Pollen host selection by predominantly alpine bee species of the genera Andrena , Panurginus , Dufourea , Megachile , Hoplitis and Osmia ( Hymenoptera , Apoidea )

The pollen host selection by 19 bee species, which have their main Central European distribution in the Alps, was assessed by microscopical analysis of the scopal contents of about 900 females from museum and private collections. The results of the pollen analyses were complemented by a literature survey as well as by field observations. The examined species widely vary in pollen host spectrum and specialization, revealing a fascinating diversity in bee host plant use. Observed patterns of host plant choice range from narrow specialists, which exclusively collect pollen from the flowers of a single plant genus, to pronounced generalists, which harvest pollen from the flowers of up to 17 different plant families. A quantitative character is given to separate the morphologically very similar females of Panurginus herzi and P. montanus.


Introduction
Bees are herbivores using nectar and pollen as the pre dominant food source for their larvae.While no floral specificity is known with respect to the collection of nec tar, many bee species restrict pollen harvesting to closely related plant taxa (Robertson 1925, Westrich 1989).The degree of host plant specialization among bees widely Kuhlmann 2008).Even in polylectic species, host plant choice is usually limited to some degree and as in more specialized taxa governed by constraints with respect to pollen digestion, flower recognition or flower handling (Sedivy et al. 2008, Haider et al. 2014).Detailed knowl edge of pollen host use is important for species conserva tion, the reconstruction of bee host range evolution and our understanding of the fascinating interrelationships between flowers and their pollinators.
The pollen host preferences of Western, Central and Northern European bee species are fairly well known thanks to several studies, which analyzed pollen loads of collected females by light microscopy (Chambers 1968, Raw 1974, Westrich 1989, Müller 1996, Michez et al. 2008, Müller and Kuhlmann 2008, Sedivy et al. 2008, 2013, Haider et al. 2014, Wood et al. 2016, Wood and Roberts 2017).Gaps of knowledge, however, exist for numerous species, which exclusively or predominantly occur in the Alps.Among the bee species, which have their main Central European distribution in the Alpine arc, pollen host preferences have been thoroughly ana lysed so far only for Colletes floralis Eversmann, Colletes impunctatus Nylander, Anthidium montanum Mora witz, Chelostoma grande (Nylander), Hoplitis lepeletieri (Pérez), Hoplitis loti (Morawitz) and Hoplitis tuberculata (Nylander) (Müller and Kuhlmann 2008, Müller 1996, 2015, Sedivy et al. 2008, 2013, Westrich 1989, 1993).For many other predominantly alpine species, information on host plant use either completely lacks, is based on field observations, which do not differentiate between nectar uptake and pollen collection, or relies on pollen analysis of female scopal loads without quantification, resulting in a list of pollen sources not considering their relative significance in the species' larval diet.
The present publication aims to fill the knowledge gap still existing on the host plant preferences of 19 predomi nantly alpine bee species, which are either endemic to the Alps or have a boreoalpine or boreomontane distribution.For that purpose, pollen stored in the scopae of females from museum and private collections originating from across the Alpine arc was removed and microscopically analysed.

Methods
The pollen host spectra of 19 bee species belonging to the six genera Andrena and Panurginus (Andrenidae), Dufourea (Halictidae) and Megachile, Hoplitis and Osmia (Megachilidae) were assessed by microscopical analysis of the scopal pollen contents of 877 female specimens deposited in museum and private collections and cap tured between the beginning of the 20 th century and 2018.For each species, the aim was to analyze a minimum of 40 and a maximum of 50 pollen loads all originating from the Alpine arc.This goal was not attained for Osmia alticola Benoist and Osmia steinmanni Müller due to their rarity and poor representation in collections.For Panurginus herzi Morawitz, the pollen samples analysed by Vögeli (2001) were included in the present study and supplemented with numerous new samples.To account for potentially deviating pollen host use by specimens from different populations, pollen samples from females collected at as many different localities as possible were analysed.Specimens with identical information on the collection label (i.e.locality, date and collector) were regarded as originating from the same locality.Before removing pollen from the female scopae, the degree to which they were filled was estimated.The amount of pol len in the scopae was assigned to five classes, ranging from 1/5 (filled to onefifth) to 5/5 (full load).The pol len grains were stripped off the scopae with a fine nee dle and embedded in glycerol gelatine on a microscopic slide.When a pollen load was composed of different pol len types, their percentages were estimated by counting the grains along two transects chosen randomly across the cover slip at a magnification of 400×.Pollen types represented by less than 5% of the counted grains were excluded to prevent a potential bias caused by contamina tion.For pollen loads consisting of two or more different pollen types, the proportion of the different types was cor rected by their volume (Buchmann andO'Rourke 1991, Silveira 1991).For that purpose, the relative volumina of all pollen types within the sample were estimated by eye and the counted numbers of each type multiplied by a factor that corresponded to its volume.After assigning different weights to scopae according to their degree of filling (full loads were weighted five times more strongly than scopae filled to only onefifth), the estimated per centages were summed up over all pollen samples for each species.To characterize the degree of host plant association, such as oligolecty, mesolecty or polylecty, definitions and methods proposed by Müller and Kuhl mann (2008) were applied.The pollen grains were identi fied down to family or, if possible, to subfamily, tribal or genus level at a magnification of 400× or 1000× with the aid of the literature cited in Westrich and Schmidt (1986), Beug (2004) and an extensive reference collection.Flow er records written on the collection labels often facilitat ed pollen identification to a taxonomic level lower than the plant family.Difficult pollen types were identified by the palynologist Katharina Bieri (Biological Institute for Pollen Analysis, Kehrsatz).Certain closely related plant genera could not be properly identified with the method applied in the present study, such as Sedum and Sempervivum (Crassulaceae), Fragaria and Potentilla (Rosaceae), Euphrasia and Rhinanthus (Orobanchaceae) and Campanula and Phyteuma (Campanulaceae).Pollen grains of these taxa, however, were assigned to genus level if altitude, phenology or habitat excluded the alternative genus or if own field observations, literature data or label records all invariably suggest only one of the two genera.Such nonmorphologically based pollen identifications are marked with "cf." in Table 1.The distinction of pol len from Rhododendron and Vaccinium (Ericaceae) and from Euphrasia and Veronica (Plantaginaceae) is diffi cult.For the former pair the thickness of the exine and for the latter pair the shape of the swollen grains was used as distinctive characters.However, these characters could only be recognised in grains that were specifically orient ed within the gelatine layer and/or completely swollen.If only typical Rhododendron (or Vaccinium) grains or typical Euphrasia (or Veronica) grains were present in a pollen sample, all grains for which the distinctive charac ters were not discernible were assigned to Rhododendron (or Vaccinium) and Euphrasia (or Veronica), respectively.Data based on a comprehensive literature survey on con firmed pollen hosts and observed flower visits as well as field observations made by the author during the last three decades were used to complement the results obtained by the microscopical analysis of scopal pollen loads.
The taxonomy of Panurginus in Central Europe is still under discussion.While the validity of P. herzi Morawitz as a biological species is undisputed, there is no consen sus yet whether P. sericatus (Warncke) is a species of its own (e.g.Schwarz andGusenleitner 1997, Amiet et al. 2010) or should rather be treated as a western subspecies of P. montanus Giraud (e.g.Warncke 1972, Ebmer 2001).
No morphological characters are known to discriminate between the females of these latter two taxa.In contrast, the males slightly differ in the shape of the gonostylus (Amiet et al. 2010).However, there are populations in eastern Switzerland, which show an intermediate gono stylus shape (Amiet et al. 2010).Due to these morpholog ical transitions, the two taxa are regarded as conspecific in the present study.The characters given by Amiet et al. (2010) to separate the females of P. herzi from those of P. montanus proved to be partly unreliable.The shape of the head, however, was recently found to be a reliable charac ter allowing for the unambiguous identification of P. herzi (Hopfenmüller 2017).To quantify this character, the ratio of head length to head width (sensu Michener 2007) was measured under a stereomicroscope for 150 Panurginus females originating from 78 different localities in Switzer land, Austria and Germany.The measurements revealed an almost nonoverlapping bimodal distribution in the head length to head width ratio (Fig. 1), corroborating the reliability of this character to separate the females of P. herzi and P. montanus.For the present study, Panurginus females with a head length to head width ratio of 0.76-0.82were identified as P. herzi and those with a ratio of 0.84-0.90as P. montanus.Pollen loads of females with a ratio of 0.83 were not considered for pollen analysis.

Andrena (Andrena) fucata Smith, 1847
Andrena fucata harvested the pollen of 17 plant fami lies, among which Rosaceae, Apiaceae and Helianthemum (Cistaceae) predominated (Tab. 1, Fig. 2).Pollen of these three plant taxa contributed 77.7% to the to tal pollen grain volume.Herbs, shrubs and trees were among the exploited hosts (Tabs 1, 2), which suggests a high flexibility of the female bees to use different vegetation strata for pollen collection.The seven plant families listed by Chambers (1968) and Westrich (1989) as pollen hosts of A. fucata were all confirmed in the present study.Rhamnaceae were recently demonstrated to be an additional pollen source (Wood and Roberts 2017), and observations of flower visits on Berberis and Lysimachia (Stoeckhert 1933, Dylewska 1987, BWARS 2018) suggest that Berberidaceae and Primu laceae might be two further plant families occasionally exploited for pollen.The high significance of Rosaceae in the host plant spectrum of A. fucata as found in the present study is confirmed by two British studies, which identified Rosaceae pollen to be the most important con stituent of the collected pollen, representing 80.3% and 57.9%, respectively, of the total pollen grain volume (Chambers 1968, Wood andRoberts 2017).According to several authors, Rubus idaeus L. is a preferred pol len host among the Rosaceae (Alfken 1913, Stoeckhert 1933, Amiet et al. 2010).

Andrena (Andrena) lapponica Morawitz, 1872
Andrena lapponica had the narrowest pollen diet among the three Andrena species of the subgenus Andrena in vestigated in the present study.Although it collected the pollen of 16 plant families (Tab.1), flowers of the Ericaceae were strongly preferred.Among the Ericace ae, both Vaccinium and Rhododendron were exploited.As revealed by field observations, Vaccinium myrtillus L. is the most important pollen host in the Swiss Alps (Fig. 3), albeit the pollen of other Vaccinium species such as V. vitis-idaea L., V. uliginosum L. or V. oxycoccos L. is occasionally also collected (Stoeckhert 1933, Westrich 1989).As the flowering period of the main host is short and hardly extends over a period of more than few weeks, many of the alternative nonEricaceae hosts might have acted as replacement pollen sources before and after the  Michener (2007).n = total number of pollen loads, N = number of pollen loads from different localities.(Gogala 2011).
Here, pollen is collected on Helianthemum (Cistaceae) and Salix (Salicaceae).Thus, the pollen specialization of A. lapponica appears to be less strict than was formerly assumed, e.g. by Westrich (1989).

Andrena (Cnemidandrena) freygessneri Alfken, 1904
Andrena freygessneri collected the pollen of 10 plant families (Tab.1).However, it exhibited a strong prefer ence for Crassulaceae.Field observations indicate that species of Sempervivum, such as S. arachnoideum L., S. montanum L. and S. tectorum L., are the main or even the exclusive hosts among the Crassulaceae (Fig. 5).As Crassulaceae pollen cannot reliably be assigned to genus level by the method applied in the present study, it can not be excluded that other Crassulaceae taxa, such as Sedum, are occasionally also exploited.The preference for Sempervivum as demonstrated by this study is supported by observations by De Beaumont (1958), Ebmer et al. (1994), Ebmer (2001Ebmer ( , 2003) ) and Amiet et al. (2010), who recorded A. freygessneri visiting flowers of Sempervivum in the Swiss and Austrian Alps.

Andrena (Oreomelissa) coitana (Kirby, 1802)
Andrena coitana harvested the pollen of 12 plant fam ilies (Tab.1).More than 95% of the pollen recorded in the female scopae originated from herbs (Tabs 1, 2), sug gesting that the species usually does not collect pollen on shrubs and trees except for Rubus (Rosaceae), but restricts pollen harvesting to the herbal layer.Important host plant taxa, whose pollen represented more than 10% of the total pollen grain volume, were Asteraceae (Fig. 6), Campanulaceae and Rosaceae.All three subfamilies of the Asteraceae were exploited and Potentilla was the most important host among the Rosaceae.Literature data, label records and field observations indicate that among the Campanulaceae flowers of Campanula, Jasione and Phyteuma are all exploited for pollen (Tab.2).Chambers (1968) and Westrich (1989) list pollen hosts belonging to five and seven plant families, respectively, among which only Caryophyllaceae were not recorded in the pollen loads analysed in the present study.

Panurginus herzi Morawitz, 1891
Panurginus herzi exclusively collected pollen on Potentilla (Rosaceae) except for two specimens, whose pollen loads additionally contained marginal amounts of pollen of Helianthemum (Cistaceae) and Vaccinium (Ericaceae), respec tively (Tab.1).Field observations revealed that Potentilla aurea L. and P. erecta (L.) Raeusch.are among the most important pollen hosts in the Swiss Alps (Fig. 7).These results are in line with Vögeli (2001), who supposed P. herzi to be a Potentilla oligolege, and with Romankova and Astafurova ( 2011), who mention a flower visiting record of P. herzi on Potentilla chrysantha Trevir. in western Siberia.

Panurginus montanus Giraud, 1861
Panurginus montanus had a distinctly broader diet than P. herzi and collected the pollen of 10 plant families (Tab. 1, Fig. 8).Potentilla (Rosaceae) and Helianthemum (Cis taceae) were by far the most important hosts.Pollen of these two taxa contributed 89.7% to the total pollen grain volume.As numerous pollen loads contained mixtures of Potentilla and Helianthemum pollen, the existence of two cryptic species each specialized to a different host can be excluded.The results of this study contradict the assump tion that P. montanus is specialized to Asteraceae (Westrich 1989).Blüthgen (1952) observed several females visiting the flowers of Geranium sylvaticum L. (Geraniaceae), sug gesting that Geranium might be an additional pollen host.

Dufourea alpina Morawitz, 1865
Dufourea alpina collected the pollen of 7 plant families (Tab.1), but exhibited a strong preference for the pollen of Campanulaceae (Fig. 9).Field observations showed that among the Campanulaceae flowers of both Campanula and Phyteuma are exploited for pollen.These results support Friese (1898) and Westrich (1989), who assumed a preference of pollen collecting females of D. alpina for Phyteuma and Campanula, respectively.

Dufourea paradoxa (Morawitz, 1867)
Dufourea paradoxa had a distinctly broader diet than D. alpina and collected the pollen of 13 plant families, among which Crassulaceae, Orobanchaceae and Lamiaceae pre dominated (Tab.1).Pollen of these three plant families contributed 63.7% to the total pollen grain volume.Lit erature data, label records and field observations indicate   Michener (2007).The plant families are arranged according to their significance in the species' pollen host spectrum (see Tab. 1), the plant genera within each family are arranged in alphabetical order.
that among these three families flowers of Sempervivum, Euphrasia and Thymus, respectively, are the most import ant pollen hosts.In strong contrast to Dufourea alpina, flowers of Campanulaceae are only exceptionally exploit ed.Friese (1898) observed females visiting the flowers of Veronica (Plantaginaceae), suggesting that this genus is probably an additional pollen source.

Megachile (Megachile) alpicola Alfken, 1924
Megachile alpicola collected the pollen of 16 plant fam ilies (Tab.1).Almost 75% of the pollen recorded in the female scopae originated from the flowers of Fabaceae and Asteraceae (Fig. 10).Among the Fabaceae, Lotus was by far the most important host, but pollen was also collected on several other genera.Among the Asterace ae, species of all three subfamilies were exploited for pollen.The four plant families listed by Westrich (1989) as pollen hosts were all confirmed in the present study.

Megachile (Xanthosarus) analis Nylander, 1852
Megachile analis had a distinctly narrower pollen diet than M. alpicola and restricted pollen collection mainly to spe cies of Campanulaceae and Fabaceae (Tab.1).Pollen of these two plant families contributed 95.3% to the total pol len grain volume.Field observations revealed that among the Campanulaceae both Campanula and Phyteuma are exploited for pollen.Among the Fabaceae, Lotus was by far the most important host, but pollen was also collected on other genera, such as Hippocrepis or Onobrychis.The strong preference of M. analis for Campanulaceae and Fabaceae as found in the present study conforms to field observations and pollen analytical studies by other authors (Alfken 1913, Benno 1952, Westrich 1989).In northern Europe, M. analis often visits the flowers of Erica tetralix L. (Ericaceae), which are forcefully exploited for nectar and possibly also serve as pollen source (Alfken 1913, Benno 1952, Haeseler 1980).In the pollen loads from the Alps, however, no pollen of Ericaceae was recorded.

Hoplitis (Anthocopa) villosa (Schenck, 1853)
Hoplitis villosa almost exclusively collected pollen on Asteraceae.In seven pollen loads, however, pollen of Helianthemum (Cistaceae) or Geranium (Geraniaceae) was recorded in addition to that of Asteraceae (Tab.1), suggesting that pollen is rarely harvested also on plant taxa other than Asteraceae.Among Asteraceae, species of the subfamily Cichorioideae were by far the most im portant pollen hosts (Fig. 11), followed by representa tives of the subfamily Carduoideae, whereas pollen of the subfamily Asteroideae was only exceptionally col lected.These results are in line with pollen analytical studies by Westrich (1989), who categorized H. villosa as an Asteraceae oligolege that preferentially exploits species of the Cichorioideae.

Osmia (Helicosmia) labialis Pérez, 1879
Osmia labialis exclusively collected pollen on Asteraceae except for one specimen, whose pollen load additionally contained marginal amounts of pollen of Helianthemum (Cistaceae) (Tab.1).Among the Asteraceae, it showed a near exclusive preference for pollen of the Carduoideae (Fig. 12) and only very rarely collected pollen on Cicho rioideae and Asteroideae.All published flower records of O. labialis refer to species of the Carduoideae (Tkalců 1975, Ebmer 2001, Herrmann 2010, Kraus 2010), sup porting the high importance of this subfamily in the spe cies' larval diet.O. labialis appears to have a narrower pollen diet than its closest relative Osmia leaiana (Kir by), which also restricts pollen collection to the Astera ceae, but often also exploits Cichorioideae in addition to Carduoideae (Raw 1974, Westrich 1989, A. Müller un published data).
and Lotus (see below), the diet of O. alticola was broad er and encompassed the pollen of additional Fabaceae genera, such as Anthyllis, Trifolium and others.Due to the low number of pollen samples available, the catego rization of O. alticola as a Fabaceae oligolege (Tab. 1) may appear premature.However, the fact that eight out of the 16 pollen loads analyzed consisted of mixtures of pollen of several Fabaceae genera clearly points to a strict pollen specialization at the family level.

Osmia (Melanosmia) inermis (Zetterstedt, 1838)
Osmia inermis collected the pollen of four plant families, but exhibited a strong preference for Loteae, particularly for Lotus and Hippocrepis (Tab.1).These results are in line with Westrich (1989), who categorized O. inermis as being narrowly polylectic with a preference for Fabaceae, but do not support Stoeckhert (1933), who assumed Vaccinium to be the preferred pollen host.In Atlantic Canada and the northeastern United States, however, the species appears to be dependent primarily on Ericaceae (Hicks 2009).There, it occasionally also collects pollen on Salix (Salicaceae) (Stubbs in Hicks 2009).

Osmia (Melanosmia) parietina Curtis, 1828
Osmia parietina harvested the pollen of five plant fami lies (Tab.1).However, it exhibited a strong preference for Loteae, particularly for Lotus and Hippocrepis (Tab. 1, Fig. 13).Among the Fabaceae, Trifolium was also reg ularly exploited; its pollen was recorded in eleven pol len loads and contributed 7.4% to the total pollen grain volume.Sedum (Crassulaceae) is listed as an additional pollen source by Westrich (1989), and Veronica (Plantag inaceae) might possibly be a further pollen host based on the observations by Blüthgen (1952).

Osmia (Melanosmia) steinmanni Müller, 2002
Osmia steinmanni had a strong affinity for Fabaceae (Tab. 1) and collected pollen mainly on Hippocrepis and Lotus, more rarely on Trifolium.One load additionally contained substantial amounts of pollen of Rhododendron (Ericace ae).In spite of the low number of pollen loads available, these findings suggest that O. steinmanni has very similar pollen host preferences as its close relative O. inermis (see above).More pollen loads are needed both to clarify the significance of Ericaceae in the pollen diet of O. steinmanni and to examine whether pollen of plant families other than Fabaceae and Ericaceae is occasionally also harvested.

Osmia (Melanosmia) uncinata Gerstaecker, 1869
Osmia uncinata harvested the pollen of eleven plant fam ilies (Tab.1), but exhibited a strong preference for Faba ceae, which however was less pronounced than in the other Osmia species of the subgenus Melanosmia inves tigated in the present study.Among the Fabaceae, Lotus, Hippocrepis and Trifolium were the predominant pollen sources, but other Fabaceae genera were also exploited.Moderately important pollen hosts were species of Rosa ceae and Lamiaceae, whose pollen contributed about 10% and 7%, respectively, to the total pollen grain volume (Fig. 14).Thus, O. uncinata is the least specialized species of the subgenus Melanosmia both in terms of the number of plant families and the number of Fabaceae genera exploit ed.Westrich (1989) lists pollen hosts belonging to ten plant families, among which Asteraceae, Cornaceae, Ericaceae and Salicaceae were not recorded in the pollen loads an alysed in the present study.Based on the observations by Stoeckhert (1933), pollen might occasionally also be col lected on Polygala (Polygalaceae) and Viola (Violaceae).

Osmia (Melanosmia) xanthomelana (Kirby, 1802)
Osmia xanthomelana exclusively collected pollen on Hippocrepis and Lotus (Fabaceae) (Tab. 1, Fig. 15), ren dering this species the most specialized among the Osmia (Melanosmia) species examined in the present study.The strict dependence of O. xanthomelana on only two Faba ceae genera of the tribe Loteae was already supposed by Westrich (1989).In contrast, the assumption of Stoeck hert (1933) that O. xanthomelana also harvests pollen on other taxa than Hippocrepis and Lotus is not supported by the results of the present study.

Discussion
The 19 bee species investigated in the present study wide ly vary in their pollen host spectra and degree of host plant specialization, revealing a fascinating diversity in bee pol len host use (Tab.1).The examined set of species encom passes i) narrowly oligolectic species, which exclusively collect pollen on a single plant genus, ii) broadly oligolec tic species, which harvest pollen on a single plant family, iii) mesolectic species, which are dependent on two plant families, iv) polylectic species, which exhibit a strong but not exclusive preference for a single plant taxon, and v) polylectic species, which do not prefer any single plant taxon and exploit up to 17 different plant families.
Comparison of pollen host use among closely related species of the same subgenus or the same monotypic ge nus reveals different patterns (see species accounts above and Tab. 1).The six Osmia species of the subgenus Melanosmia all exhibit a pronounced affinity for the pollen of Fabaceae, supporting other studies which demonstrated that host plant preferences are often conserved in clades of closely related bee species (Müller 1996, Wcislo and Cane 1996, Michez et al. 2004, Sipes and Tepedino 2005, Patiny et al. 2007, Larkin et al. 2008, Michez et al. 2008, Sedivy et al. 2008, 2013).However, the degree of depen dence on Fabaceae pollen differs among the O. (Melanosmia) species, ranging from a narrow specialization on Fabaceae as in O. xanthomelana, which collects pollen solely from the flowers of two closely related Fabaceae genera, to a moderately strong dependence on Fabaceae as in O. uncinata, which exploits the flowers of at least 14 additional plant families.In contrast to O. (Melanosmia), the three Andrena species of the subgenus Andrena wide ly differ in their pollen host use as do the two Dufourea species.While the differences between A. fucata and A. rogenhoferi are potentially due to deviating distribution and habitat selection with A. fucata restricted to forested areas of the montane and subalpine zone and A. rogenhoferi colonizing a wide spectrum of habitats from the submontane to the alpine zone (SwissBeeTeam 2018), the pronounced preference of A. lapponica for Ericaceae pollen might possibly be genetically based as it is the case for other bee species with a specialized diet (Praz et al. 2008).A genetic basis for the differing pollen host choice is also suggested for the two Dufourea species, which often colonize the same habitats in the Alps, where they encounter a similar flower supply but nevertheless collect the pollen of different plant taxa.For both Panurginus species, pollen of Potentilla (Rosaceae) plays an import ant role in the larval diet.Interestingly, P. herzi is entirely dependent on Potentilla, whereas P. montanus is capable of exploiting several additional hosts.If future studies show that the polylectic habit of P. montanus is evolu tionary derived, this would support the view that many generalist bee species that evolved from specialized an cestors had broadened their diet under maintenance of the exclusive host of their ancestors (Sedivy et al. 2008).

Figure 1 .
Figure 1.Head length to head width ratio in females of Panurginus herzi (light blue) and P. montanus (dark blue) (n = 150).Definitions for head length and head width according to Michener (2007).