Climate change is driving community composition rearrangements and biodiversity losses worldwide, currently representing a crucial theme in conservation biology (e.g., [1
]). Understanding and predicting global warming impacts on biodiversity is essential to develop effective conservation strategies (e.g., [4
Not all ecosystems display the same degree of exposure, sensitivity, and vulnerability [3
]. Mountain ecosystems host a high level of biodiversity and are especially threatened (e.g., [7
]), owing to the presence of species with small distribution areas, low dispersal ability, and high levels of physiological and ecological specialization [8
]. In the European Alps, a large number of endemic species is present, and during the last decades, many habitat types and animal and plant populations have undergone exceptional decline [11
]. Many mountain ecosystems reported warming higher than the global average [14
]: in the Italian Alps, the measured air temperature rise was of the order of 0.2–0.4 °C per decade since 1950 [18
]. For the Alps, climate projections suggest future warming of 0.5–0.7 °C per decade [20
], again higher than the global average [21
Alpine biodiversity has already responded to the temperature rise. For alpine flora, common effects are upward displacements of high altitude plants, treeline advance, the decline of arctic-alpine species, changes in community composition, and species richness (e.g., [11
]). For alpine fauna, similar responses have been measured, such as upward shifts of single species and changes at assemblage level [26
], but the number of studies is markedly lower.
Individual species are expected to respond differently to climate warming, and the most important effects will presumably be at the community level [1
]. Even though the temperature rise has been reported to impact community composition, the number of long-term standardized datasets with a sufficiently high temporal and spatial resolution is still limited, in particular for multi-taxa analyses [31
Understanding how multiple species and communities will respond to climate change is essential for management, from global to local scale [34
]. Indeed, the effectiveness of protected areas and of the Natura 2000 network under climate change has been questioned (e.g., [3
]). It has been argued that nature reserves should improve their role from pure conservation to the stewardship of the changes, representing key sites for the study of climate change responses [38
]. Conservation actions should consequently include the management of “smoother transitions” to new conditions, which necessarily requires the understanding of current dynamics and future states [39
Species distribution models (SDM) are simple tools to estimate taxa response to climate warming, assessing the potential vulnerability of individual species and communities as a whole [40
]. SDM could have large uncertainties in their outcomes, as they use a correlative approach based on the observed, realized species niche. SDM usually assume niche conservatism and the absence of evolutionary processes, ignoring biological interactions and physiological mechanisms [44
]. More sophisticated, mechanistic models are currently available, but they have high data demands and thus cannot be used for any of the studied taxa, owing to the lack of data (e.g., [46
]). Consequently, simple SDM are often considered as the main tool to estimate species and community vulnerability to climate change [40
] and fill knowledge gaps in understudied taxa and habitats (e.g., invertebrates, [31
]; altitudinal gradients and mountain ecosystems, [50
]). SDM allows us to explore different scenarios and quantify biodiversity changes for a wide range of species at the same time, obtaining useful indications for protected areas’ management [5
To obtain quantitative information on biodiversity along altitudinal gradients on spatial and temporal scales suitable for site management, a multi-taxa monitoring project was started in 2007 in the northwestern Italian Alps [52
]. Five taxa (Coleoptera Carabidae, Coleoptera Staphylinidae, Araneae, Lepidoptera Papilionoidea and Hesperiioidea, Aves) representing a wide array of ecological and evolutionary traits, have been monitored in three separated mountain protected areas. Climatic and vegetation data were collected on the same spatial scales. The results of the analyses have shown that invertebrate species’ richness and community composition are related to temperature and altitude, highlighting the potential vulnerability of the alpine communities to climate change [52
The present work uses the above dataset to assess the risk of short-term changes in Alpine animal biodiversity under climate warming. The coarse spatial resolution of current (global and regional) climate projections, however, does not allow to assess the response of mountain ecosystems, that are characterized by high topographic heterogeneity and small-scale microhabitat and microclimate variability [54
]. Short-term, fine-grained climate predictions would certainly be needed by site managers [56
], but there is still a generalized paucity of climate information for mountain areas [57
On the other hand, it is important to provide informative projections, even within simplified modeling exercises, to investigate the relationships between land-use and climate change [58
]. Our study, based on local-scale biodiversity and environmental data, is a contribution to address the challenge of estimating the response of multi-taxa distributions. We applied “what if” scenarios, increasing the maximum, minimum, and mean temperatures with respect to the increase observed in the Alps during the last decades. To assess the potential role of land cover, we compared projections with and without vegetation constraints.
Many works indicate differences in the ability of different taxa to adapt to climate change, but most of the studies are based on meta-analysis or data coming from independent studies [33
]. Our approach compared data coming from different taxa collected with the same methodologies and in the same monitoring framework. Our main objective was thus to estimate patterns of congruence in multi-taxa response to climate warming. In particular, we asked whether: (i) the number of species changing their distribution differs among taxa, degree of vulnerability and scenarios and (ii) species richness and community composition significantly change along the altitudinal gradient. This allowed us to obtain indications on both indicator selection and habitat protection priorities.
3.1. Species Distribution
The comparison between the results provided by each scenario and the corresponding baseline showed high variability in the species response to the temperature increase.
The majority of species showed no variation (42.1 ± 1.0%), some others displayed an increase (30.7 ± 0.7%), and others a decrease in the number of occupied plots (27.1 ± 0.9%). Considering only endemic and vulnerable species, the percentage of species with decreasing occupancy was higher (39.8 ± 1.1% for endemics; 57.4 ± 2.5% for vulnerable species) and in the case of the vulnerable species, the percentage of increasing species also became lower (7.5 ± 0.7%). Considering only varying species
, both the endemic and the vulnerable species showed a higher percentage of decreasing occupancies and a lower percentage of increasing species, when compared to the whole set of species (Figure 2
For the species that showed a variation, the amount of change was usually low (1st quartile = −4, median = 1, 3rd quartile = +4), even if the values ranged between −28 and +21 plots.
Comparing the three model classes, we found that the number of species with changing distributions was lower with a larger number of environmental constraints. Indeed, varying species in each scenario significantly differed between model classes (TRV = 25.7 ± 4.4, TR = 43.7 ± 4.4, T = 67.7 ± 8.4; 20.64 < χ2 < 27.82, p < 0.001).
Comparing the three scenarios, we observed a higher number of varying species in the Min scenario (Max = 37.3 ± 10.7, Min = 56.7 ± 14.6, D = 43.0 ± 11.3) in each model class, but the differences were significant only in the T model class (χ2 = 8.56, p = 0.02).
3.2. Species Richness
Changes in species richness showed considerable variations across taxonomic groups, warming scenarios, and vegetation belts, even if the amount of change was generally low, as exemplified by the effect size, ranging from −0.4 to 0.3 (Figure 3
Butterflies displayed an increase in species richness per plot for almost all models and warming scenarios (Figure 3
e); the impact of temperature increase is significantly stronger on this taxon than for all taxa pooled together (Table 1
In the case of spiders, the results showed the opposite trend, with a decrease in species richness in all cases (Figure 3
d) and significantly lower effect size than for all taxa pooled together (Table 1
). Carabids showed no significant differences in any of the scenarios (Figure 3
b). Staphylinids (Figure 3
c) and birds (Figure 3
f) showed slightly different results when considering different model classes.
The overall species richness showed a significant increase considering only the TRV model class.
For all taxa except carabids, we observed significant differences in the species richness changes predicted by the different model classes (Table 2
). For butterflies, T models indicate an increase in species richness, which is less pronounced for TR and TRV models. For spiders, T models indicate a more pronounced decrease in species richness than TRV models. For all taxa pooled together, staphylinids and birds, TRV models indicate a significantly larger effect than T and TR models.
The analysis of changes in species richness along the altitudinal gradient indicated a more pronounced effect at higher altitude, in particular above 2000 m (Figure 4
). In the Alpine belt, we observed a significant effect for all taxa pooled together, for butterflies, for birds and for spiders (Table 3
). In the case of spiders, this implies that the decrease in species richness observed in the montane and the subalpine belt is less pronounced at higher altitudes (Figure 4
; Table 3
3.3. Community Composition
In almost all cases, warming scenarios determined significant changes in community composition along the first CA axis. The Wilcoxon test was applied to assess changes in plot scores from current to warmer conditions; it is significant in 53 over 54 cases, p < 0.01, the only exception are staphylinids in the TRV model class, max scenario. The pattern along the second axis is less coherent (Wilcoxon test with p < 0.01 in 35 cases over 54).
For both the current conditions and the warming scenarios, plot community composition significantly changed from the montane to the alpine belt (Figure 5
a; current, KW Test, df
= 2, all p
< 0.0001, Figure 5
b; projected, KW Test, df
= 2, all p
< 0.0001; Figure 5
Plot scores were significantly correlated in all cases (Spearman’s rank correlation; first axis, 0.978 < ρ < 0.998; second axis, 0.894 < ρ < 0.997), indicating that species composition changed gradually and coherently along the altitudinal gradient, retaining similar patterns of species dissimilarity.
Differences in the first axis between pairs of plots (current minus
projected), if significant, showed that the amount of change in community composition was more pronounced for the montane belt compared to the others (Figure 5
d). Interestingly, we observed significant differences for all taxa pooled together (for all model classes and scenarios), and in many cases also for carabids (T and TR, all scenarios), staphylinids (only T, Min and D scenarios), and butterflies (T, all scenarios; TR, Min and D scenarios; TRV, Max scenario), always showing the lowest values (around 0) for the montane belt.
As an estimate of community homogenization, we estimated the mean Euclidean distance from the distribution centroid for current conditions and for warming scenarios, showing always lower values in the warming case (which were also significant in 29 to 54 cases, p < 0.05).
The Jaccard Index values were close to zero for all model classes and scenarios. Considering all taxa pooled together, we observed significant differences across scenarios (p < 0.0001), with the lowest values in Max, and across model classes (p < 0.0001), with the highest values in T.
For all taxa pooled together (Figure 6
a), the variation in community composition in the warming scenarios is dominated by turnover (βsim
= 0.066 ± 0.005). The nestedness component is lower (βsne
= 0.026 ± 0.002), indicating no clear pattern of species loss or gain. This global pattern has been observed also for carabids and spiders (Figure 6
b,d), while in the case of butterflies and birds, differences are less pronounced and there is a tendency toward higher values of nestedness (Figure 6
e,f). Staphylinids show no clear pattern (Figure 6
Disentangling the role of temperature increase in shaping biodiversity patterns, especially in mountain ecosystems, is a fundamental challenge of conservation biology [2
]. Species respond at individual level [1
] and there is a current knowledge gap for many taxonomic groups [31
]. Projections of potential future species distributions along elevational gradients need high-resolution spatial data [83
]. In our work, the Maxent approach allowed to analyze data collected at a fine spatial scale (plots with 100 m radius) and to take into account the response of rare and localized species. Reliable projections should also consider that ecotypes from different geographical origins can have different responses to the same climatic variations (e.g., [101
]). Our approach reduced this bias in two ways. First, we used the same set of plots to derive the empirical relationships between species and climatic/environmental variables and to project future potential distributions. Second, the limited geographic extent (northwestern Italian Alps) of the sampled plots assures that different populations of the same species will respond coherently.
In general, our simulations indicated that even for a moderate temperature increase, species richness and community composition can display subtle but significant alterations. Since different studies suggested different roles of minimum and maximum temperature increase [14
], we simulated three scenarios having different relative importance of the minimum and maximum daily temperature rise. We also adopted three different empirical biodiversity model classes, characterized by the presence of temperature alone or of other environmental variables (vegetation and geographical location) in addition to temperature. Differences between scenarios were not as marked as those emerging from the different model classes.
We detected a high variability in the response of the different species, depending on the position and breadth of the climatic niche and the species range. The majority of modeled species did not show much response to the temperature increase. On the other hand, high-altitude species were the most affected, consistently with the fact that habitat specialists are negatively influenced by environmental change (e.g., [7
]). The probability of extinction under climate change reflects the species ability to shift with suitable habitats; a scarce ability to withstand climate change impacts could depend also on a reduced niche breadth [7
In our analysis, endemic species appeared as less affected than vulnerable ones. In the altitudinal gradient explored here, the spatial distribution of endemic species is spread over the whole altitudinal and temperature range, owing to the presence of species that are characteristic of the montane belt. Our results indicate that many of the endemic species studied here are not restricted to high altitudes and can shift their occupancy area along the altitudinal gradient. These results are not in disagreement with the general vulnerability of endemic species to climate change (e.g., [7
]): thanks to the small scale of the analysis, our study indicates how local altitudinal gradients can offer a possibility of survival for otherwise declining species.
Significant differences between taxonomic groups were also observed.
Carabids did not show any relationship with temperature [52
] and, as a consequence, they did not display any significant change in species richness under temperature increase scenarios. On the other hand, they displayed the highest “temporal dissimilarity” in community composition (even if always low, with an overall Jaccard index < 0.15). This behavior was mainly determined by turnover, i.e., by the substitution of groups of species under climate warming.
In the case of staphylinids, the pattern of change was also not marked: few significant differences were observed, and they were not coherent across model classes and scenarios, hampering the possibility of drawing general considerations.
Butterflies are heliothermic and can be considered climatically sensitive [33
]. In our analysis, for many species the distribution was mainly determined by temperature. This taxon displayed the highest increase in species richness under temperature increase. This pattern is exacerbated by the model classes using temperature only as a constraint, and it is particularly strong in the alpine belt. These results indicate that butterflies can respond quickly to warming, determining new species assemblages, in particular at high altitudes and that the vegetation structure has the potential to buffer this response. Similar results were found analyzing observed temporal changes in butterfly distribution and community composition [106
], also during a short time frame [108
] and in the same mountain ranges considered here [53
On the opposite side, spiders strongly suffered in temperature increase scenarios, showing a decrease in species richness. In our monitoring program, spiders were highly localized and many species were present only in a small number of plots. Therefore, variations of microclimatic conditions can effectively reduce their areas of occurrence, determining a general decrease in species richness. Models using only temperature provide the most marked changes, indicating again that the vegetation structure can buffer the effects of warming. For spiders, the highest decrease in species richness was observed in the montane belt. Our results suggest that the montane belt will become too warm for many species. Indeed, spider species are usually influenced by the small-scale vegetation structure and are adapted to the local environmental conditions [109
]. This leads to a high level of turnover along altitudinal gradients and consequently to assemblages that are sensitive to small changes in local climatic conditions, as observed here.
Birds display a significant increase in species richness for many of the tested models and scenarios and a general positive effect of warming. Temporal analysis of bird and butterfly communities showed that both taxa respond to warming, with a faster response of butterflies [106
]. For birds, changes in community composition were dominated by the nestedness component, indicating that species losses and gains represent the principal pattern of change.
Considering “temporal” dissimilarity as a whole, we observed significant changes for all cases (model classes, scenarios, taxa). Plot assemblage composition coherently changed in warming scenarios, showing a tendency to homogenization as observed in other studies [111
As in other studies focused on plants [11
], our warming scenarios indicated an upward shift from the lower to the higher belts, implying an increase in species richness at the highest altitude. Alpine meadows had a high probability of experiencing the largest modifications in community composition, as we coherently observed for almost all taxa. Such changes in richness and composition, if effectively realized, will probably be detrimental to biodiversity: the colonization of the alpine belt by species from lower altitudes could increase competition, with negative effects for localized and specialized taxa [13
]. Consequently, the increase in species richness could be transitory, followed by a decline of strictly alpine species stressed by the growing competition and on the verge of going beyond their tolerance breadth [24
The montane belt, on the opposite, is the lower limit of the altitudinal gradient considered here and we cannot account for the colonization of species coming from still lower elevations. We classified plots along the vegetation belts considering both altitude and potential vegetation [65
]: consequently, some montane plots display temperature values that are lower than for subalpine ones, allowing for a partial increase in species richness under climate warming. In any case, the montane belt community composition was the most stable, showing lower dissimilarity than the upper two belts.
Among the different warming scenarios, the one with a larger increase of minimum temperature displayed the largest number of varying species
, indicating that the minimum daily temperature can be an important limiting factor for species distribution [65
]. Considering that the minimum temperature has increased at a faster rate than the maximum temperature during the latter half of the 20th century [117
], variations in biodiversity patterns larger than the ones simulated here could occur.
Even if uncertainty is the rule in species distribution models [40
], projections are useful to explore biodiversity patterns, identifying specific areas or groups of species that should be monitored. In this regard, the different responses displayed by different taxa confirm the importance of using a multi-taxa approach to estimate climate change effects on animal biodiversity. It is important to include taxa with potentially opposite responses to climate warming, such as butterflies and spiders. Targeted long-term field data are then essential to revisit and fine-tune estimates of biodiversity responses, comparing them with real changes in species responses, and adopting timely conservation strategies [118
Even if the alpine belt displayed the highest projected change in species richness, monitoring focused only on high altitudes could miss some of the important changes along the altitude gradient. The high level of turnover due to species shifts from the montane and to the alpine belts could happen also in the subalpine belt, and the montane belt can be influenced by the arrival of species from the surrounding lowlands.
In conclusion, our analysis suggested that moderate warming has the potential to change biodiversity patterns in mountain ecosystems, with significant differences between taxa and along the altitudinal gradient. Such changes could determine new ecological relationships, which could in turn influence ecosystem processes in an unpredictable way.