Roadsides provide refuge for orchids: characteristic of the surrounding landscape

Abstract Seminatural habitats are declining throughout the world; thus, the role of small anthropogenic habitats in the preservation of plants is becoming increasingly appreciated. Here, we surveyed the orchid flora of roadside verges in five Central European countries (Austria, Hungary, Romania, Slovakia, and Slovenia) and tested how the surrounding landscape matrix affects the overall number of species and individuals, and also different functional groups of orchids. We found more than 2,000 individuals of 27 orchid species during our surveys. According to our results, the increasing coverage of agricultural and urban areas negatively affects both the number of orchid species and individuals on roadsides. Our study further suggests that differences in the surrounding habitats affect which species are found on roadsides, since the increasing coverage of grasslands or forested areas around orchid occurrences had a significant positive effect on the number of grassland or forest‐dwelling species and individuals, respectively. Most variance in orchid numerosity and diversity was explained by the cover of the suitable habitat types of the respective taxa in the surrounding landscape of the sampling points. This highlights the importance of roadsides acting as refugia for numerous species and valuable plant communities as well as in supporting biodiversity in general.


| INTRODUC TI ON
Seminatural habitats are dominated by native flora, characterized by a typical diversity and species composition, but they bear human-induced alterations. Despite disturbance, they generally host a large number of threatened plant species and have long been considered as hotspots for biodiversity (Benton et al., 2003;Henle et al., 2008).
During the last century, intensification of human activities has led to a dramatic reduction of seminatural habitats, contributing to the severe decline of biodiversity worldwide (Butchart et al., 2010;Hooftman & Bullock, 2012;Malcolm & Markham, 2000). Seminatural habitats are subject to several threats, such as habitat degradation, destruction and fragmentation (Nascimbene et al., 2016;Tikka et al., 2000;Tilman et al., 2001), intensification of agricultural land use, and abandonment of traditional agricultural practices (Bignal & McCracken, 1996). For instance, seminatural grasslands are often converted into arable lands in response to a higher demand for food production (Hodgson et al., 2005) and afforested for timber production (Mason, 2007), or they are frequently lost to urbanization (Feranec et al., 2010). These alterations to natural and seminatural habitats urge the shift of focus of conservation-oriented research toward anthropogenic habitats that have a potential to provide refuge for native flora elements. Several anthropogenic habitats were shown to provide such refuges, including cemeteries, poplar plantations, orchards, and roadsides (Bódis et al., 2018;Djordjević & Tsiftsis, 2020;Fekete et al., 2017;Löki et al., 2015;Süveges et al., 2019).
Given that appropriate management practices are adopted, road verges might allow the persistence of valuable grassland communities, as well as the maintenance of rare species, conveying a significant conservation value to these man-made habitats (Auestad et al., 2011;Hovd & Skogen, 2005). In some regions, roadside habitats have even been regarded as "Roadside Nature Reserves" and they receive special management due to their recognized conservation priority (Dawson, 1991;Parr & Way, 1988;Spooner, 2005).
In the United Kingdom, for instance, almost half of the native plant species can be found on roadsides (Way, 1970). Moreover, even regularly mowed roadsides can serve as refuges for endangered grassland species, for instance, the highly cut-tolerant Gentianella campestris in Finland (Huhta & Rautio, 2007).
Roadsides not only provide refuge to native flora, but were shown to serve as ecological corridors for a wide range of taxa (Gustafsson & Hansson, 1997;Haddad et al., 2003). Nonetheless, at least in the case of plants, the ecological corridor role of roads is most widely demonstrated by the dispersal of alien and invasive species alongside them (Bacaro et al., 2015;Das & Duarah, 2013;Gulezian et al., 2012;Joly et al., 2011;Lin, 2007). Ecological or landscape corridors are strips of suitable habitats connecting isolated habitat patches and are considered to facilitate gene flow and the movement of species between these, thus increasing the number of native species in large-scale communities and reducing the negative effects of fragmentation (Cody et al., 1975;Damschen et al., 2006).
Creating ecological corridors recently became a widespread ecological management practice; nonetheless, some studies question their effectiveness in aiding the movement of organisms between otherwise fragmented habitats (Beier & Noss, 1998;Hobbs, 1992;Merriam & Saunders, 1993;Simberloff et al., 1992). A survey by Tikka et al. (2001) suggests that roadsides play a direct role in the dispersal of grassland communities, clearly serving as ecological corridors. Nonetheless, Fritz and Merriam (1993) found no support of the role of fencerows serving as ecological corridors of forest floor herbs. A number of recent studies suggest a rather indirect impact of linear landscape elements on the spread of taxa, by facilitating pollination and seed dispersal between suitable habitats or by promoting plant-animal interactions (Haddad et al., 2003;Tewksbury et al., 2002;Van Rossum & Triest, 2012). According to previous studies, there could be a difference between grassland and woodland plant species in the use of ecological corridors. For instance, grassland species can easily spread to alternative open habitats, such as roadsides due to regular mowing of verges which keeps the vegetation low, but in the case of woodland species, the use of corridors is more difficult (Fritz & Merriam, 1993;Tikka et al., 2001). In the case of roadsides, the dispersal of small seeds can also be facilitated by the air turbulence of cars (Ross, 1986) or by the mud attached to the vehicles, which often contain large number of seeds, especially when the roadside vegetation is well developed (Clifford, 1959). Recent studies highlight the role of vehicles in the spread of alien species (Von Khan et al., 2018han et al., 2018Nguyen, 2011;der Lippe & Kowarik, 2007).
Although roadsides are often isolated remnants of seminatural habitats, their species richness is largely dependent on the surrounding landscape (Cousins, 2006). For instance, Cousins and Lindborg (2008) noted that in the most intensively managed landscapes, the species richness of roadsides and midfield islets declines with increasing distance to seminatural grasslands.
The central aims of the current study were to (a) assess species and individual numbers using roadsides as habitats in Central European countries and (b) examine how the landscape matrix affects the species composition, diversity, and abundance of orchids generally and also different functional groups of orchids in the sampled habitats.

| Fieldwork
Field sampling was carried out in five Central European countries (Austria, Hungary, Romania, Slovakia, and Slovenia). Two types of sampling processes were adopted. First, we conducted thematic sampling by driving along asphalt roads and we stopped in every 5 km. Second, we conducted non-thematic sampling, meaning that we stopped at every road section, where orchids were spotted from the car while driving. Details of the sampling localities and sampling periods are given in Table 1. At every sampling point, we recorded geo-coordinates (WGS84 format) and altitude (m) using a Garmin eTrex Legend GPS Device. Where orchids were present, we additionally recorded the list of orchid species and the number of specimens belonging to each of these along a 50-m road section on one side of the road. The width of the surveyed area usually spanned from 0 to 10 m, being delimited by roadway on one side and ditches, walls, or taller vegetation on the other side. In some cases, identification of orchids to the species level was not possible, due to their vegetative phenological state. In the latter case, we counted the number of individuals, but these were only included in the overall count of orchid individuals. Taxa were identified following Delforge (2006), and the nomenclature used in this work follows this source. CLC database, 28 were present in the buffer zones of our sampling points. We estimated land cover for each of these categories, but we subsequently performed a categorization of these in order to reduce dimensionality in the analyses. We considered watercourses and water bodies as unsuitable places for orchids. We did not join the different forest types, since broad-leaved forest serves as habitats for some species which would not prefer shady coniferous forests; thus, different species have different forest type needs. We further considered vineyards, fruit trees, berry plantations, and land principally occupied by agriculture with significant areas of natural vegetation as "semi-agricultural areas," because when they are abandoned or extensively used, they could serve as orchid habitats. Details of the categorization are given in Table S1.

| Statistical analyses
Statistical analyses were carried out in the R statistical environment (version 3.4.1, R Core Team, 2017).
To avoid multicollinearity in the models, we performed VIF selection (Craney & Surles, 2002) using vif function in fmsb package (Nakazawa, 2017) which calculates the VIF values for all of our explanatory variables, then removes the variable with the highest value, and repeats this until all VIF values are below the threshold, which in our case was "2." Following this, the concerned variables were removed from the analyses. After testing the distribution of TA B L E 1 Number of the different sampling points and the date of the surveys carried out in the sampled countries the data, we built generalized linear mixed models (GLMM) with quasi-Poisson distribution due to significant overdispersion in the independent variables, using the glmmPQL function (MASS R package, Ripley et al., 2013). In all cases, we started by building full models containing all explanatory variables which were selected by VIF.
This was followed by model simplification, when predictors were removed from the model using a stepwise backward procedure based on the largest p values. All predictors with p < 0.1 were retained in the minimal model. In all of the models, we used "country" and "sampling type (thematic or non-thematic)" as random factors. and Natural unsuitable places for vegetation within the 1-km-radius circle (the variable Coniferous forests was highly correlated with several land cover types and was thus eliminated from multivariate analyses by VIF). Another eight models were built with the same variables using 10-km-radius circle data (Table S3).

| General results
Out of the 465 thematic sampling points, we found orchids at 83 locations, with records in all of the five surveyed countries (Figure 1.).
The ratio of the sampling points where orchids were present was the highest in Slovenia and the lowest in Hungary ( Using the non-thematic sampling protocol, at two sampling points in Austria, we found 23 individuals belonging to four species, including the rare and local G. lithopolitanica. In Hungary, 272 individuals were found belonging to 13 species at 33 non-thematic sampling points. In Romania, 53 individuals from four species were found at four sampling points. In Slovakia and Slovenia, one non-thematic locality was surveyed in both countries. In Slovakia, nine individuals were found from the species Orchis mascula, and in Slovenia, 61 individuals were found from two species (D. fuchsii and Epipactis helleborine) during non-thematic surveys.

| Landscape analyses
Multivariate models indicated a significant lower number of orchid species (Figure 2a,b) and individuals in sampling locations where the area of urban and agricultural land covers was higher within a 1 km radius (Table 3.). In the case of the total number of species, the cover of natural grasslands and pastures and broad-leaved forest also had a significant negative effect.
In the 1-km-radius circle, the cover of natural grasslands and pas-

| D ISCUSS I ON
During our extensive field surveys, we found more than 2,000 orchid individuals of 27 different orchid species in roadsides. This alone suggests that roadsides provide an important habitat for orchids in Central Europe, similarly to other regions across Europe (Fekete et al., 2017 (Rankou, 2011). The highest number of orchid individuals present on roadsides belonged to D. fuchsii, a species that is characterized by a broad ecological tolerance.
Roadsides surveyed here hosted almost twice as many grassland specialist orchid species as forest specialists (i.e., 13 and 7, respectively), which might potentially be explained by road maintenance practices, namely the regular mowing of roadsides. Due to the latter, vegetation on roadsides is usually less closed, while mowing is known to have positive effects on grassland orchids in other habitat types (Curtis, 1946;Janečková et al., 2006;Sletvold et al., 2010;Smith & Cross, 2016). Forest specialist orchid species were also present on roadsides surveyed, being represented by more than 200 individuals. This suggests that despite being less adapted for regular mowing, they are able to cope and maintain populations in these anthropogenically influenced habitats. Roadsides are narrow grassland fragments and could act as ecotones (representing mainly transitions from grasslands to forest edges). Due to their weak competitive ability, orchids are frequently found in transitional, ecotone habitats, such as mesoxeric scrubland patches and forest edges (Bray & Wilson, 1992;Djordjević et al., 2016;Duchoň, 2012;Rai et al., 2010;Slaviero et al., 2016). Furthermore, they often colonize newly created habitat patches (such as roadsides), where the abundance of dominant plant species and the cover of trees and shrubs are low (Arditti & Ghani, 2000). Based on Grime's theory, orchids are considered as stress tolerators or ruderal species, (Dressler, 1981;Hágsater & Dumont, 1996) competing for resources and space, thus orchids around roadsides might compete for favorable light conditions, that study has already indicated that the surrounding landscape matrix has a high impact on species composition of habitats along roads (Tikka et al., 2000). Moreover, a decline in species richness on linear landscape fragments with increasing distance from seminatural habitats was also reported (Cousins & Lindborg, 2008). A study discussing drainage ditches showed that the proximity of natural grasslands TA B L E 3 The eight minimal models (GLMMPQL) explaining variance in the number of species (left) and number of individuals (right) in overall, in grassland specialist and forest specialist and broad ecological tolerance orchids, respectively, in function of land cover within a 1 km radius increased the value of grassland vegetation of the ditches subsequently filled for restoration purposes, suggesting that their vegetation is highly dependent on the landscape matrix (Valkó et al., 2017).
We found that grassland specialists are mostly present on verges, where there are suitable grassland habitats adjacent to the roadside, while forest specialist is more common on roadsides where there are forests in the surrounding landscape. It is important to note that these linear landscape elements are often highly influenced by agricultural activities (e.g., use of fertilizers and herbicides) on adjacent fields (van Dorp, 1996); thus, species of nutrient-poor ecosystems (such as orchids) are particularly unlikely to migrate along these elements (van Dorp et al., 1997;Thiele et al., 2018). When the landscape matrix environment is unsuitable for the dispersal of plants along ecological corridors, or dispersal is ineffective due to a high percentage of low-quality patches, it is unclear whether they could truly function as a corridor.
Under such circumstances, roadside patches might rather serve as refugia (van Dorp, 1996;van Dorp et al., 1997). Although according to Forman (1991) plants may in theory migrate along ecological corridors, there has been little empirical support for this; thus, it is more likely that the dispersal of native and rare plants occurring along the linear landscape elements is saltatory. However, the spread of less sensitive, successful alien and invasive species along roads is a well-known phenomenon (Benedetti & Morelli, 2017;Dar et al., 2018;Vakhlamova et al., 2016). This is an especially likely scenario in the case of orchids since their microscopic seeds are effectively dispersed by the wind, even over long distances (Arditti & Ghani, 2000), thus facilitating the effective colonization of new habitat patches. This is in correspondence with another study, suggesting that corridor use is common mostly in case of plant species that lack the ability of long-distance dispersal (Thiele et al., 2018). Nonetheless, the conservation importance of roadsides is not to be underrated due to these facts, since these corridors most likely still function as linear reserves for plants (Forman, 1991). Consequently, the conservation value of these narrow fragments of seminatural habitats is becoming increasingly appreciated worldwide (Bernes et al., 2017;Hopper, 1990;Melman & Verkaar, 1991;Niu et al., 2019;Ryttäri & Kettunen, 1997;Saunders & Hobbs, 1991), and in some countries, they have been identified as Sites of Special Scientific Interest (Parr & Way, 1988).
Overall, it is becoming clear that plants are able to disperse to roadsides from the surrounding landscape, but the possibility exists that colonizations might occur the other way around as well, which could facilitate the natural restoration of degraded grasslands adjacent to roads. Moreover, roadsides with properly managed native vegetation could contribute to pollinator conservation, which is particularly important today, as we are facing a global pollination crisis (Hopwood, 2008;Hopwood et al., 2015;Wojcik & Buchmann, 2012).
Many orchids are specialist species; therefore, their conservation is important because their disappearance leads to functional homogenization in ecosystems, promoting biodiversity loss (Clavel et al., 2011). Thus, orchids might serve as general indicators of the ecological state of roadside vegetation. Moreover, considering tendencies of decline in seminatural habitats worldwide, it is possible that roadsides will serve as important refugia that could aid the maintenance of floristic diversity.

| IMPLI C ATI ON S FOR MANAG EMENT
Here, we emphasize that floristic surveys of roadsides and adjacent areas are key for planning appropriate road management.
We believe that management planning should be conducted in accordance with local and regional conservation efforts, because roadside vegetation and its importance changes along the roads,  (Iuell et al., 2003). Whenever possible and when more time is available for the stabilization of verges, during road building or broadening, the use of subsoil-instead of topsoil-would be favorable for the reduction of soil fertility, since high fertility F I G U R E 3 (a) Coverage of natural grassland and pasture within 1 km radius in the three orchid groups: species with broad ecological tolerance (BET), forest species (FS), and grassland species (GS).
(b) Coverage of broad-leaved forest within 1 km radius in the three groups: species with broad ecological tolerance (BET), forest species (FS), and grassland species (GS) negatively affects the floristic composition of natural grasslands (Gough & Marrs, 1990). Local origin of the soil used during construction is also very important, as soils from a different source can contain seeds of alien species (Greenberg et al., 1997). To facilitate floristic and pollinator diversity after construction, it is also favorable to revegetate roadsides using specific seed mixtures appropriate for adjacent vegetation. Pollinators are key factors in the maintenance of native vegetation on roadsides; thus, it is very important to reduce their collision with cars, by keeping the meadows a few meters away from the road's edge and keeping long continuous flower meadows, reducing their will to cross the road for flowering patches (Hopwood et al., 2010;Keilsohn et al., 2018). As a part of roadside management, regular mowing is a cost-efficient element of their maintenance, and it is obligatory in most of the countries for safety reasons. According to a previous study from the Mediterranean, the regularly mowed 0-2-m part of the roadside is the most suitable for orchid individuals .
However, there could be a difference in regularity of roadside mowing due to climatic differences between the Mediterranean and the European roadside verges. In the Mediterranean region, the growth of the vegetation could be slower, while the best practice for creating and maintaining species-rich meadows along European roads should be mowing twice per year (this being better compared to once a year), and the hay should be removed after each cutting (Jakobsson et al., 2018). The use of herbicides and paving of roadsides is strongly unadvised. We further urge local authorities to conduct appropriate field surveys and impact assessments before broadening roads.

ACK N OWLED G M ENTS
The authors are thankful to Zsófia Simon and Ildikó Juhász for their help during field surveys. RF was supported by NTP-NFTÖ-19 grant.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
All sample data used in the analyses are available from Dryad at https://doi.org/10.5061/dryad.tdz08 kpxv.