Multivariate morphometric analysis of the Stipa turkestanica group (Poaceae: Stipa sect. Stipa)

Based on numerical analyses of macromorphological characters (cluster analysis, principal coordinate analysis and principal component analysis), scanning electron microscopy observation of lemma and lamina micromorphology, as well as field observations, five taxa belonging to the Stipa turkestanica group have been recognized in the mountain area of Central Asia. They are S. turkestanica subsp. turkestanica, S. turkestanica subsp. trichoides, S. macroglossa subsp. macroglossa, S. macroglossa subsp. kazachstanica and S. kirghisorum. As a result of this study, we propose one new combination, S. macroglossa var. pubescens, and designate lectotypes for S. turkestanica subsp. trichoides and S. macroglossa var. pubescens, and an epitype for S. kirghisorum. Illustrations of micromorphological structures of the lemma, patterns of leaf hairiness and an identification key are provided. A taxonomic synopsis including information on nomenclatural types, synonyms, descriptions of the taxa, and, as supplementary information, a list of the specimens examined is also presented.


Introduction
The mountains of Central Asia have been recognized as being among the world's top 34 biodiversity hotspots (Mittermeier et al. 2005). There are more than 8000 vascular plant species but, despite this, it contains several regions that are still significantly underexplored (Kamelin 2002). Central Asia is also characterized by a high diversity of taxa from the genus Stipa L. Of the 150 species of feather grasses, ca. 70 species occur in Central Asia (Roshevitz 1934;Pazij 1968;Bor 1970;Tzvelev 1976Tzvelev , 2012Freitag 1985;Kotukhov 2002;Nobis 2010Nobis , 2013Nobis , 2014 and more than half of them are generally accepted as endemic or subendemic species either to particular mountain ranges or to that region (Tzvelev 1976;Nobis 2011aNobis , b, 2012Nobis , 2013. Feather grasses of Central Asia are classified into several sections containing critical groups of closely related and similar taxa. In this study, we examined the Stipa turkestanica group of Stipa sect. Stipa. The group includes seven taxa: Stipa turkestanica Hack., S. kirghisorum P.A.Smirn., S. macroglossa P.A.Smirn., S. trichoides P.A.Smirn., S. ikonnikovii Tzvelev, S. kazachstanica Kotukhov and S. nikitinae Tzvelev, which occur throughout Central Asia and in surrounding areas but are particularly abundant in the Altai, Tian-Shan, Kopet Dag, Pamir, Alai, Hindu Kush and Kunlun mountain ranges. The main, striking characters of this group are: glabrous (smooth or scabrous) column (lower part of awn), more or less scabrous leaves of the vegetative shoots and a relatively short anthecium, less than 17 mm long, without a ring of hairs at the apex. Because of morphological similarity, some of taxa from the S. turkestanica group were treated at different taxonomic levels or merged with other species (Table 1).
Due to the high phenotypic plasticity observed within the Stipa turkestanica group, narrow species concept or taxonomic splitting (e.g., Kotukhov 2002;Tzvelev 2012) may cause many difficulties in determination of species. On the other hand, too broad species concept can also create problems in understanding patterns of diversity (Freitag 1985;Gonzalo et al. 2013). Thus, well-documented delimitation of individual taxa supported by specific combination of morphological traits, habitats requirements and distribution range is highly necessary. Additionally, clarification of the boundaries between particular species of the examined group is of crucial significance in terms of correct identification and delimitation of the taxa arisen through hybridisation, such as S. 9 alaica Pazij, S. 9 manrakica Kotukhov, S. okmiri Dengub. or S. 9 talassica Pazij, for which they are one of the parental species (Nobis 2013).
Despite the uniformity of the epidermal characters, micromorphological traits such as presence and shape of long cells, silica bodies, cork cells, hooks, prickles and hairs on lemmas and/or laminas display specific patterns within Stipa and have proved to be taxonomically useful (e.g., Barkworth and Everett 1987;Romaschenko et al. 2012;Nobis 2013Nobis , 2014Nobis et al. 2014bNobis et al. , 2015. However, up to now within the S. turkestanica group, patterns of the lemma micromorphology are known only for S. turkestanica s. lato Nobis et al. 2015). Thus, we endeavored to test the micromorphological patterns of the lamina and lemma micromorphology to identify characters which would provide additional support for the taxonomical findings.
The aims of this paper were: (a) to clarify patterns of macromorphological and micromorphological variations within the Stipa turkestanica group; (b) to disclose the level of morphological differentiation for recognized taxa; (c) to indicate the most informative characters for identification of the taxa; (d) to provide a key, descriptions and notes on ecology and distribution of all members of the studied group.

Characters scored for morphometric analyses
The numerical analyses are based on 108 specimens of Stipa turkestanica subsp. turkestanica, 90 of S. turkestanica subsp. trichoides, 121 of S. macroglossa subsp. macroglossa, 6 of S. macroglossa subsp. macroglossa var. pubescens; 44 of S. macroglossa subsp. kazachstanica, 123 of S. kirghisorum, 5 of S. ikonnikovii, and 2 of Stipa nikitinae. Measurements were conducted on each welldeveloped specimen, using digital calipers or a ruler. A total of 26 most informative quantitative and qualitative morphological characters were chosen for analysis (see Table 2). Measured morphological characters of the anthecium and the awn are illustrated in Fig. 1.

Multivariate morphometric analyses
Each specimen was treated as an Operational Taxonomic Unit (OTU), in accordance with the methods used in numerical taxonomy (Sokal and Sneath 1963). The assumptions of normality were examined using the Lilliefors test. Those variables that did not meet the assumptions of normality were Box-Cox transformed to find the optimal normalizing transformation in each case. The Pearson correlation coefficients or non-parametric Spearman correlation coefficients were then calculated to check if any strong correlation ([0.90) exists between variables which could potentially affect the results of further multivariate analyses. If the correlation coefficients for the logically correlated pairs of variables exceeded r=0.90, they were excluded from the multivariate analyses. Cluster analysis was performed on all the OTUs using all 26 characters, to obtain information about general relationships and similarities between them. The similarity between two OTUs was calculated on the basis of Gower's general similarity coefficient. The dendrogram was prepared using UPGMA method.
A principal coordinate analysis (PCoA) was performed on the basis of all quantitative and qualitative characters. The goal of PCoA was positioning of objects (individuals) in a space of reduced dimensionality while preserving their distance relationships.
Subsequently, a principal component analysis (PCA) was conducted on all quantitative characters, on the basis of the correlation matrix (Sokal and Sneath 1963). The specimens were grouped with no a priori assumptions. Each specimen was then marked with the symbol on the scatter plot corresponding to a particular taxon. The analysis enabled the determination of reduced set of variables (features), which were most strongly correlated with the principal components. Factors with eigenvalues[1 were chosen according to the Kaiser criterion (Kaiser 1960). The characters which had the highest factor loadings on first three principal components (r[0.60) were selected.
Next, descriptive statistics of characters for previously recognized groups were calculated. To reveal significant differences between means of particular characters across all examined groups (after using Levene's test to assess the equality of variances), one-way analysis of variance (ANOVA) followed by Tukey's HSD test for unequal sample frequencies was calculated.

Scanning electron microscopy (SEM) observation
Using scanning electron microscopy (SEM) observation, pattern of the lemma micromorphology and hairiness of vegetative leaves (adaxial surface of blades) were studied in all of the examined taxa of the group. We analyzed 28 samples of 7 species. A list of voucher specimens used in the study is given in Table 3. For SEM observation, samples were coated with gold using a JFC-1100E Ion sputter manufactured by JEOL. Micromorphological structures of lemmas and laminas were observed and photographs taken by means of the scanning electron microscope Hitachi S-4700, at various magnifications. Lemmas (removed from mature spikelets in the middle part of the panicle) were studied from the base to the distal portions. Qualitative and quantitative characters were studied for the abaxial lemma surface: length of long cells, shape of silica bodies, presence and shape of hooks and prickles, length and distribution of macrohairs. The terminology used was adopted from Thomasson (1978Thomasson ( , 1981, Ellis (1979), Snow (1996) and Nobis (2013Nobis ( , 2014. The patterns of hairiness of leaves (adaxial surface of blades  Table 2 from the middle part of leaves) were also studied in all of the examined taxa of the group.

Numerical analysis
Cluster analysis (UPGMA) performed on the basis of all 26 quantitative and qualitative characters (Table 2), resulted in the delimitation of two main clusters (Fig. 2). The first comprises groups of OTUs belonging to S. kirghisorum and S. macroglossa and the second, OTUs of S. turkestanica. Within S. macroglossa, two subclusters are clearly distinguished: S. macroglossa subsp. macroglossa and S. macroglossa subsp. kazachstanica. A similar result was generated for the cluster of S. turkestanica, where S. turkestanica subsp. turkestanica and S. turkestanica subsp. trichoides are well differentiated. A comparable pattern is shown in the ordination diagrams from the PCoA, performed also on 26 quantitative and qualitative characters. The first axis clearly separated the OTUs of S. turkestanica subsp. turkestanica and S. turkestanica subsp. trichoides, which are positioned on the negative side of the axis, while all the remaining OTUs belonging to S. kirghisorum and S. macroglossa are located on the positive side. At the same time, axis 2 distinctly divides OTUs of S. kirghisorum from those of both subspecies of S. macroglossa (Fig. 3). For the purposes of comparison, we marked the OTUs of S. nikitinae and S. ikonnikovii within S. kirghisorum cluster (Figs. 2, 3).
The PCA diagram, performed on 22 quantitative characters, also displays a pattern similar to that described above. The first three principal components accounted for 63.6 % of the total variance: 34.9 and 21.2 % for the first and second axis, respectively. Ten characters studied displayed highest correlations with the first axis, five with the second axis and one with the third axis (Table 4). Thirteen characters had high positive factor loadings ([0.60) on the two first axes and two had high negative factor loadings (\-0.60). A plot onto these axes ( Fig. 4) revealed 3-5 neighboring or slightly overlapping groups of OTUs. The first group, located in the right upper part of the diagram, belongs to S. kirghisorum, which consists of three completely overlapping groups of OTUs belonging to S. kirghisorum, S. ikonnikovii (type) and S. nikitinae (type). The cluster in the bottom right-hand section of the diagram can be divided into two slightly overlapping subclusters. They correspond to currently recognized subspecies S. macroglossa subsp. macroglossa and S. macroglossa subsp. kazachstanica, which differ mainly in AL, CalL, DDL and LHB ( Fig. 5; Table 5, Online Resource 2, 3). Apart from the indumentum of the cauline sheaths, a character not used in the analysis, the specimens of S. macroglossa var. pubescens do not differ in any other character, from those of the type variety, and consequently, they are plotted together. The two somewhat overlapping clusters in the left-hand part of the diagram correspond to the two subspecies of S. turkestanica, subsp. turkestanica and subsp. trichoides. Specimens of these subspecies differ from each other in LigV, AwnL, SL, Col 1 L, and CalL (Table 5, Online Resource 2, 3).
The results of the one-way ANOVA revealed significant differences in all of the characters examined (Table 4). The results of the post hoc tests and Tukey's HSD test for variables with normal distribution and multiple

Adaxial surface of leaves of the vegetative shoots: patterns of hairiness
The adaxial surface of leaves of the vegetative shoots is ribbed and covered by hairs (Fig. 5). Density and length of hairs differs in particular taxa of the Stipa turkestanica group; however, patterns of hairiness are consistent. Generally, three types of hairiness can be distinguished here: (a) shortly pilose, with prickles and/or short hairs 0.05-0.1 mm long, (b) pilose, with short hairs 0.15-0.30 mm in length, and (c) pilose with mixture of short and long hairs 0.1 and 0.2-0.5 mm (Fig. 5). The first type contains specimens of S. macroglossa subsp. kazachstanica and S. kirghisorum (Fig. 5d, e), which have a pattern of hairiness unlike all the other members of the studied group. The second type comprises specimens of the three taxa: S. turkestanica subsp. turkestanica, S. turkestanica subsp. trichoides, S. macroglossa subsp. macroglossa, with the adaxial surface of leaves covered by short hairs. The third type contains specimens of S. kirghisorum, in which hairs on the adaxial surface of leaves distinctly differ in length; however, density and distribution of longer and shorter hairs is variable. This concerns in particular the long hairs, which were sometimes located along the marginal ribs or distributed across the entire adaxial leaf surface (Fig. 5).

Lemma micromorphology
The general patterns of the lemma micromorphology are typical of the genus Stipa (cf. Barkworth and Everett 1987;Romaschenko et al. 2012;Nobis 2013;Nobis et al. , 2014bNobis et al. , 2015 and are relatively uniform in all of the studied taxa. Fundamental (long) cells are elongated to rectangular in S. turkestanica s. lato and S. kirghisorum, whereas rectangular to more or less square in shape in both subspecies of S. macroglossa (Fig. 6). Side walls of long cells in all of the studied taxa are raised, thickened and undulate with sinuous Xto V-shaped curves. Silica bodies are reniform to oblong or ovate, while cork cells were either sporadic or completely absent. Hooks are common and generally morphologically similar in all four taxa, whereas prickles are very sparse and uniform, occurring High value of factor loadings ([0.6) and F are given in bold. For character abbreviations, see Table 2 Multivariate morphometric analysis of the Stipa turkestanica group (Poaceae: Stipa sect. Stipa) 143 almost exclusively in the upper part of the lemma (near the apex), sometimes being completely absent. The lemma apex is glabrous or, sometimes possess macrohairs forming a tuft of short hairs up to 0.3 mm long occurring solely on the margins of the ventral part of the lemma. On the abaxial lemma surface, macrohairs are organized in seven lines which always end below the top of lemma in all of the studied taxa (Table 5, Online Resource 2). Exceptionally, e.g., in S. turkestanica subsp. turkestanica (specimens from Iran collected by J. Soják, PR), subdorsal and subventral lines of hairs are undeveloped and, consequently, 5 or 3 lines are visible.

Discussion
The result of morphometric analyses enabled the delimitation of three well-separated sub-groups within the Stipa turkestanica group, namely 'S. turkestanica', 'S. macroglossa' and 'S. kirghisorum'. Because of morphological similarity of S. trichoides and S. turkestanica, Tzvelev (1974) reduced the former to the subspecies of the later. Whereas Freitag (1985), based on the original description, went further treating S. trichoides as a synonym of S. turkestanica. He mentioned the fact that, in the description of S. trichoides, Smirnov (1925) differentiates his new species from S. turkestanica by glabrous, rather than scabrous awn column. On the other hand, Hackel (1906) stated in the diagnosis that S. turkestanica has a relatively short awn with a glabrous or scabrous column. Unfortunately, he gave no information on the length of the ligules at the vegetative leaves. A detailed examination of type collections of S. turkestanica at W and MW has shown that its ligules are much longer than in S. trichoides. Both of these taxa were well defined on the basis of their distribution range (see below); however, in the countries such as Tajikistan, where they co-occur, their delimitation was difficult and they were sometimes misidentified. Generally, the two taxa may be distinguished using the length of ligules at the vegetative leaves, which are (0.5-)1.6-2.8(-3.7) mm long in subsp. trichoides and (2.6-)4.5-8.2(-11.5) mm long in subsp. turkestanica; the length of awn, (145-)163-196(-225) mm long  Table 2 in subsp. trichoides and (88-)118-158(-184) mm long in subsp. turkestanica (Table 5, Online Resource 2); and the surface of column, which is smooth in subsp. trichoides and scabrous, rarely slightly scabrous to smooth in subsp. turkestanica.
The specimens of S. turkestanica subsp. trichoides collected from high elevations in the Alai Mts are smaller and more gracile than specimens from lower elevations, e.g., the Zeravshan Mts and resembles S. turkestanica subsp.
turkestanica. These similarities can be explained by phenotypic plasticity associated with climatic conditions. Such plasticity is also observed in most vascular plants that have a large altitudinal range. In contrast, some specimens of S. turkestanica subsp. turkestanica collected in the southwestern Pamirs (4 June 2011, M. Nobis, KRA) are relatively tall and, apart from the scabrous column of the awns and longer ligules at the vegetative shoots, they are very similar to specimens of S. turkestanica subsp. trichoides   from the western Pamir Alai Mts (Nobis 2013). In our numerical analysis, both subspecies are well distinguished from the other members of the studied group; however, they are always grouped together. Given the several characters that separate the two taxa (Table 5, Online Resource 2), as well as their geographic isolation, we concur with Tzvelev (1974) that they should be treated on the subspecies level. Stipa turkestanica subsp. turkestanica is a more southeastern taxon of higher mountain elevations in Central Asia, occurring mainly in eastern Tajikistan, Afghanistan, Pakistan, northeastern India and in a handful of sites in Iran. Stipa turkestanica subsp. trichoides is more a northwestern taxon, occurring in Kyrgyzstan, Tajikistan, Uzbekistan, Turkmenistan, northern Iran and Afghanistan at generally lower elevations than S. turkestanica subsp. turkestanica (Ovchinnikov and Chukavina 1957;Pazij 1968;Bor 1970;Tzvelev 1976;Freitag 1985;Gonzalo et al. 2013). Kotukhov (1994) described Stipa kazachstanica from the Altai Mts. However, because of high similarity to S. macroglossa, it has recently been reduced to subspecies of the latter (Nobis 2013). Stipa macroglossa subsp. kazachstanica differs from the type subspecies in having somewhat shorter anthecium, seta, as well as in adaxial surface of blades of the vegetative shoots, which are covered by very short hairs, rather than solely by a mixture of short and long hairs (Fig. 5). These taxa were grouped close to each other in all of the analyses performed. Because of very short indument of the adaxial surface of leaves, S. macroglossa subsp. kazachstanica is slightly similar to S. kirghisorum (Fig. 5). However, they differ in the length of anthecium, awn, column, ligules at the vegetative shoots and the length of long cells on abaxial surface of lemma ( Fig. 6; Table 5, Online Resource 2). It is also possible that hybrids of S. macroglossa subsp. kazachstanica and S. kirghisorum can occur in areas where the two last taxa co-occur, such as in the eastern Tian-Shan and Altai Mts. The specimens collected by V. Goloskokov (3 July 1952, LE) in Kungei-Alatau, in the Kara-bulak valley, may be hybrids, with intermediate characters including the length of awn column, seta, and ligules at vegetative shoots (which are 1-3.5 mm in length). Gonzalo et al. (2013) recognized S. macroglossa as a subspecies of S. turkestanica. However, based on the morphological and geographical evidence, we do not share this opinion. Moreover, Gonzalo et al. (2013) treat S. kazachstanica as conspecific with S. turkestanica subsp. turkestanica. However, it is also hard to agree with that, since the OTUs of the former taxon were grouped with OTUs of S. macroglossa subsp. macroglossa with a remarkable distance from the OTUs of S. turkestanica subsp. turkestanica in all our analyses (Figs. 2, 3, 4, 5, 6; Table 5). Stipa macroglossa subsp. kazachstanica differs from both mentioned above taxa, more or less significantly, by 17 and 21 characters, respectively (Online Resource 3). Apart from the morphological characters distinguishing S. macroglossa subsp. macroglossa from S. macroglossa subsp. kazachstanica (Figs. 2, 3, 4, 5, 6), these two taxa differ also in their geographical distribution. The former is a more south-westerly, Central Asian taxon, occurring mainly in Kazakhstan, Kyrgyzstan, Uzbekistan, and Tajikistan (Lavrenenko and Nikolskaya 1965; Nobis et al. 2014c), whereas S. macroglossa subsp. kazachstanica is a more north-easterly, Central Asian taxon, occurring mainly in western Mongolia, western China, eastern Kazakhstan and eastern Kyrgyzstan (Nobis et al. 2014a).
Specimens of feather grasses with pubescent cauline sheaths are sometimes distinguished and given in different taxonomic ranks. In Stipa lessingiana Trin. & Rupr., for instance, they have been distinguished as var. brauneri (Pacz.) Roshev., or sometimes even at subspecies or species level [S. lessingiana subsp. brauneri Pacz. or S. brauneri (Pacz.) Klokov]. However, the pubescent specimens show no other difference from their glabrous relatives, nor do they occupy a distinct geographic range. Moreover, pubescent and glabrous plants may occur in the same population. For these reasons, the variety rank seems to be most appropriate for such specimens. Similar variation is observed also in S. macroglossa (Figs. 2, 3, Online Resource 2), where two varieties can be distinguished: var. macroglossa, with glabrous sheaths of cauline leaves, and var. pubescens, with shortly pubescent sheaths of cauline leaves. Such specimens were seen from Kazakhstan and Tajikistan (cf. Online Resource 1).
Together with Stipa kazachstanica, Kotukhov (1994) described S. tzveleviana Kotukhov from eastern Kazakhstan. It was recently synonymised with S. turkestanica subsp. turkestanica by Gonzalo et al. (2013). However, M. Nobis and P. Gudkova (submitted) based on field investigation and revision of herbarium materials consider S. 9 tzveleviana as a result of hybridisation between S. orientalis Trin. and S. macroglossa subsp. kazachstanica. A detailed examination of corresponding specimens revealed that it differs from S. turkestanica subsp. turkestanica in having shorter ligules at the vegetative shoots, poorly developed ring of hairs present at the top of anthecium and in scabrous column, which is covered by (0.05-)0.1-0.3 mm long spinules or hairs on its lower part and 0.2-0.5(-0.7) mm long on its upper part. In members of the S. turkestanica group, there is no ring of hairs at the apex and the column is either smooth or slightly scabrous macroglossa (e, f), S. macroglossa subsp. kazachstanica (g, h), S. kirghisorum (i, j). h hook, l long cell, mh macrohair, sb silica body owing to short prickles. For these reasons, we did not include S. 9 tzveleviana in the numerical analyses.
In the mountains of Central Asia, Stipa kirghisorum often co-occurs with S. macroglossa and the two taxa have sometimes been misidentified in that region. Within S. kirghisorum, specimens with more purple lower part of the awn were distinguished as S. kirghisorum var. violacea (E.Nikit.) Tzvelev, and recently it was renamed by Tzvelev (2012) as S. nikitinae Tzvelev. The main difference between these two taxa is the color of column, which should be pale-green in S. kirghisorum and redpurple in S. nikitinae. During the revision of herbarium collections and field research, we observed specimens with a pale-green, red-purple, or dark purple awn column occurring frequently even within one population. We also noted that the color of column is influenced by local micro-climatic conditions. Local cooling, and even frosts, causes color changes not only in awns but also in glumes, sheaths or culms. The purple tint is often observed in various species of Stipa, such as S. arabica Trin. & Rupr., S. hohenackeriana Trin. & Rupr., S. orientalis and S. turkestanica. For this reason, it is hard to believe that specimens of S. kirghisorum with redpurple awns merit taxonomic recognition.
Throughout the distribution range of Stipa kirghisorum, we observed striking variation in the indument of the leaf adaxial surface as well as in the length of ventral line of hairs on lemma. The adaxial surface can be covered by short hairs (up to 0.1 mm) and/or by mixture of short and long hairs, while ventral line of hairs on lemma may terminate (0.4-)1.4-3.1(-4.6) mm below the top of lemma. Tzvelev (1977) described Stipa ikonnikovii from the Pamir Mts, and distinguished it from S. kirghisorum by a longer ventral line of hairs on lemma, reaching almost to the top, i.e., terminating (0.3-) 0.4-1.2(-1.5) mm below the top, and longer hairs on the adaxial surface of the vegetative leaves. The first taxon is known almost exclusively from the Badakhshan region in the Pamir Mts. However, apart from these characters, the examined specimens of S. ikonnikovii do not differ from those of S. kirghisorum in any of 25 characters studied. It is worth noting that other taxa from the genus Stipa also express significant variability in hairiness of adaxial surface of vegetative leaves. The variation in the indumentum of the adaxial surface of the lamina is given strong weight by some authors, including even the rank of species (e.g., Martinovský 1980), while other authors consider this kind of variation rather of less important (e.g., Freitag 1985;Nobis 2012;Gonzalo et al. 2013;Gudkova et al. 2013). For the time being, we include S. ikonnikovii into synonymy of S. kirghisorum, but the importance of this character requires further study, including population level.
Distribution: Central-southwestern Asia: the Pamir, Hindu Kush, southern Karakorum, western Himalaya and Elburs Mts, in Tajikistan, Afghanistan, eastern Pakistan, northern India and Iran.
Distribution: Central Asia: the western Pamir Alai, western Tian-Shan and Kopet Dagh Mts, in Tajikistan, Kyrgyzstan, Uzbekistan, Afghanistan, Turkmenistan and northern Iran.
Notes: The original description of Stipa trichoides is based on D. Litvinov's and B. Fedtschenko's collections, without designation of the type (Smirnov 1925). Tzvelev (1976) treated one of the sheets with specimens of S. trichoides collected by D. Litvinov as a type (= holotype), and the other three as isotypes; however, we still do not know which sheet was treated as a holotype. Therefore, a lectotype, the specimen from the Litvinov's collection selected and labeled by M. Nobis is designated here. Habitat: Stony mountain steppes, steppe grasslands and screes, mainly at exposures of S-W, S or W rarely N-E or E and at altitudes of (1100-)1750-2400(-2800) m. Description: Leaves of vegetative shoots: ligules (2.9-) 4.0-6.5(-8.5) mm long; blades (0.4-)0.5-0.6 mm in diameter, outer surface more or less scabrous, adaxial surface of blades covered by shorter than 0.1 mm long hairs. Cauline leaves: ligules of middle leaves (1.1-) 2.3-4.0(-5.6) mm long; leaf sheaths scabrous to shortly hairy scabrous. Glumes subequal, the lower (42- Habitat: Stony mountain steppes, steppe grasslands and screes, mainly at exposures of S-W, N-NE-E and at altitudes of (1200-)1250-1750(-2000) m.
Distribution: Widely distributed in Central Asia and Southern Siberia: southern parts of Asiatic Russia, Kazakhstan, Mongolia, western China, Kyrgyzstan, Uzbekistan, Tajikistan, Afghanistan, Pakistan and northern India.
Notes: The original material (holotype) of Stipa kirghisorum at MW is a fragmentary specimen, consisting of four awns with anthecia and few leaves of the vegetative shoots. The label of the holotype was partially preprinted from the original Kossinsky's labels for plants collected during his expedition in Semipalatinsk province in 1914 (plants material is preserved in LE! and TK!). There is also a significant difference in the text on the label in MW and in LE and TK, and, what is more important, on any of four sheets from original Kossinsky's collection, there is no a name 'm. Bokaj' as it is in protologue. Whereas on two labels (in LE and TK), there is the name Togai lake. In the original diagnosis of the species, there is also very general reference to the type of S. kirghisorum, without date and number of collection. On Smirnov's revision label attached to the holotype at MW (Stipa kirghisorum m. 1924.I, det. P.A. Smirnow.), there is also his handwritten note, ''Semipalatinskaya obl.'' [oblast]. Thus, the holotype at MW is a small part of a collection taken possibly from a sheet preserved at LE. Because the holotype is not complete, a safe determination of the species is insufficient; therefore, a complete specimen from Kossinsky's collection and labeled by P.A. Smirnov as n. sp. on 18 January 1924 is designated here as the epitype.