Macro- and microhabitat preferences of eastern Hermann’s tortoise ( Testudo hermanni

. Macro- and microhabitat preference of Testudo hermanni boettgeri , the eastern subspecies of Hermann’s tortoise, was investigated utilizing modiﬁed methodology for the western subspecies which emphasized the importance of habitat heterogeneity preservation. The study objective was to explore the habitat preferences of the eastern subspecies of T. hermanni . Research was conducted within the same year at four localities in Eastern and Southeastern Serbia. Macrohabitat determination was conducted using a 0 to 5 land cover score system (coverage with herbaceous, bushy or tree vegetation) for 4 m 2 tortoise encounter surroundings. Microhabitat analysis was carried out by determining the plant species in closest contact with the tortoise in the moment of recording. Plants were classiﬁed into six groups: 1) aromatic, 2) bramble, 3) herbaceous, 4) thorny shrub, 5) tree and 6) non-thorny shrubs. X 2 test was used for comparison between expected and empirical habitat preference. Results conﬁrmed that the most attractive macrohabitats for Eastern Hermann’s tortoises in this part of the Balkans are meadows and open shrublands, with the addition of dense forest (important in wormer months), what is concordant with earlier data from the Mediterranean part of former Yugoslavia. The most attractive microhabitats were “herbaceous plants”, followed by “bramble”. anal-ysed seasons and in two different environments – natural vs. human altered), whether their distribution in the ﬁeld was random or they were more present in speciﬁc macrohabitat types. By Pearson Product-Moment correlation we tested the relation between distribution of tortoises among macrohabitats and the proportion of individual macrohabitats in the total investigated area while controlling for locality and season. All analyses were performed using Microsoft Excel 2017 and SPSS version 15.0 software.


Introduction
Detailed knowledge on habitat preferences of threatened species is necessary for establishment of efficient protection measures (Wiktander et al., 2001). This applies also to Hermann's tortoise (Testudo hermanni) from Europe which is considered Near Threatened due to declining population size across its relatively large species range . As fragmentation and loss of suitable habitats are increasing, these are now among the most significant factors of threat for Hermann's tortoise (van Dijk et al., 2004;Bertolero et al., 2011;Fernández-Chacón et al., 2011).
Numerous studies on the western subspecies (T. h. hermanni) revealed its preference toward specific habitats for specific purposes 1 -Department of Biology and Ecology, Faculty of Sciences and Mathematics, University of Niš, Niš, Serbia 2 -Department of Evolutionary Biology, Institute for Biological Research "Siniša Stanković" -Institute of National Importance for Republic of Serbia, University of Belgrade, Beograd, Serbia * Corresponding author; e-mail: zerocool.axl@gmail.com (Longepierre et al., 2001;Corti and Zuffi, 2003;Rugiero and Luiselli, 2006;del Vecchio et al., 2011;Corti et al., 2013;Berardo et al., 2015;Vilardell-Bartino et al., 2015): in some places, tortoises were selected both bushy and woody habitats for hibernation (probably due to a more stable local environmental temperature that enables them to survive the winter -see in Steen et al., 2007 andVilardell-Bartino et al., 2015) while shrubs without spines were the most visited in the hottest parts of the summer days (Vilardell-Bartino et al., 2015); during the feeding period, the tortoises showed preference to spots with herbaceous vegetation, while blackberries were predominantly chosen for hiding, shelter and during the mating season. Studies on the Eastern Hermann's tortoise (T. h. boettgeri) analyzed habitat preferences mostly on the broad scale, indicating meadows, bushes, shrubs and the edges of forests as the most preferable habitats both in Mediterranean parts and in the inland of the Balkans (Cruce and Rȃducan, 1976;Meek and Inskeep, 1981;Meek, 1985Meek, , 1988Wright et al., 1988;Rozylowicz and Dobre, 2010;Rozylowicz and Popescu, 2013;Türkozan et al., 2015;Stojadinović et al., 2017). It was also suggested that complex habitat matrices harbor relatively dense eastern Hermann's tortoise populations due to (still) low human impact e.g. modest alteration of primary habitats (Stojadinović et al., 2017). However, studies that analyze preference of eastern Hermann's tortoises on a finer scale of macro and micro habitats (see in Villardel-Bartino et al., 2015) have not been published so far.
Researchers face three problems when conducting studies for detailed determination of habitat preference and these are: characterization of the available microhabitat types, quantification of the relative abundance of different microhabitats and determination of the habitat spatial scale relevant to the focal organism (Del Vecchio et al., 2011). These authors revealed that western Hermann's tortoises choose small patches of suitable habitat in a matrix of less desirable habitats, while this study attempted to investigate macro-and microhabitat selection in eastern Hermann's tortoise where habitat degradation on a large scale is still lower (Rozylowicz and Popescu, 2013;Stojadinović et al., 2017). To get more accurate information on possible variation, we have analyzed Hermann's tortoises' macro and microhabitat choice on several localities and in two seasons (spring and summer). The main goals of the study were to assess for 1) preferable macrohabitat type(s) and 2) microhabitat preferences, taking into consideration effects of locality and season.

The study sites
The study was conducted in year 2016 at four localities in Serbia -two situated in the eastern and two in the southeastern part of the country ( fig. 1). Two localities are within the protected areas: in Eastern Serbia, it was localityČermor near Donji Milanovac, situated in "Djerdap" National Park. In Southeastern Serbia, we selected locality Kunovica, situated in "Sićevačka klusura" nature reserve (table 1). "Ðerdap" National Park has rich diversity of flora and fauna within relatively small area (Stevanović, 1996;Medarević, 2001;Crnobrnja-Isailović et al., 2015). "Sićevačka klisura" Nature Reserve is situated 15 km east of the city of Niš. The vegetation in the area of Sićevo has a very diverse zone of thermophilic oak tree forests, inhabited by relatively rich and rare flora and fauna (Lazarević et al., 2007). The other two localities are composed mostly of human altered habitats: Gonjište near Kladovo is a complex of meadows, agricultural areas and vineyards. PašinaČesma near Leskovac is the abandoned complex of vineyards previously run by the public company and situated near popular local picnic place (Stojadinović et al., 2013).

Field procedures
The monitoring program followed general temporal dynamics described in previous studies of Stojadinović et al. (2013Stojadinović et al. ( , 2017. It included two visits -one in May and other in July -considered in analysis as spring and summer (Cheylan,1981).
At every locality, researchers spent 3-7 consecutive days searching for tortoises by visual encounter survey method, from 8 a.m. to 7 p.m. during the day within defined area. At every locality eight people were involved in the fieldwork. Tortoises were determined for sex and age and permanently marked on first capture similar to Meek (1989) by notching carapace plates in a manner that every tortoise gets unique identification number by specific combination of marked carapace plates (more details in Stojadinović et al., 2013). That enabled easy identification in future studies.

Macro-and microhabitat selection
Macrohabitat types were defined following the methodology in the study of Vilardell-Bartino et al. (2015), with small modifications: we had no opportunity to apply radiotelemetry for tracking the tortoises, although it is the best  (Kenward, 2001). Instead, both at first and every subsequent encounter we were recording the location coordinates of every tortoise seen using a GPS device, handheld model GARMIN eTrex Vista ® . Every tortoise was photographed at the encounter place, with surrounding vegetation included in the square space projected around the tortoise of approximately 2 × 2 m in size. The photographs were digitized in vector format and on the area of 4 m 2 we estimated the cover percentage of following plant types: a) grass, b) shrubs -up to 2 m and c) trees -higher than 2 m (see in Etienne and Prado, 1982). After that, we classified the patches in our study, according to Roura-Pascual et al. (2005), into one of nine macrohabitat types: barren land, grassland, open shrubland, dense shrubland, open woodland, wooded grassland, wooded grassland -shrubland, open forest and dense forest (supplementary table S1). This type of classification was previously applied by Vilardell- Bartino et al. (2015) to a study on western subspecies of T. hermanni.
Also, we mapped all macrohabitat types in the field, and then used area calculator tools in Google Earth software (version 7.1.8.3036) to determine the percentage of each macrohabitat at each locality. This information allowed us to determine the expected number of tortoises in each habitat type using proportion method.
For the microhabitat analysis we visually recorded the plant species in closest contact with the tortoise in the moment of recording (the one that was touching the body of the spotted tortoise). We sorted the plant species into six groups following Vilardell- Bartino et al. (2015) systematization: 1) aromatic, 2) bramble, 3) herbaceous, 4) thorny shrub, 5) tree and 6) non-thorny shrubs.

Statistical analyses
To analyze the macrohabitat preference of Hermann's tortoises on the chosen localities, we have created data matrices using the following variables: ID of individuals, season, locality, and macrohabitat type. To avoid pseudo replication, only first records of individual tortoises during the whole study period were included in the analysis (Rugiero and Luiselli, 2006).
We calculated the expected numbers of individuals per specific macrohabitat by using empirical data and the basic expected value formula: Nh × Nl/Nt, where Nh is the total number of recorded tortoises per macrohabitat type throughout all four localities; Nl is the total number of recorded tortoises per locality, and Nt is the total number of all the recorded in certain season; by Chi-square Goodness of fit test we compared those expected values to the empirical (observed) values. In the same way, we tested (within two analysed seasons and in two different environments -natural vs. human altered), whether their distribution in the field was random or they were more present in specific macrohabitat types. By Pearson Product-Moment correlation we tested the relation between distribution of tortoises among macrohabitats and the proportion of individual macrohabitats in the total investigated area while controlling for locality and season. All analyses were performed using Microsoft Excel 2017 and SPSS version 15.0 software.

Results
The results are shown in table 2. Dense forest predominated on the localitiesČermor and Kunovica (58% and 35% of the studied areas, respectively), while open habitat type predominated on the localities Gonjište and Pašinǎ Cesma (grassland, 46% of the studied area and dense shrubland, 66% of the studied area, respectively).
We marked 302 individuals during our study, 174 in the spring and 128 in the summer. In general, the records of the tortoises were not randomly distributed among macrohabitat types. The results of the Chi-square Goodness of fit test showed statistically significant difference between recorded and expected number of individuals in specific macrohabitat types per locality per season (table 3); in addition, the distribution of tortoises among macrohabitats and the proportion of individual macrohabitats in the total overall researched area was significantly correlated in most cases (table 2), which will be explained further in the text.
AtČermor, the dominant macrohabitat type was dense forest (58% of the total area, fig. 2, table 2), with 13% of individuals recorded in the spring and 48% in the summer. In the spring, the largest number of individuals was recorded  The symbol "-" means that this macrohabitat type does not exist in a locality. * * P < 0.01; * * * P < 0.001.  on grassland habitat type -36%, vs. expected 14%, while in summer it was dense forest habitat type with 48% vs. expected 58% of individuals. The difference between observed and expected proportions of tortoise individuals was statistically significant in all cases (Chi square Goodness of fit test P < 0.0001 and P < 0.0001 for spring and summer, respectively). The distribution of tortoises was highly correlated with the proportions of individual macrohabitats only in the summer (Pearson's r = 0.96, P < 0.001). At Gonjište, the dominant macrohabitat type was grassland (46% of total area, fig. 2, table 2) with 50% of individuals recorded in the spring (the largest number of individuals among the habitat types), and 29% in the summer; the largest number of individuals recorded at Gonjište in the summer was in the dense shrubland habitat type, with 67% observed vs. 15% expected. The difference between observed and expected proportions of tortoise individuals was statistically significant in all cases (Chi square Goodness of fit test p < 0.008 and p < 0.0001 for spring and summer, respectively). The distribution of tortoises was highly correlated with the proportions of individual macrohabitats only in the spring (Pearson's r = 0.94, P < 0.01).
At the locality Kunovica, the dominant macrohabitat type was dense forest (35% of total area, fig. 2, table 2) with 24% and 13% of recorded tortoises recorded in the spring and summer, respectively. The largest number of individuals was recorded in the grassland habitat type -30% recorded vs. 8% expected in the spring and 58% recorded vs. 8% expected in the summer. The difference between observed and expected proportions of tortoise individuals was statistically significant in all cases (Chi square Goodness of fit test p < 0.0001 and p < 0.0001 for spring and summer, respectively). The distribution of tortoises was not statistically significantly correlated with the proportions of individual macrohabitats.
At the locality Pašinačesma, dense shrubland was the most common habitat type (66% of total area, fig. 2, table 2) and the most favorable for tortoises in both seasons, with 50% of individuals recorded in the spring and 45% in the summer. The difference between observed and expected proportions of tortoise individuals was not statistically significant, but the distribution of tortoises was highly correlated with the proportions of individual macrohabitats in both seasons (Pearson's r = 0.92 and r = 0.88 for spring and summer, respectively, P < 0.01).
Pearson product-moment correlation results showed that, in both seasons, distribution of tortoises among macrohabitats of PašinaČesma was concordant with their proportion in the total investigated area (r = 0.92 and r = 0.88 for spring and summer, respectively, P < 0.01); the same was approved at Gonjište in spring (r = 0.94, P < 0.01) and atČermor in the summer (r = 0.96, P < 0.001). On the contrary, distribution of tortoises in Kunovica was not correlated at all with the proportions of individual macrohabitats in the total investigated area at this locality (table 2).
At the microhabitat scale ( fig. 3) we revealed that the herbaceous plants were the most common type of plants being in the closest contact with the recorded tortoises, except for locality Gonjište in the summer, where it was "bramble". "Bramble" was also the second most common microhabitat type at all sites the tortoises were in contact with according to our results. At the localityČermor, except "herbaceous", large number of tortoises were found within microhabitats consisting of trees in the summer and aromatic plants in the spring.

Discussion
In this study, we analyzed macro-and microhabitat preference in four Hermann's tortoise populations in Serbia. Additionally, these localities also represent two types of environment inhabited by tortoises -dominated by dense or open vegetation types.
Depending on the season, the eastern Hermann's tortoises show somewhat different preferences toward specific macrohabitat types. The previous longitudinal study conducted at the locality of Kunovica (Stojadinović et al., 2017) has not confirmed that the tortoises select specific habitats for carrying out specific activities, but rather that certain particularities exist in their habitat preferences and activity patterns in the mixed landscape of oak forests, meadows, gardens, orchards and vineyards. Hermann's tortoises in our study were mostly recorded in meadows and open shrublands (table 2), while less individuals were notified in the dense forest habitat type (similar with the results of Vilardell-Bartin et al., 2015 for western subspecies). However, increase of ambient temperature during the summer could have the largest influence on the high percentage of tortoises present in the dense forests and dense shrublands in this season. The findings of Meek (1988) on thermal ecology of T. hermanni boettgeri indicated that tortoises' presence in the summer season, in different Mediterranean habitats of Montenegro, was high in shrublands, despite available wooded areas. However, dense shrubs act as effective shade enabling effective thermoregulation, which is in accordance to our finding. Additionally, Meek (1988) compared only shrub and woodland areas during different weather conditions, whilst in our study we distinguish 9 specific macrohabitats within completely different study design. Nev-ertheless, Meek's finding emphasize the importance of ambient temperatures in thermoregulation, as well as in habitat preference of eastern subspecies of T. hermanni. Tortoises' preference to closed habitat type in our study was most pronounced at the localityČermor. This locality abounds closed types of vegetation and favorable shelters for the tortoises during high summer temperatures. Kunovica locality is very similar toČermor, but tortoises there mostly preferred grassland habitat type. The study conducted in Kunovica during the period 2010-2014 (Stojadinović et al., 2017) has shown that the frequency of tortoises there varied significantly among the years and it could happen that the meteorological conditions during data collecting were not favoring closed macrohabitat types. On the contrary, the localities Gonjište and Pašinačesma lack the dense forest, and we suppose that tortoises there mostly choose dense shrubland because it provides the best available shelter from high ambient temperatures.
Our findings on the distribution of Eastern Hermann's tortoises in this range of habitats supported results of previous studies carried out by Meek (1984Meek ( , 1985Meek ( , 1988 in the Mediterranean parts of Montenegro and Croatia and, for this part of subspecies area, by Stojadinović et al. (2017), but also provided detailed insight into the habitat preferences of Eastern Hermann's tortoise on a finer scale. Despite research conducted by Rozylowicz and Popescu (2013), that confirmed selection of habitats on the rough scale of grasslands, shrubs and forest edges, it showed that tortoises haphazardly occurred in forest habitat. However, in this study, there is a statistically significant difference between the expected and the recorded number of tortoises (especially in localitiesČermor and Kunovica) in the forest habitats. We partitioned forest habitats into several types of macrohabitats, and even the dense forest habitat type was the one with a high percentage of individuals at more than one locality (table 2). The difference among the results obtained at four localities may be due to different environments i.e. different habitat matrices among the localities. This also indicates the importance of precise definition of macrohabitats, in as much details as possible, to the point on how complex are the macrohabitats preferred by Eastern Hermann's tortoises. That could be the explanation for some discrepancy between our and Stojadinović et al. (2017) study, and for difference in preferred habitats among the years of study at the Kunovica locality.
The fact that there is no statistically significant difference between the recorded and the expected number of individuals per habitat type on the Pašinačesma locality can be explained by the uniformity of the local environment. There, 66% of the habitat there consisted of dense shrubland, while the other types of habitat are significantly less present (table 2). The cause of such one habitat structure is anthropogenic: this locality represents a neglected vineyard, which has been out of use for more than a decade. At this locality, we found that the distribution of tortoises among macrohabitats was concordant with their proportion in the total investigated area (Pearson's r = 0.92 and r = 0.88 for spring and summer, respectively). Again, the reason can be the uniformity of the environment and perhaps low diversity of macrohabitats (see fig. 2) at this locality. Therefore, the largest number of tortoises there was, in both seasons, detected in the dominant habitat type -dense shrubland (fig 2.), in agreement with earlier studies in Greece, Croatia and Montenegro (Meek, 1984(Meek, , 1985(Meek, , 1988. On the other side, Gonjište andČermor exhibited the same characteristics in some seasons. At the locality Gonjište, in the spring, the largest number of tortoises (even 50%) was recorded on the grassland habitat type, which is also the dominant habitat type on this locality (table 2). As the spring represents the main period of activity and nesting for tortoises (Vilardell-Bartin et al., 2015) and temperatures are not so high, we assume that this may be the reason for such spatial arrangement of tortoises among macrohabitat types. In the summer, temperatures are higher and therefore tortoises avoid open macrohabitat types, like meadows and open shrublands. For that reason, tortoises on the locality ofČermor, in the summer, would mostly choose dense forest (r = 0.96).
Our results showed that distribution of tortoises on the territory of Kunovica was not correlated with the proportion of individual macrohabitats (r = 0.36 in the spring and r = −0.003 in the summer). Kunovica also represents the largest investigated area in this study. In addition, this is the locality where all macrohabitat types from our list are present, and the dominant habitat type occupies less than 40% of the total territory. For comparison, the dominant macrohabitat type on other localities occupied more than 45% of the total territory (table 2). The high diversity of macrohabitat types in Kunovica, together with relatively low proportion of dense forest type could impact on relatively low number of tortoises recorded specifically in this macrohabitat.
The results of our study on the microhabitat scale showed that, throughout the year, the eastern subspecies of Hermann's tortoise prefers herbaceous habitats in the both parts of season, bramble during the breeding season (spring) and slightly more bramble and trees in the summer ( fig. 3). Eastern form of Hermann's tortoises in general choose similar types of microhabitats in both seasons. In the summer, large number of tortoises at the localityČermor was found within the microhabitats predominantly consisted of trees. Tortoises at the locality Gonjište were mostly found alongside of bramble, concordant with the fact that the majority of them choose dense shrubland habitat type at this locality ( fig. 2), probably to avoid high midday temperatures in the absence of closed macrohabitats (see table 2).
This study has shown the complexity of eastern Hermann's tortoise habitat requirements and the necessity for maintaining this habitat diversity on a fine scale. Continuous monitoring is obviously necessary for better understanding of ecology of Hermann's tortoise, as this species is nowadays under tremendous anthropogenic pressure throughout its entire range. The results should contribute to efficient conservation planning related to either the preservation of already existing network of microhabitats important for the tortoises, or to their restauration where required (Rozylowicz and Popescu, 2013;Couturier et al., 2014), where the knowledge on macro-and microhabitat use has considerable practical value for sustainable management. In the long term it should lead to minimalization of decline of natural populations and prevention of species extinction because of habitat degradation and/or destruction. It could also support implementation of ex situ conservation measures.