Surface soil phytoliths as vegetation and altitude indicators: a study from the southern Himalaya

Phytoliths represent one of the few available altitudinal vegetation proxies for mountain ecosystems. This study analyzed 41 topsoil phytolith samples collected from five altitudinal zones in the southern Himalaya as far as, and beyond, the timberline, from tropical forest (up to 1,000 m a.s.l.) to subtropical forest (1,000–2,000 m a.s.l.), to temperate forest (2,000–3,000 m a.s.l.), to subalpine forest (3,000–4,100 m a.s.l.) and finally to alpine scrub (4,100–5,200 m a.s.l.). The statistical results show a good correlation between phytolith assemblages and these five altitudinal vegetation zones: the five phytolith assemblages identified effectively differentiated these five altitudinal vegetation zones. In particular, coniferous phytoliths accurately indicated the timberline. Additionally, we tested the phytolith index Ic (a proxy for estimating the percentage of Pooideae vis-à-vis the total grass content) as a quantifier of phytolith variety versus altitude. Ic increased along altitude, as expected. An investigation of phytoliths provided an initial basis for the analysis of the composition of gramineous vegetation. Furthermore, redundancy analysis and discriminant analysis also suggested a significant correlation between phytolith assemblages and altitude. Our research therefore provides an up-to-date analogue for the reconstruction of changes to palaeovegetation and palaeoaltitude in mountainous areas.

most phytoliths are naturally resistant to strong weathering and are therefore well-preserved in terrestrial sediments, where pollen-spores can be easily destroyed 10 . Second, even though phytoliths can be transported by the wind, gravity-aided in situ deposition remains their dominant modus operandi 6,10 . Third, phytoliths are more reliable indicators for differentiating varieties of grass subfamily and genera than pollen-spores. Besides, as the study of phytolith morphology progresses, the potential of phytoliths to differentiate various microenvironments, especially near the timberline, has become increasingly apparent 11,12 . There has been some limited research focusing on morphology and phytolith types in mountainous areas [13][14][15][16] , but, as far as we know, only a very few studies of the relation between phytolith composition and altitudinal vegetation belts, e.g. on tropical African mountains 17 , have been published. More international research is required. We chose the Himalaya, specifically the southern Qinghai-Tibet Plateau (QTP), for our research.
The QTP averages > 4,000 m a.s.l., rising to 8,848 m a.s.l. at Mount Everest (Qomolangma). With its typical mountain climate ecosystem, it is a key area in the scientific research into Asian, and even global, climate change 18 . The uplift of the QTP is also crucial to a better understanding of global tectonics 19 . The study of the QTP's palaeoaltitude could be significant in both regards, but there remain no direct records of the local palaeoaltitude, palaeovegetation or palaeo-timberline 20 .
The Himalayan Mountains form the southern margin of the QTP and the northern boundary of the Indo-Gangetic Plain (Fig. 1). The southern Himalaya provides environments for various vegetation types, ranging clearly from tropical rainforest to perpetual frost along the altitudinal gradient 21 . It is an optimum area for the study of variations in vegetation, as well research into indicators of QTP palaeoaltitude. Several classification systems are already available for the vegetation of Nepal [22][23][24] , which lies on the mountain range's southern flank. Notably, Dobremez et al. developed a six-category bio-climatic (altitudinal) classification of vegetation 23 . The published data have focused on flora classification and the characterization of plants 24,25 , and the area's flora and vertical vegetation zones are therefore well-documented. However, due to international borders and poor accessibility, no known work on indexing vegetation belts along the tropical rainforest to perpetual frost altitudinal gradient has been conducted, even if the vertical vegetation range and climate change gradient are most marked in this area. It is therefore imperative to establish a useful index for the reconstruction of both palaeovegetation and palaeoaltitude in this region.
We obtained a diverse collection of samples from a wide range of vegetation belts from the southern Himalaya. In this study, we aimed to explore variations in the composition of phytolith assemblages, and Red-filled circles show sampling sites. Topographic relief was obtained using a dataset provided by the Geospatial Data Cloud, Computer Network Information Center, Chinese Academy of Sciences (http://www.gscloud.cn). The figure was created by Xiaohong An using DIVA-GIS 7.5.0 (http://www.diva-gis.org/). verify our hypothesis that phytolith assemblages can indicate and differentiate vegetation zones along an altitude gradient, providing the basic data necessary for the reconstruction of palaeovegetation and palaeoaltitude in mountainous areas.

Results
Vegetation description and material. The Himalaya exhibit typical montane altitudinal vegetation belts. Based on previous work, mainly Dobremez's altitudinal classification 23 , this paper classifies local forest vegetation vertically from bottom to top into six formations (Fig. 2).
Tropical moist lowland Indo-Malayan forest (<1,000 m a.s.l.) (Fig. 2a). Shorea robusta (Sal) is predominant in this belt. Acacia catechu and Dalbergia sissoo replace Sal in riverine forests. Other dominant broadleaved evergreen forest types include Cycas pectinata, Gnetum montanum, Calamus spp., Cythea spinulosa, Arecaceae, Magnoliaceae and Pandanus nepalenses, except for some coniferous forests (Pinus roxburgii) on southern slopes at higher altitudes. In the foothills of western Nepal, Sal forest is replaced by Terminalia/Anogeissus forest. This zone consists of ~2,000 species of flowering plants and ~80 species of Pteridophytes. The common understory grasses include Zingiberaceae, Acanthaceae, Commelinaceae in addition to tall bamboos and grass. Grassland emerges where rainforest has been cut down. Grassland is composed of the tall grasses Saccharum arundinaccum, Apluda mutica, Themeda Forssk, et al. Rice, maize and banana are cultivated in this assemblage 21,26 . Subtropical forest (1,000-2,000 m a.s.l.) (Fig. 2b). This consists of species such as Schima wallichii, Castanopsis indica and Castenopsis tribuloides in relatively humid areas, and Pinus roxburghii forests in drier regions. > 1,900 flowering plant species grow vigorously in this zone 27 . Sinarundinaria nitida accounts for a large proportion of the total shrub content. The grasses Carex changmuensis, Arisaema sp., Ophiopogon bodinieri, Elatostema surculosum, Miscanthus sp. et al. are also found in these forests 26 . Cultivation is common in this belt. In some places, natural forests have been almost entirely replaced by cultivated crops such as Musa paradisiaca and Chaenomeles sinensis. Owing to the hot and rainy summer, as well as the dry and warm winter, there are many paddy fields in this zone 28 . Temperate forest (2,000-3,000 m a.s.l.) (Fig. 2c). This zone is dominated by plants such as Quercus lamillosa and Q. semicarpifolia, forming pure or mixed broadleaved evergreen forest. This latter category can be subdivided into lower temperate mixed broadleaved forest, upper temperate broadleaved forest, upper temperate mixed broadleaved and coniferous forest 22 . Bamboo grows in sunny areas in this zone, and Sinarundinaria nitida is abundant 28 . Ainsliaca latifolia, Ophoiopogon intermedia, Pilea racemosa, Arisaema, Zingiber and Thalictrum are common understory grasses 26 . Subalpine forest (3,000-4,100 m a.s.l.) (Fig. 2d). This 26 . The cultivation line reaches as high as 3,600 m a.s.l., where buckwheat is planted 28 . The highest elevation at which Abies spectabilis, Pinus wallichiana and B. utilis grow is 4,000-4,300 m a.s.l. 29  Perpetual snow (>5,200 m a.s.l.) (Fig. 2e). This zone is composed of permanent snowfields, rocks, glaciers and ice on the high Himalayan ranges to the north. The area is barren, with lichens on exposed rocky places and a few hardy flowering plants, such as Stellaria decumbens. The main vegetation type on the northern slopes of the Himalaya is grass, reflecting the cold-dry climate of the QTP. Forests can be found only in valleys below 4,100 m a.s.l. 28 . Table S1 shows the coordinates, altitudes and the principal flora of the sampling sites.
Phytolith types. Of the 47 processed samples, 41 contained > 300 phytoliths. These were identified and classified into the 27 common categories listed below, mainly following the classification system used by Lu et al. 30 , but with reference to the classification systems of Wang and Lu 31 , Kondo et al. 32 and Twiss et al. 33 , and using the International Code for Phytolith Nomenclature 1.0 34 . The woody phytolith types include globular, abbreviated stellate, cylindrical sclereid and Gymnosperm. Globular types were subdivided into globular echinate (Fig. 3a) and globular granulate (Fig. 3b), the former being produced specifically in Palmae and the latter in tropical trees and shrubs (but principally in Palmae in China) 14,35,36 . Abbreviated stellate (jigsaw) phytoliths ( Fig. 3x) are produced by evergreen broadleaved plants 13,31,37,38 . The cylindrical sclereid (Y-shape) phytolith (Fig. 3aa) is a type particular to broadleaved plants 31,38 . Gymnosperm (Fig. 3w), as the term suggests, derives mainly from Pinaceae 39,40 .
A diversity of gramineous phytolith types were observed during the course of this study. Bilobate short cell (dumbbell 30 ) (Fig. 3d) and cylindrical polylobate (multilobate) phytoliths ( Fig. 3f) are both representative of Panicoideae, which adapt to warm-humid conditions 31 . The cross-shaped 41 phytolith ( Fig. 3c) is typical of the bilobate short cell type; the form of this phytolith produced in maize can be differentiated from the type produced in wild grass by its mirror-image and greater width (usually > 12.5μ m) 42,43 . The cross-shaped type referred to in this paper specifically represents maize, with other variants classed as bilobate short cell phytoliths. Square saddle (short saddle 30 ) phytoliths ( Fig. 3h) are mainly found in Chloridoideae, with a small fraction occurring in Arundinoideae. Chloridoideae mainly grows in dry-hot conditions. Arundinoideae covers a broad range of southern China. The square saddle type is considered representative of C 4 plants 31,38 . Oblong concave saddle (long saddle 30 ) phytoliths (Fig. 3I,j) are produced in Bambusoideae, which grow in hot, moist climates throughout southern China 31,44 .
Cuneiform bulliform cell (fan-shaped 30 ) ( Fig. 3o-3q) and parallepipedal bulliform cell (square and rectangular 30 ) (Fig. 3u,v) phytoliths develop from motor cells produced in Panicoideae, Oryzoideae and Bambusoideae 31 . These plants flourish in the warm and humid climate of southeastern China. Sometimes an apparently parallepipedal bulliform cell can actually be the side-on view of a cuneiform bulliform cell 31 . Some Chloridoideae produce parallepipedal bulliform cells, but not cuneiform bulliform cells. Cuneiform bulliform cell-rice (Fig. 3o) and cuneiform bulliform cell-bamboo (Fig. 3p) phytoliths can be differentiated from other cuneiform bulliform cell phytoliths by the ornamentation of cracks and spines along their front edges, respectively 31,45 .
The hair cell (point 30  In addition to the abovementioned, the Pteridophyte phytolith type (Fig. 3ab) is particular to ferns 31 , and the sedge (papillae 17 ) type (Fig. 3m,n) is typical of Cyperaceae, a grass-like plant growing in wet places 31 .
We identified some types with no current taxonomical significance: the one-horned tower (Fig. 3k) and two-horned tower (Fig. 3l) are small phytolith types, exhibiting a diversity of shapes 48 ; the elongate type, including elongate smooth (Fig. 3z) and elongate echinate (Fig. 3y) phytoliths, develop from long epidermis cells 31 (these increase in quantity in China from south to north and from humid to dry regions); and the gobbet (nubby-irregular shape) type, a non-gramineous phytolith, appears in arid areas in China 30 . We have classified these difficult-to-categorize types as unknown and as-yet-uninvestigated (see Supplementary Fig. S5 online).
Supplementary Table S2 (online) is a summary of phytolith types, descriptions and their source plants and ecoenvironments. More plates of the aforementioned phytolith types can be found in Supplementary Figs S1-S5 (online). Phytolith assemblages. Phytoliths were divided into the following five assemblages, according to phytolith type and percentage of total composition (Fig. 4).
Assemblage I, from tropical lowland evergreen broadleaved forest (< 1,000 m a.s.l.), is characterized by a high percentage of cylindrical sclereid, globular and cuneiform bulliform cell phytoliths. The maximum contents of abbreviated stellate, cylindrical sclereid, globular and cuneiform bulliform cell phytoliths are 1.8%, 7.3%, 3.9% and 19.6% of the total phytolith content, respectively. The highest numbers of broadleaved types (cylindrical sclereid and abbreviated stellate) are found in this belt, where evergreen broadleaved plants predominate. The volumetric production of Palmaceae phytoliths implies tropical lowland, high temperature conditions. Parallepipedal bulliform cell 1, bilobate short cell, and oblong concave saddle types account for 11.6%, 5.2%, 3.4% and 4.3% of the content total, respectively. The low contents of gramineous types such as bilobate short cell and oblong concave saddle phytoliths indicate a weak growth of grass and bamboo in low altitude areas with abundant tree cover. Sedge (2.5%) is at its maximum in this zone, implying a lowland, humid environment. There is also a high quantity of unknown types (4.8%); many varieties from this wide range of phytolith morphotypes have yet to be investigated. The emergence of cuneiform bulliform cell-rice (0.5%) and maize cross (0.5%) phytoliths implies that rice and maize were cultivated in this low altitude zone.
Assemblage II, from subtropical broadleaved forest (1,000-2,000 m a.s.l.), is characterized by cylindrical sclereid, bilobate short cell and cuneiform bulliform cell phytoliths. The higher percentages of bilobate short cells (15.8%) and oblong concave saddles (11.8%) vis-à-vis total content in this assemblage compared to Assemblage I suggests a greater quantity of grasses and bamboo. Frequent and repeated human cultivation has resulted in a high grass content 28 . In contrast with gramineous types, Palmaceae (0.9%) and cylindrical sclereid (2.6%) phytoliths decrease markedly, and abbreviated stellate types decline sharply from 1.8% to 0.3% of total content, but cylindrical sclereid remains the principal woody type. Both falls in overall content are likely to be the result of a decrease in woody plants due to human cultivation.
To summarize, an abundance of phytolith types, accompanied by clear variations in phytolith percentages vis-à-vis total phytolith content, occurs along the altitudinal gradient on the slopes of the southern Himalaya. Phytolith assemblages can be clearly differentiated (Table 1).
Floristic composition (RDA results). 18 of the aforementioned 27 phytolith types account for a certain proportion of the samples and clearly indicate vegetation type. In elucidating the relation between phytolith assemblages and altitude, redundancy analysis (RDA) results (Fig. 5) show that the first and second axes account for 42% and 16% of the total variance, respectively, describing 58% of the information in total. The other axes exhibit very low values. This suggests that phytolith composition is controlled by environmental factors represented by the first two axes, and especially by altitude (the first axis).
As the RDA results reveal, low altitude-adaptable phytolith types are grouped in a positive direction toward the first axis, while those adapting to high altitude point in a negative direction. The former group includes globular, cylindrical sclereid, bilobate short cell, parallepipedal bulliform cell, cuneiform bulliform cell, cuneiform bulliform cell-bamboo, oblong concave saddle and square saddle phytoliths; the latter is composed of gymnosperm-type, hair cell, elongate smooth, elongate echinate, rondel, trapeziform, Stipa-bilobate short cell and gobbet phytoliths (Fig. 5a). Low altitude samples t1-t20 are positively grouped toward the first axis, while high altitude samples t21-t41 exhibit a negative direction (Fig. 5b). This reflects the clustering of phytolith samples from higher altitudes around negative coordinates, in contrast to samples from lower altitudes, which are clustered around positive coordinates. This therefore demonstrates a correspondence between phytolith assemblages and the altitudinal distribution of vegetation.
Floristic composition (DA results). Discriminant analysis (DA) was applied to test whether the established surface phytolith assemblages described in this paper can reliably differentiate the vertical vegetation belts in the southern Himalaya. Using a priori groups, the 41 surface samples with the 18 phytolith types used for RDA were then classified to co-validate the classification of phytolith assemblages. 39 samples (95%) were correctly classified with respect to these a priori groups ( Table 2). The first two discriminant function scores are illustrated herein: group centroids are distinctly separate (Fig. 6). The classification function coefficients are shown in supplementary Ic (%) represents an index for the total percentage of Pooideae grass cover. Ic has been shown to depend principally on altitude (SPSS curve estimation of R 2 = 0.8) (Fig. 7a). As a result, it is clear that, along with increasing altitude, Pooideae grasses increase gradually, while Chloridoideae and Panicoideae grasses decrease. The relation between altitude and Ic can be expressed using the equation:  = − y 50x 485 where x represents Ic (%) and y represents elevation (m).
Ic is a climatic index for temperature, because a high Ic represents higher quantities of Pooideae, which adapt to the cool climates prevalent at high altitudes 49 . Altitude is the most influential factor vis-à-vis temperature in the Nepal Himalaya; mean annual temperature in general gradually decreases northward as altitude increases 50 . The Ic results therefore render an approximate, linear correlation between Pooideae amount, temperature and altitude 51 .
Iph was also tested, but the result of the curve estimation was less useful, because the sig. = 0.109 for curve estimation (Fig. 7b).

Discussion
We have obtained samples from the world's highest mountain range, the Himalaya, and have derived significant results from our analysis of the material. Although the composition of preserved phytoliths in soils is influenced by translocation and dissolution 52 , phytolith assemblages in soils reflect the composition of local vegetation 10 . First, the surface soil phytoliths described in this paper are mainly bleached mountain spodosols, dark-brown mountain earths and brown mountain earths 21 . These weakly acidic  soils are suitable depositional environments for phytoliths; pH ranges between four and six, and is therefore never high enough for the sustained dissolution of phytoliths. Even if dissolved, dissolution rates are similar for most phytolith morphotypes 53,54 . Second, small phytoliths may be dispersed by the wind 55 . However, the geographical scale of transfer versus study area is considered insignificant in this paper. Most probably, therefore, phytolith composition accurately reflects phytolith composition for each corresponding vegetation belt.
It should however be recognized that the percentage content of each phytolith vis-à-vis total phytolith content cannot be equal to the percentage content of the corresponding vegetation. Some plants yield abundant phytoliths, resulting in over-representation. So in describing vegetation assemblages, it is imperative we focus on the relative variation in each phytolith type versus altitude. Furthermore, the taphonomy of phytoliths in sediments should be considered when reconstructing palaeovegetation 10 .
In view of the reliability of our samples, the phytoliths analyzed in this paper can accurately describe vegetation zones. Phytolith percentages clearly illustrate that each vegetation zone has a distinctive phytolith assemblage and some typical phytolith types. In tropical forest at low altitudes (< 1,000 m a.s.l.), the cylindrical sclereid/globular/cuneiform bulliform cell phytolith assemblage corresponds to an abundance of hot climate-adaptable woody plants, such as Shorea, Cycas, Magnoliaceae, Palmae, and Panicoideae, principally Saccharum, Apluda and Themeda. In the subtropical broadleaved forest belt (1,000-2,000 m a.s.l.), the cylindrical sclereid/bilobate short cell/cuneiform bulliform cell assemblage highlights the dominance of broadleaved woody plants and tall Panicoideae grasses (e.g. Miscanthus). In warm-temperate mixed forest (2,000-3,000 m a.s.l.), the saddle/parallepipedal bulliform cell/rondel phytolith assemblage corresponds to the mixture of warm and cold temperate zone grasses, as well as a mixture of broadleaved and coniferous plants. In the subalpine needle-leaved forest zone (3,000-4,100 m a.s.l.), the gymnosperm-type/trapeziform/Stipa-bilobate short cell phytolith assemblage suggests the dominance of Gymnosperm (Abies and Pinus) and low temperate gramineous types (Deyeuxia and Oryzopsis). In the alpine shrub zone (4,100-5,200 m a.s.l.), an abundance of rondel, trapeziform and gobbet phytoliths implies an absence of woody plants and a predominance of high altitude-adaptable Pooideae, including Carex, Festuca, Helictotrichon, and especially Ptilagrostis (belonging to Stipeae).
To corroborate the increasing/decreasing trend in different phytoliths along altitude, typical bilobate short cell, cuneiform bulliform cell and trapeziform phytolith types are shown in boxplots, with emendations (Fig. 8). It is clear that the bilobate short cell and cuneiform bulliform cell types decrease with altitude, while the trapeziform type increases.
DA results, with their clearly-separated group centroids, also demonstrate that the five phytolith assemblages can accurately delineate vertical vegetation zones, as each vegetation belt is represented by a distinctive phytolith assemblage. Our classification of phytolith assemblages suggests that phytoliths can be used as valid proxies for subdividing montane vegetation in the QTP area, and can thus be used effectively in reconstructions of palaeovegetation and palaeoaltitude.
Notably, some phytoliths exhibit a high degree of sensitivity to environment and climate. In particular, the fluctuation of the timberline is crucial to montane phytolith study. It not only reflects variations in vegetational composition, but, more importantly, indicates changes in altitude. The upper altitudinal limit of montane needle-leaved plants is commonly considered to be the same as the timberline 56 . However, most conifer pollens have two or more sacs, e.g. pine pollen. Long-distance dispersal of pollens leads to their wide representation in assemblages. Investigations of modern pollens have shown that pine forest only grows where conifer pollens account for at least 30% of the total 4 , so the existence of conifer pollens does not necessarily accurately reflect the proportion of coniferous plants. Autochthonous or proximal deposition is characteristic of phytoliths 31,37 . Coniferous phytoliths from the southern slopes of the Himalaya appear only in the planting zone of needle-leaved plants, correspondent to the distribution of coniferous forests; phytoliths can thus be used as discriminants of the timberline and the altitude of montane vegetation zones. When researching montane palaeoenvironments, both the frequency of occurrence and the variability in the content of coniferous phytoliths can provide a good basis for deducing the historical position of the timberline, as well as palaeoaltitude.
Furthermore, crop phytoliths can reflect different types of farming activity 43,45 , and indeed different crops grow at different altitudes. The cross-shaped and cuneiform bulliform cell-rice phytoliths produced in the surface soils of the Nepal Himalaya indicate the planting of maize and rice, in accordance with the area's relatively low altitude. This also implies that human disturbance has changed the composition of the local vegetation. The existence of cultivated phytolith morphotypes in sediments and/or archaeological remains would certainly indicate an historical development of this region by humans 10 .
Based upon our qualitative analysis, we conducted a quantitative analysis, with good results. The mathematical transfer function can be ideal for providing a basis for palaeoenvironmental research, especially when substantial quantities of modern data are applied. In this study, although we did not have a huge number of samples, our Ic values expressed the functional relation between phytolith assemblage variety and altitude very well, demonstrating that Ic is a perfect index for representing gramineous composition 17 in mountainous areas like the Himalaya. However, we adjusted its formulaic expression. In previous research, the Stipa-bilobate short cell phytolith was added to neither the numerator nor the denominator in the index's mathematical expression. However, Stipa belongs to Pooideae, and can survive in extremely cold and arid environments. Stipa grasses are widely distributed at high altitudes on the QTP, and are a vital component of the grass cover in our study area 57 . When Stipa-bilobate short cells are taken into account, the curve estimation of R 2 = 0.8. However, R 2 = 0.7 without the inclusion of Stipa-bilobate short cells. Certainly, Stipa-bilobate short cells, as an individual phytolith type, should be distinguished from the sum of phytoliths during identification and this should be taken into account when using the Ic index. However, when samples are collected from soil surfaces or Cenozoic stratigraphic sections in cold and arid areas like the QTP, the presence of Stipa-bilobate short cells must be fully taken into consideration.
Ic can be used to reconstruct the Quaternary palaeoelevation of the QTP during periods of relatively stable climate. We noted that in the East African tropical mountains, Ic values are ~40, ~70 and ~98 at 2000 m a.s.l., 3000 m a.s.l. and 4000 m a.s.l., respectively 17 . The corresponding values are ~50, ~70 and ~90 in our paper. This may imply that Ic has some universal applicability in low latitude areas, as both of the abovementioned areas are low latitude regions.
The humidity-aridity index Iph was also tested, but proved not obviously applicable to altitude. First, unlike temperature, precipitation in the Nepal Himalaya is affected by two major atmospheric circulation systems. The interaction of the complex topography with the monsoonal and westerly weather systems results in variations in the spatial distribution of rainfall 50 , suggesting that the annual precipitation pattern is not dominated by altitude. Second, Iph was defined as the ratio of Chloridoideae (saddle) versus Chloridoideae and Panicoideae (saddle, cross-shaped and bilobate short cell) 17 . The square saddle phytolith is typical type of Chloridoideae. Some Chloridoideae species, which have adapted to drought conditions, are prone to produce square saddle phytoliths. Other Chloridoideae species in humid environments can produce another short cell phytolith, i.e. bilobate. Iph has been shown to be applicable to the Great Plains of North America as well as to tropical savannah 17,58 ; both these environments have dry seasons which last several months. However, the northern Himalaya enjoys rainier weather. Square saddle phytoliths may therefore not indicate the presence of Chloridoideae in the area. Moreover, the presence of Bambusoideae may also render Iph inapplicable. Bambusoideae produce great quantities of oblong concave saddle phytoliths, which account for a certain proportion of all saddle phytoliths. Using Iph as a index, saddle phytoliths should present as Chloridoideae, but the saddle phytoliths in the northern Himalaya are derived from Chloridoideae, Bambusoideae and Arundinoideae. Bremond et al. (2008) also demonstrated that Iph was not a relevant proxy for Chloridoideae versus Chloridoideae and Panicoideae 17 .
Less research has been conducted into woody phytoliths than gramineous phytoliths, most probably because of the great variety in the morphology of woody phytoliths and the consequent difficulties discriminating between them. If a more detailed identification were possible, the environmental marker function would become more precise. For example, palmaceae phytoliths from the montane subtropical evergreen belt provide good indices for discriminating palm tree planting. There is a fundamental need for systematic and detailed research into woody phytoliths. Such palaeoenvironmental research would provide a basic reference tool for identifying vegetation belts, phytocoenoses, edificators, and even typical species.

Methods
In this study, 47 topsoil samples were collected at altitudes between 100 m a.s.l. and 5200 m a.s.l. from the Sino-Nepal Himalaya (Fig. 1). No sample was collected from the perpetual snow zone, because the vegetation becomes extremely sparse above 5200 m a.s.l. After clearing away any loose debris, each sample was extracted from the top 0-2 cm of surface soil and put into a valve bag. A GPS receiver was applied to keep a record of the longitude, latitude and altitude of each sampling site. Samples were dried out in the laboratory before long-term storage in order to prevent clumping.
Phytoliths were extracted from each sample by conventional heavy liquid flotation based on the method of Piperno and Pearsall 6,59 , but were not passed through a sieve, as some phytoliths in this paper (e.g. rondel and square saddle phytoliths) are very small and could be filtered off during the process.
Scientific RepoRts | 5:15523 | DOi: 10.1038/srep15523 All samples were treated in the Key Laboratory of Cenozoic Geology and Environment, Institute of Geology and Geophysics, Chinese Academy of Sciences. A subsample of ~1-3 g was taken from each sample, according to its particular composition. Each subsample was sequentially processed as follows: (1) organics were removed with 30% H 2 O 2 ; (2) a tablet of Lycopodium spores (27637 spores per tablet) was added to determine phytolith concentration (the method of Piperno and Pearsall does not include this step); (3) carbonates were dissolved with 10% HCl; (4) flotation of phytoliths was accomplished using a ZnBr 2 solution (density 2.35 g/cm 3 ); and (5) after cleaning, each subsample was dipped in resinene to facilitate slicing. Morphotypes were counted under a microscope; 41 samples were found to include > 300 phytoliths.
Canoco5 software was used to perform RDA 60 . We used SPSS 17.0 statistics software to build scatter diagrams of Ic and Iph along altitude, and make curve estimations. We also applied SPSS 17.0 to DA and to our boxplots.