NC-Clustering demonstrates heterospecificity of the cryptic ant species Temnothorax luteus ( Forel , 1874 ) and T . racovitzai ( Bondroit , 1918 ) ( Hymenoptera : Formicidae ) With 6 figures

Evidence is presented that the W Mediterranean cryptic ant species Temnothorax luteus (Forel, 1874) and T. racovitzai (Bondroit, 1918) are clearly separable by exploratory and hypothesis-driven data analyses of complex morphological data sets. Investigation was concentrated to the sympatric range of both species. 18 morphological characters were evaluated in a total of 178 workers originating from 64 nest samples. Classifications by hierarchical NC-Ward clustering, non-hierarchical NC-K-Means clustering, NC-NMDS-K-Means clustering and linear discriminant analysis coincided by 100 % after reduction of the character set to the seven most discriminative characters. The type series of T. massiliensis (Bondroit, 1918) and T. racovitzai (Bondroit, 1918) were assigned to the same cluster with posterior probabilities of p = 1.000 in the former and p = 0.9992 in the latter. The type series of T. tristis (Bondroit, 1918) and T. luteus were allocated to another, clearly separated cluster with p = 0.9999 in each case. Accordingly, T. tristis is considered as junior synonym of T. luteus whereas T. massiliensis was established as junior synonym of T. racovitzai. The classification error of the LDA was 1.7 % in 178 individuals. Despite a broad sympatric range, there are no suggestions to any nest samples containing a mixture of both phenotypes. This rejects the hypothesis that T. luteus and T. racovitzai could represent an intraspecific dimorphism. Significant interspecific differences of the two species in geographic distribution and thermal niche were demonstrated. Mean air temperature from May to August of 24 sampling sites of T. luteus was by 3.77°C cooler than in 21 sites of T. racovitzai (p<0.0001).


Introduction
The identity of Temnothorax luteus (Forel, 1874) described from Grand Salève/ France and Ticino /Switzerland has remained obscure until very recently when Plateaux & Cagniant (2012) discovered two type workers in the collection of Musée Zoologique Lausanne.These specimens carried no type labels but the argumentation of Plateaux & Cagniant, based on the original morphological description and the localities reported by Forel, was conclusive.However, because their paper was based on very few investigated specimens and poor investigation standards, they published several diffuse and problematic decisions.In a synonymic list on page 430 and following Casevitz-Weulersse & Galkowski (2009), Plateaux & Cagniant published Temnothorax massiliensis (Bondroit, 1918) as a synonym of T. luteus and they considered T. racovitzai (Bondroit, 1918) as a subspecies of T. luteus.However, results of exploratory and hypothesis-driven data analyses of complex character sets, show that T. massiliensis is no synonym of T. luteus and that T. racovitzai cannot be a subspecies of T. luteus.Furthermore, we want to assess the status of all five European taxa belonging to the T. luteus species complex and to demonstrate that there are two reliably delimited species, T. luteus and T. racovitzai, which differ significantly in morphology, distribution and in the thermal component of their climatic niches.

Material subject to numeric data recording
For both sibling species, numeric data recording was performed in a total of 64 samples containing 178 workers and originating from 38 localities.

Investigation Methods
18 morphometric characters currently being used in taxonomy of Palaearctic Temnothorax (Seifert 2006, Csösz et al. 2013) were investigated.In bilaterally recorded characters, arithmetic means of both body sides were calculated.All measurements were made on mounted and dried specimens using a pin-holding stage, permitting full rotations around X, Y, and Z axes.A Leica M165C high-performance stereomicroscope equipped with a 2.0 planapochromatic objective (resolution 1050 lines/mm) was used at magnifications of 120-384x.The mean relative measuring error over all magnifications was 0.3 %.A Schott KL 1500 cold-light source equipped with two flexible, focally mounted light-cables, providing 30°-inclined light from variable directions, allowed sufficient illumination over the full magnification range and a clear visualization of silhouette lines.A Schott KL 2500 LCD cold-light source in combination with a Leica coaxial polarized-light illuminator provided optimal resolution of tiny structures and microsculpture at highest magnifications.Simultaneous or alternative use of the cold-light sources depending upon the required illumination regime was quickly provided by regulating voltage up and down.A Leica cross-scaled ocular micrometer with 120 graduation marks ranging over 52 % of the visual field was used.To avoid the parallax error, its measuring line was constantly kept vertical within the visual field.Measurements of body parts always refer to real cuticular surface and not to the diffuse pubescence surface.
The process of species discrimination included four major routines: (a) Reduction of errors in primary data recording as described in Seifert (2002).
(b) Removal of allometric variance (RAV) performed with the procedure described by Seifert (2008).
(c) Application of the exploratory data analyses Nest Centroid Clustering (NC-Clustering).This included a hierarchical method (NC-Ward), a non-hierarchical method (NC-K-Means) and a highly flexible ordination method (NC-NMDS-K-Means).These methods were described in more detail by Seifert et al. (2013) who also provided a script written in R and freely available under the GNU / GPL license from the following website: http//sourceforge.net/projects/agnesclustering/.
(d) Checking of the demonstrated species clusters by linear discriminant analysis (LDA) also provided by SPSS 16.0.

The morphometric characters
CL -maximum cephalic length in median line; the head must be carefully tilted to the position with the true maximum.Excavations of occiput and/or clypeus reduce CL.CS -cephalic size; the arithmetic mean of CL and CW, used as a less variable indicator of body size.CW -maximum cephalic width; the maximum is found in Temnothorax and Leptothorax usually across and including the eyes, exceptionally posterior of the eyes.EYE -eye-size index: the arithmetic mean of the large (EL) and small diameter (EW) of the elliptic compound eye is divided by CS, i.e.EYE = (EL+EW)/(CL+CW).
All structurally visible ommatidia are considered.FRS -distance of the frontal carinae immediately caudal of the posterior intersection points between frontal carinae and the lamellae dorsal of the torulus.If these dorsal lamellae do not laterally surpass the frontal carinae, the deepest point of scape corner pits may be taken as reference line.These pits take up the inner corner of scape base when the scape is fully switched caudad and produce a dark triangular shadow in the lateral frontal lobes immediately posterior of the dorsal lamellae of scape joint capsule.MGr -depth of metanotal groove or depression, measured from the tangent connecting the dorsalmost points of promesonotum and propodeum; here given as per cent ratio of CS.MH -in workers: with mesosoma in lateral view and measured orthogonal to "longitudinal mesosomal axis", MH is the longest measurable section line of mesosoma at mesopleural level (not height above all)."Longitudinal mesosomal axis" in lateral view is defined as straight line from the centre of propodeal lobe to the border point between anterior pronotal shield and propleuron.In gynes it is the longest section line directed perpendicular to the straight dorsal profile line of mesosoma (formed by mesonotum and scutellum).The lower reference point is usually the lowest part of mesopleuron.MW -maximum mesosoma width; in gynes measured anteriorly of the tegulae.ML -in workers: mesosoma length from caudalmost point of propodeal lobe to transition point between anterior pronotal slope and anterior propodeal shield (preferentially measured in lateral view; if the transition point is not well defined, use dorsal view and take the centre of the dark-shaded borderline between pronotal slope and pronotal shield as anterior reference point).In gynes: length from caudalmost point of propodeal lobe to the most distant point of steep anterior pronotal face.PEH -maximum petiole height.The straight section of ventral petiolar profile at node level is the reference line perpendicular to which the maximum height of petiole node is measured.
PEL -diagonal petiolar length in lateral view; measured from anterior corner of subpetiolar process to dorsocaudal corner of caudal cylinder.PEW -maximum width of petiole.PoOc -postocular distance.Use a cross-scaled ocular micrometer and adjust the head to the measuring position of CL.Caudal measuring point: median occipital margin; frontal measuring point: median head at the level of the posterior eye margin.Note that many heads are asymmetric and average the left and right postocular distance.PP -maximum width of postpetiole.

SL
-maximum straight line scape length excluding the articular condyle as arithmetic mean of both scapes.SP -maximum length of propodeal spines; measured in dorsofrontal view along the long axis of the spine, from spine tip to a line, orthogonal to the long axis, that touches the bottom of the interspinal meniscus (Fig. 4 in Seifert 2006).This mode of measuring is less ambiguous than other methods but it results in some spine length in species with reduced spines.SPBA -the smallest distance of the lateral mar gins of the spines at their base.This should be measured in dorsofrontal view, since the wider parts of the ventral propodeum do not interfere with the measurement in this position.If the lateral margins of spines diverge continuously from the tip to the base, a smallest distance at base is not defined.In this case, SPBA is measured at the level of the bottom of the interspinal meniscus.SPST -distance between the centre of propodeal stigma and spine tip.The stigma centre refers to the midpoint defined by the outer cuticular ring but not to the centre of real stigma opening that may be positioned eccentrically.SPTI -the distance of spine tips in dorsal view; if spine tips are rounded or thick, the centres of spine tips are taken as reference points.

Removal of allometric variance (RAV)
In ant groups without extreme allometries removal of allometric variance (RAV) does not change the results of linear discriminant functions (Seifert 2008) und Geodynamik (1996).We corrected the data for an altitudinal temperature decrease of 0.661°C per 100 m according to the equation of Seifert and Pannier (2007): TAS = -0.694LAT + 0.078 LON -0.00661 ALT + 52.20 where TAS is the predicted standard air temperature in °C, LAT and LON are the geographical latitude and longitude in international decimal format, and ALT is the altitude above sea level in meters.

Diagnosis of the Temnothorax luteus complex
Members of the Temnothorax luteus species complex (Figs 3-6) can be separated from other species by the character combination described below.Numeric data are given as overall means ± standard deviation of individual workers without RAV-correction.Whole ants light yellowish to yellowish brown, lacking any blackish or blackish brown pigmentation with exception of eyes; some body parts or the entire body may be infuscated.Dorsum of head capsule rather weakly longitudinally rugulose; small to very large areas of central vertex often completely smooth and shining.Relatively small-sized and long-headed in terms of the genus (CS 608 ± 45 µm, CL/CW 1.133 ± 0.020

Clustering, classification and synonymies
NC-Ward, NC-K-Means and NC-NMDS-K-Means clustering considering 18 characters (i.e., the absolute size measure CS and 17 RAV-corrected shape characters) resulted in the demonstration of two main clusters (Fig. 1).These clusters are interpreted to represent two species Temnothorax luteus (Forel, 1874) and Temnothorax racovitzai (Bondroit, 1918).All three exploratory data analyses made a coincident assignment of the types series to either cluster: The type series of T. massiliensis (Bondroit, 1918) shared a cluster with the type series of T. racovitzai and the type of T. tristis (Bondroit, 1918) clustered with the T. luteus type series.As a consequence, we consider T. massiliensis as a junior synonym of T. racovitzai and T. tristis a junior synonym of T. luteus.NC-K-Means and NC-NMDS-K-Means provided identical classifications for each of the 64 samples but they differed from NC-Ward in the allocation of three samples to the T. luteus cluster.These samples, placed by NC-Ward in the T. racovitzai branch, are marked in Fig. 1 with arrows.Running these three samples as wildcards in a linear discriminant analysis (LDA) confirmed their allocation to T. luteus.This 3 : 1 "majority decision" suggests that NC-Ward probably made three clustering errors when using the unselected set of all 18 characters.The LDA provided the following posterior probabilities for the type series: All four type specimens of T. massiliensis were assigned with p = 1.000 to the T. racovitzai cluster, the three type specimens of T. racovitzai were allocated with p = 0.9997, 0.9998 and 0.9994 to the same cluster, the single type specimen of T. tristis was allocated with p = 0.9998 to the T. luteus cluster and the two type specimens of T. luteus with p = 0.9999 and 0.9999 to the same cluster.The classification error of the LDA was 2.2 % in 178 worker individuals and 4.5 % in a leave-oneout cross-validation LDA.
Accepting the hypotheses provided by NC-K-Means, NC-NMDS-K-means and the LDA using the 18-character system and running a stepwise LDA, reduced the number of characters to seven: SL/CS 0.62 , FRS/CS 0.62 , EYE/CS 0.62 , PEW/CS 0.62 , SPST/CS 0.62 , PEL/CS 0.62 , ML/ CS 0.62 .As a result, the classifications of all three exploratory data analyses and the LDA were 100 % congruent and the NC-Ward dendrogram became undisturbed (Fig. 2).With the number of individuals in the smallest class being 10.2 fold larger than the number of characters, an optimum condition for running an LDA was given.The type specimens of T. massiliensis were assigned with p = 0.9999, 1.000, 1.000, and 1.000 to the T. racovitzai cluster, the type specimens of T. racovitzai with p = 0.9994, 0.9996 and 0.9982 to the same cluster, the type specimen of T. tristis was allocated with p = 0.9999 to the T. luteus cluster and the two type specimens of T. luteus with p = 0.9999 and 0.9999 to the same cluster.The classification error of the LDA was 1.7 % in 178 individuals and 3.4 % in a leave-one-out cross-validation LDA.These data give another clear indication for the presence of two well-separable species and the synonymies stated above.
It might be asked if the different character syndromes of T. luteus and T. racovitzai could be the expression of a discrete intraspecific dimorphism rather than indicating different species.Emery (1924) noted that only the specimens from Algeciras matched his description fully while those from Tibidabo had a more or less brown antennal club.Plateaux & Cagniant (2012) had a look on the types from Tibidabo and allocated these to their newly erected subspecies T. luteus racovitzai (Bondroit).However, they did not see the types from Algeciras but assumed that there was a subspecies T. luteus subcingulata (Emery) in southern Spain.We cannot assess this question and consider Leptothorax luteus var.subcingulata Emery as a Nomen Dubium as long as there is no access to the types from Algeciras.
Interspecific morphological differences and a simpler method for species delimitation

Distribution and thermal niche
The presented study intended to investigate the relations in the northern part of both species' geographic ranges, focusing on the situation in the sympatric area.Accordingly, our data are far from depicting the entire geographic ranges.They exclude Spain south of 40°N where T. racovitzai should predominate and entire Italy where T. luteus should abound.Yet, even this reduced data set revealed significant interspecific differences in geographic distribution and thermal niche (Tab.2).T. racovitzai is the more southern and western species which occurs in the sympatric range at lower elevations.Its postglacial expansion most certainly started from an Iberian refuge while the Pleistocene refuge of T. luteus most probably was somewhere in the central or southern Apennine.The most striking interspecific difference exists in the thermal component of climatic niche: air temperature from May to August is by 3.77°C cooler in the sampling sites of T. luteus than in those of T. racovitzai.Two experienced ant collectors, who were not able to distinguish the cryptic species and having studied the situation in S France, independently expressed the opinion that workers of the T. luteus complex grow to a larger size (and develop a darker color) in cooler environments.These impressions are true but they were generated by the fact that the significantly larger species T. luteus (Tab. 1) occurs in significantly cooler environments (Tab.2).Linear regression analyses within the species showed no significant negative dependence of the body-size indicator CS from environmental temperature TAS: T. luteus: CS = -0.0004TAS + 0.629 (r = -0.0219,n.s., n = 38).T. racovitzai: CS = -0.0024TAS + 0.631 (r = -0.1897,n.s., n = 26).Powered by TCPDF (www.tcpdf.org)