Habitat Partitioning Among Mediterranean Tortoises (Genus Testudo) in Response to Vegetation Seasonality
Abstract
The genus Testudo is represented by 4 species in the Mediterranean ecoregion, each occupying distinct climatic niches with some peripheral overlap in their ranges. In this study, we used remote sensing vegetation indices (Enhanced Vegetation Index; EVI) to evaluate habitat partitioning patterns among tortoise species and subspecies/mitochondrial clades across the circum-Mediterranean region. Our analyses revealed low habitat niche segregation among 3 mesic species (Testudo graeca, Testudo hermanni, and Testudo marginata). Testudo graeca has the widest niche breadth, overlapping broadly with all other species, including the xeric specialist, Testudo kleinmanni. Among the 3 mesic species, T. marginata has the narrowest niche breadth, with its distribution largely falling within the niche of the sympatric T. hermanni. Shared thermal requirements and the avoidance of dense forests are key factors driving the extensive niche overlap observed in mesic species. Within species, parapatric clades generally occupy similar habitat niches, although some subspecies can show pronounced differences in habitat properties. Remote sensing data prove valuable for assessing habitat use patterns among congeneric reptile species across large geographic areas.
Resource competition is a primary driver of niche differentiation among congeneric species in sympatry (Brandl et al. 2020). Niche partitioning reduces fitness costs associated with trophic competition, interbreeding, and pathogen transmission (Jenssen 1973; González-Hernández et al. 2014). This segregation fosters species-rich assemblages with fine-scale spatial distributions, driven by subtle habitat variations (Pianka 1973; Losos et al. 1993).
Mediterranean tortoises (genus Testudo) serve as a suitable model for studying niche partitioning hypotheses. These tortoises are distributed across a broad area of the circum-Mediterranean region, with 4 species showing both sympatric and parapatric distributions (Willemsen and Hailey 1989; Ballasina 1995). The distribution patterns of these species are primarily influenced by climatic factors (Rodríguez-Caro et al. 2016; Javanbakht et al. 2017; Türkozan et al. 2021). In contrast to other reptiles, competition among tortoise species likely plays a less significant role, due to their lower species richness and low population densities (Luiselli et al. 2006).
Testudo tortoises thrive in specific climate zones, occasionally coexisting with other congeneric species in areas where conditions are transitional (Escoriza and Ben Hassine 2022). In these zones, sympatric species may show patterns of spatial segregation, potentially influenced by gradients in plant cover. For instance, Testudo hermanni has been found to preferentially select habitats with higher humidity and plant density than sympatric Testudo graeca (Willemsen and Hailey 2003). However, Testudo species can also be found in habitats with similar compositional structures, coexisting syntopically (e.g., T. hermanni– Testudo marginata and T. graeca– Testudo kleinmanni; Bringsøe and Buskirk 1998; Koppitz et al. 2016). Therefore, the influence of vegetation on species distribution limits remains unclear, though it may help to delineate or structure these boundaries (Willemsen and Hailey 2003).
Dense canopies hinder the vertical transmission of light within forested habitats, thereby limiting basking opportunities for heliothermic reptiles (Jofré et al. 2016). In contrast, open habitats with sparse vegetation provide little refuge from predators and increase the risk of overheating (Barje et al. 2005). One method for evaluating habitat properties involves measuring surrogates of vegetation structure and density through leaf spectral reflectance analysis (Huete et al. 2002). This technique enables large-scale assessment of habitat quality and its seasonal variations over extensive areas (Son et al. 2014; Wu et al. 2018). These characteristics have established this method as a valuable tool for ecologists, serving as an indirect measure of habitat suitability and aiding in the investigation of interspecific spatial segregation patterns (Palomares et al. 2016; Farwell et al. 2021; Benedetti et al. 2023).
Our study aims to investigate potential niche differentiation along a plant density gradient within the genus Testudo. We expect that species such as T. kleinmanni, a specialist in sandy desert environments, will show a contrasting response compared to those typically associated with Mediterranean shrubland–garrigue (Bour 2004; Kati et al. 2007; Werner 2016). We also expect some variation between mesic species (T. graeca, T. hermanni, and T. marginata), shaped by differences in their range extent and macroclimatic niches (Escoriza and Ben Hassine 2022). Additionally, we evaluate potential differences between subspecies (mitochondrial clades) of T. hermanni and T. graeca, as these clades inhabit ecologically diverse environments and may display adaptive divergence (Bertolero et al. 2011; Türkozan et al. 2023).
METHODS
Study Region and Species. —
The study investigated 4 native tortoise species within the Mediterranean ecoregion, which spans southern Europe, northern Africa, and southwest Asia: the Mediterranean spur-thighed tortoise (T. graeca), Hermann’s tortoise (T. hermanni), the Egyptian tortoise (T. kleinmanni), and the marginated tortoise (T. marginata) (Bonin et al. 2006). This region is characterized by a gradual latitudinal transition from cold-temperate climates in the north to subtropical desert climates in the south (Beck et al. 2018). Additionally, the study included mitochondrial clades of T. hermanni and T. graeca, based on the taxonomy proposed by Rhodin et al. (2021) and the intraspecific phylogenetic relationships outlined by Fritz et al. (2006), Türkozan et al. (2018), and Javanbakht et al. (2017). Testudo marginata and T. kleinmanni are considered monotypic species (Rhodin et al. 2021).
Habitat Data. —
We collected occurrence data from field sampling and bibliographic sources (Escoriza and Ben Hassine 2022; Escoriza et al. 2022, 2023; Türkozan et al. 2010, 2019, 2021, 2023), retaining only records with ≥2-decimal precision (Fig. 1). To assess the influence of plant cover on tortoise occurrence, we used the Moderate Resolution Imaging Spectroradiometer Enhanced Vegetation Index (EVI) as a proxy for primary vegetation productivity (Huete et al. 1994; Waring et al. 2006). EVI values range from 0 (barren terrain) to 1 (dense vegetation) (Huete et al. 2002). We used EVI data from 2014 to 2023 to provide a comprehensive view of annual vegetation changes over a 10-yr period (available at https://modis.ornl.gov). Before analysis, the data were cleaned using quality flags (Kong et al. 2019). The EVI data were then processed to create variables that capture the seasonal dynamics of vegetation cover (Palomares et al. 2016): EVIi (average of all EVI values), EVImin (average of minimum values), EVImax (average of maximum values), EVIrel (difference between maximum and minimum average values), EVIcv (average coefficient of variation within the year), and EVIcir (circular dispersion of the date of maximum plant productivity between years).
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World’s Turtle and Tortoise Journal 23, 2; 10.2744/CCB-1633.1
Data Analysis. —
The analyses assessed 2 aspects of habitat partitioning between species and subspecies: 1) their responses to seasonal EVI variations and 2) their niche breadth and interspecific overlap. We standardized the response variables and performed a principal component analysis (PCA; Pearson 1901) to visualize interspecies relationships and potential niche differentiation. We modelled the association between species and response variables using generalized least squares (GLS), which accommodates various correlation structures and corrects for spatial interdependence in the data (Beguería and Pueyo 2009). GLS models were constructed with different spatial correlation structures (exponential, Gaussian, linear, spherical, and rational quadratic) and with or without nugget effects (Zuur et al. 2009). We also included a null model assuming no spatial correlation between data points. Model selection was based on pooling candidate models using the sample-size–corrected Akaike information criterion (Burnham and Anderson 2002). For subspecies comparisons, only geographically close or parapatrically distributed pairs were examined. These analyses were conducted using the nlme package (Pinheiro et al. 2023) in R (R Core Team 2024).
Pairwise niche breadth and overlap were quantified within a probabilistic niche region for all species combinations. To account for uncertainty, we used a Bayesian inference model (Swanson et al. 2015). Niche breadth and overlap were estimated through 10,000 Monte Carlo iterations to capture variability within the niche region (α = 0.90). Niche size was measured by the variability among multivariate niche dimensions (Swanson et al. 2015). Niche overlap was evaluated by estimating the likelihood of finding an individual of species X in the habitat of species Y, and vice versa, resulting in 2 index values for each species pair (i.e., X → Y and Y → X) (Lysy et al. 2020). This method, due to its low sensitivity to sample size (Swanson et al. 2015), is particularly suitable for quantifying ecological overlap between species with substantially different distribution ranges, as observed in this study. All analyses were performed using the nicheROVER software package for R (Lysy et al. 2020).
RESULTS
EVI data indicated differences in the habitats occupied by Testudo species and subspecies (Supplemental Tables 1 and 2; all supplemental material is available at http://dx.doi.org/doi:10.2744/CCB-1633.1.s1 and http://dx.doi.org/doi:10.2744/CCB-1633.1.s2). Testudo hermanni occupied habitats with the highest EVI values (EVIi = 0.335, EVImax = 0.497), while T. kleinmanni habitats had the lowest values (EVIi = 0.098, EVImax = 0.133). The EVIi and EVImax data also showed some level of variation among some subspecies of T. graeca and a high similarity between the eastern and western clades of T. hermanni (Supplemental Table 2). The other derived variables revealed that species and subspecies differed widely in seasonal variability indicators of habitat plant productivity, with T. graeca and its subspecies showing high intra-annual variability (Supplemental Tables 1 and 2).
PCA analysis showed niche convergence among 3 Testudo species (T. graeca, T. hermanni, and T. marginata), while T. kleinmanni overlapped only with T. graeca (Fig. 2). However, T. graeca extended its niche into habitats with much lower productivity compared to T. hermanni and T. marginata (Fig. 2). Some mitochondrial clades, such as T. g. zarudnyi and T. g. cyrenaica, almost exclusively occupied these low-productivity habitats (Fig. 3). The 95% confidence ellipses generated for the eastern and western clades of T. hermanni overlapped extensively, whereas the ellipse for the western group of T. graeca was entirely included within that of the eastern group, which extended into more productive habitats, particularly the subspecies T. g. ibera (Fig. 3; Supplemental Table 2).
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World’s Turtle and Tortoise Journal 23, 2; 10.2744/CCB-1633.1
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World’s Turtle and Tortoise Journal 23, 2; 10.2744/CCB-1633.1
The GLS models were fitted with different spatial correlation structures depending on the data subsets (Supplemental Tables 3 and 4). These models revealed that the species pair T. graeca–T. hermanni differed significantly only in EVIcv (Table 1). This difference in EVIcv suggested that habitats occupied by T. graeca experienced more pronounced seasonal changes in plant productivity, as evidenced by higher EVIcv values (T. graeca = 9.96 vs. T. hermanni = 6.12; Supplemental Table 1). Testudo graeca and T. hermanni did not differ in any parameter from T. marginata (Table 1). Testudo kleinmanni differed from the other 3 species in EVIi, EVImin, EVImax, and EVIrel, but not in EVIcv and EVIcir (Table 1). Both indices were indicative of vegetation seasonality. While T. kleinmanni showed slightly higher EVIcv values (7.70) compared to T. hermanni (6.12) and T. marginata (5.87), its EVIrel values were significantly lower (0.05) compared to 0.27 (T. hermanni) and 0.18 (T. marginata) (Supplemental Table 1). These findings suggested that T. kleinmanni habitats underwent pronounced seasonal fluctuations but with less pronounced productivity peaks compared to the other species, impacting the coefficient of variation. However, EVIcir results aligned with those of the other species due to shared Mediterranean rainfall patterns.
For the main mitochondrial clades, the eastern and western clades of T. hermanni differed significantly in vegetation productivity seasonality indicators (EVIcv and EVIcir) (Table 2). The western clade showed higher EVIcv and lower EVIcir values (Supplemental Table 2), indicating it occupied habitats with greater fluctuations in plant productivity. In contrast, the eastern and western groups of T. graeca did not show significant differences in these parameters (Table 2). Testudo g. marokkensis tended to occupy significantly more productive habitats (higher EVIi) compared to other North African subspecies. However, in most pairwise comparisons, the differences between parapatric clades of T. graeca were small (Table 2).
Niche breadth analyses indicated that T. graeca had the widest niche (mean = 2.182 ± 0.184 SE) (Fig. 3), followed by T. hermanni (0.966 ± 0.087 SE), T. marginata (0.262 ± 0.063 SE), and T. kleinmanni (0.002 ± 0.001 SE) (Fig. 4). Consistently, the analyses also revealed that the niche of T. graeca overlapped with those of the other 3 species (T. hermanni → T. graeca = 74.19%, T. kleinmanni → T. graeca = 98.52%, T. marginata → T. graeca = 84.86%) (Table 4). The niche of T. marginata was almost completely nested within that of T. hermanni (T. marginata → T. hermanni = 92.59%) (Table 3).
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World’s Turtle and Tortoise Journal 23, 2; 10.2744/CCB-1633.1
Similarly, some subspecies or mitochondrial clades of T. graeca showed broad niches, with the eastern group (2.411 ± 0.297 SE) being wider than the western group (1.39± 0.162 SE) (Fig. 5). Among the subspecies of T. hermanni, the western subspecies T. h. hermanni (0.906 ± 0.113 SE) had a wider niche than the eastern subspecies T. h. boettgeri (0.632 ± 0.081 SE) (Fig. 5). Within the mitochondrial groups of T. graeca, T. g. ibera (1.470 ± 0.267 SE), T. g. nabeulensis (1.579 ± 0.568 SE), and T. g. terrestris (1.802 ± 0.353 SE) showed the greatest niche dispersion compared to T. g. armeniaca (0.165 ± 0.076 SE), T. g. graeca (0.849 ± 0.180 SE), T. g. marokkensis (0.373 ± 0.083 SE), T. g. whitei (0.457 ± 0.098 SE), and T. g. zarudnyi (0.006 ± 0.004 SE). Consequently, significant differences were observed in the overlap between parapatric clades: while both clades of T. hermanni had similar overlaps, there were important differences between some pairs of subspecies of T. graeca, such as T. g. graeca and T. g. marokkensis (T. g. graeca → T. g. marokkensis = 51.05%; T. g. marokkensis → T. g. graeca = 20.88%; Table 4).
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World’s Turtle and Tortoise Journal 23, 2; 10.2744/CCB-1633.1
DISCUSSION
The findings of this study suggest a preference for structurally similar habitats among 3 Testudo species (T. graeca, T. hermanni, T. marginata). Testudo hermanni had the highest EVI score among the mesic Testudo species, although this difference was not statistically significant. Testudo graeca is the most ecologically plastic species, with its niche overlapping widely with the other 3 species. The niches of T. kleinmanni and T. marginata are almost completely nested within that of T. graeca. Testudo hermanni and T. graeca also share a large part of their habitat niche; however, T. hermanni is found in regions with lower fluctuations in vegetation productivity, consistent with a species that occupies northern Mediterranean regions with hydrologically more stable regimes (Margeta 1997; Ivančan-Picek et al. 2014). Testudo graeca has a wider niche breadth than T. hermanni and T. marginata, extending its range into steppes—extensive, open habitats with low precipitation and sparse vegetation (EVI < 0.11–0.15; John et al. 2008; Halos and Abed 2019).
At the extreme of its niche, T. graeca entirely overlaps with T. kleinmanni, a xeric specialist (Werner 2016). The broad habitat tolerance demonstrated by T. graeca could be facilitated by its high phenotypic and genetic diversity, allowing adaptation to variable local conditions. This is particularly evident in the eastern clade, composed of 5 distinct mitochondrial lineages resulting from allopatric divergence during the Plio-Pleistocene epochs (Türkozan et al. 2018; Escoriza and Ben Hassine 2022; Ranjbar et al. 2022). This isolation may have promoted the retention of certain morphological traits that enhance persistence in harsh climates. An example is the ecotype occurring in the Eastern Anatolian montane steppes (T. g. armeniaca; Türkozan et al. 2023), which has a flattened carapace that confers an advantage for burrowing (Pieh et al. 2002; Arakelyan et al. 2018). Burrowing offers a crucial advantage in these warm–arid environments by mitigating exposure to extreme temperatures, as has been observed in tortoises from other hot and arid regions of the world (Ernst and Lovich 2009; Legler and Vogt 2013). Other Mediterranean species show less pronounced morphological variation within their populations compared to T. graeca, likely due to the less fluctuating climatic conditions in the regions they inhabit (Bour 1995; Türkozan et al. 2010; Bertolero et al. 2011; Delfino et al. 2011). Testudo kleinmanni does not have a conspicuously flattened carapace, but its smaller size (130-mm length) may facilitate entry into rodent burrows used for shelter from thermal extremes (Werner 2016).
Analysis revealed a preference for habitats with relatively sparse vegetation cover among all species studied. Forests typically show mean EVI values above 0.50 (Hess et al. 2009; Hasanah and Indrawan 2020), while the mean EVI values obtained for the 4 tortoise species ranged between 0.09 and 0.34. This finding aligns with previous literature, which suggests a preference for steppes, shrublands, and open forests, and an avoidance of dense forests among Testudo species (Cheylan 1981; Schleich et al. 1996; Attum et al. 2011; Bertolero et al. 2011; Bernheim et al. 2019; Bar et al. 2021; Escoriza et al. 2023; Türkozan et al. 2023). Testudo hermanni and T. marginata show higher mean EVI values (between 0.31 and 0.34) compared to T. graeca and T. kleinmanni, as they are commonly found in habitats characterized by denser vegetative formations (Mayol 2003; Bour 2004; Rugiero and Luiselli 2006; Rugiero et al. 2020). However, the structure and productivity of the plant community are likely not the only habitat factors determining the presence of these tortoises; the abundance of palatable plant species may also be significant, particularly at a local scale (Del Vecchio et al. 2011).
Thermal requirements are likely a critical determinant of habitat suitability for these entirely diurnal species (Wright et al. 1988; Werner 2016), which may explain their preference for open environments over dense forests (Meek 1985). Testudo species rely solely on solar radiation for thermoregulation, necessitating extended basking periods to initiate daily activity (Meek and Jayes 1982; Meek 1984, 1988; Willemsen 1991). Additionally, lower substrate temperatures in forested areas can negatively impact reptile reproductive success, affecting both embryonic development and sex ratio parity (Janzen 1994; Schlaepfer 2003). In Testudo, hatchling sex is temperature-dependent (Eendebak 1995; Bernheim 2014; Heimann 2016).
Intraspecifically, our results indicate that the level of variation in habitat properties is also high among the mitochondrial clades of T. graeca. Testudo h. hermanni and T. h. boettgeri largely overlap in their niches, but T. h. hermanni occupies habitats with greater seasonal variability in plant productivity. This could be attributed to a lower amount of precipitation in the southwestern Mediterranean region compared to the Adriatic region of the Balkan Peninsula (Margeta 1997). In T. graeca, the eastern and western clades do not show significant differences in the derived EVI variables, but the eastern (Asian) group occupies a broader niche, completely encompassing the niche of the North African clade group. The different subspecies or mitochondrial clades may or may not show differences in the niches they occupy. For instance, T. g. marokkensis appears in habitats with higher plant productivity than the parapatric clade T. g. graeca, but T. g. nabeulensis can appear in mesic habitats (with high EVIi) or steppe habitats (with low EVIi) indistinctly, as well as the Asian lineages T. g. ibera and T. g. terrestris.
Our findings highlight the value of remote sensing data in evaluating both interspecific and intraspecific habitat responses at broad geographical scales. Identifying distinct habitat attributes, such as structural features that promote higher species diversity in zoological assemblages or serve as faunal refugia during extreme periods characteristic of semiarid regions, facilitates the prioritization of habitats for threatened species (Selwood et al. 2018; Regos et al. 2020). These methods can also be applied to other reptile assemblages in the Mediterranean region, allowing for a coarse identification of favorable habitats for these reptiles, which could benefit from noninvasive reforestation or other conservation interventions (Cox et al. 2006; Mott et al. 2010).
Map of the study area showing the records used in the study (colored circles), categorized by species. Vegetation cover is displayed in the background (green = forests; brown = cultivated land and shrubs; white = deserts and alpine regions; based on Tuanmu and Jetz 2014). Upper panel: Yellow circles, Testudo graeca; blue circles, Testudo hermanni; brown circles, Testudo marginata; gray circles, Testudo kleinmanni. Lower panel: Subspecies or mitochondrial clades of Testudo graeca and Testudo hermanni.
Principal component analysis scatter plot showing niche partitioning based on the Enhanced Vegetation Index (EVI)-derived variables for the 4 Mediterranean Testudo species (left panel) and the 2 Testudo hermanni subspecies (right panel). EVIi = mean EVI value (10 yrs); EVImin = mean minimum value; EVImax = mean maximum value; EVIrel = EVImax − EVImin; EVIcv = average coefficient of variation of EVI; EVIcir = circular dispersion of the date of maximum EVI.
Principal component analysis scatter plot showing niche partitioning based on the Enhanced Vegetation Index (EVI)-derived variables for the eastern-western Testudo graeca clades (left panel) and all subspecies (right panel). EVIi = mean EVI value (10 yrs); EVImin = mean minimum value; EVImax = mean maximum value; EVIrel = EVImax – EVImin; EVIcv = average coefficient of variation of EVI; EVIcir = circular dispersion of the date of maximum EVI.
Box plots representing the niche breadth of the 4 species of the genus Testudo, based on the Enhanced Vegetation Index–derived variables. In the box plots, the bold line represents the median, the box captures the interquartile range, and the whiskers depict the lower and upper quartiles.
Box plots representing the niche breadth of the eastern-western clades of Testudo graeca and Testudo hermanni, based on the Enhanced Vegetation Index–derived variables. In the box plots, the bold line represents the median, the box captures the interquartile range, and the whiskers depict the lower and upper quartiles.
Contributor Notes
Handling Editor: Luca Luiselli
