatial distribution of the trees in the study landscape (Appendix S2, Fig S2, Supporting information). Tree-specific colonisation probabilities (Ci) were calculated for each time-step (year) as a function of connectivity (Si),

display math(1)

This function assumes that propagules are dispersed from unknown background sources at long distances with the rate Ξ, or from local (within the area) occupied patches according to a dispersal kernel, where α regulates the dispersal range and rij is the distance in metres between tree i and j. The variable pj = 1 if tree j is occupied by the species; otherwise pj = 0. The colonisation parameter Φ accounts for the rate of emigration of dispersal propagules from occupied patches, and the propagule establishment ability (Appendix S1, Supporting information, Johansson, Ranius & Snäll 2012). Local extinctions occurred deterministically as trees died. For C. phaeocephala, they also occurred stochastically from living trees (Table 1, Appendix S1, Supporting information). We assumed tree mortality rates according to Drobyshev et al. (2008) and a regeneration rate that retained the tree age structure over time (Appendix S2, Fig S1, Supporting information). For each scenario, we ran 500 replicates, and for each replicate, we randomly selected one combination of model parameters from their joint posterior distributions (Johansson, Ranius & Snäll 2012). The metapopulation extinction risk is defined as the largest percentage of replicates with zero occurrences at any time-step during 500 years. This is not necessarily the last year, as we also assumed immigration, allowing recolonisations from the surrounding landscape. We estimated the probability that the metapopulation size was larger in a scenario A than in a scenario B by calculating the proportion of the 500 draws from the probability distribution of the metapopulation size in A that was larger than the probability distribution of the metapopulation size in B.

We wrote our own code in the statistical software R 2.12.2 (R Development Core Team 2011) for running the simulations.

Results

For all species, the number of occupied trees remained relatively constant in the reference landscape without habitat decline (F1), while all except S. decolorans (for which the model was spatially implicit; Appendix S1, Supporting information) continued to decrease in the 10% (F2) and 40% landscapes (F3) after habitat decline (Fig. 1). There were time-lags of at least 250 years before new equilibriums were reached in the 10% and 40% landscapes (Fig. 1b–c). The probabilities that the metapopulations were larger in the reference landscape (F1) after 500 years than in the 10% (F2) and 40% landscapes (F3) were >96% and >83%, respectively. Species with narrow niches or large dispersal propagules (i.e. low colonisation rates; Table 1) had higher extinction risks than species with wide niches and small propagules (i.e. high colonisation rates; Table 1). The extinction risk increased with decreasing oak density and was highest for L. amylacea, but was also considerable for S. decolorans and non-negligible for B. violaceofusca (Fig. 2, Table 2). Lecanographa amylacea even had a small extinction risk in the reference landscape (Table 2). The extinction risk for the two species with wide niches and small dispersal propagules, C. corrugatum and C. phaeocephala, was 0 or very low also in the 10% landscape, that is, in the worst scenario (Table 2). For C. corrugatum, the 95% confidence limits for the minimum number of occupied trees were 3–20. For C. phaeocephala, these limits were 31–74, and because all other scenarios were less severe, we only included this species in scenarios investigating the effect of conservation actions in the 10% landscape.

Figure 1.

Mean number of occupied trees and occupancy (%) among suitable trees over time for five lichen species in (a, d) the reference landscape (F1), (b, e) the 10% landscape (F2) and (c, f) 40% landscape (F3). Vertical broken lines may indicate the situation in many current landscapes in Sweden, where oaks declined drastically approximately 180 years ago.

Figure 2.

The metapopulation extinction risk (i.e. the percentage of replicates with zero occurrences) for five lichens over time in landscapes where the oak density at year 0 declined to 10% (continuous lines; F2) and 40% (broken lines; F3) of the density in the reference landscape.

Conservation actions, that is, clearing around oaks (F4, F5) or increasing the tree regeneration rate (F6, F7), had only minor effects on species extinction risks in the 10% and 40% landscapes (Table 2). All species with low colonisation rates (Table 1) had extinction risks of magnitudes similar to scenarios without conservation actions (Table 2). Specifically, the probabilities that the minimum metapopulations of the study species during 500 years were larger with clearing than with constant tree regeneration (F2, F3) were 0–18% in the 10% landscape (F4) and 0–20% in the 40% landscape (F5). The corresponding probabilities with increased tree regeneration were 0–1% (F6) and 0–7% (F7). However, in scenarios with increased tree regeneration (F6, F7), the long-term number of suitable and occupied trees increased considerably (Fig. 3). After 500 years, the probabilities of larger metapopulations were 9–100% in the 10% landscape and 0–97% in the 40% landscapes, compared to constant regeneration (F2, F3). With clearing, the corresponding probabilities were only 0–26% (F4) and 0–25% (F5). The time-lag until the metapopulation increased after the increase in regeneration was longer for species with narrow niches than for species with wide niches (Fig. 3). Moreover, the slope of the increase was higher for species with high colonisation rates than for species with low rates.

Figure 3.

The mean number of suitable trees (a), occupied trees (b) and occupancy among suitable trees (c) for five lichens in the 10% landscape where the regeneration rate is constant (F2; continuous line), 15% more old oaks were created (i.e. simulating clearing, F4; dotted line) or the regeneration rate was increased to that of the reference landscape (F6; broken line). Vertical broken lines indicate where the conservation action was initiated.

Low tree regeneration (10% of the normal rate) in a landscape with a currently high old oak density (F8) led to a long-term decline in the number of suitable and occupied trees (Fig. 4). The time-lag until the decline was longer for species with narrow niches (i.e. that can only utilise old trees) than for species with wider niches.

Figure 4.

Mean (continuous lines) number of suitable (a, d), occupied trees (b, e) and occupancy among suitable trees (c, f) for four lichens in the reference landscape with a tree regeneration rate that from 100 years prior to year 0 and onwards is 10% of the normal rate (F8; a, b, c), and with a normal regeneration after 100 years (i.e. at year 0) of low (10% of normal) regeneration (F9; d, e, f). 95% confidence limits for the number of occupied trees are shown as broken lines. Chaenotheca phaeocephala was not included in this analysis as it persisted under worse conditions.

Conservation actions that increased the tree regeneration rate to normal after 100 years of low regeneration (F9) increased the number of occupied trees after the age gap (Fig. 4e). Moreover, the actions decreased extinction risks to 0 or very low levels (L. amylacea) during 500 years (Table 2). After 500 years, there were 84–100% probabilities that the metapopulations were larger than without any action (F8), while there were 26–42% probabilities that they were smaller compared to those in the reference landscape (F1). Creating 15% more old oaks (simulating clearing) and normal regeneration (F10) or doubling the regeneration, compared to normal regeneration, during 100 years after the age gap (F11) decreased the extinction risk of L. amylacea (Table 2), as it increased the number of suitable and occupied trees (Fig. 5). After 500 years, there was a 13% and 40% probability, respectively, of a larger metapopulation of L. amylacea than with normal regeneration (F9). The remaining species had extinction risks of 0 with normal regeneration (F9) and were therefore excluded from F10 and F11.

Figure 5.

Mean (continuous lines) and 95% confidence limits (broken lines) for the number of suitable (a), occupied trees (b) and occupancy among suitable trees (c) for Lecanographa amylacea in the reference landscape where the regeneration rate had been 10% of the normal 100 years prior to the simulation start: without conservation actions (F8), with a normal regeneration rate (F9), combining normal regeneration with creating 15% more oaks (i.e. simulating e.g. clearing; F10) or doubling the normal regeneration rate under the first 100 years (F11).

Discussion

By using Bayesian incidence function models to predict future metapopulation dynamics of epiphytic lichens confined to old oaks, we show that (i) there are time-lags of at least 250 years before new equilibriums are reached following a decrease in tree numbers, (ii) large declines in the number of host trees or low tree regeneration are increasing epiphyte extinction risks, (iii) higher extinction risks are related to narrower niche breadth and larger dispersal propagule size of lichens and (iv) conservation actions may not address extinction risks in landscapes with currently low old tree densities, but can be efficient for bridging gaps in the tree age structure in areas with high old tree densities.

Habitat decline

Our models show that following a dramatic decline in viable host trees, the number of occupied trees continues to decline. This is likely to be driven by lower connectivity reducing the colonisation rate (Hanski 1999). However, there were time-lags of at least 250 years in the number of occupied trees after the habitat decline. This agrees with relationships between current distribution patterns and the historical landscape structure that suggests time-lags of more than a century for these and other sessile species (e.g. Ellis & Coppins 2007; Reitalu et al. 2009; Johansson, Snäll and Ranius, in press). The long time-lag is because of negligible stochastic extinctions of these epiphytes (Johansson, Ranius & Snäll 2012) in combinations with their host trees being long-lived and providing suitable habitat for hundreds of years (Drobyshev & Niklasson 2010). Shorter time-lags can be expected for other sessile species with faster patch dynamics (e.g. Snäll, Ehrlén & Rydin 2005; Jönsson, Edman & Jonsson 2008), or with higher stochastic extinction rates (e.g. Fedrowitz, Kuusinen & Snäll 2012; Zartman et al. 2012).

Metapopulation extinction risks were higher for species with narrow niches or large dispersal propagules (i.e. low colonisation rates; Johansson, Ranius & Snäll 2012) than for species with wider niches and small dispersal propagules (high colonisation rates). When available habitat declines, the colonisation rates of most species decrease. However, species with already low colonisation rates are those most likely to enter a situation where colonisation cannot compensate local extinction (Hanski 1999; Hanski & Ovaskainen 2002). Moreover, species with narrow niches are likely to have smaller populations than those with wide niches, further increasing their metapopulation extinction risk (Hanski 1999).

Conservation actions to increase the tree regeneration rate should be particularly efficient at preserving lichens that can colonise younger trees at a high rate. Tree regeneration may also help the recovery of extinct lichen species if their long-distance dispersal is high enough (see below). However, such conservation measures will have little impact in the short term because of the time it takes for trees to become suitable hosts. The common conservation action of clearing vegetation around old oaks surrounded by secondary woodland is assumed to increase the number of patches without a long delivery time (Read 2000; Claesson & Ek 2009). However, our results show only minor positive effects on species persistence because the potential restoration objects are so few, only increasing the number of old trees by 15% (the mean for the focal county of Östergötland).

Low tree regeneration rate

Low patch creation rates result in declining number of patches, which decreases the number of occupied patches and increases the extinction risks (Hanski 1999). However, in our system, there are long time-lags until suitable patches are delivered, and hence, before the low tree regeneration rate affects the number of suitable and occupied trees.

Our results show that many epiphytes can bridge 100-year gaps in the tree age structure when the old tree density is relatively high. Thus, lichens can be conserved in systems with many old but few young trees by rapidly addressing browsing management (e.g. Kirby et al. 1995; Kouki, Arnold & Martikainen 2004) or fire regimes (Lankia et al. 2012), even with large gaps in tree age structure (e.g. Plieninger & Schaar 2008; Fischer et al. 2009; Ranius, Niklasson & Berg 2009). This is because landscapes with high old tree density potentially harbour large metapopulations, and because these species rarely go extinct from standing trees (Johansson, Ranius & Snäll 2012), they can persist during long periods of low tree regeneration and provide dispersal sources for new suitable trees after the age gap has been filled. However, for species with narrow niches, age gaps may constitute bottlenecks with increased extinction risk even in landscapes with high old tree densities. Moreover, for epiphytes associated with trees that have shorter lifespans (e.g. Snäll, Ehrlén & Rydin 2005), or for species with considerable stochastic extinction rates (Öckinger & Nilsson 2010; Fedrowitz, Kuusinen & Snäll 2012), gaps of 100 years in the tree age structure may have much more severe impacts on population persistence compared to our study species.

Limitations of the simulations

Our models include long-distance dispersal, which has been shown to be important for the persistence of metapopulations of sessile species (e.g. Mildén et al. 2006). However, the importance of the influx of dispersal propagules from distant sources and how it changes over time is difficult to estimate. Thus, assumptions in our simulations may affect the absolute metapopulation sizes in different scenarios; however, it will not affect our conclusions regarding the relative difference among scenarios. The background deposition of propagules from unknown dispersal sources is estimated as a mean over a long time period. Thus, the current background deposition may be overestimated because oak numbers have declined during recent centuries (Eliasson & Nilsson 2002). Note, however, that the major oak decline was during the early 19th century and that was before most of today's occupied trees had become suitable for colonisation. A second limitation is that the background deposition rate was constant in all scenarios. In reality, it may decrease in the future as the number of suitable and occupied oaks decline, resulting in an underestimation of extinction risks. This is especially likely in low-density landscapes where whole metapopulations can be considered as sinks dependent on a constant influx of dispersal propagules (Shmidal & Ellner 1984).

Synthesis and implications for conservation

We show that for species with low extinction rates, there are long time-lags before the metapopulations reach new equilibriums. Therefore, in landscapes where the amount of habitat has recently declined, current species occupancies provide an over-optimistic view of the long-term size of metapopulations that can be harboured in the landscape. Epiphytes may display time-lags of several hundred years due to their low local extinction rates, and we can therefore expect future declines and extinctions in many agricultural and forest landscapes where old trees have declined during recent decades (e.g. Nilsson 1997; Plieninger & Schaar 2008). However, low extinction rates provide an opportunity to improve long-term population persistence through habitat restoration before species reach critical thresholds (Snäll et al. 2004). The success of conservation actions though depends on current old tree density and differs between species with different traits.

Conservation actions may be inefficient in landscapes with low old tree densities, especially for species with narrow niches (i.e. only colonising very old trees). Thus, the best approach for conserving such species is to ensure regular recruitment of trees into areas with high old tree densities where lichens occur today. However, many of these areas have had little or no tree regeneration for many years (e.g. Kouki, Arnold & Martikainen 2004; Plieninger & Schaar 2008; Ranius, Niklasson & Berg 2009), resulting in gaps in the tree age structure. Our results suggest that in landscapes with many old but few young trees, epiphytes could persist if conservation actions quickly address the need to increase tree regeneration rates, while the number of old trees harbour viable metapopulations.

Successful conservation of lichens associated with old oaks requires long-term management plans that strive to achieve a balance of light availability and protection against browsing, which is necessary for successful oak regeneration (Bakker et al. 2004). Pasture fencing and grazing management, or planting oaks (Pigott 1983; Fischer et al. 2009), is now necessary to ensure long-term persistence of several rare and red-listed lichen species.

Acknowledgements

We thank two anonymous reviewers for valuable comments on the manuscript and Matthew Low for linguistic editing. The work was funded by FORMAS grant 2006-2105 to T.R. and grant 2005-933 to T.S. and by a grant from Stiftelsen Eklandskapsfonden i Linköpings kommun to V.J.

Ancillary