Impact of Climate Change on the Distribution of Four Closely Related Orchis (Orchidaceae) Species

: Long-term monitoring programs and population demographic models have shown that the population dynamics of orchids are to a large extent dependent on prevailing weather conditions, suggesting that the changes in climatic conditions can have far reaching e ﬀ ects on the population dynamics and hence the distribution of orchids. Although a better understanding of the e ﬀ ects of climate change on the distribution of plants has become increasingly important during the ﬁnal years, only a few studies have investigated the e ﬀ ects of changing temperature and precipitation on the distribution of orchids. In this study, we investigated the impact of climate change on the distribution of four terrestrial orchid species ( Orchis anthropophora , Orchis militaris , Orchis purpurea and Orchis simia ). Using bioclimatic data for current and future climate scenarios, habitat suitability, range shifts and the impact of di ﬀ erent abiotic factors on the range of each species were modelled using Maxent. The results revealed an increase in suitable habitat area for O. anthropophora , O. purpurea and O. simia under each RCP (Representative Concentration Pathway) scenario, while a decrease was observed for O. militaris . Furthermore, all four of the orchids showed a shift to higher latitudes under the three RCPs leading to a signiﬁcant range extension under mild climate change. Under severe climate change, a signiﬁcant decline in the distribution area at the warm edge of their distributions was observed. Overall, these results show that mild climate change may be beneﬁcial for the studied orchid species and lead to range expansion. However, continued warming may yet prove detrimental, as all species also showed pronounced declines at lower latitudes when temperature increases were larger than 4 ◦ C.


Introduction
Understanding the various factors that limit the distribution of plant and animal species is a central theme in ecology and evolutionary biology [1,2]. The topic has gained momentum as predicted changes in climatic conditions will most likely shift species' distributions and ecology [3,4]. There is mounting evidence that species have already shifted their ranges over recent decades and will continue to do so throughout the 21st century. However, the shifts in distribution are not necessarily associated with range contractions, because not all species will immediately show declines at rear edges [5][6][7][8], suggesting that mild climate change may be favorable for many species. Better insights into the processes that drive species distributions are therefore important for applied ecology and species conservation [9].
In general, a species can only persist at a given location when the number of individuals recruiting and surviving in a population is larger than the number of individuals dying and emigrating, i.e., when the population has an intrinsic population growth rate λ > 1. Understanding the processes governing population demography can therefore help to forecast which species can resist range shifts, which will successfully shift their ranges and which will show rapid declines as a result of continued climate change [10,11]. One way to assess the current and future performance of plant and animal populations involves performing demographic analyses across the entire range of the species and comparing the population growth rates and vital rates between the edge and core populations [5]. Detailed investigations of range-wide variation in the demography of two North American tundra plants (Silene acaulis and Polygonum viviparum), for example, has shown that the population growth rates of southern populations were not lower than those of northern ones and that compensatory changes in demographic rates were buffering southern populations against the negative effects of a warming climate [6]. Similarly, about 50% of all monitored populations of plant species with an arctic-alpine and boreal distribution across western North America remained stable or even increased in abundance across the rear edge populations occurring in the northern Rocky Mountains [7].
Such methods are, however, generally labor intensive, are usually conducted across multiple years at a limited number of sites, and do not allow for the assessment of population growth parameters at sites where the species is currently absent. An alternative approach that is currently widely applied, is environmental niche modeling [12][13][14][15]. In this case, geographic biodiversity records, typically in the form of observed presences, are related to the abiotic and/or biotic characteristics at those locations and are subsequently used to forecast the impact of global environmental change on the distribution of the species [9,10,16]. While these models lack direct information about the ecological processes governing range distributions, they have the advantage of allowing for the identification of potentially suitable areas across large spatial scales and the use of different climate change scenarios.
Orchids represent an interesting study system to investigate the impact of climatic conditions on current and future distribution ranges. Despite the evolutionary success of the family [17], most of its members tend to have restricted distribution areas and many species are currently in decline, most likely as a result of severe habitat loss due to land conversion and urban development, the habitat degradation of the remaining patches, over-collection and a loss of critical ecological interactions [18][19][20]. Although there are very few studies that have directly investigated the impact of climate change on orchid decline, it has been suggested that climate extremes further contribute to threaten orchid populations [18].
There is ample evidence that the demography of orchid species is dependent on prevailing weather conditions. For example, warm winter conditions favored the population performance of the terrestrial orchid Himantoglossum hircinum in a nature reserve in Germany [21]. Similarly, demographic modeling showed that observed changes in climatic conditions were beneficial to the long-term survival of the same species in the UK and suggested that they may have even been the driving force behind the current range expansion of the species in England [22]. Detailed demographic analyses using integral projection models also showed that milder winters and wetter springs had a positive effect on the demography of populations of the lady orchid (Orchis purpurea) at the northern edge of its distribution [23]. Although it is hard to generalize from these case studies, these results suggest that climate warming may have beneficial effects on the population dynamics of orchids at the cold edge of their distribution ranges. However, to what extent climate change affects the demography of orchids at the warm edges of their distributions remains largely unclear [20].
Here, we used species distribution modeling to assess the effect of climate change on the European distribution of four species of the orchid genus Orchis. This genus consists of 21 terrestrial orchid species and several subspecies and varieties that are widely distributed across most of Europe and some parts of Asia Minor [24]. Specifically, we tested how the suitable habitat available to the species changed when the modelling species distributions under the current climate and climate scenarios projected for 2050 and compared the importance of environmental characteristics to each species to investigate how habitat selection contributes to distribution patterns. Because of the beneficial impacts of climate change on the demography of orchids at the northern edges of their distributions, we predict that potentially suitable areas will shift pole-wards or to higher elevations as the Earth's climate warms, possibly leading to range expansions under mild climate change. However, under severe climate change, orchid populations may no longer be buffered against changes in the climate at the warm edges of their distributions and therefore may be more susceptible to decline than more central populations, ultimately leading to range contractions.

Study Species
The four species studied here (Orchis anthropophora, O. militaris, O. purpurea and O. simia) belong to the so called 'anthropomorphic' group of the genus Orchis [25]. Although O. anthropophora displays pronounced differences in flower morphology (Figure 1), phylogenetic analyses based on DNA sequences have placed the four species within the same genus [25]. Before, O. anthropophora was assigned to the genus Aceras, of which it was the sole member [24]. In general, O. anthropophora, O. purpurea and O. simia prefer somewhat warmer conditions for growth and survival than O. militaris. As a result, O. anthropophora, O. purpurea and O. simia are species with a mainly Mediterranean distribution, whereas O. militaris has a more continental distribution, occurring from the Atlantic coast to Mongolia [24]. However, in grazed dry meadows and calcareous grasslands of Western Europe they can often be found growing together.
Diversity 2020, 12, x FOR PEER REVIEW 3 of 13 changes in the climate at the warm edges of their distributions and therefore may be more susceptible to decline than more central populations, ultimately leading to range contractions.

Study Species
The four species studied here (Orchis anthropophora, O. militaris, O. purpurea and O. simia) belong to the so called 'anthropomorphic' group of the genus Orchis [25]. Although O. anthropophora displays pronounced differences in flower morphology (Figure 1), phylogenetic analyses based on DNA sequences have placed the four species within the same genus [25]. Before, O. anthropophora was assigned to the genus Aceras, of which it was the sole member [24].  [24]. However, in grazed dry meadows and calcareous grasslands of Western Europe they can often be found growing together. All species except O. anthropophora, which is mainly pollinated by beetles, are visited by generalist pollinators such as bees and bumblebees and occasionally butterflies [26]. Fruit set in the four investigated Orchis species tends to be generally low (<20%) [26,27], but within a single fruit a substantial number of seeds can be found. The seeds require a mycorrhizal fungus for germination. The four Orchis species are mycorrhizal generalists that associate with a large number of mycorrhizal fungi, most often members of the Tulasnellaceae and Ceratobasidiaceae [28][29][30][31][32]. The four species also share a substantial proportion of their fungi, and as a result hybridization can be encountered in sympatric populations [24,[33][34][35].

Occurrence Data and Predictor Variables
We acquired the GPS coordinates recorded for each species in Europe from the years 1999-2019 available from the GBIF online repository (Global Biodiversity Information Facility; www.GBIF.org). After removing the records with missing data or a spatial resolution lower than 100 m, we aggregated the locality points for each species into a 5 km grid squares to reduce the effects of spatial clustering and sampling bias using QGIS v3.4 [36]. After aggregation, we took the midpoint of each 5 km grid square, resulting in each species having between 1350 and 4315 occurrence records to be used in modelling. All species except O. anthropophora, which is mainly pollinated by beetles, are visited by generalist pollinators such as bees and bumblebees and occasionally butterflies [26]. Fruit set in the four investigated Orchis species tends to be generally low (<20%) [26,27], but within a single fruit a substantial number of seeds can be found. The seeds require a mycorrhizal fungus for germination. The four Orchis species are mycorrhizal generalists that associate with a large number of mycorrhizal fungi, most often members of the Tulasnellaceae and Ceratobasidiaceae [28][29][30][31][32]. The four species also share a substantial proportion of their fungi, and as a result hybridization can be encountered in sympatric populations [24,[33][34][35].

Occurrence Data and Predictor Variables
We acquired the GPS coordinates recorded for each species in Europe from the years 1999-2019 available from the GBIF online repository (Global Biodiversity Information Facility; www.GBIF.org). After removing the records with missing data or a spatial resolution lower than 100 m, we aggregated the locality points for each species into a 5 km grid squares to reduce the effects of spatial clustering and sampling bias using QGIS v3.4 [36]. After aggregation, we took the midpoint of each 5 km grid square, resulting in each species having between 1350 and 4315 occurrence records to be used in modelling.

Maxent Modelling
The ecological niche of each species was modelled in Maxent v3.4.1 [44] using the processed occurrence data and environmental rasters. Maxent uses the concept of maximum entropy to estimate the probability distribution of species occurrences given the environmental predictor variables of the occupied pixels, and then projects the probability of a species occurring in each pixel to the entire study site as a measure of habitat suitability [45]. Seventy five percent of the data were used to train the model and the remaining 25% to test, using a random seed, 100 bootstrap replicates and leaving the remaining settings as the default [45,46]. The output includes a map of the predicted distribution of each species, the contribution of each variable to the distribution of the species, response curves of each species' occurrence to the environmental variables, and measures of fit such as the area under the receiver operating curve (AUC).
The models were then projected onto the climatic conditions of 2050 predicted by RCP (Representative Concentration Pathway) 2.6, RCP 4.5 and RCP 8.5. In RCP 2.6, carbon dioxide (CO 2 ) emissions start declining by 2020 and reach zero by 2100. In this scenario, global temperature rise is expected to stay below 2 • C by 2100. In RCP 4.5, greenhouse gas emissions peak around 2040, then decline, while in RCP 8.5 greenhouse gas emissions continue to rise throughout the 21st century. Under RCP 4.5 and 8.5, the global temperature is expected to rise between 2 and 3 • C and between 2.6 and 4.8 • C, respectively. Therefore, these models represent moderate (RCP 2.6 and RCP 4.5) to more extreme (RCP 8.5) scenarios of the range of future climate scenarios.

Data Analysis
We used ENMTools v1.3 [47] to calculate Levins' B value for 100 replicate ecological niche maps for each species. The mean Levins' B value was calculated to represent the niche breadth of each species, with values closer to 0 indicating narrow niche and values closer to 1 indicating wide niche breadth. A value of Schoener's D was calculated for each species pair in ENMTools to investigate the niche overlap among the four Orchis species, with values closer to 1 indicating more similar habitat preferences and values closer to 0 indicating less similar habitat preferences.
To calculate the area of suitable habitat available to each species, we extracted all the pixels above the maximum training sensitivity plus the specificity threshold from each mean Maxent output map, to give binary presence/absence maps for each species at each climate scenario. We used these maps to investigate the change in available habitat from the current climate to the three RCPs in 2050, and calculated the area of current habitat that would be lost if the species was not able to track the habitat changes in the future, using QGIS v3.4. The contributions of environmental variables to each species' habitat were visualized as a heat map using the package 'pheatmap' [48] in R.

Results
The data for all four species fit the Maxent models well, with all having an AUC value of at least 0.87. The suitable habitat (in terms of climate and soil) available to the four Orchis species was predicted by Maxent to cover most of the western European mainland, extending to Romania in the east, southern Poland in the north and the Mediterranean in the south, as well as parts of Spain and the United Kingdom (Figure 2). The northern distribution of the O. militaris habitat extended into Sweden, and this species had the largest current suitable habitat (2,465,675 km 2 ) of the four species, as well as the widest niche breadth (Levins' B = 0.4687 ± 0.0012) ( Table 1 and Figure 3). Orchis simia had the smallest suitable habitat available (1,791,875 km 2 ) and the narrowest niche breadth (0.2797 ± 0.0011) ( Table 1 and Figure 3). In terms of niche similarity, O. purpurea and O. anthropophora demonstrated the highest similarity with an 85.3% overlap, while O. simia and O. militaris demonstrated the lowest (but still considerable) overlap of 64.2% (Table 2).       The high degree of niche similarity was reflected by the variable contributions output (Figure 4), with the four species reflecting similarly weak associations with bedrock, elevation and two of the soil components, while being more affected by the first two soil components (pH and potassium level, and soil texture), temperature and precipitation (Figure 4). Orchis simia had a notably strong association (36%) with temperature and O. militaris with the primary biochemical component (pH and potassium level, 32%). The mean annual temperatures and annual precipitation of the habitats predicted for the four species were similar to one another, with O. militaris occupying areas with the lowest temperature of 9.4 ± 0.04 • C and O. anthropophora the highest of 10.9 ± 0.06 • C (Table 3). Precipitation levels were also similar among the species, with O. militaris occupying areas with the highest precipitation of 800.7 ± 3.88 mm 3 Figure 3). For these species, the distribution of habitat expanded into the north, reaching southern Sweden and the northern United Kingdom, and continued in the central latitudes of mainland Europe, and disappeared in parts of the Mediterranean region and western Europe (Figure 2). The habitat of O. militaris decreased by 15,890, 507,850 and 878,550 km 2 for the respective climate scenarios (Figure 3), with the northern limits of the distribution extending further into Estonia and Latvia, but diminishing considerably in western and eastern mainland Europe. If the species were not able to track the climate northwards, the area of current habitat would decrease in 2050 for the four Orchis species by 19.3 ± 4.5, 24.5 ± 5.7 and 36.7 ± 8.0% for each climate scenario, respectively (Table 1). O. militaris would be particularly negatively affected, with its available habitat decreasing by 33.6% in the most optimistic climate scenario (RCP2.6) and 61.3% in the worst-case scenario (RCP8.5).   Figure 3). For these species, the distribution of habitat expanded into the north, reaching southern Sweden and the northern United Kingdom, and continued in the central latitudes of mainland Europe, and disappeared in parts of the Mediterranean region and western Europe (Figure 2). The habitat of O. militaris decreased by 15,890, 507,850 and 878,550 km 2 for the respective climate scenarios (Figure 3), with the northern limits of the distribution extending further into Estonia and Latvia, but diminishing considerably in western and eastern mainland Europe. If the species were not able to track the climate northwards, the area of current habitat would decrease in 2050 for the four Orchis species by 19.3 ± 4.5, 24.5 ± 5.7 and 36.7 ± 8.0% for each climate scenario, respectively (Table 1). O. militaris would be particularly negatively affected, with its available habitat decreasing by 33.6% in the most optimistic climate scenario (RCP2.6) and 61.3% in the worst-case scenario (RCP8.5).

Changes in Distribution Area
The distribution ranges of the Mediterranean species O. anthropophora, O. purpurea and O. simia were predicted to increase under mild climate change, while that of the more continental O. militaris was predicted to decrease. These results are in line with demographic studies that have shown that milder winters and wetter springs have a positive impact on the population growth rate of orchids growing at the leading edge of their distributions [22,23]. Two populations of Orchis purpurea at the northern edge of their distributions showed population growth rates that were consistently above 1.0 between 2004 and 2015, indicating that the prevailing climatic conditions during that period were favorable for orchid growth [23] (Williams et al. 2015). Similarly, climate change was predicted to increase the future population growth rates of Himantoglossum hircinum at the northern edge of its distribution in the UK [22]. Here, the species benefits from the slightly warmer climate throughout the year and less precipitation in summer. The number of populations of this species has increased from 10 to 17 between 1988 and 2000 and has continued to rise to at least 26 populations by 2015 [22].
With the climate becoming more benign, novel habitats that were predicted to become suitable for growth included large areas of Estonia, Finland and Sweden. These areas currently already support a large number of orchid species that often co-occur with the investigated species in more southern areas (e.g., Anacamptis morio, Orchis mascula, Gymnadenia conopsea), suggesting that edaphic and habitat conditions are suitable to sustain orchid growth. Under severe climate change, suitable habitats were predicted to disappear at the rear edge of the distribution, most notably in Southern Italy, Croatia and Greece. In these areas, climatic conditions were predicted to no longer sustain the growth of these orchid species. Although the precise mechanisms leading to the disappearance of these species at southern edges is unknown, it is likely that increased drought has a strong negative effect on seed germination, survival and flowering. Previous research has indeed shown that drought negatively affects the probability of flowering and hence fecundity [23].

Role of Climatic vs. Abiotic Variables
Previous research has shown that the ecological niche of orchids is a function of precipitation and temperature, altitude, soil composition, bedrock and vegetation type [49][50][51][52][53][54]. Using eight different environmental and climatic variables, our results showed that soil biochemistry and climate were the most important variables determining the distribution of the four Orchis species, whereas the underlying bedrock and altitude only played a minor role, suggesting that the species can occur both in lowland and mountainous areas. This study modeled the habitat available to species in 2050 based on abiotic environmental factors, but did not cover whether the species would actually be able to expand north into these climatically suitable areas given that there are likely anthropogenic and biotic factors that will limit this. It is, therefore, also important to look at the habitat associations of species, because although O. militaris was the only species of the four which demonstrated a loss in total habitat, whilst for the more severe climate scenarios, O. simia and O. purpurea showed the greatest latitudinal shift in habitat, and the strong association of O. simia occurrence with temperature and precipitation likely means that in the long term or with severe changes in climate, this species could be very much affected by climate change.

Colonization Potential
For the species that have limited dispersal capabilities or those with habitats that have become largely fragmented, keeping track of optimal habitat conditions may be difficult or even impossible [55], and these species may ultimately go extinct if they do not adapt to changing climatic conditions. However, seed dispersal in orchids may occasionally occur over fairly large distances [56], and once populations have established, they can quickly expand. Orchis simia, for example, managed to colonize the Netherlands in 1972 and establish a viable population [57]. The nearest natural population was thought to be 50 km away. Similarly, a single flowering plant of O. simia and two non-flowering rosettes were found in 1974 in a dune area in east Yorkshire, England, more than 250 km north of the nearest site in the Thames valley [58]. A large population of O. militaris was found in a disused chalkpit near Mildenhall, Suffolk, England, in 1955, some 100 km from its previously known site. These and other examples [56] indicate that orchids are capable of sporadically crossing large distances and establishing novel populations at large distances from existing populations.

What about Crucial Interactions?
For orchid populations to colonize new habitats and to persist in the long term, partnerships with at least two different organisms need to be established. Interactions with mycorrhizal fungi are needed to facilitate germination and subsequent establishment as a seedling, and pollinators are required to pollinate the flowers and hence to produce seeds. Failure to pollinate the flowers will negatively affect fecundity and hence the recruitment of new individuals. A recent study has modeled the distributions of two sexually deceptive orchid species and their unique pollinator under climate change [59]. Results showed that including the interaction with the pollinator led to pronounced differences in the distribution range under climate change. However, in this case, both orchid species were pollinated by a single insect species, which is unlikely to be the case in the four studied Orchis species. The four studied orchid species are pollinated by generalist pollinators such as bees and bumblebees that are attracted to the flowers by their color or form, but quickly learn to avoid them as they do not provide any reward [27]. For example, the flowers of O. militaris have been shown to be visited by at least 49 different insect species, including hoverflies, bumblebees and bee flies [60]. However, most of these insects probably did not function as an effective pollinator, and pollinia were only observed in two species (Apis mellifera and Bombus lapidarius). Nonetheless, these insect species have very broad distribution ranges and therefore are probably not limiting their fruit set at potentially novel growth locations.
The germination of orchid seeds generally requires the establishment of a mycorrhiza [61,62]. Previous research has shown that the investigated Orchis species tend to be mycorrhizal generalists that associate with a large number of fungi [27][28][29][30][31]. Moreover, the seeds of O. militaris were able to germinate at mine tailing hills where the species did previously not occur, suggesting that the mycorrhizal fungi needed to facilitate seed germination are widespread and not necessarily limiting the distribution of the orchid [63]. Similar results have been observed for several species of the orchid genus Epipactis. Seed introduction experiments using seed packages showed that local growth conditions had little influence on germination patterns, and seedlings were found in more habitats than expected from the adults' ecology [64]. However, it should be noted that most seed introduction experiments using seed packages usually do not monitor the subsequent growth to the adult stage, so it remains unclear whether some habitat patches may be only temporarily suitable or whether introduced seeds germinating in seed packets would establish and form self-sustaining populations [62].
To unambiguously know whether orchid populations can establish and persist in habitats that have been predicted to become suitable in the future, species distribution models are ideally combined with seed introduction or transplant experiments. For example, a recent study that simultaneously applied an SDM (species distribution model) using high-resolution spatial predictors, and an integral projection (demographic) model, based on a transplant experiment, in central Sweden, showed that the predicted habitat suitability and population growth rate, yielded by the two approaches, were not correlated across the transplant sites at the cold range margin [65]. These results suggest that local microclimate, light and soil conditions may have a large effect on the distribution and performance of plant species and that mere extrapolation of suitable habitat from SDMs is likely to generate some spurious results [65]. However, orchids transplant experiments are not easy to conduct, and preferably, introduction experiments should start from seeds.

Implications for Conservation and Future Perspectives
Using environmental niche modeling, our results showed that orchid populations at the warm edge of their distribution range are bound to decline under severe climate change, and even if such extreme scenarios may be unlikely in the future [66], warrant for more detailed research on the demography, population genetics and adaptive potential of range edge populations. Only a few studies have compared demographic rates across the entire distribution range of species, and some of these have shown that populations may be buffered against climatic change. However, at present no such studies are available for orchids, and we currently do not know whether demographic buffering mitigates the negative impact of climatic variation and helps orchid populations at the warm edge of their distribution to persist in an increasingly variable environment. Second, orchids critically rely on mycorrhizal fungi for germination and seedling establishment, however, we virtually know nothing about the global distribution patterns of orchid mycorrhizal fungi and how they affect the ability of orchids to shift their distribution range under climate change. We also do not know how the distribution range of the fungi themselves will be affected by climate change. Previous research has already shown that orchid mycorrhizal communities vary across large geographic distances and generally decrease in species richness with increasing latitude [67]. More extensive datasets about the distribution of orchid mycorrhizal fungi need to be generated in order to define the environmental variables determining the distribution of fungi and orchids relying on these fungi.