Global Change Biology (2015) 21, 3367–3378, doi: 10.1111/gcb.12917

Rapid adjustment of bird community compositions to local climatic variations and its functional consequences € ER " E 1 , F R ED # ER # I C J I G U E T 2 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, Universit!e Montpellier, CNRS, IRD, Place Eug"ene Bataillon, Montpellier 34095, France, 2 Mus!eum National d’Histoire Naturelle, UMR 7204 CESCO, Centre de Recherches sur la Biologie des Populations d’Oiseaux, CP 51, 55 Rue Buffon, Paris 75005, France

Abstract The local spatial congruence between climate changes and community changes has rarely been studied over large areas. We proposed one of the first comprehensive frameworks tracking local changes in community composition related to climate changes. First, we investigated whether and how 12 years of changes in the local composition of bird communities were related to local climate variations. Then, we tested the consequences of this climate-induced adjustment of communities on Grinnellian (habitat-related) and Eltonian (function-related) homogenization. A standardized protocol monitoring spatial and temporal trends of birds over France from 2001 to 2012 was used. For each plot and each year, we used the spring temperature and the spring precipitations and calculated three indices reflecting the thermal niche, the habitat specialization, and the functional originality of the species within a community. We then used a moving-window approach to estimate the spatial distribution of the temporal trends in each of these indices and their congruency with local climatic variations. Temperature fluctuations and community dynamics were found to be highly variable in space, but their variations were finely congruent. More interestingly, the community adjustment to temperature variations was nonmonotonous. Instead, unexplained fluctuations in community composition were observed up to a certain threshold of climate change intensity, above which a change in community composition was observed. This shift corresponded to a significant decrease in the relative abundance of habitat specialists and functionally original species within communities, regardless of the direction of temperature change. The investigation of variations in climate and community responses appears to be a central step toward a better understanding of climate change effects on biodiversity. Our results suggest a fine-scale and short-term adjustment of community composition to temperature changes. Moreover, significant temperature variations seem to be responsible for both the Grinnellian and Eltonian aspects of functional homogenization. Keywords: birds, community, functional, global change, homogenization, specialist, temperature Received 13 November 2014 and accepted 8 February 2015

Introduction It is now well documented that climate change has substantial ecological and evolutionary consequences in almost all major taxonomic or functional groups (Walther et al., 2002; Parmesan & Yohe, 2003; Parmesan, 2006). However, understanding climate-induced changes in community composition and ecosystem functioning is still a major issue in global change ecology. Beyond climate change, other human-induced disturbances can affect species dynamics depending on their specific characteristics (Jiguet et al., 2007) and result in biotic homogenization of ecological communities (McKinney & Lockwood, 1999; Buisson et al., 2013). For instance, the local increase in generalist species within communities has been documented for several Correspondence: Pierre Ga€ uz"ere, tel. +33 6 30 58 89 93, fax +33 4 67 14 36 22, e-mail: [email protected]

© 2015 John Wiley & Sons Ltd

taxa and interpreted as a general response of biodiversity to global changes (Clavel et al., 2011; Le Viol et al., 2012; Monnet et al., 2014). In principle, testing whether changes in community composition are related to climate variations should rely on exploring congruency between the contrasting dynamics of community and climate. In practice, this test runs into the difficulty of obtaining fine-scale and standardized data on climate and community changes over large enough areas. Therefore, only national or regional trends in community composition are generally reported (Devictor et al., 2008b; Kampichler et al., 2012; Monnet et al., 2014). Most studies have equated the overall effect of climate change as temperature increase, and omitted its spatio-temporal variability. Yet, the processes driving climate-induced variations can be blurred by other factors such as habitat composition (Barnagaud et al., 2013) and various anthropogenic activities (Clavero et al., 2011). Moreover, the intensity 3367

€ ER " E et al. 3368 P . G A UZ and direction of temperature changes are not equally distributed over large areas. A complex lattice composed of latitude, longitude, elevation, landscape, and land use creates strong spatial variability in climate change (Loarie et al., 2009). Overall, linking the spatial distribution of both climate and community temporal variations provides a more rigorous test of community responses to climate change. Moreover, several central aspects of the effect global change on local assemblages remain unclear. First, biodiversity responses to climate change are in synergy with other environmental changes and thus difficult to disentangle of the effects of global change (Brook et al., 2008). Therefore, to what extent the local dynamics of habitat specialists vs. generalists is influenced by local climate changes remains to be tested (but see Davey et al., 2012). Second, measuring changes in the taxonomical composition of local assemblages does not reflect whether the functional composition of these assemblages is also affected. Surprisingly, while a more functional concept of community dynamics has received growing interest (McGill et al., 2006; Ackerly & Cornwell, 2007; Mouillot et al., 2013b; Monnet et al., 2014), whether species with original attributes (Pavoine et al., 2005) are also filtered by climate change is yet to be explored (but see Thuiller et al., 2014). This quest for a better understanding of the impact of global changes on community composition has rapidly led to several promising approaches. These rely on the spatial and temporal monitoring of communityweighted means (CWM), which directly measure the change in the relative abundances of species with a particular trait of interest (Lep$s et al., 2011). Using traits relating species-specific sensitivity to a given pressure, CWMs can provide indices specifically built to track changes in community compositions mostly induced by climate changes (Devictor et al., 2008b) or land-use changes (Barnagaud et al., 2011; Kampichler et al., 2012). In particular, a Species Thermal Index (STI), calculated as the average temperature of a given species distribution, has been used as a simple proxy for estimating the thermal preference of that species. Averaged for a given community, a Community Temperature Index (CTI) can then be used to reflect the relative abundance of species dependent on a warm climate (i.e., those with a high STI) vs. a cold one (those with a low STI). The CTI is expected to increase following climate warming if relative species abundances are adjusted according to their temperature requirement. It has been shown to be a relevant metric to mirror how communities respond to climate change in birds (Devictor et al., 2008b, 2012; Godet et al., 2011; Barnagaud et al., 2013), butterflies (Zografou et al., 2014), and plants (Bertrand et al., 2011). Moreover, this approach

has been valuably used at continental (Devictor et al., 2012), subcontinental (Princ#e & Zuckerberg, 2014), and local scales (Lindstr€ om et al., 2012; Roth et al., 2014). However, to what extent fine-scale variations in temperature can be related to local changes in CTI has rarely been explored (Lindstr€ om et al., 2012). Similarly, the increase in the relative proportion of generalist species has been measured with a CWM based on a Species Specialization Index (SSI; Julliard et al., 2006). The SSI is generally calculated as the coefficient of variation of a given species density across habitats to reflect how the species can thrive in different habitats (Devictor et al., 2010a). The average SSI of a given assemblage (weighted by species abundance when available) provides a Community Specialization Index (CSI) of this assemblage. Following habitat disturbance, CSI should decrease due to the relatively greater success of generalists (i.e., with a low SSI). This approach has been successfully used in birds (Le Viol et al., 2012) to track directional community changes following habitat disturbance and fragmentation (Devictor et al., 2008a) or agricultural intensification (Doxa et al., 2010; Filippi-Codaccioni et al., 2010). But habitat specialization only reflects how species respond to their environment (so-called Grinnellian specialization, Devictor et al., 2010a,b), and a change in CSI does not a priori reflect changes in what species do in their environment (so-called Eltonian specialization, Devictor et al., 2010a,b). For instance, Monnet et al. (2014) have emphasized a relative asynchrony between the dynamics of ecological specialization and functional diversity, suggesting differential responses of Grinnellian and Eltonian specialists and generalists in face of global changes. To build an Eltonian version of the CSI, Godet et al. (2014) recently proposed a CWM based on the functional originality of species (Calba et al., 2014) called Community Functional Index (CFI, Godet et al., 2014). The consequences of the local loss or gain in warmor cold-dwelling species on Eltonian and Grinellian specialization within a community have never been tested. In this context, we propose hereafter to test the spatial heterogeneity and congruency of local changes in three major aspects of bird assemblages with local changes in climate. We specifically address two main questions: • Are local temporal trends in CTI and climate related? Following our prediction that CTI dynamics are a fine-scale response of community to climatic variations, we expect both local climatic and community changes to be tightly and concomitantly structured in space. Although spring temperature appears as the key climatic pressure in shaping bird population dynamics (Jiguet et al., 2006, 2010), Ill#an et al. (2014) © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

C L I M A T E C H A N G E A F F E C T S B I R D C O M M U N I T I E S 3369 recently showed a potential effect of precipitation variations on birds abundances. We therefore tested for the effect of spring temperature and spring precipitation changes on CTI dynamics. • To what extent are changes in CTI related to Grinnellian (habitat specialization) and Eltonian (functional originality) aspects of biotic homogenization? Climate change differentially affects species dynamics depending on specific traits such as niche breath or dispersal abilities (Jiguet et al., 2007; P€ oyry et al., 2009). Beyond the adjustment of the relative abundance of cold- and hot-dwelling species, we therefore expect climate change to apply additional filters to the community assembly process. We specifically predict a decrease in the relative proportion of both habitat-specialized and functionally original species within communities resulting from local climate changes (Davey et al., 2012). To test these predictions, we used a standardized protocol monitoring the long-term (between 2001 and 2012) and large spatial trends in French breeding birds coupled with an independent meteorological database documenting temperature and precipitation variations at a high resolution. We computed plot-scale climate and community metrics and then used a moving-window approach to estimate relevant local trends of these variables. We used this framework to relate the mentioned aspects of community composition with climate change in space and time.

Materials and methods

2 km*2 km plots monitored by the French Breeding Bird Survey (see Bird data below). For each of the monitored plots and years, two climatic variables known to affect the population dynamics of breeding birds were calculated (Julliard et al., 2004; Ill# an et al., 2014): the mean spring temperature (April to August, in °C) and the mean spring precipitation (April to August, in mm per month).

Bird data Data from the French Breeding Bird Survey (FBBS) were used. This is a monitoring program in which skilled volunteer ornithologists have been counting birds following a standardized protocol at the same plot between 2001 and 2012 (Jiguet et al., 2012). Each year, species abundances were recorded in each 2 km*2 km plots whose centroids were located within a 10km radius around a locality specified by the volunteer (Fig. 1). On each plot, volunteers carried out ten point counts (5 min each, separated by at least 300 m) twice each spring within three weeks around the pivotal date of May 8th to ensure the detection of both early and late breeders. To be validated, counts must be repeated at approximately the same date between years (!7 days) and at dawn (within 1–4 h after sunrise) by a unique observer in the same order. The maximum count per point for the two spring sessions is retained as an indication of point-level species abundance. Between 2001 and 2012, more than two million individuals were counted on 2133 plots surveyed over an average of 5.7 ! 2.8 years (mean ! SD; Fig. 1). The FBBS focuses on common birds that regularly breed in France. To avoid the influence of rare species not correctly monitored by the protocol (e.g., wetland species), only the 122 most common species (representing 99% of the total abundance monitored in the database) were included in our analyses.

Climatic data Climatic data (monthly mean temperature and accumulated precipitations) were extracted from the SAFRAN meteorological model (Quintana-Segu#ı, 2008). The SAFRAN system analyzes eight climatic parameters hourly including the 2-m air temperature and rainfall on an 8 km*8 km grid. SAFRAN takes into account all of the observed data from MeteoFrance’s observation network and data from some well-instrumented stations in and around the area under study. For instance, there are more than 1000 meteorological stations for the 2-m temperature and humidity measurements, and more than 3500 daily rain gauges. This meteorological model also includes an analysis of atmospheric models (ARPEGE or ECMWF from Meteo-France) and uses it to calibrate the analysis. For each variable analyzed, an optimal interpolation method is used to assign values to given altitudes within the zone. Quintana-Segu#ı (2008) presented a detailed description and assessment of the SAFRAN analysis over France. This study confirmed the relevance of this module in weather analysis and especially for precipitation and air temperature without climate bias. These high-resolution monthly temperature and precipitation data were matched with each of the 2133

© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

Fig. 1 Spatial distribution of the plots monitored by the French Breeding Bird Survey and the spatial definitions considered. Ten point counts (1) were performed within each of the 2133 plots (2). Secondly, each plot was considered the center of an 85-km radius window (3) containing at least 20 plots. Indices calculated for the 2102 regions retained were interpolated to 10*10 km pixels (4) covering the whole country to obtain continuous maps for illustration.

€ ER " E et al. 3370 P . G A UZ

Species indices

Community-weighted mean indices

The Species Thermal Index (STI, expressed in °C; Devictor et al., 2008b) is an integrative species characteristic representing the thermal distribution of the species. It is calculated as the average temperature experienced by a species across its geographical range during the breeding season. STI values were computed from 0.5°*0.5° temperature grids (April–July averages for the period 1950–2000; Worldclim database coupled with Western Palearctic distributions of species at a 0.5° resolution from the EBCC atlas of European breeding birds (Hagemeijer & Blair, 1997). STI values are higher for species breeding in ranges with a higher temperature (e.g., southern distribution). They have already been used in several studies showing the sensitivity of warm vs. cold dwellers to climate change (Jiguet et al., 2007, 2010). The Species Specialization Index (SSI, Julliard et al., 2006) ideally represents the ecological niche breadth of a species for its habitat and is calculated as follows. During FBBS monitoring, each observer classifies each point count among 18 habitat classes. Species density in each habitat class is then calculated as the average of the species abundances (calculated over the period 2001–2012) in each class divided by the number of point counts that were monitored in this habitat during this period. For a given species, the SSI is computed as the coefficient of variation of densities of a given species across habitats (Devictor et al., 2008a,b,c). It reflects how much species are narrowly or widely distributed across habitats. The SSI has been successfully used to characterize habitat specialization in birds (Devictor et al., 2010a,b; Filippi-Codaccioni et al., 2010; Barnagaud et al., 2011). The Species Functional Index (SFI, Calba et al., 2014; Godet et al., 2014) was also calculated as a proxy for the functional originality carried by a species (Pavoine et al., 2005). It is measured from 22 functional traits (extracted from Devictor et al., 2010b) related to the life history and feeding habits of species such as the quantity and the quality of resources consumed, the feeding behavior, and the activity period (Petchey et al., 2007) and identified as major functional traits in birds (Sekercioglu, 2006). We used this species–traits matrix to compute the Gower distance matrix (Gower, 1971; Mouillot et al., 2013b) using the function ‘gowdis’ from the R package ‘FD’ without weighting any traits. The Gower distance has been widely used to estimate distances of entities characterized by a set of both quantitative and qualitative attributes (Legendre & Legendre, 2001). For a given species, SFI was defined as the average of the Gower distances between the considered species and the others. Using this approach, species with low SFI are those sharing high functional similarities with others and thus considered as less original. Note that the thermal optimum, the habitat niche breath, and the originality of functions carried by species are largely independent from each other (STI, SSI, and SFI were not or only weakly correlated, R2 from linear relationship range from 0 to 0.05, see Fig. S1). Although we based our analysis on the 122 most common species, this limited commonness range still ensures a good coverage of SFI and SSI ranges (see Fig. S1).

Within a community, an aggregated metric called the community-weighted mean (CWM) represents the expected trait value of a randomly sampled community. It is calculated for a given assemblage as the averaged trait values of the species present in this assemblage, weighted by their relative abundance. For each community monitored, the Community Temperature Index (CTI, as the CWM based on STI), the Community Specialization Index (CSI, as the CWM based on SSI), and the Community Functional Index (CFI, as the CWM based on SFI) were calculated. A given CWM was calculated as follows: PNy;s ðAbi;y;s $ SXIi Þ CXIy;s ¼ i¼1PNy;s i¼1 Abi;y;s

with CXIy,s corresponding to the CWM based on the trait X (STI, SSI, and SFI, respectively) of the N species (i) recorded at a given plot (s) for a given year (y), weighted by their respective abundances in that particular plot and year (Abi,y,s). Each of these indices is expected to increase following the local relative increase in individuals belonging to species with higher X values and/or the local relative decrease in individuals belonging to species with lower X values. To test whether potential changes in community composition were driven by changes in relative abundance only or by true changes in the local species list, each of these indices was also calculated using nonweighted community mean (i.e., by replacing all Abi, y,s by 1 in the above formula, see Fig. S2).

Data analysis Assessing national and subnational temporal trends. Generalized additive mixed models (GAMM) were used to determine the national and subnational temporal variations in CTI and spring temperature over the 12-year survey at a national scale. First, a national-scale model was built in which the response variable was the plot-level CTI (n = 2133), regressed over years expressed as a factor (n = 12, 2001–2012). To avoid any violation of the independence assumption in our model, the influence of spatial gradients among plots and temporal autocorrelation among years was taken into account. For spatial autocorrelation, the plot’s geographical coordinates were fitted using a smoothing function (2-dimensional thin plate regression splines) as a trend surface of the explained variable (Dormann et al., 2007) according to the methods of Wood (2006). To handle temporal autocorrelation between years, a temporal (1st order autoregressive function) correlation structure was added to the model’s error term correlation structure. As a certain amount of variability between plots (observers, habitat, regional species pool, and bioclimatic region) adds uncontrolled variations to this analysis, the random variation of the intercept on each plot was allowed for by adding plots as a random intercept term. This model provided the year-toyear changes in CTI at the national scale and the corresponding standard error without any violation of independence assumptions. Note also that in this model, yearly variations were estimated rather than a slope of a linear trend. Therefore, © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

C L I M A T E C H A N G E A F F E C T S B I R D C O M M U N I T I E S 3371 a simple chart of the yearly changes over time could reveal nonlinear changes in CTI. To assess subnational trends, the same model was used with subsets of our national dataset, depending on the biogeographical area of the monitored sites (Continental, Mediterranean, Alpine, Atlantic; see Fig. 2).

Assessing local temporal trends. The spatial distribution of the temporal trend of each of our climatic variables and community indices was assessed using a moving-window approach (Devictor et al., 2010a). The principle is to calculate the temporal trend of a given metric within a moving window, which is delineated by a circle around a given monitored plot and must include enough plots to allow the estimation of a temporal trend. The value of the estimated trend is then attributed to the central plot of the window. Then, the same process is repeated for all plots of the studied area. Thereby, the temporal trends of each plot are estimated with values from the neighboring plots. This approach is straightforward to summarize locally spatial or temporal trends emerging from regional dynamics (Gaucherel, 2007; Gaucherel et al., 2008). More specifically, each plot was defined as the center of an 85-km radius circle, encompassing at least 20 plots (Fig. 1). This approach provided 2102 spatial windows that were of similar spatial extent (Fig. 1). Note that the determined width and the number of plots within each window result from a compromise between a sufficiently fine spatial resolution and

the highest number of regions for the best cover of the study area. However, the window size was varied to assess whether our results were robust to variations in this parameter (available as supporting information, see Fig. S5.2). The temporal trend of each variable (spring temperature, spring precipitation, CTI, CSI, and CFI) was then estimated within each window as the linear regression of the variable considered over years as a continuous variable. Spatial autocorrelation was accounted for by performing generalized least square models in which the spatial structure (exponential semivariogram) was used to model the error term correlation structure. Note that the temporal trends were not estimated with a constant sample size. The number of plots within windows indeed varied depending on their spatial distribution. However, fixing the number of plots by a random sampling of 20 plots within each window did not qualitatively change our results (results not shown). This moving window enabled the local temporal trends of each community index to be compared with the local temporal trends of abiotic variables, such as climatic ones. Finally, local trends in each community index or climatic variable were mapped by spatial interpolation of each window value over a national 10 km*10 km grid for visual representation. We used ordinary Kriging method implemented in the ‘AUTOMAP’ package in R software. This method tested different variogram models including spherical, exponential, Gaussian, and others.

(b)

(a)

(c)

(d)

(e)

Fig. 2 Yearly estimates of Community Thermal Index (blue lines) and spring temperature (red lines) at a national scale (a, gray panel), and for Atlantic (b, green panel), Continental (c, blue panel), Alpine (d, red panel) and Mediterranean (e, purple panel) biogeographical zones. ‘Scaled estimates’ refer to scaled yearly values of Community Temperature Index (blue) or spring temperature (red) estimated by a general additive mixed model (GAMM) taking into account both spatial and temporal autocorrelations. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

€ ER " E et al. 3372 P . G A UZ The best-fit variogram model was then selected automatically for each variable and used for Kriging according to Hiemstra & Pebesma (2009).

Testing the spatio-temporal congruence of local dynamics. First, generalized additive mixed models (GAMM) were performed to test the relationship between the different local trends calculated for each window. In the first model, the local temporal trend in CTI (i.e., calculated for each window, n = 2102) was the response variable and the local temporal trend in spring temperature (also calculated for each window) was the explanatory variable. To integrate the variance associated with each local temporal trend previously estimated, the standard error value of coefficients was used as a weight in this model. Biogeographical areas were declared as random effects to account for the variability in climate between these areas. To account for structural spatial gradients, geographical coordinates were integrated in isotropic smooth terms, according to the method of Wood (2006). The second model was similar, but the explanatory variable was a smooth function of the local temporal trend in spring temperature. These two models reflected whether local changes in CTI were related to local changes in temperature, either linearly or nonlinearly. Secondly, the potential effects of climate-induced community changes were described in terms of their Grinnellian (habitat specialization) and Eltonian (functional originality) consequences. For this, whether and how changes in the local CTI were related to changes in CSI and CFI was tested. Thus, the same analysis scheme previously described was performed using the local trend in CSI (or CFI) as the response variable and a smoothed function of the absolute value of the local temporal trend in CTI as the explanatory variable. The absolute value of the CTI trend was chosen as it corresponds to the thermal community adjustment to temperature over the period considered. According to our predictions, the strength of this thermal adjustment (i.e., the magnitude of the absolute value of the temporal trend in CTI) should have consequences on the average specialization of local assemblages (CSI) and/ or more functional consequences (CFI), regardless of the direction of change (increase or decrease in CTI). To test their robustness and application for monitoring data with no quantification of local species abundances, all these analyses were also repeated with nonweighted community mean (see Fig. S2). The analysis was carried out using R statistical software (R Core Team, 2013) and the following packages: NLME, MGCV, AUTOMAP, FD.

Results

Spatial distribution of temporal trends in temperature and CTI At the national scale (Fig. 2a), there was no consistent linear trend in either temperature or CTI over the period considered (2001–2012). Instead, the temporal trends in temperature and CTI showed high year-to-year

variations, with a slight increase until 2006, turning into a steady to negative trend from 2007 to 2012. However, these national trends masked large spatial heterogeneities. First, separating these national trends between biogeographical zones revealed marked and contrasting qualitative trends. For instance, temperature and CTI increased steeply in Alpine areas (Fig. 2d), while these trends were negative in the Mediterranean region (Fig. 2e). In contrast, no particular directional trend occurred in the Atlantic (Fig. 2b) and Continental (Fig. 2c) zones. More generally, a systematic and continuous spatial distribution of change in temperature and CTI showed high local variations in these trends (Fig. 3). Therefore, the nonlinear national trends actually masked important differences between subnational dynamics. Moreover, the relative thermal composition of the bird community measured by the CTI closely matched the temperature variations in both space and time. Our moving-window analysis revealed that local trends in mean spring temperature and CTI were significantly related (Fig. 4) by a positive linear relationship (t = 6.24, res.df = 2088, P < 0.001, fixed effect marginal R² = 0.29). Considering a nonlinear relationship substantially increased the model fit (F2.97, 2086 = 892.9, P < 0.001, fixed effect marginal R² = 0.41). The close match between local changes in temperature and in CTI was largely driven by windows that had undergone large changes in temperature (either positive or negative). In other words, weak temporal changes in spring temperature did not trigger substantial changes in CTI below a marked threshold (around !0.05 °C. yr&1), but a directional change in CTI was observed beyond this threshold (Fig. 4).

Spatial distribution of temporal trends in precipitation and CTI In both their temporal changes and spatial distribution, mean spring precipitation and temperature were correlated (Pearson’s correlation = &0.47, t = &24.8, df = 2117, P < 0.001, see Fig. S3.1). Based on Akaike’s information criterion (AIC; Burnham & Anderson, 2002), the best fixed effect explaining CTI changes included both temperature and precipitation changes, but temperature remained the best climatic variable when tested alone (see Table S3.1).

Effect of community thermal adjustment on functional homogenization Temporal trends in the Community Specialization Index (CSI) and Community Functional Index (CFI) were both related to change in CTI (CSI: F33, 2069 = 17.7, © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

C L I M A T E C H A N G E A F F E C T S B I R D C O M M U N I T I E S 3373

Local community temperature index trend

Fig. 3 Spatial interpolations of regional trends in spring temperature (left panel) and Community Temperature Index (right panel). Each point represents the center of a region used for interpolation. The brighter the color, the higher the value of the trend, and conversely.

0.05

0.00

−0.05 −0.2

0.0

0.2

0.4

Local spring temperature trend (C.year−1)

Fig. 4 Relationships between regional trends (calculated in each window of 85-km radius composed of 20 plots, n = 2102) in Community Temperature Index (CTI, Y-axis) and spring temperature (X-axis). Points represent estimates of slope coefficient, horizontal and vertical lines represent standard error of temperature and CTI slope coefficients, respectively.

P < 0.001, R² = 0.12; CFI: F3, 2069 = 24.7, P < 0.001, R² = 0.09, Fig. 5). In other words, communities with the strongest response to temperature change were those with a decreasing average specialization (decrease in CSI). Similarly, large changes in CTI (> !0.01 °C yr&1) were related to marked changes in the average functional originality (decrease in CFI). Interestingly, 1560 windows exhibited local increase in CFI where only 448 windows exhibited local increase in CSI. These observations argue for an overall positive trend in CFI and an overall negative trend in CSI. We tested for these trends at national scale by performing linear mixed models (LMM) over the 12-year survey. The response variable © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

was successively the plot-level scaled value of CSI or CFI (n = 2133) regressed over years expressed as a continuous variable (n = 12, 2001–2012) and plot as a random intercept term. These models confirmed a significant decrease in CSI (&0.009 ! 0.001, df = 7953, P-value 0.05). Temporal changes in the CSI and CFI were also directly and negatively related (although more weakly than above, where CTI was considered the explanatory variable) to absolute changes in spring temperature (CSI: F3, 2071 = 7.34, P < 0.001, R² = 0.09; CFI: F3, 2071 = 13.35, P < 0.001, R² = 0.05). These results suggest that temperature changes affect the ecological and functional composition of bird communities, but using the CTI directly as a signature of climate-induced community response exacerbated these functional responses even further. Moreover, these changes thus correspond to a nonrandom reconfiguration in assemblage compositions: habitat generalists and species with less original functional traits are those that seem to benefit from the marked thermal adjustment of local communities. Note, however, that our results mostly resulted from a limited number of windows experiencing both strong decrease in CTI and strong decreases in CSI or CFI (Fig. 5). This suggests that few areas are in fact exposed to sufficient changes to filter out habitat specialists and functionally original species following climate change. Our results were qualitatively similar to the analysis based on presence–absence data (see Fig. S2). Note also that varying the size of the moving window did not generally change these results (Fig. S5.2). Increasing region size mostly strengthened the relationships described above, except for that between the thermal

€ ER " E et al. 3374 P . G A UZ (b)

(a)

Fig. 5 Relationships between regional trends (calculated in each window of 85-km radius composed of 20 plots) in Community Specialization Index (a) or Community Functional Index (b) and absolute regional trends of Community Temperature Index. Points represent estimates of slope coefficients, horizontal lines represent standard errors of CSI (a) or CFI (b) slope coefficient, and vertical lines represent standard errors of CTI slope coefficient.

adjustment of assemblages (absolute trend in CTI) and CFI, which tended to be weaker or even disappear when moving windows of 125 km or more were used (see Fig. S5.2, panel 2).

Discussion In this study, we addressed two central, albeit still unresolved, aspects of community responses to climate changes. First, we assessed whether trends in the thermal composition of communities and two local climatic variables (temperature and precipitation) were congruent in space and time. Secondly, we tested the consequences of climatic variations and their induced community reshuffling (CTI changes) on average Grinnellian specialization (CSI) and Eltonian specialization (CFI) dynamics.

Using a moving-window approach, we examined the fine-scale variations in the effect of climatic changes on species assemblages. We first showed that the absence of a clear linear trend measured at the national scale masked significant spatial heterogeneity at more local scales. This result is in line with previous studies arguing for the importance of spatial heterogeneity in environmental studies conducted at large spatial scales (Forman, 1995; Gaucherel, 2007). The processes shaping community characteristics and ecosystem properties are likely to be better explained by these spatial variations than by global averaged trends (Gaucherel et al., 2007). We further described a substantial congruence in space and time between changes in spring temperature and the thermal composition of bird communities during this period. Previous studies exploring CTI temporal trends at a national scale (e.g., Devictor et al., 2008b, 2012; Godet et al., 2011) have already reported such patterns within the context of a linear increase in temperature over a given period. Lindstr€ om et al. (2012) further showed that Swedish bird communities responded to summer temperature with a lag of 1 to 3 years, suggesting a rapid adjustment of community composition to climatic variations. However, to our knowledge, a more systematic test of the local adjustment of CTI to local changes in temperature has never been reported. Here, we clearly showed that the trends in CTI and temperature were not randomly distributed. For instance, in the Mediterranean communities, the recent decrease in temperature led to a particularly rapid and localized response of communities with an increase in the relative abundance of cold dwellers (as shown by a rapid decrease in CTI). Less than 100 kilometers away from this area (in the Alpine zone), we observed both a strong increase in temperature and a corresponding increase in CTI. These results were also true when calculated with presence–absence data. This suggests that local spring temperature is probably a key pressure modulating the short-term reproductive success and between-year survival in birds (Julliard et al., 2004; Jiguet et al., 2006). To our knowledge, our study is the first to describe a clear local match between community composition and temperature fluctuation at a local scale using extensive national monitoring data. Obviously, interpreting correlations between temporal trends of indices requires caution as marked temporal trends can be correlated without any causal links. However, our results reveal contrasting situations with increasing, stable, and decreasing temperatures with corresponding trends in CTI. These close matches in the spatio-temporal variations in temperature and CTI clearly confirm that the CTI dynamic reflects changes in community composition in response to temperature changes. These © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

C L I M A T E C H A N G E A F F E C T S B I R D C O M M U N I T I E S 3375 results are in line with numerous studies documenting fine temporal responses of bird communities to climate variations (Godet et al., 2011; Lindstr€ om et al., 2012; Princ#e & Zuckerberg, 2014). Among the possible processes explaining these responses, the short-term effect of change in local temperature on life-history traits should be investigated. Previous works have shown that climate variations a given year (t) affect adult survival, breeding dispersal, and breeding success of individuals at year t. At year t + 1, the effect of climate on natal dispersal and juvenile survival could also affect the local species abundances (Julliard et al., 2004; Jiguet et al., 2006). To go further in identifying the processes governing variations in CTI in response to climate changes, a finer time series analysis relating life-history traits and climatic variations could be conducted. More interestingly, we found that changes in CTI were nonlinear, suggesting that the smallest temperature changes do not affect changes in assemblage composition. Mitigation of the timing of the breeding season and a change in migration strategy could be partly responsible for such a buffered response when temperature changes are moderate (Stenseth & Mysterud, 2002; Dunn & Winkler, 2010). When studying broad-scale temporal variations in CTI, Devictor et al. (2008b, 2012) documented a slower rate of change in CTI from what was expected given the observed change in temperature. This difference was identified as a possible ‘climatic debt’ resulting from an insufficient change in community composition to keep up with temperature increase. However, whether this debt resulted from a true delay in species dynamics or simply a lack of sensitivity of most species to changes in temperature was not clear (Rodr#ıguez-S# anchez, 2012,). Our results enable us to refine these findings. Within a range of temperature change, communities seem to be unaffected by climate change. A directional shift in community composition is observed, however, above a certain intensity of climatic variation. This threshold effect could be responsible for the apparent climatic debt observed at a national scale. Whether this abrupt response results from the limit of phenotypic and phenological plasticity reached for most species needs to be confirmed. Beyond the spatio-temporal adjustment of community composition to thermal fluctuations, the consequences of a climate-induced community reshuffling have not been explored. Climate change could be expected to apply other filters to the community assembly process than the adjustment of the relative abundance of cold and hot dwellers (Davey et al., 2012). A large number of empirical studies have reported the impacts of climate on community diversity (Klanderud © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

& Birks, 2003; Men#endez et al., 2006; Britton et al., 2009). Regarding terrestrial vertebrates, a sparser literature has documented impacts of climate change on community composition (La Sorte, 2006; Lemoine et al., 2007). Moreover, only a few studies have explored the specific impact of climate change on the potential directional homogenization of assemblages (Davey et al., 2012; Buisson et al., 2013). Changes in community composition toward enrichment in species with a specific ecological strategy (specialist vs. generalist) or functional traits have generally been treated as a response to specific land-use changes. For instance, habitat artificialization has been documented as an important driver of biotic homogenization (Olden, 2006). Habitat specialists have also been shown to be more sensitive to agricultural intensification (Filippi-Codaccioni et al., 2010). Nevertheless, separating the effects of climate-induced vs. habitatinduced changes in such observed shifts in community composition remains tricky. In fact, climate and habitat changes can be combined within a given biogeographical region. For instance, Clavero et al. (2011) have shown that changes in habitat structure and composition could benefit species with high or low thermic preferences, in a scenario of constant temperature. Other studies have emphasized the effect of habitat on shifts in species distribution ranges in response to climate change (Warren et al., 2001). More generally, Barnagaud et al. (2012) have shown that habitat and thermal niches could be intrinsically related, casting some doubt on interpreting community changes only in terms of a response to land-use or climate changes. Accordingly, the development of our work lies in the quantification of joint or synergistic effects of climate and land-use changes, providing that high-enough resolution data on habitat dynamics are available. However, our study also suggests that large changes in CTI are linked to a decrease in CSI and CFI. Climateinduced community reorganization seems to be related to both a relative enrichment of habitat-generalist species as well as a depletion of original functions within communities. These results mirror the interaction between several aspects of the biotic homogenization process. Environmental variations are filtering species at the expense of specialists (Grinnellian homogenization), narrowing the available range of species-specific responses by the loss of species functions (Eltonian homogenization). Such homogenized communities could, in turn, jeopardize the ability of ecosystems to cope with future disturbances at a regional level (Olden, 2006). Despite strong evidences at national scale (Godet et al., 2014), our results also show that locally, the communities experiencing slight changes in CTI were almost all showing an increase in functional

€ ER " E et al. 3376 P . G A UZ originality coupled with a decrease in habitat specialization. Monnet et al. (2014) showed similar results describing a national scale increase of functional c- and b-diversity over the last ten years, while community specialization steeply decreased during the same period. Our results suggest that climate change could be partly responsible for both Grinellian and Eltonian homogenization. Further investigations are, however, needed to assess the strength of this driver and whether it is also acting at larger temporal and spatial scales (Thuiller et al., 2014). The true effect of these changes on ecosystem functioning is also difficult to quantify. Several studies have emphasized the potential impact of the loss of functional diversity on stability and resistance to perturbations (Tilman et al., 2006; Gonzalez & Loreau, 2009). More recently, other works have shown that habitat specialists and rare species tend to harbor more original and vulnerable functions (Mouillot et al., 2013a; Calba et al., 2014). Our study, however, was limited to a relatively common species pool. Following Mouillot et al. (2013a), one could expect even starker responses of CSI and CFI by taking into account rarer species. Overall, similar studies conducted on other taxonomic groups, degrees of rarity, and/or coupled with measurements of ecosystem functioning would be a promising research agenda. A more practical perspective would be to test whether the changes in each of these facets of communities are similarly affected in different habitats of conservation interest (e.g., protected or not, fragmented or not). Supplementary analysis showed that the effect of climate change on Eltonian homogenization was not consistent over increasing scales of analysis. Indeed, the substantial relation between CTI and CFI changes at local scale tends to disappear when analyzed at larger scales (>125 km radius windows, an area equal to onetenth of the total area of France). This pattern could probably result from the absence of directional trend in CFI at largest scale (see national scale result above). However, this result pleads for the need in a systematic investigation of multiscale patterns in global change ecology (Dray et al., 2012). To conclude, our work shows that global changes must be investigated at several spatio-temporal scales and that even mid-term climatic variations, providing that they are large enough, can cause community composition to become more simplified. Note that while our results still rest on averaged climatic and species characteristics, we show that any strong change in CTI (positive or negative) correlates with decreased CSI and CFI. Therefore, regions with no climate trend over the years (i.e., varying CTI) can still result in ever decreasing CSI and CFI. In other words, climatic variations per se (rather than directional trends) is likely to also act as

an environmental filter for habitat specialists and functionally original species. Moreover, many studies have demonstrated that climate is characterized by an increase in extreme events (Min et al., 2011; Rahmstorf & Coumou, 2011). Therefore, it would be relevant to focus on the consequences of temporal variance of climate on biodiversity, and not only on the dynamics of temperature averages. Our results also showed that both spring temperature and precipitation changes were included in the most efficient model to predict community temperature changes. In agreement with other studies (Dunn & Winkler, 2010; Ill#an et al., 2014), this argues for the integration of several climatic variables to predict or project the upcoming effects of climate change on communities. Overall, this study shows that climate change biology can benefit from a more functional approach in biogeography (Violle et al., 2014) whereby integrating changes in functional composition within and among communities can enhance our capacity to describe and predict biodiversity dynamics.

Acknowledgements We sincerely 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. We also thank three anonymous referees for providing valuable comments to improve an early version of this paper, and Carol Robins for English editing. This is publication ISEM 2015-040.

References Ackerly D, Cornwell W (2007) A trait based approach to community assembly: partitioning of species trait values into within-and among-community components. Ecology Letters, 10, 135–145. Barnagaud JY, Devictor V, Jiguet F, Archaux F (2011) When species become generalists: on-going large-scale changes in bird habitat specialization. Global Ecology and Biogeography, 20, 630–640. Barnagaud JY, Devictor V, Jiguet F, Barbet-Massin M, Le Viol I, Archaux F (2012) Relating habitat and climatic niches in birds. PLoS ONE, 7, e32819. Barnagaud JY, Barbaro L, Hampe A, Jiguet F, Archaux F (2013) Species’ thermal preferences affect forest bird communities along landscape and local scale habitat gradients. Ecography, 36, 001–009. 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. Britton AJ, Beale CM, Towers W, Hewison RL (2009) Biodiversity gains and losses: evidence for homogenisation of Scottish alpine vegetation. Biological Conservation, 142, 1728–1739. Brook BW, Sodhi NS, Bradshaw CJA (2008) Synergies among extinction drivers under global change. Trends in Ecology and Evolution, 23, 453–460. Buisson L, Grenouillet G, Vill#eger S, Canal J, Laffaille P (2013) Toward a loss of functional diversity in stream fish assemblages under climate change. Global Change Biology, 19, 387–400. Burnham KKP, Anderson DRD (2002) Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach (2nd edn). Springer, New-York. Calba S, Maris V, Devictor V (2014) Measuring and explaining large-scale distribution of functional and phylogenetic diversity in birds: separating ecological drivers from methodological choices. Global Ecology and Biogeography, 23, 669–678. Clavel J, Julliard R, Devictor V (2011) Worldwide decline of specialist species: toward a global functional homogenization? Frontiers in Ecology and the Environment, 9, 222–228.

© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

C L I M A T E C H A N G E A F F E C T S B I R D C O M M U N I T I E S 3377 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.

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 characteris-

Davey CM, Chamberlain DE, Newson SE, Noble DG, Johnston A (2012) Rise of the generalists: evidence for climate driven homogenization in avian communities. Global Ecology and Biogeography, 21, 568–578. Devictor V, Julliard R, Jiguet F (2008a) Distribution of specialist and generalist species along spatial gradients of habitat disturbance and fragmentation. Oikos, 117, 507– 514. Devictor V, Julliard R, Clavel J, Jiguet F, Lee A, Couvet D (2008b) Functional biotic

tics 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(Suppl), S490–S492.

homogenization of bird communities in disturbed landscapes. Global Ecology and Biogeography, 17, 252–261. Devictor V, Julliard R, Couvet D, Jiguet F (2008c) Birds are tracking climate warming, but not fast enough. Proceedings of the Royal Society, 275, 2743–2748. Devictor V, Clavel J, Julliard R et al. (2010a) Defining and measuring ecological specialization. Journal of Applied Ecology, 47, 15–25.

Julliard R, Clavel J, Devictor V, Jiguet F, Couvet D (2006) Spatial segregation of specialists and generalists in bird communities. Ecology Letters, 9, 1237–1244. 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. Klanderud K, Birks H (2003) Recent increases in species richness and shifts in altitu-

Devictor V, Mouillot D, Meynard C, Jiguet F, Thuiller W, Mouquet N (2010b) 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. (2012) Differences in the climatic debts of birds and butterflies at a continental scale. Nature Climate Change, 2, 121–124.

dinal distributions of Norwegian mountain plants. The Holocene, 13, 1–6. La Sorte FA (2006) Geographical expansion and increased prevalence of common species in avian assemblages: implications for large-scale patterns of species richness. Journal of Biogeography, 33, 1183–1191. Le Viol I, Jiguet F, Brotons L et al. (2012) More and more generalists: two decades of changes in the European avifauna. Biology Letters, 8, 780–782. Legendre P, Legendre L (2001) Numerical Ecology (2nd edn). Elsevier, Amsterdam, the

Dormann C, McPherson JM, Ara# ujo BM et al. (2007) Methods to account for spatial autocorrelation in the analysis of species distributional data: a review. Ecography, 30, 609–628. Doxa A, Bas Y, Paracchini ML, Pointereau P, Terres J-M, Jiguet F (2010) Low-intensity agriculture increases farmland bird abundances in France. Journal of Applied Ecology, 47, 1348–1356.

Netherlands. Lemoine N, Bauer H-G, Peintinger M, B€ ohning-Gaese K (2007) Effects of climate and land-use change on species abundance in a Central European bird community. Conservation Biology, 21, 495–503. $ P, Dole$zal J (2011) Community trait response to environLep$s J, de Bello F, Smilauer ment: disentangling species turnover vs. intraspecific trait variability effects. Ecog-

Dray S, P#elissier R, Couteron P (2012) Community ecology in the age of multivariate multiscale spatial analysis. Ecological Monographs, 82, 257–275. Dunn POP, 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. Filippi-Codaccioni O, Devictor V, Bas Y, Clobert J, Julliard R (2010) Specialist response to proportion of arable land and pesticide input in agricultural land-

raphy, 34, 856–863. % Green M, Paulson G, Smith HG, Devictor V (2012) 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, 462, 1052–1055. McGill BJ, Enquist B, Weiher E, Westoby M (2006) Rebuilding community ecology

scapes. Biological Conservation, 143, 883–890. Forman R (1995) Some general principles of landscape and regional ecology. Landscape Ecology, 142, 133–142. Gaucherel C (2007) Multiscale heterogeneity map and associated scaling profile for landscape analysis. Landscape and Urban Planning, 82, 95–102. Gaucherel C, Burel F, Baudry J (2007) Multiscale and surface pattern analysis of the

from functional traits. Trends in Ecology and Evolution, 21, 178–185. McKinney M, Lockwood J (1999) Biotic homogenization: a few winners replacing many losers in the next mass extinction. Trends in Ecology and Evolution, 14, 450– 453. Men#endez R, Meg#ıas AAG, Hill JK et al. (2006) Species richness changes lag behind climate change. Proceedings of the Royal Society, 273, 1465–1470.

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. 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.

Min S-K, Zhang X, Zwiers FW, Hegerl GC (2011) Human contribution to moreintense precipitation extremes. Nature, 470, 378–381. Monnet A-C, Jiguet F, Meynard CN, Mouillot D, Mouquet N, Thuiller W, Devictor V (2014) Asynchrony of taxonomic, functional and phylogenetic diversity in birds. Global Ecology and Biogeography, 23, 780–788. Mouillot D, Bellwood DR, Baraloto C et al. (2013a) Rare species support vulnerable functions in high-diversity ecosystems. PLoS biology, 11, e1001569.

Godet L, Ga€ uzere P, Jiguet F, Devictor V (2014) Dissociating several forms of commonness in birds sheds new light on biotic homogenization. Global Ecology and Biogeography, 24, 416–426. Gonzalez A, Loreau M (2009) The causes and consequences of compensatory dynamics in ecological communities. Annual Review of Ecology, Evolution, and Systematics, 40, 393–414.

Mouillot D, Graham NAJ, Vill#eger S, Mason NWH, Bellwood DR (2013b) A functional approach reveals community responses to disturbances. Trends in Ecology and Evolution, 28, 167–177. Olden JD (2006) Biotic homogenization: a new research agenda for conservation biogeography. Journal of Biogeography, 33, 2027–2039. Parmesan C (2006) Ecological and evolutionary responses to recent climate change.

Gower J (1971) A general coefficient of similarity and some of its properties. Biometrics, 27, 857–871. Hagemeijer W, Blair MJ (1997) The EBCC atlas of European breeding birds: their distribution and abundance. Hiemstra P, Pebesma E (2009) Real-time automatic interpolation of ambient gamma dose rates from the Dutch radioactivity monitoring network. Computers and Geosciences, 35, 1711–1721.

Annual Review of Ecology, Evolution, and Systematics, 37, 637–669. Parmesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts across natural systems. Nature, 421, 37–42. Pavoine S, Ollier S, Dufour A-B (2005) Is the originality of a species measurable? Ecology Letters, 8, 579–586. Petchey OL, Evans KL, Fishburn IS, Gaston KJ (2007) Low functional diversity and no redundancy in British avian assemblages. The Journal of Animal Ecology, 76, 977–

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. Jiguet F, Julliard R, Thomas CD, Dehorter O, Newson SE, Couvet D (2006) Thermal range predicts bird population resilience to extreme high temperatures. Ecology

985. P€ oyry J, Luoto M, Heikkinen RK, Kuussaari M, Saarinen K (2009) Species traits explain recent range shifts of Finnish butterflies. Global Change Biology, 15, 732– 743. 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–

Letters, 9, 1321–1330. Jiguet F, Gadot A-S, Julliard R, Newson SE, Couvet D (2007) Climate envelope, life history traits and the resilience of birds facing global change. Global Change Biology, 13, 1672–1684.

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.

© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378

€ ER " E et al. 3378 P . G A UZ R Core Team (2013) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.

Wood SN (2006) Generalized Additive Models: An Introduction with R. Chapman and Hall/CRC, Boca Raton, FL.

Rahmstorf S, Coumou D (2011) Increase of extreme events in a warming world. Proceedings of the National Academy of Sciences of the United States of America, 108, 17905–17909. Rodr#ıguez-S#anchez F (2012) Uncertainty in thermal tolerances and climatic debt. Nature Climate Change, 2, 636–637. 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 alti-

Zografou K, Kati V, Grill A, Wilson RJ, Tzirkalli E, Pamperis LN, Halley JM (2014) Signals of climate change in butterfly communities in a Mediterranean protected area. PLoS ONE, 9, e87245.

tude. PLoS ONE, 9, e82490. Sekercioglu CH (2006) Increasing awareness of avian ecological function. Trends in Ecology and Evolution, 21, 464–471. 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. Thuiller W, Pironon S, Psomas A et al. (2014) The European functional tree of bird life in the face of global change. Nature Communications, 5, 3118. Tilman D, Reich PB, Knops JMH (2006) Biodiversity and ecosystem stability in a decade-long grassland experiment. Nature, 441, 629–632. Violle C, Reich PB, Pacala SW, Enquist BJ, Kattge J (2014) The emergence and promise of functional biogeography. Proceedings of the National Academy of Sciences, 111, 13690–13696. Walther G-R, Post E, Convey P et al. (2002) Ecological responses to recent climate change. Nature, 416, 389–395. Warren M, Hill J, Thomas J, Asher J (2001) Rapid responses of British butterflies to opposing forces of climate and habitat change. Nature, 414, 65–69.

Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. Pair correlations and limitation of range in species indices for the 122 species studied. Figure S2. Analysis based on presence–absence data. Figure S3. Analysis of the effect of precipitation on CTI trend. Figure S4. Effect of Positive and negative CTI changes in CFI and CSI changes. Figure S5. Analysis of the effect of the windows radius sizes on results.

© 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 3367–3378