Assessing the role of habitat, climate, and anthropization gradients on terrestrial mammal diversity in the western Mediterranean basin
Ferran Sayol and Roger Puig-Gironès have contributed equally to the supervision and conceptualization of the study.
Abstract
Mammal species globally exhibit distribution patterns conditioned by environmental conditions and human impact. The Mediterranean basin provides an ideal system to study these effects due to its diverse climate, and habitat conditions. In this work, we aim to assess the impact of landscape heterogeneity and anthropization degree on terrestrial mammal diversity in this region. Accordingly, we deployed over 300 camera traps across 28 sites for 3 months. Detected mammal species (weighing more than 1kg) were classified as domestic carnivores, domestic ungulates, wild carnivores, wild ungulates, lagomorphs, and large rodents. Alpha and beta diversity were calculated for each group and all wild mammals. Simple linear regressions and multimodal analysis were conducted between mammal diversities and climate, environmental conditions, landscape heterogeneity, and anthropization degree variables. Redundancy analyses were performed to identify variables and species determining the mammalian community composition. Indexes measuring landscape heterogeneity, anthropization degree, and its 30-year change did not correlate with mammal diversity. However, the difference in elevation within sites and domestic carnivore abundance showed a significant positive correlation with some of the diversity indexes. Nonetheless, rainfall and mean elevation factors generally showed the highest correlation with mammal diversity. Instead, a few influential species, including generalists and open-habitat specialists, highlighted the importance of conserving open areas, as well as the importance of the Pyrenees region as a key habitat for certain species. Therefore, climatic variables emerged as the key determinants of mammal diversity, highlighting climate change as a potential threat to mammal diversity in this area.
INTRODUCTION
The great diversity of mammals has long fascinated biologists, inspiring a quest to unravel the underlying factors that shape the distribution of species across the planet, ranging from polar regions to deserts, and from mountain tops to the sea. Although several studies have focused on explaining the distribution of mammal diversity across space based on natural covariable gradients (Stevens et al. 2019), there is increasing evidence that human impacts need to be taken into account to explain current biodiversity patterns (Faurby & Svenning 2015).
Previous studies have shown that both climatic and habitat heterogeneity could determine mammal diversity. While some studies emphasize the influence of environmental factors, particularly climate, on species richness (Owen 1990; Andrews & O'Brien 2000; Oliveira et al. 2016), others highlight the importance of habitat heterogeneity to determine mammalian diversity across space (Ceballos & Brown 1995; Kerr & Packer 1997; Fløjgaard et al. 2010). Nevertheless, some studies highlight the importance of accounting for human activity as another driver of mammal diversity (Dorph et al. 2020; Remmers et al. 2023; Suárez-Tangil & Rodríguez 2023). Although it is clear that human impacts have caused the extinction of hundreds of mammal species (Faurby & Svenning 2015; Andermann et al. 2020), the relationship between human activities and mammal diversity is complex. For instance, some studies suggest that low levels of human disturbance in the habitat can sometimes increase species richness (Pita et al. 2009; Santos et al. 2020), as it can increase habitat heterogeneity, with a positive effect on mammal diversity (August 1983; Tews et al. 2004). In contrast, other studies show that human impacts can affect different types of species diversity, such as alpha, beta, and gamma diversities (Swan et al. 2020).
The Mediterranean Basin (MB) is considered a biodiversity hotspot, harboring 770 species of vertebrates, of which 235 are endemic (Myers et al. 2000; Blondel et al. 2010). The northeastern corner of the Iberian Peninsula provides a great system to study the effects of environmental factors on the alpha and beta diversity of mammals, as it contains a wide range of climatic and habitat conditions within a very narrow area. This area includes the Pyrenees and Pre-Pyrenees region, which is considered a mammalian hotspot in the Iberian Peninsula (López-López et al. 2011), and also offers an anthropization gradient. Previous studies in the region have found that mammal diversity was primarily influenced by climatic factors, such as temperature, precipitation, and humidity, finding a positive relation between water availability and mammalian richness (Ferrer-Castán et al. 2016). Furthermore, Varga et al. (2019) reported higher mammal species richness in non-urban areas, especially in mixed forests. Nonetheless, there is a lack of studies focusing on the combined effects of environmental and anthropic variables on mammal diversity. A possible reason for this gap is that some of these mammal species are nocturnal or avoid direct contact with humans (Rogala et al. 2011; Teixeira et al. 2023), making it challenging to detect their presence and estimate their abundance at a large scale. This gap might be fulfilled by the use of camera trapping, a rising non-invasive method in the field of studying terrestrial mammals since the late decade of the 20th century (O'Connell et al. 2011; Burton et al. 2015).
Here, we make use of an extensive network of camera traps in Catalonia (NE Iberian Peninsula) across 28 study sites to test how factors mentioned above can affect the diversity of large terrestrial mammal species (above 1 kg). Specifically, we use the relative abundance data from camera traps to test which are the main climatic, habitat, and anthropic factors that can affect the alpha and beta diversity of carnivores and ungulates. We explored the hypotheses that a higher diversity of mammals will be (a) positively associated with habitat and climatic heterogeneity and (b) negatively associated with anthropic disturbances. We also tested whether mammal diversity is correlated not only with present conditions but with changes in habitat heterogeneity and anthropogenic disturbance in the last decades.
MATERIAL AND METHODS
Study area
Our study area is Catalonia, a region in the northeastern corner of the Iberian Peninsula in the southwest of Europe (Fig. 1). This 32 100 km2 region has great environmental heterogeneity due to its steep climatic and topographical gradients. There is a gradient of aridity, temperature, and elevation, finding the wettest and coldest zones in the north and west, where the elevation is higher (especially in the Pyrenees), while the driest and hottest areas are found in the south and east, generally zones with a lower elevation (Puig-Gironès et al. 2017). The human presence since prehistoric times has resulted in significant changes in the distribution of apex predator species throughout history, like the extinction of large carnivores (Pushkina & Raia 2008). Nowadays, land cover chiefly consists of forests (31%), evergreen shrublands (29%), and agricultural land (33%) (Vallecillo et al. 2013). Most of our study sites are placed within natural protected areas (included in the Plan of Natural Interest Areas or PEIN) that occupy 32% of the region (Generalitat de Catalunya 2022). However, there is still a high heterogeneity of landscape among the different sites, finding various levels of coverage among coniferous, broad-leaved and mixed forests, mountain areas, grasslands, pastures, and urban areas.
Sampling design and target species
Data of macro-mammals (considering larger mammalian species like deer), as well as meso-mammals (including medium-sized mammal species like foxes, heavier than 1 kg, following Montalvo et al. (2017) criteria), were obtained from 28 permanent monitoring plots (PMPs), which were active for a minimum of 3 months between December 2021 and June 2022, as part of the “Wildcat Project” in Catalonia. Each PMP consisted of at least 12 deployed camera traps spaced 1.5–2 km apart (Fig. 1) forming a continuous sampling area. The distance between camera traps corresponds to the theoretical minimum average size of the territories of the mesocarnivore target species (Rosalino et al. 2004; Santos-Reis et al. 2004; Monterroso et al. 2009). These distances between camera traps may be small for large mammals but adequate for studying habitat and climatic selection for the species considered in this study (MacKenzie et al. 2004). The initial locations were selected randomly, and subsequently, the camera network was established by placing them on suspected wildlife trails, up to 150 m from the initial point, to increase the probability of detecting mammal species. All camera traps were Browning Strike Force units or other infrared models with similar capabilities (i.e. Browning Recon Force Edge and Bushnell Core S4K). Each camera trap was installed in a location that optimized the likelihood of detecting the target species without the use of lures or baits to avoid biases in the relative detection of different mammal species. Cameras were positioned 20–50 cm high above the ground, to effectively capture both large and medium-sized animals. Cameras operated 24 h a day, capturing three-photo bursts with a maximum of 10-s time between consecutive bursts, depending on camera settings. Although most of the cameras were recorded for more than 90 days (mean ± SD of 117.55 ± 75.17 days per camera trap), data were only considered from cameras that were active for at least 40 days. This resulted in a data set of 11.61 ± 2.75 cameras per PMP and 108.60 ± 30.94 days per camera trap. To avoid double-counting, detections of the same potential individual species within the same camera were only counted as independent detections (contacts) if they were separated by a minimum time interval of 1 h. This interval is commonly used in camera trapping studies and enables the maximization of the number of theoretically independent contacts, allowing a more precise estimation of relative activity and abundance (Azevedo et al. 2018).
We only considered terrestrial mammal species weighing more than 1 kg (Fig. 1). Consequently, small mammals (such as hedgehogs, squirrels, or weasels) and semi-aquatic mammals (such as otters, Lutra lutra (Linnaeus, 1758), or American mink, Neogale vison (Schreber, 1777)) were excluded from our analysis. The target species were categorized into distinct groups for further diversity analysis. The first group consists of wild mammals, further classified into carnivores, ungulates, lagomorphs, and large rodents (Table S2, Supporting Information). The second group is composed of domestic mammals, which was subdivided into domestic ungulates (“Cattle”) and domestic carnivores (“Domestics”). In a few cases, we were not able to distinguish the species, only the genus. This was the case of six detections of Felis sp. individuals, which were not possible to attribute to either wildcat (F. silvestris Schreber, 1777) or domestic cat (F. catus Linnaeus, 1758). These detections were not included in the subsequent analyses.
Environmental and anthropic variables
Environmental and anthropic variables were recorded at each study site (PMP). First, to determine the coverage of each landscape in the different PMPs, we created a 2-km buffer around each of the 12 cameras within the PMP. This approach resulted in a total coverage area of 12.5 km2 per camera. We used a 2-km buffer size because it has been considered an appropriate buffer size to cover the average home range of the target species (Rosalino et al. 2004; Santos-Reis et al. 2004; Monterroso et al. 2009) and adequate for habitat and climatic selection. Moreover, it commonly overlapped with buffers of nearby cameras. Subsequently, we extracted several climatic variables, including annual mean temperature, annual rainfall, and thermal amplitude from the Climatic Atlas of Catalonia (Martin-Vide et al. 2008) and the Hipermapa (Generalitat de Catalunya 2023). For this purpose, we used polygon covers that represent temperature within a margin of 1°C and rainfall within a span of 50 mm. For each PMP, we calculated a composite value by averaging the values within each 2 km buffer. This was achieved by calculating the weighted average based on the area of each polygon within the buffer. For example, if a polygon had a mean temperature or thermal amplitude ranging from 8°C to 9°C, the value used was 8.5°C. Similarly, if a polygon represented annual rainfall from 200 to 250 mm, the value used was 225 mm. These values were then multiplied by the proportion of the PMP area occupied by the given polygon. It is essential to note that the Climatic Atlas's coverage is limited to Catalonia. Therefore, for PMPs with buffer zones extending beyond Catalonia's borders, we adjusted the results to consider only the surface area situated within Catalonia, as shown in the Climatic Atlas. We assumed that the climate in the adjacent areas (France or Aragon region) was similar to Catalonia's climate, enabling this adjustment. The mean elevation and the difference in elevation of each PMP, that is, the variation between the maximum and the minimum elevation was extracted using the Digital Elevation Map of Catalonia (Institut Cartogràfic i Geològic de Catalunya 2018).
Buffers were overlaid with a simplified CORINE Land Cover map from 2018 (European Union 2018), categorized into 16 landscape categories (Table S3, Supporting Information). This simplification was done by joining similar habitats following a methodology proposed by Puig-Gironès and Real (2022) and were selected based on their potential influence on the presence and abundance of mammal species in the study area. The landscape diversity index of Shannon (hereafter, landscape index) was calculated for each PMP using the reclassified map. This analysis was performed using the LecoS Landscape Ecology Statistics 3.0.0 plugin (Jung 2016). The anthropization degree of each location was assessed using the hemeroby index. This index is commonly used to estimate the degree of human perturbations in a given location (Eurostat 2017) and has been widely used in essays and reports that measure the extent of deviation from the potential natural vegetation caused by human activities (Paracchini & Capitani 2011; Walz & Stein 2014; Szilassi et al. 2017). It assumes that a change in natural vegetation is directly related to a more anthropized landscape. This index assigns a value between 1 (more natural vegetation) and 7 (more artificial vegetation) to each landscape category (Table S3, Supporting Information). Thus, to determine the hemeroby value for each CORINE landscape, the scoring was based on Paracchini and Capitani (2011) and Walz and Stein (2014). The hemeroby final value is the result of adding all these values multiplied by the proportion of area they occupy within the 2-km buffer zone of the cameras within the PMP (the same buffer that was used to calculate landscape indexes).
Hereafter, to determine the habitat changes over time, we repeated the explained process for the CORINE Land Cover of 1990 and 2006. With these data, relative differences between the Shannon diversity index and the hemeroby index were calculated per PMP. This was considered as the fraction of the subtraction between 2006 and 2018 values with respect to the 2006 value (hereinafter referred to as short-term changes in the hemeroby index), and as the fraction of the subtraction between 1990 and 2018 values with respect to the 1990 value (hereinafter, long-term changes in the hemeroby index). QGIS 3.28.2 software from Open Source Geospatial Foundation (seated in Beaverton, USA) was used to analyze the landscape data.
Statistical treatment for species diversity
We initiated our analysis by calculating diversity parameters for each PMP (Table S1, Supporting Information). Relative abundance indexes (RAIs) were computed at camera level for each detected species to assess both domestic and wild species relative abundance. RAIs were calculated as the number of independent contacts (detections) per 100 days of activity (obtained by multiplying the number of detections by 100 and dividing it by the number of days each camera was active). RAIs were calculated at the PMP level by averaging the RAIs from all cameras within the PMP. To assess species diversity, we used the Shannon index to measure alpha diversity per PMP while beta diversity was quantified using the modified Whittaker (1972) equation: (1/((α/γ) − 1)) following Puig-Gironès et al. (2023), where α is the site species richness and γ corresponds to gamma diversity. These indexes were calculated for wild mammals as a whole (taking into account the regional richness of the 19 species, Fig. 1), as well as for wild carnivores (regional richness of 8 species, Fig. 1) and wild ungulates (regional richness of 7 species, Fig. 1) separately (Table S2, Supporting Information). Therefore, the three lagomorphs and one rodent were only considered in the index including all wild mammals. We assumed that the diversity index calculation had not been significantly affected by the species detectability, since within a 3-month census period and 12 camera traps, all the species present in the plot should have been detected (Rowcliffe et al. 2008).
Simple linear regressions (SLRs) were initially employed to assess the specific impact of each environmental and anthropogenic variable on mammal diversity. Here, the response variables utilized were the Shannon and the Whittaker indexes for carnivores, ungulates, and wild mammals. Independent variables included landscape diversity index, hemeroby index, short and long-term changes of both landscape and hemeroby index, annual rainfall, annual mean temperature, mean elevation, differences in elevation within the PMP, and RAIs of domestic carnivores and domestic ungulates. Subsequently, if models did not meet assumptions of normality and homoscedasticity, we applied logarithmic transformations or opted for a generalized linear model.
Subsequently, linear models with a multi-model inference approach were employed to determine which variables influenced mammal diversity indexes the most. This method involved generating all possible submodels from a set of explanatory variables and ranking them based on Akaike information criteria (AIC) (Burnham & Anderson 2002), subject to hierarchical steps: (1) analyzed diagnostic plots of fitted values, residuals, and splines; (2) calculated the variance inflation factor for each independent variable, considering an arbitrary threshold of ≤10 to identify collinearity (Fox & Weisberg 2019). Temperature and mean elevation were correlated, so we excluded the second one from further analysis, considering temperature as the more ecologically significant variable; (3) calculated the AIC weight (AICω) for each model (Wagenmakers & Farrell 2004); (4) calculated the model average when there were multiple best models (ΔAIC less than two; Burnham & Anderson 2002); (5) calculated the AIC weight (ω+) of each independent variable, known, as relative variable importance (RVI), being the sum of the AICω of each model where the variable was present (Table S4, Supporting Information). The RVI ranges from 0 to 1, so the explanatory variable was considered robust if it had an RVI > 0.9; a moderate effect between 0.6 and 0.9; a weak effect between 0.5 and 0.6; and no effect below 0.5 (Kennedy et al. 2013; Chiaradia et al. 2016). Analyses were conducted with RStudio software (R Development Core Team 2022; RStudio Team 2022), using the lme4 (Bates et al. 2014), readxl (Wickham et al. 2022), car (Fox & Weisberg 2019), lmtest (Zeileis & Hothorn 2002), and MuMln (Bartoń 2016) packages.
Last, redundancy analyses (RDAs) were conducted to collectively examine the relationship between a set of response variables (abundance of wild mammal species) and a set of explanatory variables (environmental and anthropogenic factors). RDA models were run separately for all wild mammals, wild carnivores, and wild ungulates. To assess the significance of the RDA results and identify significant axes and variables, permutation tests were performed. This involved conducting 10 ANOVAs with 10 000 permutations for each group (all wild mammals, wild carnivores, and wild ungulates). The RDA models were fitted using the vegan package (Oksanen et al. 2022) in R software (R Development Core Team 2022).
RESULTS
We detected eight wild carnivores, seven wild ungulates, three lagomorphs, and one large rodent species (Fig. 1). Each species was present in at least four PMPs, except for the Iberian ibex (Capra pyrenaica Schinz, 1838) and the brown bear (Ursus arctos Linnaeus, 1758), which were present in two and one PMP, respectively. The most frequently detected species were the wild boar (Sus scrofa Linnaeus, 1758), the European roe deer [Capreolus capreolus (Linnaeus, 1758)], and the red fox [Vulpes vulpes (Linnaeus, 1758)]. In 20 PMPs, ungulates presented higher RAIs compared to carnivores (with a range of 5.88–131.58 ungulates per 100 days compared to 2.38–40.58 carnivores per 100 days), while carnivores had higher RAIs in seven PMPs (with a range of 32.35–72.71 carnivores per 100 days compared to 25.79–67.56 ungulates per 100 days). Furthermore, one PMP showed the same RAI for carnivores and ungulates (24.97 individuals of both groups per 100 days). Lagomorphs always constituted the least common group of wild mammals, with a few exceptions. In some PMPs, no lagomorph, rodent, domestic carnivore, or domestic ungulate species were detected.
SLRs showed that climatic variables, such as mean temperature (coefficient = −0.061 and P-value = 0.018 for alpha diversity, and −0.078, P = 0.002 for beta diversity) and annual rainfall (1.29 × 10−3, P < 0.001 for alpha diversity, and 1.14 × 10−3, P = 0.009 for beta diversity), were the most significant variables affecting both alpha and beta wild mammal diversity (Table S5, Supporting Information). Specifically, annual rainfall exhibited significant positive correlations with all diversity variables (5.96 × 10−4, P = 0.026 for alpha diversity of carnivores, 1.85 × 10−3, P = 0.002 for beta diversity of carnivores, and 8.67 × 10−4, P = 0.049 for alpha diversity of ungulates), except for the beta diversity of ungulates (1.21 × 10−3, P = 0.335). Similarly, mean temperature showed significant negative correlations with all calculated diversities (−0.115, P = 0.005 for beta diversity of carnivores, and −0.078, P = 0.006 for alpha diversity of ungulates), except for the alpha diversity of carnivores (−0.027, P = 0.136) and the beta diversity of ungulates (−0.081, P = 0.337). These two explanatory variables were the only ones to significantly affect the alpha diversity of some groups. Last, the thermal amplitude only showed significant negative correlation with the global alpha diversity (−0.122, P = 0.035). Apart from climatic factors, differences in elevation were positively correlated with all measured alpha diversities (9.44 × 10−4, P < 0.001 for wild animals, 7.57 × 10−4, P < 0.001 for carnivores, and 8.66 × 10−4, P = 0.010 for ungulates), while short-term changes of the hemeroby index (the last 12 years) were positively associated with the alpha diversities of both carnivores (0.580, P = 0.043) and ungulates (0.975, P = 0.036) (Table S5, Supporting Information). This indicates that areas with increased anthropization in the last decade exhibited higher diversity within both groups. The hemeroby index, as well as its long-term changes (the last 28 years), showed a negative correlation with the global alpha diversity of wild mammals (−0.482, P = 0.015 for hemeroby index, and −1.839, P = 0.001 for long-term changes). On the other hand, the landscape diversity index and its variation in time showed no significant effects to any species diversity variables (P-values between 0.212 and 0.738 for landscape diversity, between 0.086 and 0.887 for its variation in the short-term, and between 0.298 and 0.903 for its variation in the long-term). Additionally, RAI of domestic animals did not show any significant regression either (P-values between 0.170 and 0.648 for domestic carnivores and between 0.596 and 0.956 for domestic ungulates; Table S5, Supporting Information).
Results from both SLR (Table S5, Supporting Information) and multimodal analysis (Table 1) were consistent. One of the most influential variables determining alpha diversities was the difference in elevation within a PMP (coefficient = 5.91 × 10−4 and P = 0.035 for wild mammals, 7.29 × 10−4, P < 0.001 for carnivores, and 8.65 × 10−4, P = 0.009 for ungulates, in all cases with an RVI = 1), while beta diversities exhibited more heterogeneity in the relationships with key explanatory variables. Specifically, for the alpha diversity of wild mammals, significant positive relationships were observed in multimodal analyses with rainfall (9.70 × 10−4, P = 0.018, RVI = 1), difference in elevation (5.91 × 10−4, P = 0.035, RVI = 1), and temperature (0.064, P = 0.047, RVI = 0.23) (Fig. 2). However, the beta diversity of wild mammals maintained a significant negative relationship with temperature (−0.082, P = 0.006, RVI = 0.37), while difference in elevation (8.24 × 10−4, P = 0.029, RVI = 1), RAI of domestic carnivores (0.091, P = 0.012, RVI = 1), rainfall (9.89 × 10−4, P = 0.026, RVI = 0.63), and long-term changes in landscape diversity (0.518, P = 0.041, RVI = 0.35) presented positive relationships.
| Mammal category | Alpha diversity | Beta diversity | |||||
|---|---|---|---|---|---|---|---|
| Independent variables | Coefficient ± SE | P-value | RVI | Coefficient ± SE | P-value | RVI | |
Wild mammals |
Intercept | 0.917 ± 0.745 | 0.218 | 0.166 ± 1.014 | 0.870 | ||
| Change of hemeroby index from 2006 to 2018 | 0.735 ± 0.407 | 0.071 | 0.33 | ||||
| Change of hemeroby index from 1990 to 2018 | −0.652 ± 0.625 | 0.297 | 0.12 | ||||
| Change of Landscape diversity from 1990 to 2018 | 0.197 ± 0.207 | 0.343 | 0.11 | 0.518 ± 0.254 | 0.041 | 0.35 | |
| Difference in elevation | 5.91 × 10−4 ± 2.80 × 10−4 | 0.035 | 1.00 | 8.24 × 10−4 ± 3.76 × 10−4 | 0.029 | 1.00 | |
| Annual rainfall | 9.70 × 10−4 ± 4.08 × 10−4 | 0.018 | 1.00 | 9.89 × 10−4 ± 4.45 × 10−4 | 0.026 | 0.63 | |
| RAI of domestic carnivores | 0.091 ± 0.036 | 0.012 | 1.00 | ||||
| RAI of domestic ungulates | −0.017 ± 0.009 | 0.065 | 0.67 | ||||
| Landscape diversity | 0.299 ± 0.191 | 0.117 | 0.49 | 0.324 ± 0.167 | 0.052 | 0.28 | |
| Mean temperature | 0.064 ± 0.032 | 0.047 | 0.23 | −0.082 ± 0.030 | 0.006 | 0.37 | |
| Carnivores | Intercept | 1.379 ± 0.147 | <0.001 | −2.946 ± 1.321 | 0.026 | ||
| Change of hemeroby index from 2006 to 2018 | 0.441 ± 0.224 | 0.049 | 0.69 | ||||
| Change of landscape diversity from 2006 to 2018 | −0.925 ± 0.689 | 0.180 | 0.13 | −7.603 ± 4.443 | 0.087 | 0.43 | |
| Change of landscape diversity from 1990 to 2018 | 1.925 ± 0.967 | 0.047 | 0.77 | ||||
| Difference in elevation | 7.29 × 10−4 ± 1.69 × 10−4 | <0.001 | 1.00 | ||||
| Annual rainfall | 5.73 × 10−3 ± 1.40 × 10−3 | <0.001 | 1.00 | ||||
| RAI of domestic carnivores | 0.263 ± 0.103 | 0.011 | 1.00 | ||||
| RAI of domestic ungulates | −4.91 × 10−3 ± 4.20 × 10−3 | 0.242 | 0.19 | ||||
| Mean temperature | 0.015 ± 0.016 | 0.359 | 0.14 | ||||
| Ungulates | Intercept | 0.041 ± 0.408 | 0.919 | −0.361 ± 2.180 | 0.869 | ||
| Thermal amplitude | 0.203 ± 0.215 | 0.344 | 0.16 | ||||
| Change of hemeroby index from 2006 to 2018 | 1.477 ± 0.428 | <0.001 | 1.00 | 1.291 ± 1.318 | 0.328 | 0.17 | |
| Change of hemeroby index from 1990 to 2018 | 1.767 ± 0.863 | 0.041 | 0.79 | ||||
| Difference in elevation | 8.65 × 10−4 ± 3.29 × 10−4 | 0.009 | 1.00 | 1.95 × 10−3 ± 1.02 × 10−3 | 0.057 | 0.84 | |
| Annual rainfall | 7.57 × 10−4 ± 4.27 × 10−4 | 0.076 | 0.45 | ||||
| RAI of domestic ungulates | −0.020 ± 0.025 | 0.410 | 0.14 | ||||
| Landscape diversity | 0.623 ± 0.211 | 0.003 | 1.00 | ||||
- The table shows the model parameter coefficient, its standard error (±SE), and the associated P-values. The relative importance of each variable (RVI) ranges from 0 to 1, so the explanatory variable was considered robust if it had an RVI > 0.9, a moderate effect between 0.6 and 0.9, a weak effect between 0.5 and 0.6, and no effect below 0.5. Gray cells represent non-selected model independent variables for a specific analysis. RAI, relative abundance index; RVI, relative variable importance.
In the case of the alpha diversity of carnivores, differences in elevation (7.29 × 10−4, P < 0.001, RVI = 1) was the most explanatory variable, while changes of hemeroby index in the short-term (0.441, P = 0.049, RVI = 0.69) was the second most relevant variable. Rainfall (5.73 × 10−3, P < 0.001, RVI = 1), RAI of domestic carnivores(0.263, P = 0.011, RVI = 1), and changes of landscape diversity in the long-term played (1.925, P = 0.047, RVI = 0.77) a significant role in determining beta diversity of carnivores (Fig. 2). Last, the most important explanatory variables for the alpha diversity of ungulates were the difference in elevation (8.65 × 10−4, P = 0.009, RVI = 1), the landscape diversity index (0.623, P = 0.003, RVI = 1), and the short- (1.477, P < 0.001, RVI = 1) and long-term change of hemeroby index (1.767, P = 0.041, RVI = 0.79). However, for beta diversity of ungulates, only the difference in elevation presented a moderate effect, although nonsignificant (1.95 × 10−3, P = 0.057, RVI = 0.84) (Table 1).
The RDA of wild mammals showed a higher explanatory variability than carnivores and ungulates, with a constrained proportion of 0.77 (R2 adjusted of 0.59), compared to the 0.55 for carnivores (R2 adjusted of 0.20) and 0.48 for ungulates (R2 adjusted of 0.07). The difference between RDA1 and RDA2 was particularly pronounced for wild mammals, where RDA1 was more than 10 times more explanatory than RDA2, and for carnivores, where RDA1 was almost 8 times more explanatory than RDA2 (Table S6, Supporting Information).
The RDA of wild mammals showed a dominant X-axis (64.4% of the variability explained) largely influenced by open-habitats, with some species specialized in this habitat such as the European rabbit (Fig. 3). The X-axis also correlated with the hemeroby index and thermal amplitude, indicating the rabbit's preference for areas with high anthropization and temperature variations. Other species, such as the red fox and the European badger, appeared slightly shifted to the left on this axis, indicating that their abundance is favored by highly anthropized areas, higher relative abundance of domestic carnivores, and long-term changes in the hemeroby index, but low landscape diversity. The Y-axis (6.0% of the variability) separated the European roe deer from all other species, although it was not driven by a single specific variable. In the RDA for carnivores, the X-axis (44.4% of the variability) clearly distinguished the red fox and the European badger from other species (Fig. 4) associating them with high RAI of domestic carnivores, thermal amplitude, and hemeroby index, especially the red fox. The Y-axis (5.6% of the variability) mainly separated the stone marten and the common genet from the rest of the species, as they tend to inhabit areas with low landscape diversity and short-term changes in the hemeroby index. Regarding the RDA for ungulates, the X-axis (29.4% of the variability) was mainly correlated with areas showing higher rainfall, lower temperature and lower thermal amplitude, higher RAI of domestic ungulates, higher differences in elevation, and with short-term changes in the hemeroby index (although these areas present a low hemeroby index overall), which influenced species such as the European roe deer, the southern chamois, and the red deer (Fig. 5). The Y-axis (15.1% of the variability) mainly differentiates the wild boar from the rest of species due to their preference for areas with a low landscape diversity and a loss of landscape diversity in both the short and the long-term, as well as in places with relatively high RAI of domestic carnivores.
DISCUSSION
Overall, our results suggest that the diversity of mammals in our study area is mostly associated with climatic variables and landscape diversity. The differences in elevation were highly correlated with species diversity measurements, supporting the hypothesis that greater habitat heterogeneity is associated with higher mammal diversity. In contrast, other measurements of habitat heterogeneity, such as landscape Shannon diversity, have resulted in a lower effect on species diversity. This suggests that the relationship between species diversity and habitat heterogeneity is complex, and that it is challenging to measure it with simple indexes.
Landscape diversity showed a significant influence in multimodal analyses of alpha diversities for both wild ungulates and wild mammals. Additionally, differences in elevation were correlated with the alpha diversity of the three groups. In fact, differences in elevation are often considered as an alternative measure of landscape diversity (Kerr & Packer 1997). In our case, the greater explanatory capacity of differences in elevation over landscape diversity index for alpha diversities might be attributable to the presence of species exclusively inhabiting the Pyrenees (López-López et al. 2011; IUCN 2022; López & Martín 2022), which exhibit significantly higher elevation differences in a relatively small area compared to the plain coastal regions. Our findings are overall in line with our first hypothesis, showing that species diversity in Catalonia is concentrated in areas of high elevational amplitude, such as the Pyrenees. In contrast, differences in elevation did not exhibit strong correlations with beta diversities, contrary to Melo et al. (2009) findings in South America. This disparity is probably due to the higher mammalian diversity and elevation range in South America compared to Europe, apart from the differences in scales used in the two studies.
The second hypothesis of this work, which stated that less anthropic disturbances would be related to a higher species diversity, is not fully supported, as no clear correlation has been found between the hemeroby index and the species diversity variables. One possibility to explain these results is that the response to anthropization varies highly among species (Suraci et al. 2021). For instance, in our study area, the sites with higher values of hemeroby still contain a relatively diverse group of species, including the wild boar, the red fox, and the stone marten. These species are often considered generalists in terms of habitat and diet requirements, which might allow them to take advantage of anthropogenic resources and adapt its habits and activity patterns to human activity (Lowry et al. 2013; Santini et al. 2019). Previous works have shown that some of these species can even benefit from some human activities (Bonnot et al. 2012; Stillfried et al. 2017; Alexandre et al. 2019). However, it is also clear that not all species can persist in more anthropized environments. In our study, the European wildcat, the pine marten and the chamois are species that tended to be found in less anthropized environments (See section of RDAs for more details).
Another possible explanation for the lack of association between anthropization effects and mammal diversity is that the hemeroby index might not be a relevant measure of anthropization effects on mammals. The hemeroby index is based on the deviation of the natural potential vegetation and only presents 7 degrees of human disturbance on the landscape, from 1 to 7. Furthermore, all types of forests present the same degree of hemeroby, and the index does not take into account other sources of human disturbance rather than landscape changes (i.e. human activities, number of visitors to natural areas, traffic, etc.). Finally, this lack of effect might also be explained by the human demographic trends in Catalonia, similar to the ones of most parts of Europe and the western countries. More specifically, the concentration of population in the cities (Wolff et al. 2018), the abandonment of crops or pasture lands due to the decline of the agriculture and livestock (Rey Benayas et al. 2007) and the consequent forest expansion (Keenan et al. 2015) have led to a homogenization of most of the territory, especially of those zones included in mountain protected areas (Ameztegui et al. 2021). Therefore, anthropogenic effects might be relatively small in most of the study area and only appear in transition zones, such as those near crops or urban areas. In order to gain a better understanding of the effects of humans on mammalian diversity, other variables could be used in future analysis, such as the distance to the nearest urban centre, percentage of urbanized surface, human population density or density of roads and traffic inside our PMPs. Nonetheless, changes in the hemeroby index have been found to have a significant positive effect on the measured alpha diversity of ungulates. This might be due to the previously stated fact that the most common ungulate species in Catalonia are generalists that might take profit of human resources, like the wild boar (Stillfried et al. 2017) and, to a lesser degree, the roe deer (Bonnot et al. 2012).
Another indirect effect of human disturbance is the presence of cattle. In our study, even though we found that cattle had a general positive effect on alpha diversities of the three groups, it had contrasting effects on beta diversities. RAI of domestic ungulates showed a negative effect on the global range of wild mammals and ungulates, but a positive effect on that of carnivores. This seems to show that the effects of cattle depend on the group of mammals, as previous studies have found in other parts of the world (Hibert et al. 2010; Cuyckens et al. 2022). The positive association between livestock abundance and the diversity of carnivores is not entirely clear. Previous studies in other regions have shown that mesocarnivores are often perceived as predators of livestock by farmers (Blaum et al. 2010; Martínez-Jauregi et al. 2017). In our study region, instead, the negative attitude towards carnivores seems to be more focused on the reintroduction of large carnivores such as brown bears (Herrero et al. 2021). However, more studies on mesocarnivores persistence on human-dominated landscapes are needed, as most research tends to be biased towards larger species (Lozano et al. 2019). Therefore, our hypothesis, expecting higher diversity with less habitat heterogeneity loss over 30 years, was unsupported. Instead, environmental and climatic variables appeared to be more significant for mammal diversity. Elevation differences and the climatic factors (mean annual rainfall and mean temperature) mainly explain mammal diversity, in line with previous studies (Owen 1990; Andrews & O'Brien 2000; Oliveira et al. 2016). However, ungulates showed divergences, as climatic variables did not significantly influence their alpha diversity, which lacked significant relation with any variable, suggesting other unaccounted variables or historical reintroductions’ influence since the end of the last century.
The RDA showed that the measured variables explained a relatively low proportion of the variability of the wild mammal communities, not reaching 50% in the case of ungulates. That means that there might be other variables not included in the present study that could influence them, such as humidity or water presence. Most of RDAs are clearly dominated by a few species, like the European rabbit, the European roe deer or the red fox. Even though the European rabbit is an open-field specialist, the European roe deer and the red fox are considered more generalist species that tend to better tolerate human presence (Tixier et al. 1997; Dell'Arte et al. 2007; Suárez-Tangil & Rodríguez 2023). For example, the red fox is detected frequently in zones with high RAI of domestic carnivores and hemeroby index, both indicators of a higher anthropization degree. This can be explained because the red fox often uses human resources as part of its diet (Alexandre et al. 2019), as probably other generalist species do as well. Nonetheless, the association observed between foxes and rabbits could also be caused by a predator-prey relationship (Fernández-de-Simón et al. 2015).
The European rabbit, one of the few species included in this work that mainly inhabits open-habitats, strongly influences the relative abundance of wild mammals in our study. Open areas, such as croplands and grasslands, have a higher hemeroby index than forests, so this can explain why this index appears to be one of the most important variables determining the distribution of wild mammals. However, as it has been previously stated, this could be an effect of the homogeneity of the index among most of the PMPs, with the exception of the PMPs link to those open-habitats. Nevertheless, these results emphasize the importance of conserving landscape's heterogeneity with open-habitats and grassland, as the European rabbit is a key element for the equilibrium of the food chains in the Mediterranean ecosystems (Delibes-Mateos et al. 2008; Gálvez et al. 2008). The European rabbit occurs in places with open and cultivated land, a type of areas that are declining throughout Europe (Donald et al. 2001; Stoate et al. 2009; Reif & Hanzelka 2020). Specifically, in Catalonia, cropland areas have decreased an 11.9% between 2001 and 2021 (Institut d'Estadística de Catalunya 2022). On the other hand, wild boars are frequently found in places with a reduced landscape diversity. In Mediterranean Basin western countries, the loss of landscape diversity is commonly related to forest expansion (Keenan et al. 2015; Ameztegui et al. 2021), which might be one of the main reasons for the recent population increases of wild boars (Tack 2018).
Consistent with our results, other authors found that environmental factors had the greatest impact on mammal diversity, much more than the anthropization degree (Real et al. 2003; Oliveira et al. 2016). The fact that climatic variables have such a significant effect on mammal diversity highlights the importance of the consequences that climate change might have on the mammal community. To monitor these effects, similar analyses could include more years of data and could use annual precipitation and temperature, instead of using the averages of the past years. Furthermore, habitat specialists will probably be more affected than generalists, which have higher dispersal ability and adaptability, as has been the case in previous periods of climate change (Teacher et al. 2011). In this line and, according to most climate predictions, the Iberian Peninsula will be one of the European regions most ravaged by climate change, especially by severe droughts (Pörtner et al. 2022), that will probably also affect wildlife. In the case of mammals, the Iberian Peninsula is usually considered a “hotspot inside the Mediterranean Basin hotspot” and could become the European region with the greatest loss of species (Maiorano et al. 2011), especially the Pyrenees, which is a key area for biodiversity (Maiorano et al. 2013). This scenario, added to the previously stated fact that some of the mammal species present in Catalonia only inhabit this area, makes the Pyrenees a priority area for conservation of mammals. If further research confirms the results of this study, conservation actions should then be focused on mitigation and adaptation of the impacts of climate change in Catalonia, especially in the Pyrenean ecosystem, in order to prevent the loss of diversity in an iconic and key group such as large and mid-sized mammals.
ACKNOWLEDGMENTS
We want to thank all the naturalists, rangers, and volunteers (Table S7, Supporting Information) that have helped to collect data, including fieldwork and picture identifications, as part of the mesocarnivore monitoring scheme from FELIS-ICHN. We also thank Mar Unzeta for advice on the landscape analyses. This study received funding from Generalitat de Catalunya, Diputació de Barcelona, the Barcelona Zoo Foundation, Institució Catalana d'Història Natural, and Institut d'Estudis Catalans. Joan Real and Roger Puig-Gironès received funding from “Centre de Monitoratge de la Biodiversitat de Muntanyes Mediterrànies (CMBMM).”
CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.
