Expansion of Himantoglossum robertianum ( Orchidaceae ) in Madrid : a case study on environmental variables and geographical distribution

The increase of the localities of Himantoglossum robertianum in the Community of Madrid region in the last years is studied. We have visited the new discovered populations and characterized their habitats, finding little agreement with the basophilic preferences attributed to the orchid. A model of potential distribution of the taxon has been elaborated showing the southern river basins of the region as the areas of greatest suitability. The lack of data prevents from concluding that the increase of the registered area corresponds with a climatic or land-use change in Madrid region, but the work developed has allowed to properly characterize the distribution and the state of conservation of the species.


Introduction
Orchids have been highlighted as indicators of fragile and diverse ecosystems in the Community of Madrid, which has led to propose the entire Orchidaceae family to be included in the legal catalogue of protection for regional flora (Blanco, 1999).This proposal was the base for the study of the 40-orchid species from Madrid region at the initiative of the body of forest rangers to determine the IUCN categories of risk and the most adequate conservation measures for each taxon (Anon., 2014, CAFCM).This study was aimed at the revision and updating of the catalogue of protected species of the Community of Madrid, very obsolete after 25 years of validity (Anon., 1992, BOCM) ;Moreno Saiz & al., 2011).
Himantoglossum robertianum (Loisel.)P. Delforge, the giant orchid, is a species distributed throughout the Mediterranean basin, from the Balkans to Iberia and the Maghreb.Its traditional treatment included the species under the name of Barlia robertiana (Loisel.)Greuter (Aedo, 2005), but recent taxonomic studies supported by molecular and morphological data led to its combination within the genus Himantoglossum Spreng (Bateman & al., 2003(Bateman & al., , 2017)).In the Iberian Peninsula, it has a peripheral distribution, present in almost all the coastal provinces and only in some inland localities, with preference for the areas of basic substrates (Delforge, 2002;Aedo, 2005).In the Community of Madrid the presence of the plant was unknown for revisions of its flora (Cutanda, 1861;Abajo & al., 1982;Morales Valverde, 2003) until it was detected some years ago (Galán Cela & al., 2003).Since then it has been mentioned in more localities scattered mainly in the south of the region (López Jiménez, 2007;García Román & López Colón, 2011;Grijalbo Cervantes, 2016).
The only study on its conservation status had considered the giant orchid as an endangered species in the Community of Madrid (IUCN category EN D), with few locations involving reduced number of plants (Anon., 2014, CAFCM).However, its number of citations has been increasing in recent years, often through amateur publications on social networks or new discoveries by forest rangers.These findings had the advantage of the plant height and attractiveness, with stems up to 100 cm, together to an early flowering in Madrid region (February to April), which made the plant easily recognizable.
With this background, a review of the H. robertianum status in the Community of Madrid was addressed with several objectives: i) to update its distribution and occupation area in Madrid region; ii) analyze the ecological requirements of the plant through field work in their populations; iii) develop habitat suitability models in order to determine the environmental variables most related to their distribution.These objectives would serve to raise the hypothesis of whether H. robertianum is an expanding species in the center of the Iberian Peninsula as a result of the new findings, and if a pilot study like this could be extended to other orchids and threatened plant taxa in the Community of Madrid.

Data collection
Distributional data referred to Himantoglossum robertianum in Madrid region was compiled taking advantage of published articles, technical reports and personal communications.No herbarium sheet referring to Madrid region and correctly identified could be incorporated after the review of the main Spanish herbaria.The search in public electronic databases of common use (e.g.Global Biodiversity Information Facility, GBIF; Information System about Plants in Spain, Anthos) did not reveal new references to those already known.Based on this preliminary database, the largest possible number of known locations was visited during the spring of 2017, in order to make a contemporary sampling of the populations.In this fieldwork, the localities were georeferenced, the plants were counted as flowering or vegetative individuals, and the plant communities was inventoried.

Species distribution modeling
A dataset composed by the sampled populations spanned to 56 presences at 1 km 2 grid resolution was used.We relied on an accurate sampling of species presences so we relied on reliable true absence data, using six times more absences than presences in the model to weigh the same.These sets of presence and absence points were used to calibrate the models.
The bioclimatic variables were obtained from CHELSA dataset, which provides high resolution monthly mean temperature and precipitation patterns for the time period 1979-2013 (Karger & al., 2017).For its part, soil characteristics were obtained from the Harmonized World Soil Database (FAO).The six most meaningful species distribution models (SDMs) were selected from the 19 bioclimatic and 20 edaphic variables available.The selection was conducted through hierarchical partitioning approach with the hier.partpackage (Walsh & Mac Nally, 2003) in R software.Niche model analysis was conducted with Biomod2 package (Thuiller & al., 2009) implemented in R software (R Development Core Team, 2014).The chosen models were generalized additive model (GAM), stepwise generalized linear models (GLM), boosted regression trees (GBM), artificial neural network (ANN), multiple adaptive regression splines (MARS) and random forests (RF).Each algorithm was executed 5 times so that 90% of the randomly selected presence/absence data were used to calibrate the model and 10% to evaluate the quality of the prediction of habitat suitability using the TSS and ROC indices.Thus, 5x5x6=150 models of potential distribution were generated.To construct the unified model of potential distribution of the species, the individual models were filtered excluding all those with a quality lower than 0.8 in any of the two indices considered and were weighted according to the TSS index.The suitability threshold to consider each pixel as part of the potential distribution of the species was established as that value that maximizes the TSS value.The ensemble modeling technique implemented in Biomod2 as a weighted average of all models was utilized to build a single prediction.

Range of H. robertianum in Madrid region
From the 62 orchid populations included in the database, 51 have been visited and 48 of them found during the fieldwork.So, it has been possible to confirm the increase in the number of localities, as well as the extension of known occupation of the giant orchid in the Madrid region (Figure 1).From a single population discovered in 2003, it has gone to 53 cells occupied in 2017, spread over the southern half of the territory studied.The complete relation of coordinates and localities appear in Annex 1.
The censuses carried out raise up to 300 mature individuals the whole population, to which 507 vegetative rosettes would have to be added.As no previous demographical sampling had been performed, it cannot be determined to what extent the census has followed similar increase during this period.Nevertheless, this census together with the calculation of the Area Of Occupancy and the Extent Of Presence, has allowed to establish a new regional IUCN risk category for H. robertianum downgrading it to Vulnerable (Martínez Labarga & al., 2017).

Habitat characteristics
All populations of H. robertianum in the Community of Madrid are found at altitudes between 470 to 940 m asl, and are usually presented in E and S orientations.Notably, a preference for arcosas is observed in half of their inventories, followed by clays, marls and limestones.Thus, the predominantly basophilic nature of the distribution of this orchid throughout the Mediterranean and the Iberian Peninsula would not be confirmed in Madrid region (Delforge, 2002).However, the pH of these sands (arkoses) has not been measured in situ during our work, which makes it impossible to know the true level of acidity or neutrality of such soils.
The orchids mainly developed in clearings of forest formations (60% of inventories), followed by grasslands and nitrophilous pastures (29%), and shrublands (11%).The main woodlands were holm oaks forests (Quercus ilex subsp.ballota (Desf.)Samp.) with as much as 41%, followed by semi-natural pine forests of different species (Pinus pinea L. and P. halepensis Mill.).Plant inventories were heterogeneous in line with the variety of formations detected with presence of the giant orchid.
134 from 150 conducted models scored over 0.8 for both TSS and ROC indices.The 6 selected variables were: annual precipitation (0.828), precipitation of the coldest quarter (0.525), total amount of base exchange on the top of the soil (0.467), precipitation of the coldest quarter (0.377), temperature range (0.152), and temperature of the coldest quarter (0.075).No topographic variables were considered.Threshold suitability was 475 over 1000.
According to the model, the most suitable areas for H. robertianum in the Community of Madrid extend through the middle and lower basins of the Alberche, Guadarrama and Manzanares rivers, in the S and SW of the region (Figure 1).The consensus model shows in these basins, and even along the easternmost Henares and Tajuña valleys (Figure 2), large areas not known or occupied in spite of being highly suitable for the orchid.Even though these models can over-predict the potential niche of the species (Guisan & al., 2017), such valleys would be areas where planning prospections in future campaigns in search of new populations.
There are not enough climate data to analyze if there has been a significant change in potential areas modeled before the first appointment in the Community of Madrid of H. robertianum was discovered, to this modeled nowadays.This prevents empirically determining whether the appearance of the species in Madrid and the expansion of its known area is related to climate changes or variations in other variables that have occurred in recent decades.In any case, the quasiexponential growth of discoveries in recent years, given the ease of recognizing the species, must be explained more by a range expansion of the orchid than by the net increase of floristic knowledge of the territory, being Madrid a region with numerous studies during the last century.Variations in rainfall and temperature contrasts, operating synergistically with changes in land use, could be favoring the expansion of this Mediterranean plant to widen its area of occupation into Iberian inland.
A study like the one addressed here, combining data mining, field work and species distribution modeling, has resulted very useful in determining the conservation status of a threatened plant in a given region.In a territory like Madrid region, where it is urgent to renew the catalogue of threatened flora based on updated studies (Moreno Saiz & al., 2011) a pilot study of these characteristics can be a quick and resolutive tool to complete a task like this with means and terms in accordance with the importance of the project needed.

Figure 1 .
Figure 1.Distribution range of Himantoglossum robertianum in Madrid.Every green square represents a UTM cell of 1 km side (datum ETRS89).