Oecologia

, Volume 175, Issue 3, pp 861–873 | Cite as

Climate conditions and resource availability drive return elevational migrations in a single-brooded insect

Population ecology - Original research

Abstract

Seasonal elevational migrations have important implications for life-history evolution and ecological responses to environmental change. However, for most species, particularly invertebrates, evidence is still scarce for the existence of such migrations, as well as for the potential causes. We tested the extent to which seasonal abundance patterns in central Spain for overwintering (breeding) and summer (non-breeding) individuals of the butterfly Gonepteryx rhamni were consistent with three hypotheses explaining elevational migration: resource limitation (host plant and flower availability), physiological constraints of weather (maximum temperatures) and habitat limitation (forest cover for overwintering). For overwintering adults, abundance was positively associated with host plant density during two intensive survey seasons (2007–2008), and the elevational distribution was relatively stable over a 7-year period (2006–2012). The elevational distribution of summer adults was highly variable, apparently related both to temperature and habitat type. Sites occupied by adults in the summer were on average 3 °C cooler than their breeding sites, and abundance showed negative associations with summer temperature, and positive associations with forest cover and host plant density in 2007 and 2008. The results suggest that the extent of uphill migration in summer could be driven by different factors, depending on the year, and are mostly consistent with the physiological constraint and habitat limitation hypotheses. In contrast, the patterns for overwintering adults suggest that downhill migration can be explained by resource availability. Climate change could generate bottlenecks in the populations of elevational migrant species by constraining the area of specific seasonal habitat networks or by reducing the proximity of environments used at different times of year.

Keywords

Climate change Elevational distribution Gonepteryx rhamni Lepidoptera Seasonal movements 

Introduction

Animal migration, i.e. the movement of animals from one area to another at different times of the year, involves costs and benefits to the migrating animal. Potential costs include energetic expenses, increased predation risk and reproductive costs, whereas potential benefits include increased survival and reproduction due to exploitation of new resources and avoidance of adverse environmental conditions (Rankin and Burchsted 1992). Migration is expected to evolve only under circumstances in which remaining in the natal habitat would be detrimental relative to migrating to a different site (Southwood 1977; Rankin and Burchsted 1992).

Migration has been classified in different ways depending on the organism and on spatial or temporal attributes (Dingle and Drake 2007). Typical migrations involve seasonal displacements of more than hundreds of kilometers between high-latitude summer habitats and low-latitude spring habitats (e.g. Chapman et al. 2012). Depending on the duration of the life cycle, organisms may breed multiple times within a season, producing several generations as populations migrate polewards during spring and summer and equatorwards during autumn or, alternatively, the same individuals may perform the complete round trip (Ramenofsky and Wingfield 2007). However, organisms also show seasonal movements over elevational gradients involving shorter distances, referred to as elevational migrations. Elevational migrations have been reported from nearly all continents and from several vertebrate and insect taxa, but most evidence is based on birds (McGuire and Boyle 2013). Nevertheless, potential hypotheses explaining bird elevational migration are applicable to other taxa since they involve resource availability (e.g. Boyle 2010), physiological constraints of weather (e.g. Boyle et al. 2010), predation risk (e.g. Boyle 2008), habitat limitation and competition for mates (McGuire and Boyle 2013).

Elevational migration has been documented through different methods with different discriminatory power (McGuire and Boyle 2013). Ideally, telemetry of animals will provide spatially and temporally detailed movement information at an individual level (e.g. Norbu et al. 2013). This approach is feasible for some vertebrates, but in the case of many insects, the combination of small size, short life cycle, large population and relatively high flying speed make it difficult to follow individuals (Osborne et al. 2002; but see Urquhart and Urquhart 1978; Wikelski et al. 2006). Alternatively, elevational migration can be documented through abundance or distribution surveys at multiple elevations and time-periods (McGuire and Boyle 2013). Once such data are available, correlative distribution models can be applied to assess the importance of different variables in determining elevational shifts between time-periods. This approach has been successfully applied to several species and has provided important clues relating to the drivers of elevational migrations (e.g. Brambilla et al. 2012; Marini et al. 2013).

The importance of seasonal migration in the life-cycle of some members of order Lepidoptera is well established (e.g. Williams 1930). The monarch butterfly Danaus plexippus exemplifies the typical insect engaged in mass latitudinal migrations over long distances (e.g. Urquhart and Urquhart 1978), and latitudinal migrations of other Lepidoptera species at temperate latitudes have been well documented (e.g. Mikkola 2003; Stefanescu et al. 2007, 2013; Brattström et al. 2010; Chapman et al. 2012). Elevational migrations have been reported for several Lepidoptera species (e.g. Larsen 1976, 1982; Shapiro 1973, 1974a, b, 1975, 1980; Stefanescu 2001), but in most cases, no explicit assessment of potential hypotheses explaining the phenomenon has been made. One exception is a study on the butterfly Vanessa atalanta, where the results support the resource availability hypothesis (Stefanescu 2001).

In the study reported here, we tested the resource availability, physiological constraints of weather and habitat limitation hypotheses as potential explanations for uphill and downhill movements of the brimstone butterfly Gonepteryx rhamni (L.) in a mountain area in Spain. This species is particularly appropriate to studies on elevational migrations because it has a relatively high mobility (Gutiérrez and Thomas 2000), and regional movements between hibernating and breeding areas have been hypothesised to occur along elevational gradients (Larsen 1976, 1982; García-Barros et al. 2013). Gonepteryx rhamni is a single-brooded species which develops from egg to adult in the spring and has a non-reproductive period in summer until hibernation, following which the mating season occurs the following spring (Wiklund et al. 1996). In our study, we first determined the extent of the migration by examining abundance patterns over elevation for overwintering (breeding) and summer (non-breeding) individuals. We then modelled separately the abundance of overwintering and summer individuals based on environmental resources and conditions to assess the relevance of the three proposed hypotheses. We expected stronger support for the resource availability hypothesis (based on host plant abundance) for overwintering individuals because of the requirement for breeding sites. In contrast, the distribution of summer (non-breeding) individuals could have been driven by physiological constraints (temperature), habitat limitation (e.g. overwintering sites) and/or resource availability (e.g. flower abundance).

Materials and methods

Study system

Gonepteryx rhamni (L.) is a widespread species in Europe whose larvae feed on shrubs from the family Rhamnaceae. It has one adult generation per year (emerging in June–August) and hibernates as an adult (resuming activity in late winter) (Tolman and Lewington 1997). Gonepteryx rhamni is a relatively common species throughout Spain, but its populations are more frequent in mountainous areas in the southern half of the country (García-Barros et al. 2004). There are no detailed records of overwintering habitats for G. rhamni in the study area, but wooded areas have been suggested elsewhere in Europe (Pollard and Hall 1980).

The Sierra de Guadarrama (central Spain) is an approximately 100 × 30-km mountain range located at 40°45′N 4°00′W. The mountain range is bordered by plains at elevations of approximately 700 m (to the north) and approximately 500 m (to the south) and reaches a maximum elevation of 2,428 m [Electronic Supplementary Material (ESM) Fig. S1]. The main regional host plants reported for G. rhamni are Frangula alnus Mill. and Rhamnus cathartica L. (based on oviposition and larval records; Gutiérrez and Wilson, unpublished data), although at least two other species from the family Rhamnaceae occur in the study area (Rhamnus lycioides L. and R. alaternus L.).

Elevational abundance of G. rhamni

To study the elevational patterns of Gonepteryx rhamni abundance, we counted butterflies (including G. rhamni, if present) in 2007 and 2008 on standardised transects (length 500 m, width 5 m) at 40 sites ranging from 558 to 2,251 m a.s.l. (ESM Fig. S1). A subsample of 24 sites were also visited following identical methodology in 2006 and 2009–2012 to examine the temporal variability of elevational patterns. The line transects were walked during suitable conditions for butterfly activity (sunshine, no more than a light wind, between 10:00 and 17:00 hours European Summer Time; Pollard and Yates 1993), every 2 weeks from April to October in 2006, and from March to October since 2007. Due to snow cover and unsuitable weather, the transect walks started later than March or April in some years at sites above 1,700–1,800 m a.s.l.

During counts, we explicitly distinguished overwintering G. rhamni adults from summer adults based on wing condition. In the Sierra de Guadarrama, overwintering adults fly from early March (with some occasional records of flying in February in warm winters) to June, whereas summer adults fly from June to October. Because overwintering adults show signs of increasing wear over the season, they are easily distinguishable from recently emerged summer adults (in excellent condition) over the potential overlap in the flight period of both generations in June. Sexes were recorded separately in the field but were pooled together for analyses because of no obvious sex-related patterns and a much smaller sample size for females. For each site and year, overwintering and summer adult abundances were calculated as the sum of all counts of overwintering and summer individuals over the season. During transect walking and additional visits to the sites, information on reproductive (oviposition) and feeding (nectaring) behaviour was also recorded.

Environmental variables

Universal transverse mercator (UTM) coordinates were recorded approximately every 100 m along transects using a handheld Garmin GPS unit and were used to plot transects in a geographic information system (ArcGIS) (ESRI 2001). The average elevation of 100-m cells intercepted by transects was determined using a digital elevation model (Farr et al. 2007).

To examine potential determinants of the elevational distribution of overwintering and summer adults, we collected environmental variables from the field (spring and summer temperatures, host plant density, summer flower abundance) and from digital layers (forest cover). The biological significance of each variable is detailed in Table 1.
Table 1

List of environmental variables included in the present study, classified by their biological significance

Environmental variable (units)

Code

Mean (min–max)

Conditions: adult thermoregulation and larval development

  Spring mean maximum temperature (°C)

    2007

Sprtmax

13.26 (5.67–20.54)

    2008

13.24 (5.49–21.44)

  Summer mean maximum temperature (°C)

    2007

Sumtmax

24.52 (16.74–34.45)

    2008

24.91 (17.54–35.90)

Resources: larval host plants and adult nectar sources

  Host plant density (number of plants 0.25 ha−1)

Hostpl

2.43 (0–29)

  Summer flower abundance (percent cover)

    2007

Flowab

3.08 (0–11.25)

    2008

1.64 (0–5.50)

Conditions and resources: conditions and sites for adult overwintering

  Forest cover (proportion cover)

Forest

0.63 (0–1)

Spring and summer mean maximum temperatures, and summer flower abundance were recorded separately in 2007 and 2008. Host plant density was square-root transformed for analyses to avoid influential effects of individual sites (Zuur et al. 2007)

For the period 2006–2012, hourly air temperature was recorded using HOBO H8 Pro HR/Temp and HOBO U23 HR/Temp loggers (both Onset Computer Corp, Cap Code, MA) in semi-shaded conditions at each of the 40 sampling sites (one logger per site). Twenty data loggers were placed in position (20 of the 40 sampling sites) and started recording data in spring 2004; another 20 started recording data in spring 2006. Mechanical failure or damage to some data loggers due to snow, animals or human interference generated gaps of variable duration in the data set. Therefore, daily average and maximum/minimum temperature time-series had to be interpolated for some loggers to subsequently estimate spring and summer temperatures (ESM Appendix S1).

Interpolated daily maximum temperature data were used to calculate monthly mean maximum temperatures and, subsequently, to calculate spring (March–May) and summer (June–August) mean maximum temperatures. Seasonal temperatures were based on maximum rather than mean daily temperatures because the former more likely represent the conditions experienced by a daylight flying insect such as G. rhamni (e.g. Wiklund et al. 1996). Spring and summer mean maximum temperatures were highly correlated with spring and summer mean temperatures, respectively, in 2007 and 2008 (range for all four Spearman’s rank correlation coefficients rs 0.92–0.97).

The abundance of host plants was estimated at each of the 40 transect sites in 2006, with some additional records in 2008. The route of the 500-m transect was followed in August–September 2006, and the number of plants of Frangula alnus, Rhamnus cathartica and R. lycioides (R. alaternus was absent from sampling sites) that occurred in the 5-m wide butterfly transect was recorded to obtain a density of each species per 0.25 ha (500 × 5 m). If none of the plant species were present in the 5-m-wide transect, then the transect was walked again at an increasing width (10, 20 m, and up to a maximum of 50 m; i.e. 25 m on either side of the recorder), in which case, host plant density per 0.25 ha was estimated based on the increased transect width. Host plant species were considered present at a site if they were found in transects of ≤50 m wide (Merrill et al. 2008). To test for temporal variability in host plant density, we counted the number of plants that occurred in the 5-m wide transect again in 2009.

Summer flower abundance was estimated during the summer adult flight peak in 2007 and 2008 by making twenty 0.25-m2 quadrats (50 × 50 cm) at 25-m intervals along each transect. We set quadrats in late July 2007 and late July–early August in 2008, and flower abundance was estimated by counting the number of 2.5 × 2.5-cm sub-quadrats (100 per quadrat) containing a more than 4-cm2 surface area of flowers. Data from elsewhere in Spain show that G. rhamni is not a specialist flower visitor (Stefanescu and Traveset 2009), and therefore we considered that all flowering species could be potential nectar sources.

Forest cover was estimated from regional land-cover maps obtained in vector format at the 1:50,000 scale (Ministerio de Medio Ambiente, 2000, 2002a, b, 2003). These maps showed good agreement for all transects with our own field observations of general vegetation type (Gutiérrez Illán et al. 2010). Vector data from the land-cover maps (minimum cartographic unit 2.25 ha) were used to determine the proportional contribution of total forest cover to each 100-m grid cell. Forest cover for each site was estimated as the mean for 100-m grid cells intercepted by each transect.

Elevational abundance models

To analyse G. rhamni abundance, we applied generalised linear models (GLMs) with a quasi-likelihood estimation of regression coefficients using a log-link and setting the variance equal to the mean (quasi-Poisson regression; McCullagh and Nelder 1989). The information-theoretic approach (Burnham and Anderson 2002) was used to model G. rhamni abundance. First, elevational trends in abundance for both overwintering and summer individuals in 2007 and 2008 were analysed; second, more complex models including environmental variables were tested to explain the observed trends in abundance over the elevational gradient. For analysing elevational trends in abundance, we fitted linear and quadratic models only including elevation. The more complex models for potential explanatory factors included three candidate variables (spring mean maximum temperature, host plant density, forest cover) for overwintering individuals and four variables (summer mean maximum temperature, host plant density, flower abundance, forest cover) for summer individuals (Table 1). Pair-wise correlations between the independent variables had absolute values of <0.7 (commonly applied threshold when testing for collinearity; Dormann et al. 2013). Linear and quadratic terms for the environmental condition variables and only linear terms for the strictly resource variables were included (Table 1). Confidence sets were based on the Akaike Information Criterion for small sample size (QAICc; Burnham and Anderson 2002; ESM Appendix S2).

Following model selection, model-averaging was used to obtain model coefficients based on the confidence sets (Burnham and Anderson 2002). This approach incorporates model selection uncertainty while weighting the influence of each model by the strength of its supporting evidence. Model-averaged coefficients were calculated by weighting using Akaike weights (QAICcw) and then averaging coefficients over all models in the confidence sets. Averaging over all models means that in those cases in which a given variable was not included in a particular model, its coefficient value was set to zero. Relative variable importance (parameter lying in the range 0–1, which provides evidence about the relevance of each variable relative to the others) was calculated as the sum of Akaike weights across all models in the confidence set that contain that variable. Model selection and averaging were performed using ‘MuMIn’ package (Bartoń 2012; R Development Core Team 2012).

Spatial autocorrelation can influence the reliability of ecological analyses and potentially generates models containing a relatively larger number of predictors in information theoretic approaches (e.g. Diniz-Filho et al. 2008). To test for spatial autocorrelation, we generated all-directional correlograms (Legendre and Legendre 1998) for the abundance data of overwintering and summer adults in 2007 and 2008 by plotting values of Geary’s c coefficient (recommended for variables departing from normality) against Euclidean distances between sites. Geary’s c calculation and testing for significance were performed using 4999 Monte Carlo permutations in Excel add-in Rookcase (Sawada 1999). No correlogram was globally significant, suggesting that spatial autocorrelation in G. rhamni abundance data was negligible.

After identifying the model confidence sets for G. rhamni abundance, hierarchical partitioning (HP) was used to assess the independent and joint effects of each parameter in single models with all parameters (Chevan and Sutherland 1991; Mac Nally 1996). Poisson regression and log-likelihood as the goodness-of-fit measure were used for HP calculations, and statistical significance of the independent contributions was tested by a randomization routine (1,000 permutations) based on Z scores (Mac Nally 2002). HP was conducted using the ‘hier.part’ package (Mac Nally and Walsh 2004). One of the limitations of HP as currently implemented in the ‘hier.part package’ is that it depends on monotonic relationships between the response and predictor variables. However, all relationships of abundances of overwintering and summer adults against environmental variables were monotonic (see below) and hence this was not a major problem.

Temporal variability in elevational patterns

To examine variability in elevational abundance patterns over time, we used G. rhamni abundance data collected at 24 sites over a 7-year period (2006–2012). Mean elevation was calculated separately for overwintering and summer individuals each year by averaging the elevations of all sites where G. rhamni was present, weighted by abundance at each site. To test the potential dependency over time of elevational abundance patterns on climate conditions, mean elevations of overwintering and summer adults were compared with spring and summer mean maximum temperatures, respectively, using Spearman’s rank correlation coefficients (rs). To examine the importance of temperature, host plant density and forest cover on G. rhamni abundance in different years, we performed quasi-Poisson regressions based on the 24-site data set and using the same approach as for the 40-site data set.

To examine the extent to which G. rhamni adults maintain the temperatures experienced from spring to summer, we calculated weighted mean temperature separately for overwintering (March–May temperatures) and summer (June–August temperatures) individuals using the same approach as for weighted mean elevation. Finally, to determine the extent to which the breeding sites have greater temperatures in summer than those experienced by adult butterflies, we calculated mean temperatures at sites where larval host plants were present.

Results

A total of 212 overwintering and 116 summer G. rhamni individuals were counted in 2007, and 238 overwintering and 96 summer individuals were counted in 2008. The phenology of overwintering adults was similar in 2007 and 2008, whereas that of summer adults was delayed in 2008 relative to 2007 (Fig. 1). Eight females were recorded ovipositing on F. alnus and R. cathartica at transect sites; the earliest oviposition record was on 1 April 2011 and the latest on 12 June 2008. A total of 70 nectaring records were collected over 2007–2012 from 16 plant genera belonging to 13 different families, supporting the notion that G. rhamni adults are not specialist flower visitors (Stefanescu and Traveset 2009). Mean maximum and minimum temperatures peaked in July or August depending on site (ESM Fig. S2). The warmest mean maximum temperature was in July at the lowest site (approx. 38 °C) and in August at the highest site (approx. 20 °C).
Fig. 1

Phenology of overwintering (dashed lines) and summer (solid lines) Gonepteryx rhamni adults throughout the seasons 2007 (thick lines) and 2008 (thin lines). Phenology data are shown as the sum of all individuals counted at all transects during a given transect 2-week time-interval. Dates were calculated as the mean date for all transect counts in a given 2-week period

Elevational patterns

In 2007, overwintering adults were recorded at 23 sites (739–1,635 m a.s.l.) and summer adults at 25 sites (1,020–2,251 m a.s.l.); in 2008, overwintering adults were recorded at 23 sites (844–1,925 m a.s.l.) and summer adults at 21 sites (1,056–1,976 m a.s.l.). Maximum local abundances were 39 overwintering (at 1,270 m a.s.l.) and 22 summer (at 1,499 m a.s.l.) individuals in 2007, and 54 overwintering (at 960 m a.s.l.) and 14 summer (at 1,270 and 1,327 m a.s.l.) individuals in 2008 (Fig. 2).
Fig. 2

Abundance of G. rhamni and density of its host plants with elevation. a, b Overwintering (open circles, dashed line) and summer (filled circles, solid line) G. rhamni adults in 2007 (a) and 2008 (b). c Host plants (sum of plants of Frangula alnus, Rhamnus cathartica and R. lycioides). Lines are plotted based on equations in Table 2. Vertical dashed thin line maximum elevation at which host plants were recorded

There were quadratic relationships between abundance and elevation for overwintering and summer individuals in both study years (Table 2; Fig. 2). In all four cases, models including only the linear term for elevation had a QAICc difference of more than 6 from the quadratic (best) model (indicating that they were not well supported; Burnham and Anderson 2002; Richards 2008). Modelled maximum abundance for overwintering individuals peaked at 1,214 m a.s.l. in 2007 and at 1,152 m a.s.l. in 2008; for summer individuals, it peaked at 1,604 m a.s.l. in 2007 and 1,387 m a.s.l. in 2008. Abundances of overwintering and summer adults were not significantly correlated in 2007 (rs = 0.23, P = 0.158, N = 40), but they were correlated in 2008 (rs = 0.46, P = 0.003, N = 40). No significant correlation was found between summer adult abundance in 2007 and overwintering adult abundance in 2008 (the same generation at different times) (rs = 0.14, P = 0.377, N = 40).
Table 2

Generalised linear models (quasi-Poisson error and log-link) for the abundance of overwintering and summer Gonepteryx rhamni adults in 2007 and 2008 with elevation

Modela

Total no. of individuals

Elevation of site (km)

Elevation2

Intercept

Overwintering adults 2007

212

33.52 (8.89)

−13.81 (3.67)

−17.63 (5.32)

Summer adults 2007

116

13.54 (5.28)

−4.22 (1.73)

−9.14 (3.95)

Overwintering adults 2008

238

21.10 (9.57)

−9.16 (4.05)

−9.50 (5.54)

Summer adults 2008

96

26.35 (7.26)

−9.50 (2.62)

−16.52 (5.00)

Values given in parenthesis are the standard error (SE)

N = 40 sites in all cases

aIn the four cases, the quadratic models showed QAICc (Akaike Information Criterion for small sample size corrected for over-dispersed count-data) values which were lower by more than 6 units from those for the linear models (not shown)

A total of 16 transect sites included potential larval host plants in 2006, with the three host plants (F. alnus, R. cathartica and R. lycioides) present at nine, nine and one of the transect sites, respectively. The distribution of R. lycioides was restricted to the lowest site (558 m a.s.l.) of the study area (additional field searches showed that R. cathartica also occurs below 600 m a.s.l.), but all host plants were absent from the highest elevations (maximum elevation 1,504 m a.s.l. for F. alnus). Additional field searches at 90 sites included in a related study (Gutiérrez Illán et al. 2010) did not encounter any of the host plants at elevations of >1,504 m a.s.l.. Hence, there was approximately a. 750 m elevational gap (1,504–2,251 m a.s.l.) without host plants in the study area (Fig. 2). Mean elevation was 1,208 m a.s.l. for all sites containing host plants. There was a highly significant positive correlation between host plant density (based on 5-m-wide transect data) in 2006 and 2009 (rs = 0.96, P < 0.001, N = 40), suggesting that this variable was relatively constant over time.

Spring and summer mean maximum temperatures were highly negatively correlated with elevation in both study years. Summer flower abundance had no apparent elevational pattern in any year, and forest cover declined significantly with increased elevation (ESM Fig. S3).

Weighted mean elevation within each season (at 4-week intervals) was relatively stable for overwintering adults in 2007 and 2008 (Fig. 3; see ESM Fig. S4 for results for males and females separately). However, for summer adults, weighted mean elevation increased over the season in 2007, and increased up to August thereafter decreasing in September 2008.
Fig. 3

Weighted mean elevation of overwintering (dashed lines) and summer (solid lines) G. rhamni adults in the 2007 season (thick lines) and 2008 season (thin lines). Horizontal thin lines mean (solid ) and maximum (dashed) elevation (km) at which host plants were recorded. Numbers next to symbols are sample sizes

Abundance models

To assess the abundance of overwintering adults, we assigned model confidence sets consisting of two models each in 2007 (Table 3) and 2008 (Table 4). The final averaged models included positive relationships with host plant density and forest cover (in 2007), and with host plant density and spring mean maximum temperature (in 2008). Host plant density had a relative variable importance of 1 in both years, indicating that it was the most important variable explaining overwintering adult abundance. Spring mean maximum temperature was also well supported in 2008 (variable importance ≥0.9).
Table 3

Confidence set generalised linear models (quasi-Poisson error and log-link) for the abundance of overwintering and summer G. rhamni adults in 2007

Overwintering adults 2007

Rank

Hostpl

Forest

Intercept

K

QAICc

ΔQAICc

QAICcw

1

+

+

+

4

56.84

0.00

0.71

2

+

 

+

3

58.63

1.78

0.29

Imp

1

0.71

     

Coef

0.49

0.88

0.27

    

SE

0.09

0.66

0.63

    

Dispersion parameter = 6.05

    

Summer adults 2007

Rank

Sumtmax

Hostpl

Forest

Intercept

K

QAICc

ΔQAICc

QAICcw

1

 

+

+

4

58.44

0.00

0.81

2

+

 

+

4

62.53

4.09

0.10

3

  

+

3

62.80

4.36

0.09

Imp

1

0.10

0.81

     

Coef

−0.18

0.02

1.24

4.35

    

SE

0.08

0.15

0.67

1.55

    

Dispersion parameter = 4.49

    

N = 40 sites in all cases

The table indicates the variables included in the model and the direction of their coefficients (±; codes in Table 1), the number of parameters (K, including one extra parameter for over-dispersion factor in QAICc), QAICc, ΔQAICc (the difference in QAICc between the current and best model) and QAICcw (Akaike weight). Relative importance (Imp), model-averaged coefficients (Coef) and unconditional standard errors (SE) for each variable are also shown. Dispersion parameter is for the full model

Table 4

Confidence set generalised linear models (quasi-Poisson error and log-link) for the abundance of overwintering and summer G. rhamni adults in 2008

Overwintering adults 2008

Rank

Sprtmax

Hostpl

Intercept

K

QAICc

ΔQAICc

QAICcw

1

+

+

4

59.72

0.00

0.94

2

 

+

+

3

65.09

5.37

0.06

Imp

0.94

1

     

Coef

0.15

0.62

−1.41

    

SE

0.06

0.08

1.12

    

Dispersion parameter = 5.61

    

Summer adults 2008

Rank

Sumtmax

Hostpl

Flowab

Forest

Intercept

K

QAICc

ΔQAICc

QAICcw

1

+

+

+

6

79.28

0.00

0.74

2

 

+

+

5

82.61

3.33

0.14

3

+

 

+

+

5

83.02

3.74

0.11

Imp

0.86

1

0.89

1

     

Coef

−0.09

0.46

−0.28

1.94

1.36

    

SE

0.05

0.10

0.14

0.65

1.40

    

Dispersion parameter = 2.19

     

N = 40 sites in all cases

For explanation of terms/variables, see footnote to Table 3

For summer adults, the confidence sets consisted of three models in both 2007 and 2008. In 2007, the final model included positive relationships with host plant density and forest cover and a negative relationship with summer mean maximum temperature (Table 3). In 2008, the final model included these three variables plus a negative relationship with flower abundance (Table 4). The most important variables explaining summer adult abundance were summer mean maximum temperature (variable importance 0.86–1) and forest cover (variable importance 0.81–1). Nevertheless, host plant density and flower abundance showed relatively high variable importance values in 2008 (>0.8).

In hierarchical partitioning, the independent effect of host plant density was the only statistically significant variable for overwintering adults in 2007 and 2008 (Fig. 4). The independent contributions were not significant for the two remaining variables in 2007 or for forest cover in 2008. For summer adults, summer mean maximum temperature made the only statistically significant independent contribution in 2007, but there were significant effects of host plant density and forest cover in 2008. The negative joint contributions of summer mean maximum temperature, forest cover and flower abundance for summer adults (Fig. 4b, d) indicate that the joint action of other variables suppresses or masks the independent contribution of those particular predictors (Chevan and Sutherland 1991; Mac Nally 1996).
Fig. 4

The independent (black bars) and joint contribution (white bars) (given as the percentage of the total variance explained by the model) of the environmental variables estimated from hierarchical partitioning for G. rhamni abundance of overwintering adults in 2007 (a), summer adults in 2007 (b), overwintering adults in 2008 (c) and summer adults in 2008 (d). Asterisks significant (P < 0.05) independent contributions from randomisation tests. Variable codes as in Table 1. N = 40 sites in all cases. Note the different y-axis scales

Temporal variability in elevational patterns

Weighted mean elevation was consistently higher and more variable for summer than for overwintering adults over the 7-year period (Fig. 5). Summer mean maximum temperatures were on average approximately 12 °C higher than spring mean maximum temperatures over the elevation gradient (Fig. 5). Spring and summer mean maximum temperatures tended to follow a similar pattern over the 7-year period, but the correlation was not significant (rs = 0.68, P = 0.094, N = 7). No correlation of overwintering or summer mean elevation of butterfly adults against the corresponding seasonal temperatures was significant (P > 0.5). However, summer mean elevation was significantly positively correlated with summer mean maximum temperature if 2007 (the coldest summer) was excluded from analysis (rs = 0.83, P = 0.042, N = 6), suggesting that G. rhamni adults generally occurred at higher elevations in warmer summers (Fig. 5).
Fig. 5

Relationship between G. rhamni mean elevation and mean temperature over the period 2006–2012 for overwintering (a) and summer (b) adults. Numbers next to symbols are years. Horizontal thin lines mean (solid) and maximum (dashed) elevation (km) at which host plants were recorded

Models based on data from the 24 sites revealed that host plant density was the most important variable explaining overwintering adult abundance in all years, but spring mean maximum temperature was also included in all models with a positive effect (6 years) or quadratic effect (1 year) (ESM Table S1). Summer mean maximum temperature was the most important variable associated with summer adult abundance, with negative (3 years) or quadratic (2 years) effects. Host plant density was positively associated with summer abundance in 4 years, three of which were relatively cold (ESM Table S1; Fig. 5).

Weighted mean temperatures experienced by overwintering adults in spring ranged from 15.4 to 18.9 °C, and by summer adults from 23.7 to 27.4 °C. Mean temperatures at host plant sites in the spring ranged from 15.4 to 18.4 °C, and in summer from 26.5 to 30.4 °C. The mean difference between weighted mean temperatures for butterflies and mean temperatures for host plants were +0.5 °C in spring and −3 °C in summer.

Correlations (rs) between overwintering and summer adult abundances within the same year ranged from −0.01 to 0.58 (N = 24 in all cases), with the only significant coefficient in 2008 (P = 0.003; consistent with the analysis with the larger sample size above). Correlations between summer adult abundance in 1 year and overwintering adult abundance in the immediately following year ranged from 0.04 to 0.48 (N = 24), and they were significant for summer adults 2008-overwintering adults 2009 (P = 0.026) and summer adults 2011-overwintering adults 2012 (P = 0.017), corresponding to years with relatively cold summers (Fig. 5).

Discussion

Our results show marked differences between the elevational abundance patterns for overwintering and summer G. rhamni adults. In both intensive study years (2007 and 2008), summer adults were on average at higher sites than overwintering adults, and this pattern was maintained over five additional years in which a smaller number of sites were sampled.

The differences in abundance patterns for overwintering and summer adults were consistent with seasonal elevational migration by G. rhamni. Our oviposition records, albeit relatively limited, were consistent with the univoltine life cycle reported for G. rhamni with spring breeding (García-Barros et al. 2013). As a result, the same individuals emerging in summer that migrate uphill must migrate back down to breed the next spring. The steadily higher weighted mean elevation of summer adults until late August suggested that uphill migration was a gradual phenomenon over summer (Fig. 3). The decreased weighted mean elevation of summer adults in September 2008 was consistent with downhill migration in autumn, but this was not supported by the 2007 data. Occasional observations of adult G. rhamni flying down in October and late winter suggest that downhill migration could occur in both periods, but this point requires further research.

One intriguing result (found also in the UK, Pollard and Yates 1993) is that the abundance of overwintering individuals based on all sites was larger than that recorded in the previous summer (Table 2; ESM Table S1). Such a situation is clearly ecologically impossible (assuming that there is no significant immigration from outside the study area at some point in the season) because the number of individuals must decrease during hibernation. We cannot explain this finding, but it could be related to differences in behaviour in spring and summer, leading to differences in detectability (Pollard and Yates 1993). Also, quantification of potential downhill movement in late summer could be difficult due to reduced summer activity.

Abundance models and hierarchical partitioning suggested that explanations for migration in one direction may not explain return movements in the opposite direction. We tested for the effects of resource availability, physiological constraints of weather and habitat limitation (McGuire and Boyle 2013) and found that these were differentially supported for uphill and downhill migrations.

Several hypotheses could explain uphill migration of G. rhamni. In 2007, the strongest effect explaining summer adult abundance was summer mean maximum temperature (Table 3; Fig. 4), with G. rhamni more abundant at cooler sites during the summer period. This finding is in line with the physiological constraints of the weather hypothesis, wherein climatic factors may pose direct challenges to survival (McGuire and Boyle 2013). Central Spain is characterised by a continental Mediterranean climate, with extremely hot temperatures in summer (exceeding 35 °C at the lower sites), but much cooler (approx. 20 °C) higher up in the mountains (ESM Figs. S2, S3). Extremely hot temperatures could affect the survival and flight willingness in G. rhamni, but demonstrating this would require experimentation (e.g. Pruess 1967). In an experiment with caged individuals, Wiklund et al. (1996) observed that Swedish G. rhamni showed higher flight willingness at 23–29 °C than at 14–20 °C, but these authors did not test temperatures above 29 °C. In our study, the main variables explaining summer adult abundance in 2008 were host plant density and forest cover. This year showed the smallest difference in elevation between overwintering and summer adults, suggesting reduced elevational migration. The effect of forest cover was consistent with the habitat limitation hypothesis, which in the case of G. rhamni could be associated with the availability of overwintering sites (Pollard and Hall 1980). Weighted mean elevations for summer adults were much more variable than those for overwintering adults (Fig. 5). This result, along with the different contributions of explanatory variables to abundance models based on 40- and 24-site data sets, suggests that uphill migrations could be driven by different factors depending on the particular year. Specifically, the physiological constraints of the weather hypothesis would be expected to be more important in hotter summers. This fact was supported by the positive trend between the weighted mean elevation for summer adults and the summer mean maximum temperature (Fig. 5) (but we do not have any plausible explanation for the year 2007 outlier) and the negative or quadratic effects (based on 24 sites) of the summer mean maximum temperature in the warmest years (ESM Table S1). Surprisingly, the resource availability hypothesis for adults was not supported at all because summer flower abundance was only included in some models, but with a negative effect (Table 3).

The hypotheses explaining uphill migration by G. rhamni strongly contrast with those supported for other butterfly species. For instance, uphill migration by V. atalanta has been suggested as a strategy to track larval resources through space and time (resource availability hypothesis) (Stefanescu 2001). Although based on less detailed information, the same hypothesis has also been invoked to explain uphill migration by other species (e.g. Shapiro 1974a, 1975, 1980).

Resource availability for early stages apparently drives return downhill migration (either in autumn or following hibernation) by G. rhamni individuals before spring breeding. Host plant density was the most important variable explaining overwintering adult abundance in 2007 and 2008 and in 2006–2012 based on the reduced 24-site data set (Tables 3 and 4; Fig. 4; ESM Table S1). In addition, the weighted mean elevation for overwintering adults was relatively constant over the 7-year study period and close to the mean elevation for host plants, regardless of climatic conditions (Fig. 5). Given the relative host plant specialism of G. rhamni, the search for larval host plants is probably one of the strongest evolutionary pressures favouring downhill migration in this species. Failure to do so will result in the highest fitness cost of all—no breeding.

One further hypothesis to explain elevational migration not considered in our study is predation risk, which states that migration has evolved in response to elevational differences in predation pressure (e.g. Boyle 2008). In the case of butterflies, this hypothesis has been discussed in terms of larval parasitism (e.g. Stefanescu et al. 2012). However, this is not applicable to G. rhamni because adults do not reproduce in summer. Evaluation of the predation risk hypothesis would require the difficult task of collecting information on adult predators.

Hilltopping (a mating strategy of some insect species in which males occupy prominent topographic features due to female scarcity) has been suggested as a component of uphill migration by some Hymenoptera (Hunt et al. 1999). However, G. rhamni is a spring-breeding species with patrolling behaviour, so hilltopping cannot explain the observed distribution of migrant adults in summer.

A final possibility is that differential survival might contribute to the elevational shifts shown in this study. Greater adult abundance at higher sites in summer relative to spring could arise from the increasing survival of G. rhamni juvenile stages or adults with increasing elevation. However, three findings do not support this possibility as the main explanation: (1) substantial numbers of adults were recorded at elevations above the elevation range of larval host plants; (2) large numbers of individuals were recorded at lower elevations in spring following hibernation; (3) no significant correlation was found between overwintering and summer abundance within the same year for all years but one (2008); if in situ survival made a substantial contribution to the abundance of summer adults, some degree of correlation between overwintering and summer abundance would be expected (e.g. Pollard and Greatorex-Davies 1998).

One specific trait of G. rhamni is that individuals make a return migration to the area from which they bred. This is the most common type of migration in birds and mammals, but it has rarely been documented in insects (Holland et al. 2006; but see Samraoui et al. 1998). The best-known case of return latitudinal migration (associated with overwintering areas) is that performed by the best studied migratory insect, the monarch butterfly (Danaus plexippus), but successive broods are involved during the progression northwards (Flockhart et al. 2013). Possible return elevational migrations have been reported for some butterfly species in other areas with hot and dry summers, including G. rhamni (Larsen 1982). In this case, individuals were thought to breed at high elevations in summer and then to migrate downhill in autumn to overwinter (Larsen 1976). These findings pose questions regarding the extent to which variability in regional climates, resource distributions and seasonality may drive divergent elevational migration patterns within the same species, and the extent to which they may be subject to change in a changing climate. Our results suggest phenotypic plasticity in the extent and timing of return elevational migration by G. rhamni, probably linked to the fact that it is a univoltine species with long-lived adults. In this context, the study of possible migration patterns in other species with similar life cycles could shed light on the life-history and evolution of elevational migration in insects.

The results presented here have some implications in the context of climate change. We found that G. rhamni summer adults occurred at sites which were on average 3 °C cooler than breeding (host plant) sites; that summer abundance was sometimes negatively associated with summer mean maximum temperatures; and that higher sites may have been occupied in warmer summers. Furthermore, numbers of post-overwintering adults per site were only significantly correlated with numbers of pre-hibernating adults after two relatively cool summers (2008 and 2011). Assuming that temperature is an important determinant of summer elevational distribution, this evidence suggests that a warming climate could eventually generate a bottleneck in G. rhamni populations in the Sierra de Guadarrama through the constraint of its summer habitat network. This is in line with the notion that climate change could affect elevational migrants (Inouye et al. 2000), but through a completely different mechanism from that of phenological synchrony with resources.

Notes

Acknowledgments

J. Gutiérrez Illán and S.B.Díez assisted with fieldwork and S. Nieto-Sánchez and T. Izquierdo helped with climate data. The research was funded by Universidad Rey Juan Carlos/Comunidad de Madrid (URJC-CM-2006-CET-0592), the Spanish Ministry of Economy and Competitiveness (REN2002-12853-E/GLO, CGL2005-06820/BOS, CGL2008-04950/BOS and CGL2011-30259), the British Ecological Society and the Royal Society. Access and research permits were provided by Comunidad de Madrid, Parque Natural de Peñalara, Parque Regional de la Cuenca Alta del Manzanares, Parque Regional del Curso Medio del Río Guadarrama, Patrimonio Nacional and Ayuntamiento de Cercedilla.

Supplementary material

442_2014_2952_MOESM1_ESM.pdf (415 kb)
Supplementary material 1 (PDF 415 kb)

References

  1. Bartoń K (2012) MuMIn: multi-model inference. R package version 1.6.6. Available at: http://CRAN.R-project.org/package=MuMIn. Accessed 19 Jan 2012
  2. Boyle WA (2008) Can variation in risk of nest predation explain altitudinal migration in tropical birds? Oecologia 155:397–403PubMedCrossRefGoogle Scholar
  3. Boyle WA (2010) Does food abundance explain altitudinal migration in a tropical frugivorous bird? Can J Zool 88:204–213CrossRefGoogle Scholar
  4. Boyle WA, Norris DR, Guglielmo CG (2010) Storms drive altitudinal migration in a tropical bird. Proc R Soc B 277:2511–2519PubMedCentralPubMedCrossRefGoogle Scholar
  5. Brambilla M, Falco R, Negri I (2012) A spatially explicit assessment of within-season changes in environmental suitability for farmland birds along an altitudinal gradient. Anim Conserv 15:638–647CrossRefGoogle Scholar
  6. Brattström O, Bensch S, Wassenaar LI, Hobson KA, Åkesson S (2010) Understanding the migration ecology of European red admirals Vanessa atalanta using stable hydrogen isotopes. Ecography 33:720–729CrossRefGoogle Scholar
  7. Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-theoretic approach, 2nd edn. Springer, New YorkGoogle Scholar
  8. Chapman JW, Bell JR, Burgin LE, Reynolds DR, Pettersson LB, Hill JK, Bonsall MB, Thomas JA (2012) Seasonal migration to high latitudes results in major reproductive benefits in an insect. Proc Natl Acad Sci USA 109:14924–14929PubMedCentralPubMedCrossRefGoogle Scholar
  9. Chevan A, Sutherland M (1991) Hierarchical partitioning. Am Stat 45:90–96Google Scholar
  10. Dingle H, Drake VA (2007) What is migration? Bioscience 57:113–121CrossRefGoogle Scholar
  11. Diniz-Filho JAF, Rangel TFLVB, Bini LM (2008) Model selection and information theory in geographical ecology. Global Ecol Biogeogr 17:479–488CrossRefGoogle Scholar
  12. Dormann CF, Elith J, Bacher S, Buchmann C, Carl G, Carré G, García Marquéz JR, Gruber B, Lafourcade B, Leitão PJ, Münkemüller T, McClean C, Osborne PE, Reineking B, Schröder B, Skidmore AK, Zurell D, Lautenbach S (2013) Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography 36:27–46CrossRefGoogle Scholar
  13. Environmental Systems Research Institute Inc. (ESRI) (2001) ArcGIS 8.1. ESRI Inc., RedlandsGoogle Scholar
  14. Farr TG, Rosen PA, Caro E, Crippen R, Duren R, Hensley S, Kobrick M, Paller M, Rodriguez E, Roth L, Seal D, Shaffer S, Shimada J, Umland J, Werner M, Oskin M, Burbank D, Alsdorf D (2007) The shuttle radar topography mission. Rev Geophys 45:RG2004Google Scholar
  15. Flockhart DTT, Wassenaar LI, Martin TG, Hobson KA, Wunder MB, Norris DR (2013) Tracking multi-generational colonization of the breeding grounds by monarch butterflies in eastern North America. Proc R Soc B 280:20131087PubMedCrossRefGoogle Scholar
  16. García-Barros E, Munguira ML, Cano JM, Romo H, Garcia-Pereira P, Maravalhas ES (2004) Atlas of the butterflies of the Iberian Peninsula and Balearic Islands (Lepidoptera: Papilionoidea and Hesperioidea). Sociedad Entomológica Aragonesa, ZaragozaGoogle Scholar
  17. García-Barros E, Munguira ML, Stefanescu C, Vives Moreno A (2013) Lepidoptera Papilionoidea. In: Ramos MA et al (eds) Fauna Ibérica, vol 37. Museo Nacional de Ciencias Naturales, CSIC, MadridGoogle Scholar
  18. Gutiérrez Illán J, Gutiérrez D, Wilson RJ (2010) The contributions of topoclimate and land cover to species distributions and abundance: fine resolution tests for a mountain butterfly fauna. Global Ecol Biogeogr 19:159–173CrossRefGoogle Scholar
  19. Gutiérrez D, Thomas CD (2000) Marginal range expansion in a host-limited butterfly species Goneptery rhamni. Ecol Entomol 25:165–170CrossRefGoogle Scholar
  20. Holland RA, Wikelski M, Wilcove DS (2006) How and why do insects migrate? Science 313:794–796PubMedCrossRefGoogle Scholar
  21. Hunt JH, Brodie RJ, Carithers TP, Goldstein PZ, Janzen DH (1999) Dry season migration by Costa Rican lowland paper wasps to high elevation cold dormancy sites. Biotropica 31:192–196Google Scholar
  22. Inouye DW, Barr B, Armitage KB, Inouye BD (2000) Climate change is affecting altitudinal migrants and hibernating species. Proc Natl Acad Sci USA 97:1630–1633PubMedCentralPubMedCrossRefGoogle Scholar
  23. Larsen TB (1976) The importance of migration to the butterfly faunas of Lebanon, East Jordan, and Egypt (Lepidoptera, Rhopalocera). Notulae Entomol 56:73–83Google Scholar
  24. Larsen TB (1982) The importance of migration to the butterfly fauna of Arabia (Lep., Rhopalocera). Atalanta 13:248–259Google Scholar
  25. Legendre P, Legendre L (1998) Numerical ecology, 2nd English edn. Elsevier, AmsterdamGoogle Scholar
  26. Mac Nally R (1996) Hierarchical partitioning as an interpretative tool in multivariate inference. Aust J Ecol 21:224–228CrossRefGoogle Scholar
  27. Mac Nally R (2002) Multiple regression and inference in ecology and conservation biology: further comment on identifying important predictor variables. Biodiv Conserv 11:1397–1401CrossRefGoogle Scholar
  28. Mac Nally R, Walsh CJ (2004) Hierarchical partitioning public-domain software. Biodiv Conserv 13:659–660CrossRefGoogle Scholar
  29. Marini MA, Barbet-Massin M, Lopes LE, Jiguet F (2013) Geographic and seasonal distribution of the cock-tailed tyrant (Alectrurus tricolor) inferred from niche modelling. J Ornithol 154:393–402CrossRefGoogle Scholar
  30. McCullagh P, Nelder JA (1989) Generalized linear models, 2nd edn. Chapman and Hall/CRC, Boca RatonCrossRefGoogle Scholar
  31. McGuire LP, Boyle WA (2013) Altitudinal migration in bats: evidence, patterns and drivers. Biol Rev 88:767–786PubMedGoogle Scholar
  32. Merrill RM, Gutiérrez D, Lewis OT, Gutiérrez J, Díez SB, Wilson RJ (2008) Combined effects of climate and biotic interactions on the elevational range of a phytophagous insect. J Anim Ecol 77:145–155PubMedCrossRefGoogle Scholar
  33. Mikkola K (2003) The red admiral butterfly (Vanessa atalanta, Lepidoptera: Nymphalidae) is a true seasonal migrant: an evolutionary puzzle resolved? Eur J Entomol 100:625–626CrossRefGoogle Scholar
  34. Ministerio de Medio Ambiente (2000) Mapa forestal de España. Escala 1:50000. Provincia de Madrid. Ministerio de Medio Ambiente, MadridGoogle Scholar
  35. Ministerio de Medio Ambiente (2002a) Mapa forestal de España. Escala 1:50000. Provincia de Ávila. Ministerio de Medio Ambiente, MadridGoogle Scholar
  36. Ministerio de Medio Ambiente (2002b) Mapa forestal de España. Escala 1:50000. Provincia de Guadalajara. Ministerio de Medio Ambiente, MadridGoogle Scholar
  37. Ministerio de Medio Ambiente (2003) Mapa forestal de España. Escala 1:50000. Provincia de Segovia. Ministerio de Medio Ambiente, MadridGoogle Scholar
  38. Norbu N, Wikelski MC, Wilcove DS, Partecke J, Ugyen, Tenzin U, Sherub, Tempa T (2013) Partial altitudinal migration of a Himalayan forest pheasant. PLoS One 8(4):e60979PubMedCentralPubMedCrossRefGoogle Scholar
  39. Osborne JL, Loxdale HD, Woiwod IP (2002) Monitoring insect dispersal: methods and approaches. In: Bullock JM, Kenward RE, Hails RS (eds) Dispersal ecology. Blackwell Science Ltd, Malden, pp 24–49Google Scholar
  40. Pollard E, Greatorex-Davies JN (1998) Increased abundance of the red admiral butterfly Vanessa atalanta in Britain: the roles of immigration, overwintering and breeding within the country. Ecol Lett 1:77–81CrossRefGoogle Scholar
  41. Pollard E, Hall ML (1980) Possible movement of Gonepteryx rhamni (L.) (Lepidoptera: Pieridae) between hibernating and breeding areas. Entomol Gaz 31:217–220Google Scholar
  42. Pollard E, Yates TJ (1993) Monitoring butterflies for ecology and conservation. Chapman and Hall, LondonGoogle Scholar
  43. Pruess KP (1967) Migration of the army cutworm, Chorizagrotis auxiliaris (Lepidoptera: Noctuidae). I. Evidence for a migration. Ann Entomol Soc Am 60:910–920Google Scholar
  44. Ramenofsky M, Wingfield JC (2007) Regulation of migration. Bioscience 57:135–143CrossRefGoogle Scholar
  45. Rankin MA, Burchsted CA (1992) The cost of migration in insects. Annu Rev Entomol 37:533–559CrossRefGoogle Scholar
  46. Richards SA (2008) Dealing with overdispersed count data in applied ecology. J Appl Ecol 45:218–227CrossRefGoogle Scholar
  47. Samraoui B, Bouzid S, Boulahbat R, Corbet PS (1998) Postponed reproductive maturation in upland refuges maintains life-cycle continuity during the hot, dry season in Algerian dragonflies (Anisoptera). Int J Odonatol 1:119–135CrossRefGoogle Scholar
  48. Sawada M (1999) Rookcase: an Excel 97/2000 visual basic (VB) add-in for exploring global and local spatial autocorrelation. Bull Ecol Soc Am 80:231–234CrossRefGoogle Scholar
  49. Shapiro AM (1973) Altitudinal migration of butterflies in the central Sierra Nevada. J Res Lepid 12:231–235Google Scholar
  50. Shapiro AM (1974a) Altitudinal migration of central California butterflies. J Res Lepid 13:157–161Google Scholar
  51. Shapiro AM (1974b) Movements of Nymphalis californica (Nymphalidae) in 1972. J Lepid Soc 28:75–78Google Scholar
  52. Shapiro AM (1975) Why do California tortoiseshells migrate? J Res Lepid 14:93–97Google Scholar
  53. Shapiro AM (1980) Mediterranean climate and butterfly migration: an overview of the California fauna. Atalanta 11:161–188Google Scholar
  54. Southwood TRE (1977) Habitat, the templet for ecological strategies. J Anim Ecol 46:337–365CrossRefGoogle Scholar
  55. Stefanescu C (2001) The nature of migration in the red admiral butterfly Vanessa atalanta: evidence from the population ecology in its southern range. Ecol Entomol 26:525–536CrossRefGoogle Scholar
  56. Stefanescu C, Traveset A (2009) Factors influencing the degree of generalization in flower use by Mediterranean butterflies. Oikos 118:1109–1117CrossRefGoogle Scholar
  57. Stefanescu C, Alarcón M, Ávila A (2007) Migration of the painted lady butterfly, Vanessa cardui, to north-eastern Spain is aided by African wind currents. J Anim Ecol 76:888–898PubMedCrossRefGoogle Scholar
  58. Stefanescu C, Askew RR, Corbera J, Shaw MR (2012) Parasitism and migration in southern Palaearctic populations of the painted lady butterfly, Vanessa cardui (Lepidoptera: Nymphalidae). Eur J Entomol 109:85–94CrossRefGoogle Scholar
  59. Stefanescu C, Páramo F, Åkesson S, Alarcón M, Ávila A, Brereton T, Carnicer J, Cassar LF, Fox R, Heliölä J, Hill JK, Hirneisen N, Kjellén N, Kühn E, Kuussaari M, Leskinen M, Liechti F, Musche M, Regan EC, Reynolds DR, Roy DB, Ryrholm N, Schmaljohann H, Settele J, Thomas CD, van Swaay C, Chapman JW (2013) Multi-generational long-distance migration of insects: studying the painted lady butterfly in the Western Palaearctic. Ecography 36:474–486CrossRefGoogle Scholar
  60. R Development Core Team (2012) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Available at: http://www.R-project.org/. Accessed 20 Jan 2012
  61. Tolman T, Lewington R (1997) Butterflies of Britain and Europe. HarperCollins, LondonGoogle Scholar
  62. Urquhart FA, Urquhart NR (1978) Autumnal migration routes of the eastern population of the monarch butterfly (Danaus p. plexippus L.; Danaidae; Lepidoptera) in North America to the overwintering site in the Neovolcanic Plateau of Mexico. Can J Zool 56:1759–1764CrossRefGoogle Scholar
  63. Wikelski M, Moskowitz D, Adelman JS, Cochran J, Wilcove DS, May ML (2006) Simple rules guide dragonfly migration. Biol Lett 2:325–329PubMedCentralPubMedCrossRefGoogle Scholar
  64. Wiklund C, Lindfors V, Forsberg J (1996) Early male emergence and reproductive phenology of the adult overwintering butterfly Gonepteryx rhamni in Sweden. Oikos 75:227–240CrossRefGoogle Scholar
  65. Williams CB (1930) The migration of butterflies. Oliver and Boyd, EdinburghGoogle Scholar
  66. Zuur AF, Ieno EN, Smith GM (2007) Analysing ecological data. Springer, New YorkGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  1. 1.Área de Biodiversidad y Conservación, Escuela Superior de Ciencias Experimentales y TecnologíaUniversidad Rey Juan CarlosMóstolesSpain
  2. 2.College of Life and Environmental SciencesUniversity of ExeterExeterUK

Personalised recommendations