Turnover of Plant Species on an Ecological Gradient in Karst Dolines Is Reﬂected in Plant Traits: Chorotypes, Life Forms, Plant Architecture and Strategies

: We analyzed plants and their traits in dolines, which are characteristic enclosed terrain depressions on carbonate (karst) plateaus. These landforms range from a few meters to over 100 m in diameter, their depth generally varying from a few meters to a few tens of meters. A pronounced ecological gradient can be found from the bottom to the top, starting from humid, cool and shaded bottoms to sunny, dry and warm slopes and tops. We sampled dolines of various depths and analyzed the distribution of plant species on the gradient and how this distribution is reﬂected in plant traits: chorotypes, life forms, plant architecture and strategies. We used the transect method and sampled the ﬂoristic composition from the doline bottom to the top. We collected information about plant traits from various literature sources. The results show life forms and plant architecture explain this gradient well and, to a lesser extent, also chorotypes, but functional strategies have a low explanatory power. Life forms and plant architecture are the result of adaptation of species to the environment, and chorotypes are deﬁned as species with an overlapping geographical distribution pattern due to their distribution and environmental histories. Functional strategies, which have evolved to enable plants to succeed in various environments, unexpectedly have a low explanatory power.


Introduction
Enclosed terrain depressions termed dolines (sinkholes in American literature) can be found on carbonate (karst) plateaus representing a characteristic karst surface feature. Environmental factors in dolines deviate markedly from the surroundings, and this reflects also in the appearance of plants species and their communities.
The significance of dolines for local biodiversity has been recognized in the past. Beck-Mannagetta [1] stated that alpine plants in dolines in the Dinaric Alps are remnants of the glacial period. Horvat [2] presented vegetation zonation within various dolines over the whole Dinarides. Bátori and his team [3][4][5][6][7][8][9][10] elaborated on many aspects of dolines. They could be refugia for endangered cool-adapted plant species that have found safe havens here. Besides vascular plants, many other important diagnostic groups of species (e.g., ants) have been found in doline. They used the same transect method as in the present study. Kermavnar and his team [11][12][13][14] studied the development of vegetation in the dolines after cutting. They evaluated the microclimatic conditions and initial responses of floor vegetation (plant species and traits) after cutting. All referenced studies identified different ecological conditions in dolines, caused primarily by geomorphology.
Dolines are formed by dissolution of sedimentary carbonate bedrock (limestone, dolomite) and appear in different sizes and shapes. They are commonly sub-circular in plan According to the presented theoretical backgrounds, the following research questions were the focus of the study. (a) Is the ecological gradient from the bottom of a doline to its top the most important for plant species turnover? (b) Do only deeper dolines possess the whole gradient? (c) Can we expect more species with competitor strategy in the bottom of dolines? (d) Are there more species originating from the northern chorotypes in the bottom, whereas more species of southern chorotypes on the top? (e) Can we expect more geophytes in the bottom and more hemicryptophyte on the top? (f) How are ecological conditions reflected in the architecture of plants? (g) Can functional strategies best explain the species turnover?

Study Area
The research took place on Kras Plateau, a limestone karstic plateau lying above the Bay of Trieste in the northernmost part of the Adriatic Sea at an altitude of 300-500 m. It is an example of an uplifted and slightly leveled corrosion plain surface in the NW part of the Dinaric Mountains [41].
The climate is transitional between Mediterranean and continental (i.e., sub-Mediterranean), with rainy, cool winters and hot summers. The precipitation amount is around 1400 mm and the average annual temperature around 11 • C. Kras Plateau consists of karstified Mesozoic limestone covered predominately by Rendzinas and Cambisols. The main forest tree species in the area are hop hornbeam (Ostrya carpinifolia) and pubescent oak (Quercus pubescens) [39,42,43]. Kras Plateau is on the border between continental and Mediterranean biogeographic regions [44]. Since disturbance has a negative impact on vegetation and could impact the natural gradient, we sampled floristic composition only in forested dolines, which are the least disturbed habitats in the landscape, with high conservation values [9]. The many dolines (over 14,000) on Kras Plateau are quite diverse, having various depths, diameters, shapes, and soil depths (on average 3 m, while on slopes, the soil depth is related to the type of doline) [45]. Oak-hornbeam, or even beech forests can be found at the bottom of dolines. On slopes, there are ravine forests dominated by sycamore and Norway maple (Acer pseudoplatanus, A. plantanoides), little-leaved and large-leaved limes (Tilia cordata, T. platyphyllos), wych elm (Ulmus glabra), and common ash (Fraxinus excelsior). On slopes exposed to the south that are warmer due to environmental conditions, forests dominated by Turkey oak (Quercus cerris) and hop hornbeam (Ostrya carpinifolia) appear [26,[46][47][48].

Vegetation Sampling
We sampled vegetation (floristic composition) by transects. Each transect consisted of adjoining 2 m × 2 m large plots, positioned from south-north, from one edge of the doline through the bottom to the other edge [32,49,50]. The transects were of different lengths depending on the size of the doline (Table 1, Figure 1). We thus recorded the greater part of (vascular plant) biodiversity. Distance from the edge of doline was measured and depth of individual sample plot was determined from Lidar data. All ground floor vascular plants were identified, and we estimated the cover of individual plants visually in each plot by the 7-degree scale proposed by Braun-Blanquet [51]. We also visually estimated the proportion of bare rock on the surface in each plot. The plots were stored in the TURBOVEG database program [52] and entered in the JUICE program for analysis [53].We obtained a matrix of 286 plots × 124 plant species that were used in the analysis.
Detrended correspondence analysis (DCA) was performed on the plot matrix (Table S1). The original cover values of plants (Br.-Bl. values) were transformed to percentages and subjected to square root transformation. We extracted the first two axes of the analysis (Axis 1 and Axis 2) and used them in further analyses. The first axis represents the main floristic gradient. The calculations were done using the mass module in the vegan program package [54] run R program environment [https://cran.r-project.org, accessed 20 December 2021]. (Axis 1 and Axis 2) and used them in further analyses. The first axis represents the main floristic gradient. The calculations were done using the mass module in the vegan program package [54] run R program environment [https://cran.r-project.org, accessed 20 December 2021]. We calculated unweighted bioindicator (Ellenberg) values (EIV) [55,56] for each plot using the JUICE program [53] and passively projected them onto an ordination plane.

Analyzing Plant Composition
We calculated the chorotypes, plant functional traits and strategies of the studied vegetation plots by community-weighted means (CWM) of each trait. This can be considered to be the average trait value in the vegetation plot reflecting the relative abundances of species. As abundance of the species, we took the estimated cover value transformed into percentage [51]. We used unweighted indicator values. The results were visualized by their projection onto the ordination plane. We calculated unweighted bioindicator (Ellenberg) values (EIV) [55,56] for each plot using the JUICE program [53] and passively projected them onto an ordination plane.

Analyzing Plant Composition
We calculated the chorotypes, plant functional traits and strategies of the studied vegetation plots by community-weighted means (CWM) of each trait. This can be considered to be the average trait value in the vegetation plot reflecting the relative abundances of species. As abundance of the species, we took the estimated cover value transformed into percentage [51]. We used unweighted indicator values. The results were visualized by their projection onto the ordination plane.
We calculated plant functional traits according to the morphological features of plants [57], based on Raunkier's life forms [33], as nanophanerophytes, hemicryptophytes, therophytes and geophytes, and architecture of growth defined by Pignatti [25] as caespitose, rosulate, suffrutescens, liane, crawling and rhizomatous, to mention only the most significant ones.
We also calculated functional strategies, for which we used Grime's model of CSR [35], as competitors, stress tolerators and ruderals. This model and determination of functional strategies are the result of long-term research conducted in field surveys, laboratories, monitoring of permanent plots and manipulative experiments. There also exist methods for rapid determination of strategy based on canopy height, dry matter content, flowering period, flowering start, lateral spread, leaf dry weight and specific leaf area [58]. The position of each plot can be determined in the CSR triangle. The community thus obtains a functional signature [59]. The data were provided by the Biolflor database [60] and are shown in Table S1.
We tested the effect of the explanatory power of variables on floristic composition using CCA ordination analysis in CANOCO using the Monte Carlo permutation test with 9999 permutations. Since the results of variation partitioning overlapped, they were displayed in a diagram using Venn circles [61,62].

Ordination
Detrended correspondence analysis showed that the main floristic gradient along the slope from the bottom of dolines to their top is reflected on the first axis (Axis 1), while the second axis (Axis 2) represents the proportion of bare rock on the surface (Figures 2 and 3). The highest species richness was found on the bottom, while the most moderate was on steep and stony slopes ( Figure 2). Humid and nutrient-rich sites can be found in the doline bottoms, whereas the tops are warm, dry, and sunny. Rocky sites can be found on some doline slopes (Figures 2 and 3).
We calculated plant functional traits according to the morphological features of plants [57], based on Raunkier's life forms [33], as nanophanerophytes, hemicryptophytes, therophytes and geophytes, and architecture of growth defined by Pignatti [25] as caespitose, rosulate, suffrutescens, liane, crawling and rhizomatous, to mention only the most significant ones.
We also calculated functional strategies, for which we used Grime's model of CSR [35], as competitors, stress tolerators and ruderals. This model and determination of functional strategies are the result of long-term research conducted in field surveys, laboratories, monitoring of permanent plots and manipulative experiments. There also exist methods for rapid determination of strategy based on canopy height, dry matter content, flowering period, flowering start, lateral spread, leaf dry weight and specific leaf area [58]. The position of each plot can be determined in the CSR triangle. The community thus obtains a functional signature [59]. The data were provided by the Biolflor database [60] and are shown in Table S1.
We tested the effect of the explanatory power of variables on floristic composition using CCA ordination analysis in CANOCO using the Monte Carlo permutation test with 9999 permutations. Since the results of variation partitioning overlapped, they were displayed in a diagram using Venn circles [61,62].

Ordination
Detrended correspondence analysis showed that the main floristic gradient along the slope from the bottom of dolines to their top is reflected on the first axis (Axis 1), while the second axis (Axis 2) represents the proportion of bare rock on the surface (Figures 2  and 3). The highest species richness was found on the bottom, while the most moderate was on steep and stony slopes ( Figure 2). Humid and nutrient-rich sites can be found in the doline bottoms, whereas the tops are warm, dry, and sunny. Rocky sites can be found on some doline slopes (Figures 2 and 3).    It can be observed that only plots from the deepest dolines are positioned along the whole gradient, while all plots from shallow dolines are positioned only in the left-hand part of the gradient (Figure 4).  Only plots belonging to A4 and A8 are presented for clarity (A4 lines in yellow and orange shades, A8 lines in blue and purple shades). Plots belonging to the same transect are connected. Each doline has two transects, from the center to the north and from the center to the south. The difference between shallow (A4) and deep (A8) dolines is evident.

Explanatory Power of Chorotypes, Functional Traits and Strategies
Chorotypes (Figure 5b) reflect the origin of vegetation, and most species can be found with Eurocaucasian and Northern chorotypes (e.g., Corylus avellana, Galanthus nivalis, Galeobdolon montanum) in the humid, nutrient rich and shaded bottom of the dolines. The central European and Mediterranean chorotype appear on stony slopes (Campanula pyramidalis, Frangula rupestris, Moehringia muscosa). On less stony slopes with deeper soils (reflecting lower pH), many Balkan species (Cnidium silaifolium, Iris graminea, Symphytum tuberosum) appear. Eurasian, SE European and Mediterranean chorotypes appear on top plots that are dry and warm with high light availability (Bromopsis erecta, Carex humilis, Vincetoxicum hirundinaria).
Plant functional strategies (Figure 5a) reflect the functioning of communities. Most species with a ruderal strategy can be found, coping with disturbances (e.g., wild boars, past land usage) in the bottom of dolines. A stress tolerator strategy is somewhat more pronounced on stony slopes, where site conditions are the hardest. A competitor strategy is slightly more pronounced on top, where disturbances and hard conditions are less pronounced. The CSR signature ( Figure 6) shows that there is not much variation in functional strategies, and this also explains the low explanatory power of this trait. Only plots belonging to A4 and A8 are presented for clarity (A4 lines in yellow and orange shades, A8 lines in blue and purple shades). Plots belonging to the same transect are connected. Each doline has two transects, from the center to the north and from the center to the south. The difference between shallow (A4) and deep (A8) dolines is evident.

Explanatory Power of Chorotypes, Functional Traits and Strategies
Chorotypes (Figure 5b) reflect the origin of vegetation, and most species can be found with Eurocaucasian and Northern chorotypes (e.g., Corylus avellana, Galanthus nivalis, Galeobdolon montanum) in the humid, nutrient rich and shaded bottom of the dolines. The central European and Mediterranean chorotype appear on stony slopes (Campanula pyramidalis, Frangula rupestris, Moehringia muscosa). On less stony slopes with deeper soils (reflecting lower pH), many Balkan species (Cnidium silaifolium, Iris graminea, Symphytum tuberosum) appear. Eurasian, SE European and Mediterranean chorotypes appear on top plots that are dry and warm with high light availability (Bromopsis erecta, Carex humilis, Vincetoxicum hirundinaria).
Plant functional strategies (Figure 5a) reflect the functioning of communities. Most species with a ruderal strategy can be found, coping with disturbances (e.g., wild boars, past land usage) in the bottom of dolines. A stress tolerator strategy is somewhat more pronounced on stony slopes, where site conditions are the hardest. A competitor strategy is slightly more pronounced on top, where disturbances and hard conditions are less pronounced. The CSR signature ( Figure 6) shows that there is not much variation in functional strategies, and this also explains the low explanatory power of this trait.

High Species Richness in Dolines
Our research confirmed the results of previous studies, that dolines possess different ecological conditions than surrounding vegetation. They offer favorable climatic conditions for cool-adapted plant species and are often treated as safe havens in the event of foreseen climatic changes [49], although it should be taken into consideration that proba-

High Species Richness in Dolines
Our research confirmed the results of previous studies, that dolines possess different ecological conditions than surrounding vegetation. They offer favorable climatic conditions for cool-adapted plant species and are often treated as safe havens in the event of foreseen climatic changes [49], although it should be taken into consideration that proba-

High Species Richness in Dolines
Our research confirmed the results of previous studies, that dolines possess different ecological conditions than surrounding vegetation. They offer favorable climatic conditions for cool-adapted plant species and are often treated as safe havens in the event of foreseen climatic changes [49], although it should be taken into consideration that probably some species (e.g., Sesleria autumnalis or Paeonia officinalis) might not be capable of retreating from increasing heat and drought to the bottom of dolines due to other factors, such as the lack of light and soil conditions at such sites. Dolines already offer space for various biota [8,63] that would otherwise be excluded from, or be much rarer, in the local species pool (e.g., Dentaria enneaphylos, which needs moist and cool conditions and Campanula pyramidalis, which needs rocky crevices). It is important to note that the species richness at the bottoms of the deep dolines is consistently higher compared to shallow dolines and tops on the plateau. This is especially significant when accounting for the much smaller area covered by doline bottoms.

Plant Changes along the Ecological Gradient in Dolines
We established that moist and nutrient-rich habitats appear in the bottom of dolines, dominated by oak-hornbeam forests (Quercus petraea-Carpinus betulus) [64]. They are characterized by a high proportion of geophytes [65]. Geophytes are early spring perennial plants that are common in broad-leaved deciduous forests, where they appear before spring leafing takes place [66]. There are also harsh conditions, with thermal inversion and snow accumulation [2], and plants invest carbohydrates into below-ground organs of perennation instead of above-ground biomass. This increases protection from freezing, and plants are additionally protected by plant litter and snow cover [67]. Many wild boars appear in the region, and they are active in rooting and looking for food [68], which supports the ruderal functional strategy.
Since our results show an increased tendency for stress-tolerant plants in the bottom, compared to the plateau area, it may be that the main stress factor for the herb layer over all is not drought or heat, which are the expected stressors on the plateau, but the lack of light and activity of wild animals, both more prominent at the bottom of dolines [9,68]. However, since the strategies alone explained very little of the total variance, further studies are needed to discover the significance of the pattern of strategies distribution.
On slopes, ravine forests can be found, dominated by sycamore (Acer pseudoplatanus), little-leaved lime (Tilia cordata), and large-leaved lime (Tilia platyphyllos) [48]. These forests are adapted to stony slopes with moving rocks on the surface. These habitats are characterized by scapose phanerophytes, woody species that are adapted to moving bedrock [69]. Many of these scapose phanerophytes (e.g., Fraxinus ornus) are of SE European origin. In cases in which wood species cannot survive these hard conditions, chasmophytic vegetation composed of rosette, biennial and scapose hemicryptophytes appears [70], with which the dominant functional strategy is stress tolerance [71,72].
Plots on top can be classified within sub-Mediterranean hop hornbeam-pubescent oak (Ostrya carpinifolia-Quercus pubescens) forests [73] dominated by caespitose phanerophytes that often build coppiced forests [74]. Due to the open canopy, a well-developed herb layer can be found dominated by caespitose hemicryptophytes (graminoids). Since many graminoids belong to the Eurasian chorotype, originating from non-forested areas situated in the eastern part of Europe and further in Asia [75], this chorotype is well represented in the region. It seems that the competitor functional strategy is not characteristic in these plots, but it is only diagnostic, since the stress tolerance strategy is more pronounced on rocky slopes and ruderal in deep humid plots in the bottom of dolines.

High Explanatory Power of Functional Traits (Life Forms, Plant Architecture) in Relation to Functional Strategies
This is the first study to have compared the explanatory power of chorotypes, functional traits and strategies in relation to ecological changes in karst dolines. The results show that functional traits best reflect ecological changes from the bottom to the top of dolines. Functional traits (life forms and architecture of plants) explain 68% of explained variance.
Chorotypes are defined as species with overlapping distribution patterns [24] and they cannot, therefore, fully reflect ecological differentiation along the gradient. Some differences can be found, since it can be hypothesized that species from a certain chorotype appear in specific site conditions (e.g., Mediterranean species appear in warm and xeric site conditions), but it is not a basis for a definition of chorotype.
The functional strategies do not, in practice, reflect ecological differences and explain very little variation. Plots are positioned mainly along the CS gradient that is characteristic for forest communities [29]. We expected a stress tolerance strategy to prevail on slopes and the top; however, the current results show this strategy is more prevalent on slopes and at the bottom. It is not yet clear whether this is an artefact due to the low explanatory power of the strategies or a real result. If it is not an artefact, one possible explanation for the prevalence of stress-tolerant plants at the bottom of dolines is that factors not yet fully considered should be taken into account. Bottoms and north facing slopes of dolines receive less light, which could be a contributing factor, as well as possibly less favorable, clayey soil conditions at the bottom and the presence of rocks and moving slopes there. Although the trees appear to show that heat and water stress are the main stress factors for the absence of otherwise big and dominant trees such as Fagus sylvatica, Carpinus betulus from the plateau, which are replaced by Quercus pubescens and Ostrya carpinifolia, and by decreased sizes of the trees, the herb layer looks as if it may have a different response to the same macroclimatic conditions.

Conclusions
Our research found that functional traits best reflect the ecological gradient from the bottom to the top of dolines. We confirmed that the main floristic gradient is from the bottom of a doline to its top, and reflect the change of ecological conditions and only the transects of deeper dolines extend over the whole gradient. We found that species from the bottom originate from cooler regions and species on the top from warmer ones. We found more geophytes in the bottom and hemicryptophytes on the top. Concerning architecture, we found rhizomatous and crawling species in the bottom, liane, and rosulate on rocky slopes, whereas the tops are dominated by caespitose and suffrutescens plants. We could not confirm the hypothesis that competitor strategy dominates in the bottom, and that strategies are among the traits that best explain the species turnover. However, this pattern should be verified in other regions with different macroclimatic conditions.
The research contributed to the identification of dolines with a high conservation value, as only deeper dolines can provide shelter for rhizomatous species originating from cooler regions. On rocky slopes, we found therophytic and rosulate species that also contribute to the biodiversity of the karstic landscapes.