Global Change Biology (2017) 23, 2218–2229, doi: 10.1111/gcb.13500

Where do they go? The effects of topography and habitat diversity on reducing climatic debt in birds € ER " E1, KARINE PRINCE # 2 , 3 and V I N C E N T D E V I C T O R 1 P I E R R E G A UZ 1 Institut des Sciences de l’Evolution, CNRS, IRD, Universit!e Montpellier, Place Eug"ene Bataillon, 34095 Montpellier cedex 05, France, 2Centre de Recherches sur la Biologie des Populations d’Oiseaux, UMR7204 MNHN-CNRS-UPMC, Mus!eum National d’Histoire Naturelle, CP 51, Paris, France, 3Department of Forest and Wildlife Ecology, University of Wisconsin-Madison, Madison, WI 53706, USA

Abstract The spatial tracking of climatic shifts is frequently reported as a biodiversity response to climatic change. However, species’ range shifts are often idiosyncratic and inconsistent with climatic shift predictions. At the community scale, this discrepancy can be measured by comparing the spatial shift in the relative composition of cold- vs. warmadapted species in a local assemblage [the community temperature index (CTI)] with the spatial shift in temperature isotherms. While the local distribution of climate change velocity is a promising approach to downscaling climate change pressure and responses, CTI velocity has only been investigated on a continental or national scale. In this study, we coupled French Breeding Bird Survey data, collected from 2133 sites monitored between 2001 and 2012, with climatic data in order to estimate the local magnitude and direction of breeding season temperature shift, CTI shift, and their spatiotemporal divergence – the local climatic debt. We also tested whether landscape characteristics that are known to affect climate velocity and spatial tracking of climate change mediated the climatic debt on the local scale. We found a clear spatial structure, together with heterogeneity in both temperature and CTI spatial shifts. Local climatic debt decreased as the elevation, habitat diversity, and the naturalness of the landscape increased. These results suggest the complementary effects of the local topographic patterns sheltering more diverse microclimates and the increasing permeability of natural and diversified landscape. Our findings suggest that a more nuanced evaluation of spatial variability in climatic and biotic shifts is necessary in order to properly describe biodiversity responses to climate change rather than the oversimplified descriptions of uniform poleward shifts. Keywords: climate change, community temperature index, directionality, habitat, isotherm shifts, temperature, velocity Received 28 July 2016 and accepted 5 September 2016

Introduction Faced with climate change, species can adopt and combine two strategies: maintaining its original geographic distribution by tolerance, acclimation, or evolutionary adaptation or to keep up with its original niche by tracking the variations in space (Parmesan, 2006; Lavergne et al., 2010) and/or in time (Root et al., 2003; Dunn & Winkler, 2010; Dunn & Møller, 2013). The fate of most species facing climate changes will therefore be determined by their ability to cope with local climate changes, which in turn depends on their capacity to adjust their physiology, phenology, and/or spatial distribution. Although resistance processes and rapid evolutionary adaptation have been documented for some taxa (Norberg et al., 2012), spatial tracking of climatic shifts seems to be a common strategy, and the distribution of many species appears to have already been subject to strong poleward and upward shifts Correspondence: Pierre Ga€ uz"ere, tel. +33 6 30 58 89 93, e-mail: [email protected]

2218

(Hickling et al., 2006; La Sorte & Thompson, 2007; Tingley et al., 2012). However, while some movements are consistent with climatic shift (Chen et al., 2011; Pinsky et al., 2013; Poloczanska et al., 2013), others are not (La Sorte & Jetz, 2012; Zhu et al., 2012; Comte & Grenouillet, 2015). The concept of climatic debt encapsulates this discrepancy (Kuussaari et al., 2009) and can be defined as the spatial lag accumulated by a species or a community when its spatial response is not consistent with the experienced climatic shift. At the species scale, several formal estimates of the lag between a spatial shift in climatic conditions and species’ range shifts have been calculated (Gillings et al., 2015; Lenoir & Svenning, 2015). One common approach compares the relative composition of cold- vs. warm-dwelling species in a local community [known as the community temperature index (CTI)] with the velocity of climatic shift (Devictor et al., 2008). The ratio between the spatial CTI gradient (°Ccti km!1) and its temporal trend (°Ccti yr!1) is used to generate an estimate of the velocity of change in the composition of a community (km yr!1). Similarly, the velocity of isotherm climatic shift (km yr!1) is © 2016 John Wiley & Sons Ltd

L A N D S C A P E D R I V E R S O F B I R D ’ S C L I M A T I C D E B T 2219 a climate change metric calculated from the ratio between spatial (°C km!1) and temporal (°C yr!1) thermal gradients (Loarie et al., 2009). The direct comparison of climatic change and biotic responses can highlight whether the composition of a community is lagging behind an isotherm shift at the national (Devictor et al., 2008) or continental (Devictor et al., 2012a,b) scale. The CTI has been shown to be a reliable indicator of the re-composition of a community in response to local climatic variations (Lindstr€ om et al., 2013; Princ#e & Zuckerberg, 2014; Ga€ uz"ere et al., 2016). It has been applied to various groups such as birds, plants, butterflies, and fishes (Cheung et al., 2013; Roth et al., 2014; Zografou et al., 2014; Nieto-S# anchez et al., 2015; Savage & Vellend, 2015) and adopted as an indicator of climate change impacts on biodiversity by both governmental and nongovernmental conservation organizations in Europe (Van Swaay et al., 2010). However, source of variability in climatic and community shifts may have been overlooked, due to the oversimplification of estimations (Ohlem€ uller, 2011; Dobrowski et al., 2013). Local variations in climate are both highly heterogeneous and dynamic at the regional (Walther et al., 2002; Burrows et al., 2011) and local (Ga€ uz"ere et al., 2015) spatial scales. The complex lattice created by latitude, longitude, elevation, landscape, and land use leads to high variability in the rate and direction of isotherm shifts in space and time (Loarie et al., 2009), and recent studies have shown that finescale climate variability may drive distribution dynamics (Early & Sax, 2011; Bennie et al., 2013; Lenoir & Svenning, 2015). In practice, any measure of climatic debt is highly dependent on the scale used to measure the spatial thermal gradients, and the addition of the spatial distribution of climate change velocity (rather than monotonic changes for a given area) is a promising approach that can help to downscale climate change pressure and responses, and estimate the speed and direction of climate characteristics (Ohlem€ uller, 2011). Adopting a spatial analysis that integrates the systematic quantification of climatic conditions and communities is expected to refine the climatic debt concept and improve its local relevance. Although the concept of climatic debt has proved to be a useful yardstick to measure the delay of biodiversity responses on a large spatial scale (Visser, 2012), its magnitude and directionality have never been quantified on a local scale. The study of these factors in the context of a spatial shift of a community therefore represents a crucial step toward a better understanding of the key factors that determine the biodiversity responses to climatic variations (VanDerWal et al., 2012). A few studies have compared the shifts in species’ distributions with climate change velocities (Pinsky © 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

et al., 2013; Comte & Grenouillet, 2015; Gillings et al., 2015), but none, to our knowledge, have also related velocities to landscape characteristics. In the case of birds, the optimal biotic spatial shift (i.e., the shift required to pay off climatic debt) is probably limited by individual traits such as dispersal ability or ecological specialization (Anderson et al., 2009; Hof et al., 2012) and landscape particularities such as topography, composition, and structure. For instance, in France, alpine biogeographical domain is expected to shelter slower velocities, whereas faster shifts are expected to be seen in flatter regions such as the French Atlantic coast (Loarie et al., 2009). On an even smaller scale, many mechanisms potentially mediate the spatial response. First, steep and complex topography promotes the efficient tracking of climatic shifts by reducing the distance between isotherms. Secondly, spatiotemporal variability in microclimates can create the gaps in climate paths (Early & Sax, 2011) and create the local climatic refugia (Lenoir et al., 2013). Conversely, an inhospitable matrix of heavily managed landscapes dominated by intensive agriculture or forestry is thought to make the colonization of new areas more difficult (Thomas et al., 2012). Other human imprints, such as roads or artificial areas, have been shown to increase the habitat fragmentation and affect the dispersal of birds (Andrews, 1992; Spellerberg, 1998). Therefore, the presence of natural ‘corridors’ or ‘stepping stones’ via which species can easily disperse and settle are thought to play a key role in the process of climatic niche tracking (Huntley et al., 2008). As more diversified landscape increases provide greater probability that a given species will find a suitable habitat, habitat diversity is expected to promote dispersion and settlement during range expansion (Vos et al., 2008). Overall, whether and how the local climatic debts are accelerated or mitigated by specific habitats, landscape structures, and/or composition are still largely unknown (Bennie et al., 2013; Feeley et al., 2014; Jarzyna et al., 2015). This study combines a climate space analysis with the recent advances in community ecology in order to estimate the spatial lag between local climatic and community isotherm shifts. We use data from the French Breeding Bird Survey (FBBS) over an 12-year period (2001–2012) to estimate, visualize, and test (i) for congruence between the direction and velocity of spatial shifts in climate and community isotherms and (ii) whether spatial congruence between climatic and bioclimatic shifts is mediated by the landscape composition and configuration. The prediction is that greater congruence between climate and community changes will be found in areas with steep elevation gradients and in diverse and natural habitats.

€ ER " E et al. 2220 P . G A UZ Materials and methods

Data Birds. Data were taken from the FBBS. This is an annual monitoring program in which skilled volunteer ornithologists count birds, following a standardized protocol, at the same site, year after year since 2001 (Jiguet et al., 2012). Species’ abundance has been recorded for 2133 sites measuring 2 9 2 km. Each volunteer provides a location, and the sites are randomly selected from within a 10 km radius around the residence of each volunteer. This random selection ensured both the survey of varied habitats and representative sampling. On each plot, volunteers carried out 10-point counts (5 min each, separated by at least 300 m) twice each spring within three weeks around the pivotal date of May 8 to ensure the detection of both early and late breeders. Counts must be repeated on approximately the same date from year to year ("7 days) and at dawn (within 1–4 h after sunrise) by the same observer. We retained the maximum per-point count from these two sessions as the measure of point-level species’ abundance. The final sample included data for 124 bird species representing 99% of the total abundance. Temperature. Data were extracted from the SAFRAN meteorological model (Quintana-Segu#ı, 2008). This provides monthly means of 2-m air temperature on an 8 9 8 km grid over France. These high-resolution data were matched with the sites monitored by the FBBS, and the mean breeding season temperature (March to August, in °C) for each site and each year (Julliard et al., 2004) was calculated to estimate the average temperature experienced by birds during breeding.

Community temperature index First, a species temperature index (STI, in °C; Devictor et al., 2008) was calculated for each bird species. This reflected the long-term average temperature experienced by a species across its geographic range during the breeding season. Therefore, STI values were computed using independent dataset of the long-term and large-scale temperature data coupled with continental-scale atlas of bird distributions. We specifically used 0.5° 9 0.5° temperature grids (March–August averages for the period 1950–2000; Hijmans et al., 2005; http://www. worldclim.org) coupled with their Western Palearctic distributions at 0.5° resolution given in The EBCC Atlas of European Breeding Birds (Hagemeijer & Blair, 1997). Note that the calculation of absolute values of STI could vary according to the spatial scale considered. For example, limiting its estimation to French, European, or Western Palearctic atlas temperature cells could affect the STI value of specific species. However, this variation should not affect our analysis because a temporal variation in CTI depends on the relative species’ STI rather than on the absolute value of each specific STI, and French, European, or even Palearctic STIs are all highly correlated (Devictor et al., 2008, 2012b). For example, STI values estimated with Western Palearctic (WorldClim dataset) or French (Safran dataset) distributions were strongly correlated (see Fig. S1).

We then calculated, for each community (i.e., for each site monitored in a given year), the abundance-weighted mean of the STIs across all species present in each assemblage. This community-weighted mean called CTI (in °C, Devictor et al., 2008) has been valuably used with numerous taxa and at different spatial scales to describe how the thermal composition of communities tracks the temperature changes. It should be noted that as it is used here, the CTI is a measure of the community response to spring temperature variations and does not measure other climatic-induced change in community such as precipitation. Notwithstanding that the multidimensionality of climate can be crucial when studying the bird responses to climate change (Dunn & Winkler, 2010; Ill# an et al., 2014), European common breeding bird responses to climatic variations are largely driven by the mean temperature experienced during the breeding season (Julliard et al., 2004; Jiguet et al., 2006, 2010). Moreover, previous studies have shown that in France, the temporal variations in temperature and precipitation were roughly correlated during the study period (Comte & Grenouillet, 2015; Ga€ uz"ere et al., 2015). Note that focusing on a 12-year period seems rather short to relate climate change and community responses in birds. In fact, earlier studies have shown that CTI was highly sensitive to the temperature changes. For instance, Lindstr€ om et al. (2013) showed that the yearly fluctuation in temperature and CTI was synchronous with a 2-year delay only. Over the period 2001–2012, Ga€ uz"ere et al. (2015) also showed that in France, bird communities have adjusted their composition rapidly to the local climate change. Therefore, we think that testing and mapping how the local changes in temperature and CTI are associated over 12 years is relevant.

Community and climatic isotherm shifts The analytical framework was inspired by the climatic velocity approach developed by Loarie et al. (2009). This framework aims to derive the local isotherm shifts from the ratio between temporal and spatial gradients (Loarie et al., 2009; Burrows et al., 2011) in CTI, and mean breeding season temperature. The framework is shown in Figure 1. A moving window analysis estimated the spatial and temporal CTI trends and the mean breeding season temperature on a spatial continuum covering the whole study area. The analysis consisted in consecutively defining each plot as the center of an 80-km-radius circle encompassing at least 20 sites (average 88.6 " 42, maximum = 207). This generated 2094 similarly sized spatial windows, each containing enough temporal and spatial repetition to estimate the reliable linear trends in variables. It is a straightforward way to summarize the spatial or temporal local trends emerging from regional dynamics (Gaucherel et al., 2007, 2008) and has previously been used successfully with this dataset (Devictor et al., 2010; Godet et al., 2015; Ga€ uz"ere et al., 2016). Within each window, and for both CTI and the mean breeding season temperature, linear mixed-effects models consecutively estimated the temporal temperature trends as the linear trend in average temperature across time (the years 2001–2012 were considered as fixed continuous linear effects, while the site as the random factor effect) and the spatial thermal gradient as the trend in average temperature across the whole © 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

L A N D S C A P E D R I V E R S O F B I R D ’ S C L I M A T I C D E B T 2221

Fig. 1 Scheme of the analytical framework used to estimate the climatic, community, and residual distance shift (RDS). Data were taken from French Breeding Bird Survey sites and coupled with the climatic data from the SAFRAN meteorological model. The next step was to define 2094 80-km-radius circular windows. Isotherm shifts were calculated for each window based on the mean spring temperature, and the community temperature index was calculated as the ratio between temporal change (°C yr!1) and spatial gradient (°C m!1). RDS was estimated as the vector of the difference between the climatic and community isotherm shifts. [Colour figure can be viewed at wileyonlinelibrary.com].

80 km window (latitude and longitude based on the RGF93/ Lambert-93 projection in meters were considered as fixed continuous effects and the years 2001–2012 as random factor effects). As both CTI and temperature can be expressed in °C, we extracted, for each model, their respective fixed-effect coefficients in order to estimate the temperature or CTI gradient over time (°C yr!1), latitude (°C kilometer_latitude!1), and longitude (°C kilometer_longitude!1). Secondly, we derived biotic and climatic spatial isotherm shifts from the ratio of temporal trends to spatial (latitudinal and longitudinal) temperature or CTI gradients (°C yr!1/°C km!1 = km yr!1). Isotherm shift vectors (temperature or CTI) were then defined for each window as the combination of its latitudinal and longitudinal components.

length of these vectors as the index of the local spatial delay of the community response to temperature shifts, called the ‘residual distance shift’ (RDS). The RDS estimates the local climatic debt as it quantifies the spatial discrepancy between local shift in temperature and community composition. Finally, we represented climatic isotherm shifts, CTI shifts, and RDS for each window by mapping vectors over the entire country and plotting velocity distributions. When spatial gradients tend to zero, the value of the vector shift tends to infinity and so skews the distributions of latitudinal and longitudinal shifts. Therefore, to simplify the representation and to meet the assumption of normality in the subsequent analyses, we chose to log-10-transform the magnitude of vectors.

Landscape characteristics Residual isotherm shifts In order to estimate the local climatic debt accumulated by birds over the 12 years of monitoring, we calculated the lag in the response of bird communities to the spatial shift of climatic isotherms as the vectorial difference between the climatic isotherm shift and the shift in CTI. Estimating the difference between two vectors allows us to consider both the divergence of direction and magnitude between vectors. We retained the © 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

Landscape composition variables were calculated from CORINE Land Cover 2006 raster data at 100 9 100 m resolution (CLC 2006, Bossard et al., 2000) for each of the 80-km-radius window. In order to be able to consider the large-scale biogeographical gradients, CLC 2006 classes were aggregated according to their top-level nomenclature into artificial surfaces; agricultural areas; forest and seminatural areas; wetlands; and water bodies. We then computed the percentage land cover

€ ER " E et al. 2222 P . G A UZ (calculated as the habitat class area, in square meters over the total area of the window) for the first three of these classes, in order to define three variables that represented landscape composition (% artificial area, % agricultural area, and % natural area). Road density was calculated using data from the French National Geographic Institute (http://professionnels.ign.fr/ route500). The length of roads within each window was used to estimate the density of the road network, calculated as the ratio of road length to the total area of the window (road_density, in linear-meters m!2). Landscape structure was estimated using the Shannon diversity index, which is based on the percentage cover of each of the CLC 2006 classes aggregated by their second-level nomenclature. Finally, we computed the elevation gradients for each window as the difference between the elevation of the highest and lowest FBBS site.

Effect of landscape characteristics on RDS A linear model was developed to test the effect of landscape characteristics on the ability of the community to track the climatic isotherm shifts. The log10-transformed RDS was regressed over the set of landscape descriptors. We accounted for the spatial autocorrelation in the residuals of the model by fitting our model using generalized least-squares model. Here, the spatial structure of the data was used to model the error term correlation structure. Spatial structures were best described by an exponential semivariogram (nugget ~ 0, range ~18–21 km, depending on the model). Depending on their geographical distribution, the number of sites in each window varied (from 20 to 207, mean = 88 " 42). We took this variability into account by creating a fixed variance structure based on the number of sites and integrating it as a weight into the model. To identify a meaningful set of landscape characteristics most likely to explain the observed RDS, we then defined the best predictive model on the basis of their AICc (Burnham & Anderson, 2002). We fitted all possible subsets of fixed-effect structures nested within a maximum model (i.e., including all landscape variables) and ranked them by AICc values. The average of all models differing by less than two AICc units was considered to be the optimal fixed-effects model. While criterion-like approaches are useful for locating the optimal functional and predictive model, hierarchical partitioning makes it possible to consider the entire web of relationships between predictor variables (MacNally, 2002). We therefore adopted this approach and performed a hierarchical variance partitioning analysis in order to compute the independent contribution of each landscape variable to the correlation between variables (the hier.part package of the R statistical software, R Core Team, 2013). It should be noted, however, that this analysis does not take into account the spatial autocorrelation and variability in the sample. We therefore

simply used this analysis as a way to estimate the amount of variance explained by independent landscape descriptors in a linear multivariate model.

Results The framework made it possible to map the local shifts in the climatic and biotic temperature of each window (Fig. 2). All isotherm shifts estimated for the period 2001–2012 exhibited a strong heterogeneity in space. Climatic isotherm shifts velocity showed a geographical pattern, with values ranging from 0 in mountainous areas to 125 km yr!1 in the western and northern part of the territory; the national average velocity reached 11 km yr!1. The spatial distribution of climatic shifts’ directions is highly heterogeneous, although the average direction of the shifts was heading north–northwest (Fig. 2a). CTI isotherm shifts were generally shorter than climatic ones and ranged from 0 to 112 km yr!1 with a national average of 5 km yr!1. The average direction of CTI isotherm shifts headed southwest. Both velocities and directions exhibited a strong spatial variability (Fig. 2b). The residual isotherm shifts resulting from the vectorial difference between climatic and CTI shifts were also spatially heterogeneous in both velocities and directions. However, the velocities were generally lower in the southeast and higher in the central part of the territory. The average residual shift was heading north and ranged from 0 to 187 km yr!1 with an average velocity of 14 km yr!1. A univariate analysis of the influence of landscape descriptors on RDS (GLS models) revealed a linear relationship (Fig. 3). RDS decreased as habitat diversity increased (!0.197 " 0.030, P < 0.0001), elevation gradient increased (!0.139 " 0.023, P < 0.0001), and the extent of seminatural and forest habitats increased (!0.136 " 0.033, P < 0.0001). Conversely, RDS increased as the extent of agricultural areas increased (0.135 " 0.033, P < 0.001). Finally, no significant relationship was found between RDS and the extent of artificial areas and road density (0.05 " 0.08, P = 0.5 and 0.059 " 0.031, P = 0.056, respectively). It should be noted, however, that univariate analyses did not take into account colinearity of variables describing the composition and structure of the landscape (Fig. S2). The results from the multivariate model that took into account all landscape descriptors (Table 1) showed that RDS decreased as habitat diversity (!0.181 "

Fig. 2 Spatial distribution of isotherm shifts. Each arrow represents the local vector of mean spring climatic isotherm shifts in blue (b), community temperature index isotherm shifts in orange (c), and the residual isotherm shifts in green (a). The velocity of shifts is represented by the log10-transformed length of arrows and their luminosity, the darker the shorter. Note that the residual isotherm shifts resulted from the vectorial difference between the climatic and community isotherm shifts and were centered on each FBBS site. The length of the residual isotherm shifts corresponds to the residual distance shift (RDS) used in further analysis. [Colour figure can be viewed at wileyonlinelibrary.com]. © 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

L A N D S C A P E D R I V E R S O F B I R D ’ S C L I M A T I C D E B T 2223

(a)

(b)

© 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

(c)

€ ER " E et al. 2224 P . G A UZ

Residual isother m shifts (km.year−1)

0.047, t = !3.85, P < 0.001) and steeper elevation gradients (!0.100 " 0.030, P < 0.001) increased. No significant effects were found for other predictors. The model selection procedure (based on testing for all possible fixed-effect structures nested within the global model including all landscape variables, and ranking them on the basis of AICc) showed that the averaged model (Table 2) that best predicted RDS included elevation gradients (!0.103 " 0.023, P < 0.001) and habitat diversity (!0.140 " 0.048, P = 0.004) with a relative variable importance (i.e., the sum of AIC weights across models including the variable, Burnham & Anderson, 2002) of 1. Natural areas (relative importance = 0.49), agricultural areas (relative importance = 0.30), and artificial areas (relative importance = 0.14) were also retained in the averaged model, but their effects on RDS were nonsignificant (see Table 2). Hierarchical variance partitioning performed on a linear model that incorporated all landscape descriptors and longitude–latitude coordinates as covariates (see Fig. S3) confirmed that the independent effect of

Discussion This study examined the impact of local temperature shifts in France in the period 2001–2012 on the estimated composition of the bird community at the same sites and during the same period. We estimated the local climatic debt of bird communities, and further analysis investigated whether it was mediated by landscape characteristics. The average velocity of CTI isotherm shifts were surprisingly close to the nationalscale shift in CTI reported in France by Devictor et al. (2008), which found a northward shift of 91 " 11 km for 1989–2006 (i.e., 5.35 km yr!1).

1000

1000

1000

100

100

100

10

10

10

1

1

1

1.5

2.0

2.5

0.00050

0.00075

0.00100

1

0.00125

Road density (m.m–2)

Habitat diversity Residual isother m shifts (km.year−1)

elevation gradient explained the most RDS variability (10% of independent explained variance), followed by habitat diversity (6%, see Fig. S3). Supplementary analyses of a dataset restricted to the sites situated below 800 m (see Tables S4a–c and Fig. S4) showed that habitat diversity became the only predictor that reduced the local climatic debt (!0.12 " 0.05, t = !2.09, P = 0.03).

1000

1000

1000

100

100

100

10

10

10

1

1

1

0

25

50

Natural area (%)

75

25

50

75

Agricultural area (%)

2

3

4 5

10

20

Altitudinal Gradient (log m.m–1)

4

8

12

Artificialized area (%)

Fig. 3 Scatter plots of the relationship between landscape predictors and residual distance shift, calculated as the magnitude of the vector resulting from the difference between the climatic and community isotherm shifts. Slopes and the shaded area represent the coefficient " SE (standard error) associated with the GLS model used to test an individual predictor. Circle size represents the number of sites in each window. [Colour figure can be viewed at wileyonlinelibrary.com]. © 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

L A N D S C A P E D R I V E R S O F B I R D ’ S C L I M A T I C D E B T 2225 Table 1 Results of the linear multivariate GLS regression of all landscape descriptors. The dependent variable was log10transformed RDS. GLS was used to model the error term spatial correlation structure. Sample variability was taken into account by creating a fixed variance structure based on the number of sites and integrating it as a weight Variable

Coefficient

SE

t-value

P-value

Habitat diversity Road density Altitudinal gradient Natural area Agricultural area Artificial area

!0.181 !0.00852 !0.100 0.0558 !0.00847 0.0206

0.047 0.046 0.030 0.32 0.33 0.078

!3.855 !0.182 !3.332 0.174 !0.025 0.264

0.0001 0.8549 0.0009 0.8612 0.9794 0.7917

Unlike the earlier work, however, here the average shift was not expected to be oriented northward, as the national mean spring temperature did not increase significantly between 2001 and 2012 (Ga€ uz"ere et al., 2015). Consequently, there was no overall, large-scale directional change in CTI. Instead, the local temporal variations were particularly marked in the southeast (see Fig. S5). On the national scale, the standard deviation of the mean velocity was high, reflecting the high spatial heterogeneity shown in Figure 2. Specific climatic variations at the local scale also explained the high heterogeneity in the direction and magnitude of isotherm shifts. Climate change is highly variable in both space (Ackerly et al., 2010; Burrows et al., 2011) and time (Karl et al., 1995), particularly in the period and locations considered in our study (Ga€ uz"ere et al., 2015). This climatic variability is exacerbated by the topographic determinant of climatic velocity, resulting in highly divergent vectors between contiguous local areas (Dobrowski et al., 2013). This complex mosaic is likely to be crucial for understanding the variability in spatial responses observed between species and areas (Ackerly et al., 2010; Tingley et al., 2012; Pinsky et al., 2013). These results confirm the previous findings that document the fine-scale sensitivity of species’ range shifts to climate change for various taxa (Early & Sax, 2011; Bennie et al., 2013; Lenoir et al., 2013). The complex spatial

patterns found here offer support for the argument that a more nuanced evaluation of spatial variability in climatic and biotic shifts is necessary to properly describe the responses of biodiversity to climate change – in contrast to the oversimplified descriptions of uniform poleward shifts (Lenoir & Svenning, 2015). A promising avenue for such analyses lies in quantitative estimates of the local delay and direction of biotic responses to climatic shift. To the best of our knowledge, only a few studies have compared the climatic shifts with speciesscale biotic responses (VanDerWal et al., 2012; Pinsky et al., 2013; Comte & Grenouillet, 2015; Ash et al., 2016). For instance, Gillings et al. (2015) modeled the species’ distribution to estimate the direction of British breeding birds’ range centroid shift, but did not find a consistent correlation with the spatial shift of various climatic variables. One of the main hypotheses that could explain this discrepancy lies in the availability and distribution of suitable habitats that foster species’ dispersal and colonization, and which probably partially drive the climate-related species’ range shifts at the local scale in the short term (e.g., Hill et al., 1999). Beyond its ability to highlight the fine-scale heterogeneity of climatic and biotic isotherm shifts, our framework also made it possible to test the specific predictions regarding landscape characteristics that may underlie the ability of communities to track the local climatic changes. In a national-scale analysis back-crossing the temporal and spatial CTI gradient, Devictor et al. (2008) defined climatic debt as the accumulated delay in species’ spatial responses to temperature change, which could be attributed to an inability to track the climatic change. Here, we estimated a local climatic debt over the entire country, allowing us to test for the effect of landscape characteristics expected to affect the ability of birds to track the climate change. We found that increased elevation gradient and habitat diversity were associated with the reduced local climatic debt for French bird communities. Elevation gradient is known to reduce the lag in biodiversity response to spatial climatic variations (Lenoir et al.,

Table 2 Model-averaged estimates of the effect of landscape predictors on RDS. All possible subsets of fixed-effect structures nested within the maximum model (i.e., including all landscape variables) were ranked on the basis of AICc. The average of all models differing by less than two AICc units (n = 5) was found to give the best fixed-effects model. Rel.Imp is the relative importance of each variable within the averaged model. It is calculated as the sum of variable AICc in the models in which the variable occurs Variable

Coefficient

SE

z-value

P-value

Rel.imp

Altitudinal gradient Habitat diversity Natural area Agricultural area Artificial area

!0.103 !0.140 0.0391 !0.0226 0.0020

0.023 0.048 0.051 0.042 0.014

4.355 2.879 0.756 0.527 0.140

0.0001 0.004 0.44 0.59 0.88

1.00 1.00 0.49 0.3 0.14

© 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

€ ER " E et al. 2226 P . G A UZ 2008). Large-scale studies investigating the changes in community composition related to the climate change have most always assumed uniform spatial gradient in temperature. However, one expects the effect of a local change on biodiversity in temperature to be highly dependent on the topography. In this respect, Bertrand et al. (2011) have shown that plant communities responded differently in lowland and highland areas, with much faster response in highlands. Here, we integrated the local topography in the estimate of the local climatic debt (that we called RDS) to account for the spatial distribution of the local gradient in temperature. First, steep topography and rough terrain create significant temperature variations over a small area, which decrease the velocity of climatic change (Loarie et al., 2009) and so the climatic debt. Secondly, complex topography shelters more diverse microclimates for species within a close vicinity, which in turn are likely to lower risks of population decline due to gradual or violent climatic variations (Luoto & Heikkinen, 2008). Finally, the diversity of climatic conditions in mountainous areas increases the local co-occurrence of both cold- and warm-dwelling species, leading to biotic assemblages composed of a wider range of species’ thermal preferences (Lenoir et al., 2013). For example, alpine plant communities have been found to track the climate warming more quickly than the lowland assemblages (Bertrand et al., 2011). As mountainous ecosystems are particularly vulnerable to climate change (Nogu#es-Bravo et al., 2007), the mitigation of climatic debt due to elevation gradient is good news. However, steep elevation gradients are spatially limited, and mountaintop extinction will inevitably follow accelerating temperature increases (Walther et al., 2002; Colwell et al., 2008). Our analysis also provides the reliable evidence for the positive effect of landscape diversity in mitigating the spatial lag between biotic and climatic isotherm shifts. Although the descriptor is very integrative, the diversity of habitats within a landscape is known to provide a mosaic of microhabitats and microclimates for species and communities to track the global changes. For example, studies based on range boundary adjustment in response to climate change have shown that the successful dispersal and establishment of the butterfly Pararge aegeria is positively related to habitat availability (Warren et al., 2001). While the ability of corridors and stepping stones to increase the ‘connectivity’ of landscapes is controversial, ‘landscape permeability’, defined as the ability of a given habitat mosaic to foster species’ dispersal into new, climatically suitable areas, is likely to partially explain the climatic debt patterns (Vos et al., 2008; Hole et al., 2009; Mawdsley et al., 2009; Hole & Huntley, 2011).

The thermal reshuffling of a community, represented by a given change in CTI, can result from two concomitant processes. First, the local extirpation and colonization of individuals in response to climate changes is affecting the occurrence of specific species. Secondly, the changes in local abundance can result from the changes in demographic processes such as reproductive success and survival without immigration or emigration events. Studies and supplementary analysis based on occurrences have previously shown that among these two processes at stake, extirpation and colonization events are likely to drive a large part of CTI dynamics (Devictor et al., 2008; Barnagaud et al., 2012; Ga€ uz"ere et al., 2015; Tayleur et al., 2016). Beyond the specific dispersal abilities that limit the species’ range shifts (Brooker et al., 2007; Anderson et al., 2009; Sandel et al., 2011; Schippers et al., 2011), the probability that common bird species will become successfully established in a given area is likely to be influenced by the probability of finding a suitable habitat (Opdam & Wascher, 2004; Vos et al., 2008). Habitat diversity is therefore expected to enhance the community-level isotherm tracking. Moreover, greater spatial heterogeneity tends to support greater taxonomic, functional, and genetic diversity (Vellend & Geber, 2005; Devictor et al., 2010). These metacommunity processes enhance the adaptive variation pool, increasing the likelihood that an appropriate phenotype adapts to new conditions during the colonization events. In the same way, taxonomic diversity and functional diversity are expected to foster the environmental affinities, which provide new species’ assemblages that are better adapted to the particular environmental conditions. Note that habitat diversity was, to some extent, related to the altitudinal gradient (see Fig. S2), which intrinsically affects the velocity of isotherms. Consequently, the observed relationships between habitat diversity and RDS could be considered as a spurious effect arising from our estimation of climatic and CTI isotherms. We therefore performed a supplementary analysis with a dataset limited to the sites below 800 m in elevation. Interestingly, habitat diversity was also found to have a significant effect (see Fig. S3). The elevation gradient was too limited for isotherm tracking of birds in lowland areas; however, the effect of landscape diversity was exacerbated. While the closely spaced isotherms that are found in mountainous landscapes intrinsically help birds to track the climatic change, this result suggests that a diverse mosaic of habitats could be even more crucial for climate tracking in plain ecosystems. Our results are clearly scale dependent. In particular, they refer to a circular window with a radius of 80 km, which is the result of a trade-off between the need for wide spatial coverage and the number of sites © 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

L A N D S C A P E D R I V E R S O F B I R D ’ S C L I M A T I C D E B T 2227 in each window. Although we are confident that the relationship between CTI and breeding season temperature is consistent over different spatial scales (radiuses ranging from 30 to 200 km, see Ga€ uz"ere et al., 2015), we argue that finer spatial resolutions provide a better understanding of the effect of habitat composition and configuration on climatic debt. However, spatial tracking of climatic shift is not the only response available for species facing climate change. Facing temperature changes, birds can indeed shift their phenology, migration strategy, physiology, or behavior to keep track with their climatic niche (Stenseth & Mysterud, 2002; Dunn & Winkler, 2010). Therefore, climatic debt could encompass both spatial delay in species’ dynamics and nonspatial responses of individual responses, particularly when local temperature changes are weak (Ga€ uz"ere et al., 2015). Moreover, synergies are likely between climate and land-use dynamics because habitats can affect the shifts in species’ distribution ranges in response to the climate change (Warren et al., 2001). For instance, while we could not detect the specific effect of road densities and artificialized areas, human settlement and habitat fragmentation are probably influencing species’ or communities’ responses at finer scales (Andrews, 1992). More generally, Clavero et al. (2011) have emphasized that the spatial correlations between the particular biogeographical gradient of land use and thermal composition of community can emerge beyond the direct effect of climate on the community composition. At the species level, Barnagaud et al. (2012) have shown that habitat and thermal niches could be intrinsically related, casting some doubts on the relevance of interpreting community changes only in terms of a response to land-use or climate changes. Although the recent analysis on the local-scale congruency between temperature and CTI temporal changes showed that such bias is limited (Ga€ uz"ere et al., 2015), where, and to what extent, each of these processes contributes to the specific effect of landscape diversity on community is difficult to disentangle. It is even more likely that climate and land-use change and occupancy act jointly on species’ and communities’ responses (Hockey et al., 2011; Kampichler et al., 2012). Accordingly, the development of our work lies in the quantification of the synergistic effects of climate and specific land-use changes (e.g., following landscape disturbance, forest recovery, or conservation plans) providing that high-enough-resolution data on habitat dynamics are available (Ga€ uz"ere et al., 2016; Tayleur et al., 2016). Our results show that the local changes in the composition of bird communities are not perfectly consistent with the changes in breeding season temperature. However, these results were based on the entire bird © 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

community, and the findings might be different when considering the individual species. Species-specific responses to the climatic variation are inherent in their sensitivity to different aspects of climate, which can shift in different directions at the local level (Dobrowski et al., 2013; Gillings et al., 2015) and dramatically influence species’ responses (Nadeau & Fuller, 2015). Consequently, community-aggregated responses (such as CTI) are likely to reflect a limited set of species, which are particularly affected by a particular climatic change (breeding season temperature in this particular case). Further investigation of spatial shifts at the community level should be carried out on functional groups. These investigations should be based upon species-specific sensitivities to other aspects of climate, such as precipitation during egg laying or nest provisioning. Habitat and, more broadly, ecological preferences also discriminate the ability of a species to track the climatic shifts, and mediate the effect of these shifts on their distribution, notably through range contractions (Jetz et al., 2007; Hockey et al., 2011). Earlier work has documented the greater ability of generalist species to track the climatic variations at the expense of more specialized species. Although our results suggest that habitat diversity reduces the spatial lag between biotic and climatic shifts, the effect is likely to be mediated by species’ specialization. For instance, forest habitat specialist species can track the climate change more easily in homogeneous, forested landscapes than in more diverse habitats (Opdam & Wascher, 2004). Therefore, an interesting extension of our work would repeat the analysis on species as a function of their habitat preferences. This would highlight synergies between the diversity/fragmentation and composition of the landscape on particular groups (such as forest-, agricultural-, or urban-dwelling birds). Overall, our work pleads for the finer investigation of high-resolution and multidimensional aspects of global changes and their effects on the spatial distribution of biodiversity. We here downscaled the concept of biotic community isotherm shifts in order to test the ability of landscape characteristics to shape the climatic debt of French breeding birds. Coupling community, species, and population responses and even finer spatial resolution should allow us to better understand the ecological processes at play in the biodiversity tracking of climate changes.

Acknowledgements We sincerely and warmly thank the hundreds of volunteers who took part in the French Breeding Bird Survey (STOC EPS Program) to collect the valuable data used in our analysis. The national coordination is financially supported by the National Museum of Natural History, the CNRS, and the French

€ ER " E et al. 2228 P . G A UZ Environment Ministry. The manuscript benefited greatly from the comments provided by anonymous reviewers.

References Ackerly DD, Loarie SR, Cornwell WK, Weiss SB, Hamilton H, Branciforte R, Kraft NJB (2010) The geography of climate change: implications for conservation biogeography. Diversity and Distributions, 16, 476–487. Anderson BJ, Akc!akaya HR, Ara# ujo MB, Fordham DA, Martinez-Meyer E, Thuiller W, Brook BW (2009) Dynamics of range margins for metapopulations under climate change. Proceedings. Biological Sciences/The Royal Society, 276, 1415–1420. Andrews A (1992) Fragmentation of habitat by roads and utility corridors: a review. Biological Conservation, 59, 77. Ash JD, Givnish TJ, Waller DM (2016) Tracking lags in historical plant species’ shifts in relation to regional climate change. Global Change Biology, 23, 1305–1315. Barnagaud J-Y, Devictor V, Jiguet F, Barbet-Massin M, Le Viol I, Archaux F (2012) Relating habitat and climatic niches in birds. PLoS One, 7, e32819. Bennie J, Hodgson JA, Lawson CR et al. (2013) Range expansion through fragmented landscapes under a variable climate. Ecology Letters, 16, 921–929. Bertrand R, Lenoir J, Piedallu C et al. (2011) Changes in plant community composition lag behind climate warming in lowland forests. Nature, 479, 9–12. Bossard M, Fernec J, Otahel J (2000) CORINE Land Cover Technical Guide – Addendum 2000. European Environment Agency, Copenhagen. Brooker RW, Travis JMJ, Clark EJ, Dytham C (2007) Modelling species’ range shifts in a changing climate: the impacts of biotic interactions, dispersal distance and the rate of climate change. Journal of Theoretical Biology, 245, 59–65. Burnham KKP, Anderson DRD. (2002) Model Selection and Multimodel Inference: a Practical Information-Theoretic Approach (2nd edn) Vol. 20. Springer-Verlag, New York, NY, USA. Burrows MT, Schoeman DS, Buckley LB et al. (2011) The pace of shifting climate in marine and terrestrial ecosystems. Science (New York, N.Y.), 334, 652–655. Chen I-C, Hill JK, Ohlem€ uller R, Roy DB, Thomas CD (2011) Rapid range shifts of species associated with high levels of climate warming. Science (New York, N.Y.), 333, 1024–1026. Cheung WWL, Watson R, Pauly D (2013) Signature of ocean warming in global fisheries catch. Nature, 497, 365–368. Clavero M, Villero D, Brotons L (2011) Climate change or land use dynamics: do we know what climate change indicators indicate? PLoS One, 6, e18581. Colwell RK, Brehm G, Cardelus CL, Gilman AC, Longino JT (2008) Global warming, elevational range shifts, and lowland biotic attrition in the wet tropics. Science, 322, 258–261. Comte L, Grenouillet G (2015) Distribution shifts of freshwater fish under a variable climate: comparing climatic, bioclimatic and biotic velocities. Diversity and Distributions, 21, 1014–1026. Devictor V, Julliard R, Couvet D, Jiguet F (2008) Birds are tracking climate warming, but not fast enough. Proceedings of the Royal Society, 275, 2743–2748. Devictor V, Mouillot D, Meynard C, Jiguet F, Thuiller W, Mouquet N (2010) Spatial mismatch and congruence between taxonomic, phylogenetic and functional diversity: the need for integrative conservation strategies in a changing world. Ecology Letters, 13, 1030–1040. Devictor V, van Swaay C, Brereton T et al. (2012a) Differences in the climatic debts of birds and butterflies at a continental scale. Nature Climate Change, 2, 121– 124. Devictor V, van Swaay C, Brereton T et al. (2012b) Uncertainty in thermal tolerances and climatic debt. Nature Climate Change, 2, 638–639. Dobrowski SZ, Abatzoglou J, Swanson AK, Greenberg JA, Mynsberge AR, Holden ZA, Schwartz MK (2013) The climate velocity of the contiguous United States during the 20th century. Global Change Biology, 19, 241–251. Dunn PO, Møller AP (2013) Changes in breeding phenology and population size of birds. The Journal of Animal Ecology, 83, 729–739. Dunn PO, Winkler D. (2010) Effects of climate change on timing of breeding and reproductive success in birds. Effects of Climate Change on Birds, pp. 113–126. Oxford University Press, Oxford. Early R, Sax DF (2011) Analysis of climate paths reveals potential limitations on species range shifts. Ecology Letters, 14, 1125–1133. Feeley KJ, Rehm EM, Stroud J (2014) There are many barriers to species’ migrations. Frontiers of Biogeography, 6, 63–65. Gaucherel C, Burel F, Baudry J (2007) Multiscale and surface pattern analysis of the effect of landscape pattern on carabid beetles distribution. Ecological Indicators, 7, 598–609.

Gaucherel C, Alleaume S, Hely C (2008) The comparison map profile method: a strategy for multiscale comparison of quantitative and qualitative images. Transactions on Geoscience and Remote Sensing, 46, 2708–2719. Ga€ uz"ere P, Jiguet F, Devictor V (2015) Rapid adjustment of bird community compositions to local climatic variations and its functional consequences. Global Change Biology, 21, 3367–3378. Ga€ uz"ere P, Jiguet F, Devictor V (2016) Can protected areas mitigate the impacts of climate change on bird’s species and communities? Diversity and Distributions, 22, 625–637. Gillings S, Balmer DE, Fuller RJ (2015) Directionality of recent bird distribution shifts and climate change in Great Britain. Global Change Biology, 21, 2155–2168. Godet L, Jaffr#e M, Devictor V (2011) Waders in winter: long-term changes of migratory bird assemblages facing climate change. Biology Letters, 7, 714–717. Godet L, Ga€ uzere P, Jiguet F, Devictor V (2015) Dissociating several forms of commonness in birds sheds new light on biotic homogenization. Global Ecology and Biogeography, 24, 416–426. Hagemeijer W, Blair MJ (1997) The EBCC Atlas of European Breeding Birds: Their Distribution and Abundance. Poyser, London. Hickling R, Roy DB, Hill JK, Fox R, Thomas CD (2006) The distributions of a wide range of taxonomic groups are expanding polewards. Global Change Biology, 12, 450–455. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25, 1965–1978. Hill JK, Thomas CD, Huntley B (1999) Climate and habitat availability determine 20th century changes in a butterfly’s range margin. Proceedings of the Royal Society B: Biological Sciences, 266, 1197–1206. Hockey PAR, Sirami C, Ridley AR, Midgley GF, Babiker HA (2011) Interrogating recent range changes in South African birds: confounding signals from land use and climate change present a challenge for attribution. Diversity and Distributions, 17, 254–261. Hof C, Brandle M, Dehling DM, Munguia M, Brandl R, Araujo MB, Rahbek C (2012) Habitat stability affects dispersal and the ability to track climate change. Biology Letters, 8, 639–643. Hole DG, Huntley B (2011) Toward a management framework for networks of protected areas in the face of climate change. Conservation Biology, 25, 305–3015. Hole DG, Willis SG, Pain DJ et al. (2009) Projected impacts of climate change on a continent-wide protected area network. Ecology Letters, 12, 420–431. Huntley B, Collingham YC, Willis SG, Green RE (2008) Potential impacts of climatic change on European breeding birds. PLoS One, 3, e1439. Ill#an JG, Thomas CD, Jones JA, Wong W-K, Shirley SM, Betts MG (2014) Precipitation and winter temperature predict long-term range-scale abundance changes in Western North American birds. Global Change Biology, 20, 3351–3364. Jarzyna MA, Porter WF, Maurer BA, Zuckerberg B, Finley AO (2015) Landscape fragmentation affects responses of avian communities to climate change. Global Change Biology, 21, 2942–2953. Jetz W, Wilcove DS, Dobson AP (2007) Projected impacts of climate and land-use change on the global diversity of birds. PLoS Biology, 5, e157. Jiguet F, Julliard R, Thomas CD, Dehorter O, Newson SE, Couvet D (2006) Thermal range predicts bird population resilience to extreme high temperatures. Ecology Letters, 9, 1321–1330. Jiguet F, Gregory RD, Devictor V, Green RE, Vorisek P, Van Strien A, Couvet D (2010) Population trends of European common birds are predicted by characteristics of their climatic niche. Global Change Biology, 16, 497–505. Jiguet F, Devictor V, Julliard R, Couvet D (2012) French citizens monitoring ordinary birds provide tools for conservation and ecological sciences. Acta Oecologica, 44, 58–66. Julliard R, Jiguet F, Couvet D (2004) Evidence for the impact of global warming on the long-term population dynamics of common birds. Proceedings. Biological Sciences/The Royal Society, 271, S490–S492. Kampichler C, van Turnhout CAM, Devictor V, van der Jeugd HP (2012) Large-scale changes in community composition: determining land use and climate change signals. PLoS One, 7, e35272. Karl TR, Knight RW, Plummer N (1995) Trends in high-frequency climate variability in the twentieth century. Nature, 377, 217–220. Kuussaari M, Bommarco R, Heikkinen RK et al. (2009) Extinction debt: a challenge for biodiversity conservation. Trends in Ecology and Evolution, 24, 564–571. La Sorte FA, Jetz W (2012) Tracking of climatic niche boundaries under recent climate change. Journal of Animal Ecology, 81, 914–925. La Sorte FA, Thompson FR (2007) Poleward shifts in winter ranges of North American birds. Ecology, 88, 1803–1812.

© 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

L A N D S C A P E D R I V E R S O F B I R D ’ S C L I M A T I C D E B T 2229 Lavergne S, Mouquet N, Thuiller W, Ronce O (2010) Biodiversity and climate change: integrating evolutionary and ecological responses of species and communities.

Spellerberg IF (1998) Ecological effects of roads and traffic: a literature review. Global Ecology and Biogeography Letters, 7, 317.

Annual Review of Ecology, Evolution, and Systematics, 41, 321–350. Lenoir J, Svenning JC (2015) Climate-related range shifts – a global multidimensional synthesis and new research directions. Ecography, 38, 15–28. Lenoir J, G#egout JC, Marquet PA, de Ruffray P, Brisse H (2008) A significant upward shift in plant species optimum elevation during the 20th century. Science (New York, N.Y.), 320, 1768–1771. Lenoir J, Graae BJ, Aarrestad PA et al. (2013) Local temperatures inferred from plant

Stenseth NC, Mysterud A (2002) Climate, changing phenology, and other life history traits: nonlinearity and match-mismatch to the environment. Proceedings of the National Academy of Sciences of the United States of America, 99, 13379–13381. Tayleur CM, Devictor V, Gaüzère P, Jonzén N, Smith HG, Lindström Å (2016) Regional variation in climate change winners and losers highlights the rapid loss of cold-dwelling species. Diversity and Distributions, 22, 468–480.

communities suggest strong spatial buffering of climate warming across Northern Europe. Global Change Biology, 19, 1470–1481. $ Green M, Paulson G, Smith HG, Devictor V (2013) v) Rapid changes in Lindstr€ om A, bird community composition at multiple temporal and spatial scales in response to recent climate change. Ecography, 35, 001–010. Loarie S, Duffy P, Hamilton H, Asner G (2009) The velocity of climate change. Nature,

Thomas CD, Gillingham PK, Bradbury RB et al. (2012) Protected areas facilitate species’ range expansions. Proceedings of the National Academy of Sciences of the United States of America, 109, 14063–14068. Tingley MW, Koo MS, Moritz C, Rush AC, Beissinger SR (2012) The push and pull of climate change causes heterogeneous shifts in avian elevational ranges. Global Change Biology, 18, 3279–3290.

462, 1052–1055. Luoto M, Heikkinen RK (2008) Disregarding topographical heterogeneity biases species turnover assessments based on bioclimatic models. Global Change Biology, 14, 483–494. MacNally R (2002) Multiple regression and inference in ecology and conservation biology: further comments on identifying important predictor variables. Biodiversity and Conservation, 11, 1397–1401.

Van Swaay CAM, Harpke A, Van Strien A et al. (2010) The Impact of Climate Change on Butterfly Communities 1990–2009. Butterfly Conservation Europe & De Vlinderstichting, Wageningen. VanDerWal J, Murphy HT, Kutt AS, Perkins GC, Bateman BL, Perry JJ, Reside AE (2012) Focus on poleward shifts in species’ distribution underestimates the fingerprint of climate change. Nature Climate Change, 2, 1–5. Vellend M, Geber MA (2005) Connections between species diversity and genetic

Mawdsley JR, O’Malley R, Ojima DS (2009) A review of climate-change adaptation strategies for wildlife management and biodiversity conservation. Conservation Biology: The Journal of the Society for Conservation Biology, 23, 1080–1089. Nadeau CP, Fuller AK (2015) Accounting for multiple climate components when estimating climate change exposure and velocity. Methods in Ecology and Evolution, 6, 697–705.

diversity. Ecology Letters, 8, 767–781. Visser ME (2012) Biology: birds and butterflies in climatic debt. Nature Climate Change, 2, 77–78. Vos CC, Berry P, Opdam P et al. (2008) Adapting landscapes to climate change: examples of climate-proof ecosystem networks and priority adaptation zones. Journal of Applied Ecology, 45, 1722–1731.

Nieto-S#anchez S, Gutierrez D, Wilson RJ (2015) Long-term change and spatial variation in butterfly communities over an elevational gradient: driven by climate, buffered by habitat. Diversity and Distributions, 21, 950–961. Nogu#es-Bravo D, Ara# ujo MB, Errea MP, Mart#ınez-Rica JP (2007) Exposure of global mountain systems to climate warming during the 21st Century. Global Environmental Change, 17, 420–428. Norberg J, Urban MC, Vellend M, Klausmeier CA, Loeuille N (2012) Eco-evolutionary

Walther G-R, Post E, Convey P et al. (2002) Ecological responses to recent climate change. Nature, 416, 389–395. Warren M, Hill JK, Thomas J, Asher J (2001) Rapid responses of British butterflies to opposing forces of climate and habitat change. Nature, 414, 65–69. Zhu K, Woodall CW, Clark JS (2012) Failure to migrate: lack of tree range expansion in response to climate change. Global Change Biology, 18, 1042–1052. Zografou K, Kati V, Grill A, Wilson RJ, Tzirkalli E, Pamperis LN, Halley JM (2014)

responses of biodiversity to climate change. Nature Climate Change, 2, 747–751. Ohlem€ uller R (2011) Running out of climate space. Science (New York, N.Y.), 334, 613–614. Opdam P, Wascher D (2004) Climate change meets habitat fragmentation: linking landscape and biogeographical scale levels in research and conservation. Biological Conservation, 117, 285–297.

Signals of climate change in butterfly communities in a Mediterranean protected area. PLoS One, 9, e87245.

Ordonez A, Williams JW (2013) Projected climate reshuffling based on multivariate climate-availability, climate-analog, and climate-velocity analyses: implications for community disaggregation. Climatic Change, 119, 659–675. Parmesan C (2006) Ecological and evolutionary responses to recent climate change. Annual Review of Ecology, Evolution, and Systematics, 37, 637–669. Pinsky ML, Worm B, Fogarty MJ, Sarmiento JL, Levin SA (2013) Marine taxa track local climate velocities. Science (New York, N.Y.), 341, 1239–1242. Poloczanska ES, Brown CJ, Sydeman WJ et al. (2013) Global imprint of climate change on marine life. Nature Climate Change, 3, 919–925. Princ#e K, Zuckerberg B (2014) Climate change in our backyards: the reshuffling of North America’s winter bird communities. Global Change Biology, 21, 572–585. Quintana-Segu#ı P (2008) Analysis of near-surface atmospheric variables: validation of the SAFRAN analysis over France. Journal of Applied Meteorology and Climatology, 47, 92–107. R Core Team (2013) R: A Language And Environment For Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Available at: http:// www.R-project.org/ (accessed 24 September 2016). Root T, Price J, Hall K, Schneider S (2003) Fingerprints of global warming on wild animals and plants. Nature, 421, 57–60. Roth T, Plattner M, Amrhein V (2014) Plants, birds and butterflies: short-term responses of species communities to climate warming vary by taxon and with altitude. PLoS One, 9, e82490. Sandel B, Arge L, Dalsgaard B, Davies RG, Gaston KJ, Sutherland WJ, Svenning J-C (2011) The influence of Late Quaternary climate-change velocity on species endemism. Science, 334, 660–664. Savage J, Vellend M (2015) Elevational shifts, biotic homogenization and time lags in vegetation change during 40 years of climate warming. Ecography, 38, 546–555. Schippers P, Verboom J, Vos CC, Jochem R (2011) Metapopulation shift and survival of woodland birds under climate change: will species be able to track? Ecography, 34, 909–919.

© 2016 John Wiley & Sons Ltd, Global Change Biology, 23, 2218–2229

Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. Linear correlation between STI values used in the study and calculated with WORDCLIM data (STI_WORDCLIM, y axis) and STI values calculated with FBBS and SAFRAN data (STI_SAFRAN, x axis). Figure S2. Landscape descriptor correlations. Figure S3. RDS variance attributable to landscape variables as a joint or independent effect. Table S4. Analysis of sites below 800 m. Table S4a. Results of linear univariate GLS regressions of individual landscape descriptors for a dataset limited to sites below 800 m. Table S4b. Parameter estimates from the linear multivariable GLS regression including all landscape descriptors for a dataset limited to sites below 800 m. Table S4c. Model-averaged estimates of the effect of landscape predictors on RDS for a dataset limited to sites below 800 m. Figure S4. RDS variance attributable to each landscape variable considered jointly or independently for a dataset limited to sites below 800 m. Figure S5. Temporal and spatial gradient maps used to estimate biotic and climatic shifts.