Biological Conservation 206 (2017) 120–131
Contents lists available at ScienceDirect
Biological Conservation journal homepage: www.elsevier.com/locate/bioc
Population dynamics and future persistence of the clouded Apollo butterfly in southern Scandinavia: The importance of low intensity grazing and creation of habitat patches Victor Johansson a,⁎, Jonas Knape a, Markus Franzén b,c a b c
Department of Ecology, Swedish University of Agricultural Sciences, Box 7044, SE-75007 Uppsala, Sweden Department of Community Ecology, UFZ, Helmholtz Centre for Environmental Research, Halle, Germany Center for Ecology and Evolution in Microbial Model Systems, EEMiS, Department of Biology and Environmental Science, Linnaeus University, SE-391 82 Kalmar, Sweden
a r t i c l e
i n f o
Article history: Received 8 June 2016 Received in revised form 17 December 2016 Accepted 23 December 2016 Available online 4 January 2017 Keywords: Colonization Extinction Habitat quality Insects Metapopulation Population viability Restoration
a b s t r a c t We investigated the population dynamics and future persistence of the last remaining Clouded Apollo butterfly metapopulation in southern Scandinavia. Based on three decades of surveys (1984–2015), we modelled colonization-extinction dynamics and local population sizes using habitat patch characteristics and connectivity, while accounting for imperfect detection and uncertainty in the local population sizes. The colonization probability increased with increasing connectivity and the local extinction probability decreased with increasing local population size in accordance with metapopulation theory. The local population size increased with increasing patch area, and was also affected by grazing intensity. Light grazing resulted in larger local populations compared to heavy grazing or no grazing at all. The butterfly population has decreased considerably during the study period and according to projections of future dynamics the estimated extinction risk within the coming 10 years is 17%. However, it is possible to change the negative trends and decrease the extinction risk considerably by conservation actions. By optimizing the grazing pressure in existing patches the extinction risk was reduced to 11% (a reduction with 35% compared to the scenario with no conservation action). If a few new patches are created close to the occupied ones the extinction risk can be reduced further. In conclusion, there is a large risk that the Clouded Apollo butterfly will go extinct from southern Scandinavia within the coming decade. However, conservation measures that are focused to the core area of the current distribution and applied soon can considerably improve the situation for the butterfly. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction Semi-natural grasslands are important for biodiversity in agricultural landscapes (e.g. Duelli and Obrist, 2003). However, the intensified use of agricultural land and abandonment of less productive land in Europe during the last century has led to a major loss and fragmentation of habitat for species associated with these grasslands (Cousins et al., 2015; Krauss et al., 2010; WallisDeVries et al., 2002). For species to persist in the remaining grasslands one important factor is that habitat quality is maintained over time (Öckinger and Smith, 2007). Many grasslands rely on local grazing regimes to remain open, and abandonment leads to succession that may have negative impacts on many grassland specialists (Luoto et al., 2003; Weiss, 1999). On the other hand, too intense grazing may have negative impacts on some species (Kruess and Tscharntke, 2002; McLaughlin and Mineau, 1995). However, for most species we do not know the specific habitat requirements and how ⁎ Corresponding author. E-mail address:
[email protected] (V. Johansson).
http://dx.doi.org/10.1016/j.biocon.2016.12.029 0006-3207/© 2016 Elsevier Ltd. All rights reserved.
e.g. grazing intensity affects the population dynamics, which is essential for efficient conservation. Grassland butterflies is an example of a relatively well studied group of species that has experienced negative population trends due to the loss and changed management of semi-natural grasslands (e.g. Maes and Van Dyck, 2001; Warren et al., 2001). Although population dynamics and mobility are species-specific most butterflies seem to have a very limited dispersal ability, and their colonization probability therefore often increases with increasing connectivity to surrounding occupied patches (e.g. Hanski, 1994), in accordance with metapopulation theory (Hanski, 1999). The local extinction probability of butterflies is known to decrease with increasing local population sizes, which usually increase with patch area (e.g. Harrison et al., 1988). Therefore, patch area is often used as a rough proxy for the local population size (e.g. Hanski, 1994; Wahlberg et al., 2002) even if many suggest a more resource-based approach (e.g. Dennis et al., 2006; Turlure et al., 2010). Empirical evaluations of the patch size – population size paradigm suggest that observed local extinctions are more effectively predicted as a direct function of local population size than by patch size (Pellet et al.,
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
2007). One reason could be that the local population size, and hence the local extinction probability, is also affected by the quality of the habitat (e.g. Thomas et al., 2001). Since habitat quality is difficult to measure and differs between species it has often been neglected and less studied compared with patch area and isolation that are comparably easy to measure. However, habitat quality can contribute more to species persistence than patch area or isolation (Franzén and Nilsson, 2010; Thomas et al., 2001). For the majority of the grassland species, habitat quality may increase when less intensive late grazing is applied on managed semi-natural grasslands (Batáry et al., 2011; Dover et al., 2010; Nilsson et al., 2013) as this increases the amount of flowering plants and the availability of host plants. Effective conservation management of threatened species depends on our ability to understand and predict their local population dynamics. Projections of future dynamics are an important tool to evaluate population persistence and compare different management scenarios to find efficient conservation strategies (e.g. Beissinger and McCullough, 2002). For reliable projections, predictive models should preferably be based on long term data of local population dynamics and patch characteristics, as short time series may create misleading results that can underestimate extinction threats (Thomas et al., 2002). When comparing the effect of different management scenarios on future population sizes it is important to include model uncertainty, not only caused by stochasticity in the projected dynamics but also in the model parameters (Heard et al., 2013; McGowan et al., 2011). This can be achieved by using the Bayesian modelling framework, which is also very flexible in accounting for e.g. imperfect detection (Royle and Kéry, 2007; Sutherland et al., 2014) and handling missing data (O'Hara et al., 2002). Therefore, the Bayesian framework is very suitable for risk analysis for the management of threatened species and is increasingly used for population viability analysis (e.g. Heard et al., 2013). The aim of this study was to analyze the colonization-extinction dynamics and local population size of the Clouded Apollo butterfly (Parnassius mnemosyne), which is highly specialized species associated with semi-natural grasslands. We do this using a long term data set of a Clouded Apollo metapopulation in southern Sweden over 32 years (1984–2015) and the Bayesian modelling framework. We hypothesized that the colonization probability increases with increasing connectivity to occupied patches and that the local extinction probability decreases with increasing local population size, in accordance with metapopulation theory. Further we hypothesized that the local population size increases with increasing patch area and is related to habitat quality as measured by grazing intensity. We also aimed at evaluating future persistence of the butterfly under different scenarios of management. Specifically, we compare the population size and extinction risk until 2025 when the present conditions remain the same with scenarios of changed grazing regimes and when new patches are created by conservation actions. 2. Material and methods 2.1. Study species The Clouded Apollo (Parnassius mnemosyne) is a Palearctic butterfly that is classified as endangered (EN) in Sweden (Gärdenfors, 2015) and is also threatened elsewhere in Europe (van Swaay and Warren, 1999). The butterfly has non-overlapping generations. It is active in May–June and hibernates as an egg. The species is monophagous on host plants belonging to the genus Corydalis. In the study area it is C. pumila and C. intermedia, which both seem to have rather stable populations in southern Sweden and are therefore classified as least concern (LC) on the Swedish red-list (Gärdenfors, 2015). The Clouded Apollo is a rare inhabitant of flower-rich meadows and semi-natural grasslands, and this type of habitat is dependent on a management of either grazing or mowing. The butterfly has become extinct in many parts of its previous southern Scandinavian distribution range including Denmark (extinct
121
in 1961) and the Swedish province Scania (extinct in 1954) (Franzén and Imby, 2008). In 2015 only three local populations remained of the southern Scandinavian population, situated in the province Blekinge. Isolated populations also exist in Norway and central Sweden but they have been described as other subspecies. The decline of the Clouded Apollo has been attributed to the ceasing of traditional management regimes, grazing and mowing of semi-natural grasslands, and coppiced woodlands (Väisänen and Somerma, 1985). During the last century management of grasslands has drastically changed in southern Sweden and grazing of natural pastures has decreased, leading to gradual transformation of most grasslands into either intensively used agricultural fields or forests (e.g. Cousins et al., 2015). At the same time the few remaining grasslands have been more intensively managed and more intensively grazed by cattle (e.g. Nilsson et al., 2008). 2.2. Study area and patch definitions The field work was conducted from 1984 until 2015 in the surroundings of Ronneby in Blekinge province in southeast Sweden (long: 15.2780, lat: 56.2170). The major feature of the landscape is a mixed open-forested area with arable fields, semi-natural grasslands, deciduous forests, glades and rocks. In the study-area, all potentially suitable habitat patches (24 in total) were surveyed and mapped in 1984 (Fig. 1). Suitable patches were defined as open grasslands with N0.5 m2 cover (or N 50 shoots, see Välimäki and Itämies, 2003) of the host plant (Corydalis spp.) anywhere in the patch, and presence of the major nectar plant Lychnis viscaria (N 50 flowering individuals). Habitat patches were defined as separate if the borders were situated 150 m apart or more. This slightly more coarse patch separation, compared to other studies (e.g. Välimäki and Itämies, 2003), may result in more ‘unsuitable’ habitat being included in the patch, the absence of short inter-patch dispersal distances, and lower population densities. On the other hand, it may be more user friendly for practitioners, as our patches most often are well-defined grasslands that are naturally delimited by neighbouring arable fields or forests. By using whole grasslands both larval (host plants) and adult resources (nectar plats) are included, which has been suggested to be important when assessing patch suitability (Fred et al., 2006). However, we did not quantify these resources over time, and could therefore not model them explicitly, but they are indirectly included under the assumption that they increase with increasing patch area. 2.3. Butterfly survey During the study period we collected data on butterfly occurrences and local population sizes in the 24 patches. However, the quality of the data varied between the years because Capture Mark Release (CMR) studies were not performed each year (Appendix A). For the following 18 years: 1984 to 1987, 1991 and 2003 to 2015 the quality of the data was rather good as most patches were visited at least six times each year with standardized CMR methods used (see below). Especially, species occurrences in these years are rather reliable as several visits reduce the risk of observing false zeros (i.e. that the species was not observed when it in fact occurred in the patch). The butterfly is very easy to observe from long distances (up to 50 m) as it is a large and charismatic species exposing itself on flowers and thus has a presumably high detectability. However, we did not have information about in how many of the visits the species was detected for all patches and years. We therefore used a plug-in estimate of species detection probability in our model. The plug-in estimate was computed from three of the occupied patches that were visited more intensively (at least 9 times per year) during six consecutive years (2004–2009). During this period the species was on average found in 80% of the visits and the detection probability was thus estimated to 0.80 per patch and visit. Hence, already after three visits in one patch the detectability is extremely high (N99%). Even if the detection probability was very high we included it in the
122
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
Fig. 1. The location of the patch network in Southern Sweden (a), the spatial distribution of the 24 suitable patches (b), and the core area (c) in 2015 with the three remaining patches occupied by the butterfly in 2015 (black dots) situated in a cluster of eight suitable patches (filled dots), and twelve potentially suitable patches (empty circles) that can be made suitable for the Clouded Apollo by restoration actions.
occurrence model to account for the differences in detectability because the number of times a patch was visited differed between years (see Modelling approach below). For the periods 1988–1989 and 1992– 2002 the patches were visited more irregularly and standardized CMR studies were not performed. Therefore, from these periods observed absences are more uncertain and we lack reliable population estimates. However, by using an occupancy modelling approach (see below) we were able to utilize information on species occurrence (presence only) from these years. In the years we applied the CMR method butterflies were marked and recaptured along irregular routes through each patch. Each adult caught was marked individually using a Staedtler felt-tip pen (Lumocolor 313 Permanent) and immediately released at the point of capture. The surveys were performed between 9 a.m. and 5 p.m. Surveys were not performed in unfavourable weather conditions such as rain (within 1 h after rainfall), temperatures lower than 17 °C, or when winds were stronger than approximately 4 m/s. As movements between patches were observed the Jolly-Seber method, as applied to open populations, was used to estimate the local population sizes (e.g. Seber, 1982). For this purpose we used Jolly-Seber models as implemented in the POPAN module (Arnason and Schwarz, 1999) in Program MARK (White and Burnham, 1999). In the model, all the three components (survival, capture, and recruitment probabilities) were time dependent (i.e. they varied over time). Survival (which ranged from 0.46 to 1.00) and capture (which ranged from 0.20 to 1.00) probabilities are likely to vary over the season because of e.g. vanishing food supply and changing weather conditions (e.g. Schtickzelle et al., 2003). The reason for keeping also the recruitment rate (which ranged from 0.00 to 0.62) time dependent was that a constant rate is unlikely in species with non-overlapping generations (e.g. Schtickzelle et al., 2002). The selected model was also the one that fitted our data best based on the lower AIC compared to models with all possible combinations of constant and varying probabilities of the three components. The model output is daily estimates of the local population sizes and a total yearly local
population size N (with associated uncertainty), of which the latter is used in the further modelling (see below). If the model did not converge (due to low number of individuals) population sizes were estimated using the method described by Craig (1953). There are also other methods for obtaining estimates of population sizes from CMR data (e.g. Schtickzelle et al., 2002), that sometimes may be preferred (e.g. Schtickzelle et al., 2003), and the further modelling approach (described below) may just as well use estimates of local population sizes from such models. However, regardless of what method is used, in patch networks with high dispersal rates between patches a correction of the total number of individuals in the entire network may be necessary (if many individuals are counted in several patches). In our network the dispersal rate between patches is low (see below), and we therefore used the uncorrected local population sizes.
2.4. Grazing intensity The grazing intensity in each suitable patch was categorized by recording the number of animals per hectare and the starting date for the grazing. If grazing occurred before 15 June, independent of the number of grazing animals as early grazing is detrimental for many invertebrates (e.g. Van Noordwijk et al., 2012), or if there were more than ten grazing animals per hectare during a period of 8 weeks or more the grazing intensity was categorized as heavy. If grazing started later than 15 June and one to nine animals per hectare were present we classified the grazing as light. If no grazing animals were present it was classified as no grazing. Grazing intensity was estimated for each patch and year. Data for the grazing pressure was obtained from field observations, interviews with farmers and grazing pressure statistics for valuable semi-natural grasslands (from the Swedish board of agriculture). We measured the grass height at 15 randomly chosen patches in July 2003 (Appendix B) and it decreased with increasing number of grazing animals per hectare (P b 0.001, r = −0.89, Fig. B1).
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
123
2.5. Modelling colonization-extinction dynamics
2.6. Modelling local population size
We used a Bayesian modelling approach to simultaneously model butterfly occurrence, colonization and extinction probabilities, and local population size while accounting for imperfect detection and uncertainty in observed local population sizes. This approach utilizes the available empirical data to estimate model parameters, which are simultaneously used to predict missing data. The general model formulation is very similar to Sutherland et al. (2014), where patch-specific occurrence probability (Ψi,t) in a given year t is conditional on the occurrence state (Occi,t) in t-1 and the colonization probability (Ci,t) when unoccupied and extinction probability (Ei,t) when occupied as:
The local population size (Ni,t), given that the species occurred in a patch, was modelled based on patch characteristics. We assumed a negative binomial distribution as the data on local population size was over-dispersed. Specifically, the local population size was modelled based on patch area and grazing intensity as:
Ψi;t ¼ 1−Occi;t−1 C i;t þ Occi;t−1 1−Ei;t
ð1Þ
We modelled the colonization probability as a sigmoid function of connectivity to surrounding occupied patches as: C i;t ¼
S2i;t S2i;t
ð2Þ
þ y2
which assumes a positive relationship between connectivity (Si) and the colonization probability, where the colonization parameter y determines the slope of this relationship. Connectivity (Si) was further modelled as: n
−αdij
Si;t ¼ ∑ e j¼1
N j;t−1
ð6Þ
where β0 is the intercept, β1 the effect size parameter for the log-transformed area, which was standardized to zero mean and unit sd. The categorical variable grazing intensity had three levels where “no grazing” (GrazNi,t with associated parameter β2) and “heavy grazing” (GrazHi,t with parameter β3) are compared to the reference level “light grazing”. We also tested to include Ni,t − 1 as an explanatory variable in the model, as plots of the local population sizes over time in some patches indicated such relationship. However, when doing so the model did not converge properly, and we therefore decided to remove it. The model had a hierarchical structure assuming patch specific random intercepts, and the parameter εi is the stand specific random error modelled as: εi ~ Normal(0, σε). To include uncertainty in the local population size (Ni,t) from the Jolly-Seber estimations (see above) when fitting the model we used the estimated mean (Nesti,t) and standard error (SECMRi,t): Nest i;t Normal Ni;t ; SECMR2i;t
ð7Þ
ð3Þ
where dij is the distance in meters between the focal patch i and source patch j, α is a parameter setting the spatial scaling (which is related to dispersal distance, Appendix C, Fig. C1), and Nj,t − 1 the stage-specific local population size (see below) in source patch j. The spatial scaling parameter α can be estimated based on the spatial occurrence pattern when fitting the model (e.g. Hanski, 1999; Johansson et al., 2012). However, it most often becomes strongly correlated with the colonization parameter y (Appendix C), and it may therefore be better to estimate it based on separate data on observed dispersal distances (Hanski, 1999). As we have such information from the mark-recapture data we chose the latter approach and fitted a negative exponential function to the observed inter-patch dispersal distances (N = 24, mean = 1266 m, range = 343–2045 m). This resulted in α = 0.00079 (Fig. C1), which was entered as a fixed parameter when modelling colonization probability. When treating α as a free parameter to be fitted from the spatial occurrence pattern the model did not converge properly. However, we also performed a sensitivity analysis for the value of α, where we tested values that corresponded to both half and double of the observed mean dispersal distance (Appendix C). The extinction probability was modelled as a function of local population size as: logit Ei;t ¼ e0 þ e1 Ni;t−1
log Ni;t ¼ β0 þ β1 logðAreai Þ þ β2 GrazNi;t þ β3 GrazHi;t þ εi
ð4Þ
where e0 is the intercept and e1 the is the parameter relating extinction probability to local population size (Ni,t − 1). We also included a rescue effect (i.e. that the local population avoid extinction by immigration from surrounding occupied patches; Brown and Kodric-Brown, 1977) by multiplying Ei,t with (1 − Ci,t). To account for imperfect detection the occurrence (Occi,t), which was used in the models described above (Eq. (1)), was estimated based on the observed occurrences (Mi,t = 1 if the butterfly was found, and Mi,t = 0 otherwise), the plug-in estimate of the detection probability (p = 0.8, see above) and the number of visits (nvisit) as: ð5Þ Mi;t Bernoulli 1−ð1−pÞnvisit Occi;t
The models were fitted using JAGS 3.4.0. We ran two MCMC chains for 25,000 iterations. We kept 5000 iterations from each chain (i.e. 10,000 in total) for estimation after removal of the first 5000 iterations (burn-in) and thinning by 4. Convergence was assessed by GelmanRubin diagnostic (R), and all parameters had R b 1.05. We used vague priors (Table 2). 2.7. Projections of future dynamics under different scenarios of management To investigate the future population development under different management strategies we used the fitted models to project future dynamics during 10 years (i.e. until 2025) in four different management scenarios (Table 1). The reason for the relatively short time period of the future projections is that risk estimates for longer periods become increasingly imprecise (Beissinger and Westphal, 1998; Fieberg and Ellner, 2000). Moreover, this time-frame also fits well to the real conservation planning work. In the first scenario (S1), we assumed that the management continued as today and that current patch condition did not change. Second, we simulated that the grazing intensity was optimized to the most favorable conditions with light grazing in all 24 patches (S2). This required a change from no grazing in ten patches and from heavy grazing in five patches. Third, we simulated grazing that was only optimized in the eight patches that constitute the core area of the metapopulation today (S3). This only required a change from no grazing in three patches and from heavy grazing in two patches. The two remaining scenarios (S4 and S5) tested the effect of creating new suitable patches close the occupied ones. Based on inventories of the surrounding landscape we know that the host plant exists in closed forest patches that are currently not suitable for the butterfly (as it requires open habitats with glades or grasslands). However, these patches have the potential to become suitable if the tree cover is reduced by restoration actions. In the fourth scenario (S4), we simulated restoration that focused on patches N1 ha in size (4 in total) and in the fifth scenario (S5) we simulated restoration of all potential (12 in total, Fig. 2) patches regardless of size within 2 km from the occupied ones (i.e. the four large patches from S4 plus eight smaller ones, Table 1). We assumed light grazing in all new patches.
124
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
Table 1 The five different management scenarios. Scenario Management
Grazing intensity in the core areaa Grazing intensity outside the core area New area created
S1 S2 S3 S4 S5
Mixed Light grazing Light grazing Light grazing Light grazing
Conditions continue as today Light grazing in all patches Light grazing only in the patches of the core area Creation of four new patches N1 ha within 2 km from the occupied ones Creation of twelve new patches within 2 km from the occupied ones
Mixed Light grazing Mixed Mixed Mixed
0 ha 0 ha 0 ha 7.46 ha 12.43 ha
a The core area is the cluster of eight patches within 2 km from the currently occupied ones (Fig. 1c). Creating new patches within the core area can be done by restoration of forest patches that are today not suitable for the butterfly but where we know that the host plant exists.
Simulations of future dynamics were done in JAGS with 10,000 replicates of each scenario. Each replicate had a unique combination of the model parameters, randomly drawn from their joint posterior distribution, which thus include parameter uncertainty into the projections. The parameters were fixed over the 10 years of simulation for each draw from the posterior. In all scenarios we calculated the number of occupied patches, estimated total population size, and the metapopulation extinction risk over time. The extinction risk was defined as the proportion of replicates, at a given time-step, for which the butterfly went extinct from all patches in the whole landscape. We also estimated the quasi-extinction risk (e.g. Akçakaya, 2000; Schtickzelle and Baguette, 2004), here defined as the risk that the entire metapopulation declines below 20 individuals at any time during the 10-year projections.
3. Results During the 18 years of CMR studies a total of 1948 butterflies were marked and 4133 were released after (re)capture, meaning that each individual was caught on average 2.12 times. The estimated local population sizes varied between patches and years (Fig. A1). The predictions from the models, fitted to the empirical data (Fig. 2), showed that the colonization probability increased with increasing connectivity and the local extinction probability decreased with increasing local population size (Fig. 3, Table 2). The decrease in local extinction probability was relatively steep as the local population size increased, and already at 20 individuals the posterior extinction probability was on average b2%. The local population size increased with increasing patch area,
Fig. 2. Empirically observed data on a) colonizations in relation to connectivity (t − 1), b) extinctions in relation to the local population sizes (t – 1), c) mean local population sizes during the study period in relation to the logged patch areas (ha), and d) the local population sizes in relation to grazing intensity. In b) and d) the mean (short thin line), 50% percentiles (thick line), and 95% percentiles (long thin lines) are shown for local population sizes and in a) for connectivity. For data on local population sizes we here used the yearly estimates from the JollySeber models. When calculating connectivity we included also the predicted local population sizes for patches without data (based on our population model), as we otherwise would underestimate the total population that can contribute to colonization of empty patches.
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
125
Fig. 3. Colonization probability in relation to connectivity (a), local extinction probability in relation to local population size (b), and local population size in relation to patch area and grazing pressure (c). Thick continues lines represent means and thin broken lines the 95% confidence limits.
and was also affected by grazing intensity (Table 2, Fig. 3). Heavy grazing resulted in smaller local population sizes than light grazing (Table 2), and there was a 89% posterior probability that the population size was smaller also in patches with no grazing compared to patches with light grazing (89% of the posterior of β2 was negative). The number of occupied patches (metapopulation size) and the total number of butterflies were strongly correlated (Fig. A2) and have fluctuated during the study period, but the general pattern is a strong decrease (Fig. 4). From 1984 until 1989, 14 to 20 patches were occupied and the total population was between 400 and 600 individuals. However, during the 1990s the metapopulation declined considerably, and from 2005 to 2015 the number of occupied patches has fluctuated between 10 and 3 with a total yearly population size from ~240 individuals to b 40 individuals. In 2015 the species only occupied three patches in the eastern part of the study area, which are situated in a cluster of eight suitable patches located b 2 km from each other (Fig. 1). Projections of future population dynamics, assuming that the current patch conditions remained the same (S1), suggest that the number of occupied patches will on average be around four with an average total population size of roughly 80 individuals (Fig. 5). However, there were large uncertainties associated with these estimates. It was unlikely that the butterfly will be able to (re-)colonize any other patches than the eight closest within the coming ten years; the highest occurrence probability outside the cluster was 6.6% for the closest patch and 1.5% for the second closest (and the upper 95% credible interval is eight, Fig. 5a). The cumulative extinction risk of the entire metapopulation increased over time and the posterior probability of the population having gone extinct from southern Scandinavia until 2025 was 17% (S1, Fig. 5c). The risk that the entire population declined below 20 individuals
(quasi-extinction) at some point during this time was almost 60%. If light grazing intensity were applied in all 24 patches (S2) in the future there was a 67% probability that the total population would be larger compared to S1, with an average of ~150 individuals (Fig. 5), and the extinction risk decreased to 11% (i.e. a reduction with 35% compared to S1). The quasi-extinction risk decreased even more, to 34%. However, a qualitatively identical result was obtained when the grazing was only changed in the cluster of eight patches closest to the occupied ones (S3, Fig. 5). Restoring potential patches within 2 km from the currently occupied patches further increased the mean population size to on average ~ 230 and decreased the extinction risk to 8% (a reduction with 53% compared to S1), if focusing on the four largest patches (S4). The quasi-extinction decreased to 26%. If restoring all potential patches within 2 km (S5) the number of occupied patches increased considerably and the total mean population size was ~300 (Fig. 5), but the extinction risk in 2025 only decreased to 7% (a reduction with 59% compared to S1). However, the quasi-extinction risk decreased more from 26% to 20%. The probability that the population size was larger after ten years in S4 and S5 compared to S1 was 81% and 87%, respectively. The sensitivity analysis of the parameter α showed that our results remained qualitatively similar when the spatial scaling changed (Appendix C). 4. Discussion Based on data over 32 years from the entire Clouded Apollo butterfly metapopulation in southern Scandinavia we show (i) that butterfly population dynamics are affected by patch area, habitat quality and connectivity; and (ii) that the population has had a strong negative trend
Table 2 Parameter estimates for models of colonization probability (C), extinction probability (E) and the local population size (N) and their prior distributions. Parameter
Model
Description
Posterior mode (95% credible interval)
Prior distributiona
Y Α e0 e1 β0 β1 β2 β3 σε
C C E E N N N N N
Colonization parameter Spatial scaling parameter Intercept Slope for local population size Mean intercept Slope for patch area Slope for no grazing Slope for heavy grazing Standard deviation of the common distribution of intercepts
52.6 (38.4–70.5) 0.00079 0.08 (−0.84–1.21) −0.21 (−0.42 to −0.09) 2.47 (2.03–2.91) 0.72 (0.40–1.00) −0.39 (−0.83–0.22) −1.20 (−1.94 to −0.44) 0.54 (0.29–0.89)
Exp (0.001) Fixed U (−10,10) U (−10,10) N (0,0.01) N (0,0.01) N (0,0.01) N (0,0.01) U (0,10)
Exp, U, and N denotes the use of exponential, uniform and normal priors respectively. a The spatial scaling parameter was estimated based on observed movements between patches and included as a fixed parameter in the model.
126
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
Fig. 4. The number of occupied patches (a) and the total population size (b) over time for the Clouded Apollo in southern Scandinavia. Thick continues lines represent means and thin broken lines the 95% confidence limits.
Fig. 5. The projected number of occupied patches from 2015 to 2025 (a), total metapopulation size (b), the extinction risk of the entire metapopulation (c), and the risk that the entire metapopulation goes below 20 individuals; Quasi-extinction risk (d) for the five different conservation strategies. Thick lines represent means and thin broken lines the 95% confidence limits.
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
during the study period. Moreover, based on projections of future dynamics, we show (iii) that there is a large risk (17%) that the whole butterfly metapopulation will go extinct within 10 years if the management continues as today; but (iv) it may be possible to increase the population size and lower the extinction risk by optimizing the grazing pressure and creating new potential patches in its current core area. 4.1. Population dynamics We show that the local extinction probability of the Clouded Apollo decreases with increasing local population size, which agrees with our hypothesis and metapopulation theory (e.g. Hanski, 1999). By explicitly modelling the local population size we can show that the local population size increases with patch area, which often is assumed when having no data on population sizes (e.g. Hanski, 1994). However, compared to studies using area as a proxy for the local population size (e.g. Hanski, 1994; Thomas et al., 2001; Wahlberg et al., 2002), we can make statements about the local extinction probability in relation to the actual number of individuals. Our results suggest that only 20 individuals are associated with a rather low local extinction probability (b2%), which means that the butterfly may persist locally in low numbers during several years. However, such small populations are of course very sensitive to environmental and demographic stochasticity (e.g. Hanski, 1999), which are not explicitly included in our models (but may increase the uncertainty in model projections even more if included). We also show that the local population size was affected by grazing intensity. Hence, it is not only the patch size but also the quality of the patch that determines the local population size of the Clouded Apollo, which has also been suggested for other invertebrates (Franzén and Nilsson, 2010; Thomas et al., 2001). With intense grazing the reason often is that the host plants are eaten or trampled down (e.g. Van Noordwijk et al., 2012). However, in the case of Clouded Apollo the reason is rather that the eggs or pupae, hidden in the vegetation close to the ground, are killed by grazing animals and that nectar plants are eaten (Bubová et al., 2015), as the host plants are only present in vegetative phase early in spring (until mid May) when no livestock are normally grazing the patches. In contrast, abandoned grazing ultimately leads to an increasing cover of shrubs and trees, which is known to change the microclimate resulting in detrimental effects on many butterfly populations (Balmer and Erhardt, 2000; Öckinger et al., 2006b). However, the process can be rather slow and most species seem to be able to persist for at least ten years after abandonment (Öckinger et al., 2006a). This may explain why we see no clear effect of no grazing on the Clouded Apollo populations, compared to heavy grazing which can be expected to give a more instant effect on the population size. With longer timeframes we would expect a stronger negative effect of abandoned grazing. The colonization probability of the Clouded Apollo increased with increasing connectivity to surrounding occupied patches. Even if a positive relationship is built into the model formulation based on theory (Hanski, 1999), such relationship is also clearly supported by the estimated size of the colonization parameter y. One important reason is the restricted dispersal ability of the Clouded Apollo, which has also been emphasized in earlier studies of inter-patch movements for this species (Kuussaari et al., 2015; Välimäki and Itämies, 2003). Dispersal events N2 km seem rather rare, but occasionally happen (Välimäki and Itämies, 2003; Kuussaari et al., 2016), which agrees with our observed inter-patch dispersal distances, and the estimated parameter for spatial scaling (Fig. B1). Thus it is important that suitable patches in a metapopulation network are situated close to each other and preferably within about 2 km from an occupied patch. 4.2. Population trend We show that the Clouded Apollo population in southern Scandinavia has declined considerably during the study period. The major
127
decline seems to have occurred during the early 1990s, most likely because two large local populations in the western part of the area went extinct due to intensified grazing pressure. This may have led to a clear reduction of connectivity and thus reduced rescue effect (Brown and Kodric-Brown, 1977) in the remaining western local populations. The widely scattered distribution of the suitable patches (the average distance to the three closest patches is 1680 m) in general makes the population very vulnerable to e.g. deteriorated habitat quality, because if it becomes locally extinct recolonization is highly unlikely. The changed grazing pressure in a few patches during the early 1990s may thus have pushed the entire metapopulation over the extinction threshold (Bascompte and Sole, 1996). Then the consequence has been a relatively slow population decline, which is typical for populations close to this threshold (Hanski and Ovaskainen, 2002), and the butterfly has managed to survive in small local populations in a few patches concentrated to the more well connected eastern part of the patch network during the last decade. However, it is unlikely that it will persist in the long run without any conservation efforts (see below), and the remaining population may today thus constitute an extinction debt (Kuussaari et al., 2009). 4.3. Future persistence under different management scenarios The extinction risk for the entire metapopulation is high within the coming ten years if the patches remain as today, due to the currently small metapopulation (only three occupied patches with b40 individuals in total). Hence, the population is at present clearly not viable (Hanski et al., 1996). Even if the future projections are rather uncertain they suggest that it may be possible to change the negative trend and decrease the extinction risk by conservation actions. Just by optimizing the grazing pressure in existing patches (S2) the extinction risk was reduced to 11% (i.e. a reduction with 35%) and the quasi-extinction risk from 57% to 34% (a reduction with 40%). However, an optimized grazing regime in the cluster of eight patches (S3) was equally good as optimizing the grazing for all 24 patches in the entire patch network (S2). The reason is that recolonization of any patch outside the cluster was very unlikely, due to their low connectivity. Hence, it may be most efficient to focus conservation efforts to the patches closest to the currently occupied patches rather than scatter patches across the entire landscape (e.g. Huxel and Hastings, 1999; Johansson et al., 2016; Schultz and Crone, 2005). Apart from optimizing the grazing in existing patches, another possibility to reduce the extinction risk is to create new suitable patches by restoration actions (e.g. Öckinger et al., 2006a). Our results suggest that it is most cost-efficient to focus on creating a few large patches (S4). An additional number of small ones (S5) does not seem to decrease the extinction risk very much further and requires a larger total restoration area (Table 1). 4.4. Data and modelling approach The 32-year time-series we utilize in this study is rather long compared to most studies of butterfly metapopulations (but see e.g. Baguette et al., 2011, McLaughlin et al., 2002), which should increase our possibility of developing reliable models (Thomas et al., 2002). Moreover, few have data on local population sizes, as this is more time-consuming to collect compared to occupancy data. Many studies of butterflies have used the classical Incidence Function Model (IFM, Hanski, 1994) that can be fitted to a single occupancy snap-shot. However, this model assumes quasi-equilibrium between colonizations and local extinctions, and it may be difficult to assess whether this assumption is met. If this assumption is violated, there will be a bias in the parameter estimates and subsequent projection simulations (Baguette, 2004). For instance, our metapopulation is clearly not in equilibrium, even if it may have appeared so the first few years. Hence, even when having data from several years, which gives an opportunity for other modelling approaches (e.g. Sjögren-Gulve and Ray,
128
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
1996), results can be misleading and there is a risk of underestimating extinction threats (Thomas et al., 2002). Our study is of general relevance to the conservation of metapopulations of rare species in that we handle several data complexities that regularly arise in such cases, while making use of metapopulation theory and statistical projections of scenarios to evaluate potential conservation strategies. Specifically, we do not only model colonizationextinction dynamics, but also the local population sizes, and we account for imperfect detection, missing data and uncertainties around estimates of local population sizes. Moreover, we use our fitted model to project future persistence under realistic scenarios of changes in habitat quality and patch creation, over time-frames that fit well to the real conservation planning work. Our projections include parameter uncertainty, and not only uncertainty caused by stochasticity in the projected dynamics, which is important to not underestimate extinction risks (McGowan et al., 2011). Projections with large uncertainty often lead to larger estimated extinction risks. Differences in extinction risks between scenarios may thus not be an effect of one scenario being ecologically better than the other, but can also be caused by differences in uncertainty about scenario projections. For including parameter uncertainty the Bayesian methods that we use are very suitable (Heard et al., 2013), and this framework also enables us to estimate e.g. the probability that one scenario leads to a larger population size compared to another. This is useful when ranking management alternatives relative to each other, which is considered more robust to modelling assumptions than estimates of extinction risks (e.g. Beissinger and Westphal, 1998). 5. Conservation implications There is a large risk that the Clouded Apollo butterfly will go extinct from southern Scandinavia within the coming decade, especially if not applying any conservation measures soon. Even if our results imply
that the butterfly may persist in small local populations for quite some time, the metapopulation is at present very vulnerable. Restoration of existing habitat patches appears initially to be the most obvious way of improving the situation for the butterfly. Then, creation of new patches could considerably improve its persistence, the more patches the better, and the closer the patches are situated the better. Our results suggest that conservation efforts (at least initially) should be spatially focused to the core area of the current distribution, as a re-colonization of patches outside this area is highly unlikely at present due to their isolation. Re-colonization of patches outside the core area may require translocation, which has been shown to be successful for this species in other patch networks (Kuussaari et al., 2015). In practical terms, restoration of habitat quality implies late and low intensity grazing to decrease egg and pupae mortality, and clearing of shrubs and trees to enhance lighting of the herbaceous vegetation that may increase the accessibility and abundance of the host (Corydalis spp.) and nectar plants. Maintaining high quality habitats, with an optimal balance of an open vegetation structure and a large abundance of the host and nectar plants, requires a regular management with the appropriate (light) grazing regime, and is essential for the long-term persistence of the Clouded Apollo. Acknowledgements We thank Nicolas Schtickzelle and two anonymous reviewers for valuable comments on the manuscript, and Anders Brattström, Fredrik Bjerding, Benny Henriksson, Olle Hammarstedt and Erland Lindblad for help in the field and fruitful discussions. Thanks also to the provincial government of Blekinge, farmers and land owners that was very supportive to our study. The work was funded by a grant from Stiftelsen Oscar och Lili Lamms Minne (FO2013-0097) to VJ. JK was funded by grant 621-2012-4076 from the Swedish Research Council VR.
Appendix A. Data quality and total population size in relation to number of occupied patches During the study period we collected data on butterfly occurrences and local population sizes in the 24 patches. However, the quality of the data varied between the years (Fig. A1a). For the years where we used standardized Capture Mark Release (CMR) methods to estimate the local population sizes (Fig. A1b), recorded presences and absences are reliable and we have rather few NA:s (except in 1991). However, for the remaining years the data is poor and we only have presence-only data with many NA:s (Fig. A1a). For these years the population sizes are based on model predictions, which is the reason for the wide confidence intervals in the total population size, which is calculated as the sum of all individuals in the entire patch network (Fig. A1c). The recorded population sizes tend to be close to the lower confidence limit in years having a few NA:s, which is most clear in the first years. The reason is that the model predicts population sizes in patches with no data, and hence adds extra individuals. In 1991, only nine patches were visited and therefore the observed population size is even outside the confidence interval. The number of occupied patches and the total number of butterflies were strongly correlated (Fig. A2), and had in general decreased considerably during the study period (Fig. A1).
Fig. A1. The data quality over time with a) number of patches having recorded presences, absences and no data (i.e. NA:s), b) the estimated average local population sizes based on Capture Mark Release, and c) the observed total population size (big dots) in relation to the predicted.
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
129
Fig. A2. The mean total population size in relation to the mean number of occupied patches during the 32 year study period (R2 = 0.76, p b 0.001).
Appendix B. Sward height in relation to the number of grazing animals
Fig. B1. Sward height in relation to the number of grazing animals per hectare (r = −0.89) measured at 15 randomly chosen patches in July 2003. The patches were grazed by cattle, sheep and
Appendix C. The spatial scaling parameter and observed inter-patch dispersal distances
The parameter α is setting the spatial scaling in the connectivity measure (Eq. (3)). We used a fixed value of α = 0.00079 based on observed interpatch dispersal events from the mark-recapture data (N = 24, mean = 1266 m, range = 343–2045 m, Fig. C1); α was estimated as 1 / (the mean distance). Inter-patch dispersal was based on data from 1991 to 2015, as we lack information on inter-patch dispersal for 1984–1987, and during this period we recaptured 941 individuals in total giving an emigration rate of 2.5%. We did not observe short inter-patch dispersal distances (Fig. C1b) because of the spatial configuration of patches in the network; short distances are lacking (Fig. 1) and the average distance from any patch to its three closest patches is 1680 m. However, using the negative exponential function (see Eq. (3) and Fig. C1a) is still reasonable as short dispersal should be more likely than long dispersal, if there are patches available at short distance.
130
V. Johansson et al. / Biological Conservation 206 (2017) 120–131
We also tested the sensitivity to this value of α by fitting the model and running the simulations with α = 0.0016 and α = 0.0004, i.e. roughly half and double of the observed mean dispersal distance (Fig. C1). This changed the colonization parameter y from 52.6 (38.4–70.5) to 24.8 (17.5–34.7) and 92.0 (68.2–120.10), respectively, as there is a correlation between α and y. The extinction risk, mean population size and mean number of occupied patches changed somewhat with different values of α (Table C1). In general, smaller α (i.e. larger dispersal distances) resulted in a larger mean number of occupied patches and a larger mean posterior population sizes (but the 95% credible intervals for the population size were very similar), and also somewhat larger extinction risks, while larger α (i.e. shorter dispersal distances) showed the opposite pattern. However, the difference were rather small, especially for the estimated extinction risks, and the relative ranking among the four scenarios were very robust to different α. Therefore, we would make the same conclusions and recommendations regardless of what α we use.Fig. C1a) Weight of the contribution of dispersal source patch j to the connectivity of focal patch i, given three different values of α, as a function of distance (dij) between i and j (see Eq. (3)). The estimated value of α, based on observed mean dispersal distances, was 0.00079, and the other two represent roughly half (α = 0.0016) and double (α = 0.0004) of this distance. In b) shows the distribution of the 24 observed inter-patch dispersal distances. Table C1 The extinction risk (ER), the mean number of occupied patches (OP), and the mean total population size (PS) in 2025 given three different values of the spatial scaling parameter α (Fig. B1) and five management scenarios (described in Table 1). For OP and PS the 95% credible intervals are presented with parentheses. Scenario
Half mean distance α = 0.0016
Estimated value α = 0.00079
Double mean distance α = 0.0004
ER
OP
PS
ER
OP
PS
ER
OP
PS
S1
16%
S3
11%
S4
7%
S5
6%
4.0 (0–8) 5.7 (0–9) 5.7 (0–9) 9.4 (0–14) 16.6 (0−22)
83 (0–289) 151 (0–425) 149 (0–426) 233 (0–550) 300 (0–647)
20%
11%
62 (0–236) 111 (0–369) 111 (0–376) 187 (0–520) 242 (0–609)
17%
S2
3.1 (0–8) 4.3 (0–8) 4.3 (0–8) 7.3 (0–12) 13.3 (0−20)
4.0 (0−10) 6.4 (0−12) 6.32 (0−11) 10.8 (0–17) 18.2 (0–25)
85 (0–289) 166 (0–481) 165 (0–461) 257 (0–607) 327 (0–681)
11% 11% 8% 7%
References Akçakaya, H.R., 2000. Population viability analyses with demographically and spatially structured models. Ecol. Bull. 48, 23–38. Arnason, A.N., Schwarz, C.J., 1999. Using POPAN-5 to analyse banding data. Bird Study Suppl. S157–S168. Baguette, M., 2004. The classical metapopulation theory and the real, natural world: a critical appraisal. Basic Appl. Ecol. 5, 213–224. Baguette, M., Clobert, J., Schtickzelle, N., 2011. Metapopulation dynamics of the bog fritillary butterfly: experimental changes in habitat quality induced negative density-dependent dispersal. Ecography 34:170–176. http://dx.doi.org/10.1111/j.1600-0587. 2010.06212.x. Balmer, O., Erhardt, A., 2000. Consequences of succession on extensively grazed grasslands for central european butterfly communities: rethinking conservation practices. Conserv. Biol. 14:746–757. http://dx.doi.org/10.1046/j.1523-1739.2000.98612.x. Bascompte, J., Sole, R.V., 1996. Habitat fragmentation and extinction thresholds in spatially explicit models. J. Anim. Ecol. 65:465–473. http://dx.doi.org/10.2307/5781. Batáry, P., Báldi, A., Kleijn, D., Tscharntke, T., 2011. Landscape-moderated biodiversity effects of agri-environmental management: a meta-analysis. Proc. Biol. Sci. 278: 1894–1902. http://dx.doi.org/10.1098/rspb.2010.1923. Beissinger, S.R., McCullough, D.R., 2002. Population Viability Analysis. University of Chicago Press, Chicago, USA. Beissinger, S.R., Westphal, M.I., 1998. On the use of demographic models of population viability in endangered species management. J. Wildl. Manag. 62:821–841. http://dx. doi.org/10.2307/3802534. Brown, J.H., Kodric-Brown, A., 1977. Turnover rates in insular biogeography: effect of immigration on extinction. Ecology 58:445–449. http://dx.doi.org/10.2307/1935620. Bubová, T., Vrabec, V., Kulma, M., Nowicki, P., 2015. Land management impacts on European butterflies of conservation concern: a review. J. Insect Conserv. 19:805–821. http://dx.doi.org/10.1007/s10841-015-9819-9. Cousins, S.A.O., Auffret, A.G., Lindgren, J., Tränk, L., 2015. Regional-scale land-cover change during the 20th century and its consequences for biodiversity. Ambio 44:17–27. http://dx.doi.org/10.1007/s13280-014-0585-9. Craig, C.C., 1953. On the utilization of marked specimens in estimating populations of flying insects. Biometrika 40:170–176. http://dx.doi.org/10.2307/2333106. Dennis, R.L.H., Shreeve, T.G., Dyck, H.V., 2006. Habitats and resources: the need for a resource-based definition to conserve butterflies. Biodivers. Conserv. 15:1943–1966. http://dx.doi.org/10.1007/s10531-005-4314-3. Dover, J.W., Spencer, S., Collins, S., Hadjigeorgiou, I., Rescia, A., 2010. Grassland butterflies and low intensity farming in Europe. J. Insect Conserv. 15:129–137. http://dx.doi.org/ 10.1007/s10841-010-9332-0. Duelli, P., Obrist, M.K., 2003. Regional biodiversity in an agricultural landscape: the contribution of seminatural habitat islands. Basic Appl. Ecol. 4:129–138. http://dx.doi.org/ 10.1078/1439-1791-00140.
12% 12% 9% 7%
Fieberg, J., Ellner, S.P., 2000. When is it meaningful to estimate an extinction probability? Ecology 81:2040–2047. http://dx.doi.org/10.1890/0012-9658(2000)081[2040: WIIMTE]2.0.CO;2. Van Noordwijk, C.G.E., Flierman, D.E., Remke, E., WallisDeVries, M.F., Berg, M.P., 2012. Impact of grazing management on hibernating caterpillars of the butterfly Melitaea cinxia in calcareous grasslands. J. Insect Conserv. 16:909–920. http://dx.doi.org/10. 1007/s10841-012-9478-z. Franzén, M., Imby, L., 2008. Åtgärdsprogram för mnemosynefjäril 2008–2012 (No. 5829). Naturvårdsverket, Stockholm. Franzén, M., Nilsson, S.G., 2010. Both population size and patch quality affect local extinctions and colonizations. Proc. R. Soc. Lond. B Biol. Sci. 277:79–85. http://dx.doi.org/10. 1098/rspb.2009.1584. Fred, M.S., O'Hara, R.B., Brommer, J.E., 2006. Consequences of the spatial configuration of resources for the distribution and dynamics of the endangered Parnassius apollo butterfly. Biol. Conserv. 130:183–192. http://dx.doi.org/10.1016/j.biocon.2005.12.012. Gärdenfors, U., 2015. The 2015 Red List of Swedish Species. ArtDatabanken, Uppsala, Sweden. Hanski, I., 1994. A practical model of metapopulation dynamics. J. Anim. Ecol. 63, 151–162. Hanski, I., 1999. Metapopulation Ecology. Oxford University Press, Oxford. Hanski, I., Ovaskainen, O., 2002. Extinction debt at extinction threshold. Conserv. Biol. 16: 666–673. http://dx.doi.org/10.1046/j.1523-1739.2002.00342.x. Hanski, I., Moilanen, A., Gyllenberg, M., 1996. Minimum viable metapopulation size. Am. Nat. 147, 527–541. Harrison, S., Murphy, D.D., Ehrlich, P.R., 1988. Distribution of the Bay Checkerspot butterfly, Euphydryas editha bayensis: evidence for a metapopulation model. Am. Nat. 132, 360–382. Heard, G.W., McCarthy, M.A., Scroggie, M.P., Baumgartner, J.B., Parris, K.M., 2013. A Bayesian model of metapopulation viability, with application to an endangered amphibian. Divers. Distrib. 19:555–566. http://dx.doi.org/10.1111/ddi.12052. Huxel, G., Hastings, A., 1999. Habitat loss, fragmentation, and restoration. Restor. Ecol. 7, 309–315. Johansson, V., Ranius, T., Snäll, T., 2012. Epiphyte metapopulation dynamics are explained by species traits, connectivity, and patch dynamics. Ecology 93:235–241. http://dx. doi.org/10.1890/11-0760.1. Johansson, V., Felton, A., Ranius, T., 2016. Long-term landscape scale effects of bioenergy extraction on dead wood-dependent species. For. Ecol. Manag. 371:103–113. http:// dx.doi.org/10.1016/j.foreco.2015.10.046. Krauss, J., Bommarco, R., Guardiola, M., Heikkinen, R., Helm, A., Kuussaari, M., Lindborg, R., Öckinger, E., Pärtel, M., Pino, J., Pöyry, J., Raatikainen, K., Sang, A., Stefanescu, C., Teder, T., Zobel, M., Steffan-Dewenter, I., 2010. Habitat fragmentation causes immediate and time-delayed biodiversity loss at different trophic levels. Ecol. Lett. 13:597–605. http://dx.doi.org/10.1111/j.1461-0248.2010.01457.x. Kruess, A., Tscharntke, T., 2002. Grazing intensity and the diversity of grasshoppers, butterflies, and trap-nesting bees and wasps. Conserv. Biol. 16:1570–1580. http://dx.doi. org/10.1046/j.1523-1739.2002.01334.x.
V. Johansson et al. / Biological Conservation 206 (2017) 120–131 Kuussaari, M., Bommarco, R., Heikkinen, R., Helm, A., Krauss, J., Lindborg, R., Öckinger, E., Pärtel, M., Pino, J., Roda, F., Stefanescu, C., Teder, T., Zobel, M., Steffan-Dewenter, I., 2009. Extinction debt: a challenge for biodiversity conservation. Trends Ecol. Evol. 24, 564–571. Kuussaari, M., Heikkinen, R.K., Heliölä, J., Luoto, M., Mayer, M., Rytteri, S., von Bagh, P., 2015. Successful translocation of the threatened clouded Apollo butterfly (Parnassius mnemosyne) and metapopulation establishment in southern Finland. Biol. Conserv. 190:51–59. http://dx.doi.org/10.1016/j.biocon.2015.05.011. Kuussaari, M., Rytteri, S., Heikkinen, R.K., Heliölä, J., von Bagh, P., 2016. Weather explains high annual variation in butterfly dispersal. Proc. R. Soc. B 283:20160413. http://dx. doi.org/10.1098/rspb.2016.0413. Luoto, M., Pykälä, J., Kuussaari, M., 2003. Decline of landscape-scale habitat and species diversity after the end of cattle grazing. J. Nat. Conserv. 11:171–178. http://dx.doi.org/ 10.1078/1617-1381-00052. Maes, D., Van Dyck, H., 2001. Butterfly diversity loss in Flanders (north Belgium): Europe's worst case scenario? Biol. Conserv. 99:263–276. http://dx.doi.org/10.1016/S00063207(00)00182-8. McGowan, C.P., Runge, M.C., Larson, M.A., 2011. Incorporating parametric uncertainty into population viability analysis models. Biol. Conserv. 144:1400–1408. http://dx.doi.org/ 10.1016/j.biocon.2011.01.005. McLaughlin, A., Mineau, P., 1995. The impact of agricultural practices on biodiversity. Agric. Ecosyst. Environ. 55:201–212. http://dx.doi.org/10.1016/0167-8809(95)00609-V. McLaughlin, J.F., Hellmann, J.J., Boggs, C.L., Ehrlich, P.R., 2002. Climate change hastens population extinctions. Proc. Natl. Acad. Sci. 99:6070–6074. http://dx.doi.org/10.1073/ pnas.052131199. Nilsson, S.G., Franzén, M., Jönsson, E., 2008. Long-term land-use changes and extinction of specialised butterflies. Insect Conserv. Divers. 1:197–207. http://dx.doi.org/10.1111/j. 1752-4598.2008.00027.x. Nilsson, S.G., Franzén, M., Pettersson, L., 2013. Land-use changes, farm management and the decline of butterflies associated with semi-natural grasslands in southern Sweden. Nat. Conserv. 6:31–48. http://dx.doi.org/10.3897/natureconservation.6.5205. O'Hara, R., Arjas, E., Toivonen, H., Hanski, I., 2002. Bayesian analysis of metapopulation data. Ecology 83, 2408–2415. Öckinger, E., Smith, H.G., 2007. Semi-natural grasslands as population sources for pollinating insects in agricultural landscapes. J. Appl. Ecol. 44:50–59. http://dx.doi.org/10. 1111/j.1365-2664.2006.01250.x. Öckinger, E., Eriksson, A.K., Smith, H.G., 2006a. Effects of grassland abandonment, restoration and management on butterflies and vascular plants. Biol. Conserv. 133:291–300. http://dx.doi.org/10.1016/j.biocon.2006.06.009. Öckinger, E., Hammarstedt, O., Nilsson, S.G., Smith, H.G., 2006b. The relationship between local extinctions of grassland butterflies and increased soil nitrogen levels. Biol. Conserv. 128:564–573. http://dx.doi.org/10.1016/j.biocon.2005.10.024. Pellet, J., Fleishman, E., Dobkin, D.S., Gander, A., Murphy, D.D., 2007. An empirical evaluation of the area and isolation paradigm of metapopulation dynamics. Biol. Conserv. 136:483–495. http://dx.doi.org/10.1016/j.biocon.2006.12.020. Royle, J.A., Kéry, M., 2007. A Bayesian state-space formulation of dynamic occupancy models. Ecology 88, 1813–1823. Schtickzelle, N., Baguette, M., 2004. Metapopulation viability analysis of the bog fritillary butterfly using RAMAS/GIS. Oikos 104:277–290. http://dx.doi.org/10.1111/j.00301299.2004.12825.x. Schtickzelle, N., Le Boulengé, E., Baguette, M., 2002. Metapopulation dynamics of the bog fritillary butterfly: demographic processes in a patchy population. Oikos 97, 349–360.
131
Schtickzelle, N., Baguette, M., Boulengé, É.L., 2003. Modelling insect demography from capture–recapture data: comparison between the constrained linear models and the Jolly–Seber analytical method. Can. Entomol. 135, 313–323. Doi: 10.4039/n02-045 Schultz, C.B., Crone, E.E., 2005. Patch size and connectivity thresholds for butterfly habitat restoration. Conserv. Biol. 19:887–896. http://dx.doi.org/10.1111/j.1523-1739.2005. 00462.x. Seber, G.A.F., 1982. The Estimation of Animal Abundance and Related Parameters. MacMillan Press, New York, USA. Sjögren-Gulve, P., Ray, C., 1996. Using logistic regression to model metapopulation dynamics: large-scale forestry extirpates the Pool frog. Metapopulations and Wildlife Conservation. Island Press, pp. 111–137. Sutherland, C.S., Elston, D.A., Lambin, X., 2014. A demographic, spatially explicit patch occupancy model of metapopulation dynamics and persistence. Ecology 95:3149–3160. http://dx.doi.org/10.1890/14-0384.1. Thomas, J.A., Bourn, N.A., Clarke, R.T., Stewart, K.E., Simcox, D.J., Pearman, G.S., Curtis, R., Goodger, B., 2001. The quality and isolation of habitat patches both determine where butterflies persist in fragmented landscapes. Proc. R. Soc. B Biol. Sci. 268: 1791–1796. http://dx.doi.org/10.1098/rspb.2001.1693. Thomas, C.D., Wilson, R.J., Lewis, O.T., 2002. Short-term studies underestimate 30-generation changes in a butterfly metapopulation. Proc. R. Soc. B Biol. Sci. 269:563–569. http://dx.doi.org/10.1098/rspb.2001.1939. Turlure, C., Choutt, J., Dyck, H.V., Baguette, M., Schtickzelle, N., 2010. Functional habitat area as a reliable proxy for population size: case study using two butterfly species of conservation concern. J. Insect Conserv. 14:379–388. http://dx.doi.org/10.1007/ s10841-010-9269-3. Väisänen, R., Somerma, P., 1985. Status of Parnassius mnemosyne (Lepidoptera, Papilionidae) in Finland. Not. Entomol. 65, 108–118. Välimäki, P., Itämies, J., 2003. Migration of the clouded Apollo butterfly Parnassius mnemosyne in a network of suitable habitats: effects of patch characteristics. Ecography 26, 679–691. van Swaay, C.A.M., Warren, M.S., 1999. Red Data Book of European Butterflies (Rhopalocera), Nature and Environment. Council of Europe Publishing, Strasbourg. Wahlberg, N., Klemetti, T., Hanski, I., 2002. Dynamic populations in a dynamic landscape: the metapopulation structure of the marsh fritillary butterfly. Ecography 25, 224–232. WallisDeVries, M.F., Poschlod, P., Willems, J.H., 2002. Challenges for the conservation of calcareous grasslands in northwestern Europe: integrating the requirements of flora and fauna. Biol. Conserv. 104:265–273. http://dx.doi.org/10.1016/S00063207(01)00191-4. Warren, M.S., Hill, J.K., Thomas, J.A., Asher, J., Fox, R., Huntley, B., Roy, D.B., Telfer, M.G., Jeffcoate, S., Harding, P., Jeffcoate, G., Willis, S.G., Greatorex-Davies, J.N., Moss, D., Thomas, C.D., 2001. Rapid responses of British butterflies to opposing forces of climate and habitat change. Nature 414:65–69. http://dx.doi.org/10.1038/35102054. Weiss, S.B., 1999. Cars, cows, and Checkerspot butterflies: nitrogen deposition and management of nutrient-poor grasslands for a threatened species. Conserv. Biol. 13: 1476–1486. http://dx.doi.org/10.1046/j.1523-1739.1999.98468.x. White, G.C., Burnham, K.P., 1999. Program MARK: Survival estimation from populations of marked animals. Bird Study Suppl. 46, 120–138.
