Surface Pollen Distribution from Alpine Vegetation in Eastern Tibet, China

We explore the relationship between modern pollen spectra and vegetation patterns in the Eastern Tibet, China in order to provide information on the representation of pollen taxa and improve the general knowledge of vertical pollen transport. Forty-two modern pollen samples collected in surface soil along two altitudinal transects allowed conclusions on vertical pollen dispersal from the alpine region of Dingqing County, Changdu district in Tibet. Discriminant analyses and detrended correspondence analysis (DCA) of 24 pollen taxa were used to further discuss the difference of modern pollen spectra in these alpine vegetation zones. The surface pollen assemblage is divided into three pollen zones, such as subalpine shrub meadow, montane coniferous forest and shrub steppe with sparse trees. Altitude and precipitation are two primary factors contributing to changes in surface pollen assemblage from alpine vegetation in the eastern Tibet. Large amounts of spruce pollen at higher elevations above the timberline might be introduced from lower elevations by upslope winds. Therefore, the interpretation of spruce pollen in the fossil record must take into account long distance upward wind transport. Moreover, the destruction of coniferous forest in the study area is well illustrated in the modern pollen rain.

mainly from Central Tibet and seldom from Eastern Tibet. Because of mountain uplift and stream trenching, there are a lot of high mountains and valleys with great differences in altitude in the Changdu District, Tibet Autonomous Region, resulting in a distinct vertical distribution of vegetation and soil types. Moreover, the Tibetan Plateau is the least anthropologically disturbed region in China and provides a unique opportunity to study the relationships between surface pollen and modern vegetation. Therefore, it is an appropriate area where pollen rain is expected to more closely represent the "natural" vegetation, and the modern pollen-vegetation relationship derived from pollen-rain studies can provide a more realistic basis for paleovegetational reconstruction.
In this paper, we want to discuss altitudinal changes of surface pollen and vegetation in Eastern Tibet. Modern pollen samples were obtained on two altitudinal transects from the alpine area of Dingqing County, Changdu District, Eastern Tibet (Fig. 1). These data were then subjected to statistical analyses to further discuss the environmental factors that affect surface pollen assemblages in the region.

Results
Surface Pollen Assemblage of Different Vegetation Zones. There were 16 796 grains identified as terrestrial pollen within 49 pollen taxa. Among them, dominant arboreal pollens were Picea and Betula, but Pinus, Cupressaceae, Abies and Salix were occasionally found. Ericaceae was the common shrub type. Artemisia was the main mesic-xerophytic shrub and herb pollens. Polygonum, Compositae, Leguminosae and Cyperaceae dominated in mesic and hygrophyte herb pollens. Fern pollen types were Polypodiaceae and Selaginella.
Sporopollen spectrum in this alpine area could be divided into three surface pollen zones based on modern vegetation investigation results (Fig. 2). And each zone is summarized in the following text.
Montane coniferous forest (4 320-3 690 m) (Zone II): Coniferous forest with Picea likiangensis var. rubescens and Abies squamata, occupies the moist areas on the north-facing slope, which might be a little changed to secondary forests of Populus davidiana and Betula platyphylla after the destruction of the forest. Juniperus tibetica is distributed mainly on the south-facing slope in this zone. The zone was dominated by arboreal pollen (maximum value of 99.8%) with a large proportion of Picea. The highest Betula percentage (11.4%) was shown in this zone. Shrub pollen percentage was low (<2%). Higher percentages of Leguminosae (17.1%) and Artemisia (18.8%) were found at a middle elevation in this zone; however, their values were reduced to 2% at lower elevations. Therefore, the pollen composition in this zone can be regarded as a Picea-Artemisia-Leguminosae combination.
Shrub steppe with sparse trees (3 690-3 510 m) (Zone III): Some arboreal trees with Picea likiangensis, Betula platyphylla and Juniperus tibetica and shrubs consisting of Spiraea spp., Lonicera rupicola and Potentilla glabra occupy this zone. Percentages of arboreal pollen were still very high in this pollen zone, such as Picea (mean proportion of 90.1%). Mean percentage of Compositae accounted for 7.50%. Artemisia pollen was only 7.97%. It reflects that a Picea-Artemisia-Compositae pollen assemblage dominates this zone.
The surface pollen spectra indicate that altitude and precipitation were the primary factors affecting in surface pollen assemblage along this transect. In some zones, there are different pollen spectra in different vegetation zones. Zone II was characterized by the maximum percentage of Picea. Its vegetation cover was also much higher in this zone. The mean vegetation cover of Caragana sinica was >20% in Zone II, and their maximum pollen percentage was 21%. Surface pollen percentages provided a good indication of the vegetation types; however, surface pollen couldn't reflect modern vegetation compositions perfectly in some zones. In Zone I, the maximum percentage of Picea was >90% from 4 740 to 4 320 m, which is beyond the spruce forest line. DCA Ordination Analysis. Figure 3 suggested that eigenvalues of the first two axes were 0.209 and 0.105 (78.4% of the cumulative variance), while those of the second two axes were 0.07 and 0.045, which account for only 21.6% of the cumulative variance. Thus, the first two axes are the most important environmental factors.
Ordination of the 42 surface samples suggested that the two group centroids were entirely separated (Fig. 3). Samples of subalpine shrub meadow zone were divided from those of the other zones. But there are some overlapping between montane coniferous forest and shrub steppe with sparse trees zones. Some of the overlapping samples were from the ecotonal area between these zones. It reflects the vegetation and pollen composition have the transitional character.
With respect to pollen taxa, Ericaceaea showed lower scores on the first axis, whereas Picea showed higher scores, indicating that the first axis represented the gradients of altitude change. Pollen types of typical desert vegetation (Chenopodiaceae) were located on one end with lower scores and those of moisture-favoring taxa (Cyperaceae and Picea) were located on the other end of the second axis with higher scores. The ordination result  showed that the second axis represented the gradients of humidity changes; therefore, we concluded that there were two dominant factors that affect surface pollen assemblage in the region: elevation and humidity. Discriminant Analysis. Based on the regional and local vegetation characteristic in the area, we divided the 42 modern pollen data into three major priori vegetation groups, such as subalpine shrub meadow group, montane coniferous forest group and shrub steppe with sparse trees group. Results suggest that 79.7% and 20.3% of the total variance could be interpreted by two discriminant functions (Table 1). Figure 4 further shows that the three   Table 2 suggests that samples from subalpine shrub meadow and shrub steppe with sparse trees are correctly classified. Only two out of the 22 montane coniferous forest samples are misclassified into subalpine shrub meadow vegetation and shrub steppe with sparse trees.

Discussion
As one of the dominant tree species in boreal and cold-temperate evergreen coniferous forests, Picea, consisting of about 16 species and 9 variables in China, is continuously distributed in the northeastern region and discontinuously distributed in the high mountains of the northern, northwestern, southwestern, and eastern regions, including Taiwan 20 . Spruce pollen has received more attention in studies of both modern pollen-vegetation relationships and paleovegetation and paleoclimate reconstructions 6,21 ; however, Picea pollen in surface soil is an over-representative taxa due to its bisaccate. Impacts from wind, topography, and water on the number of Picea pollen in surface soils has been studied on the basis of data analysis in different vegetation zones of China 21-23 . On the basis of 131 surface pollen records of spruce in Xinjiang, Northwestern China, including the Tianshan Mountains, the Altay Mountains, the Kunlun Mountains, the Tarim Basin, and the Junggar Basin, Yan et al. 24 proposed that the pollen had a mean value of >30% within the spruce forest. However, the value is still >20% above the upper forest treeline, indicating an impact from the uplift of the mountain-valley wind. The strong wind-driven uphill transport of arboreal pollen to higher elevations because of atmospheric processes was first discovered in the European Alps 25,26 and has also been recorded in the mountains of Ecuador 27 , Switzerland 13 , Canada 28 , and Argentina 29,30 . In China, the phenomenon was also observed in Yunnan, Southwestern Sichuan Province 31 and even on the Tibetan Plateau 32 .
Picea was dominant in modern pollen data in the study area. The mean pollen percentage of spruce was >55% beyond the upper spruce forest limit. It was thought to reflect upslope wind-transported pollen transportation, which has been discussed in other studies in Tibetan 32,33 . Because of westerly jet of the plateau and the topography of alpine gorges in the Changdu area, mountain-valley winds are very strong in this area. Mountain-valley wind is characterized as a down-valley wind at night and ascending flows during the day. The Tibetan Plateau has a well-developed, thermally driven, mountain-valley wind system 34  The meteorological data (1970-2009) of wind speeds and wind directions were obtained from the Dingqing weather station (31°25′N, 95°36′E). The monthly average wind velocity at the station varied from 1.5 m/s to 2.1 m/s, but strong winds were frequently observed in the area. Figure 5 shows the maximum total number of days, with strong winds from 1970 to 2009, as recorded from April through June, which accounted for 45% of number of days in the entire year. Second, extreme wind speeds >18 m/s were also observed from April to June. Moreover, the dominant wind direction of extreme winds was northwest. Hence, in this area, northwest daytime upslope wind dominates in spring and summer. In addition, modern pollen data from the Tianchi Weather Station (1 980 m), the Fukang Research Station of the Chinese Academy of Sciences (460 m), and the Beishawo Field Station (400 m) in Xinjiang of China from July 2001 to July 2002 demonstrated that the pollination season for spruce is early June 14 . Therefore, during the day, the prevailing wind direction is northwest and wind speed is stronger than that at night, resulting in more spruce pollen grains at higher elevations being carried from the lowlands. However, night-time wind speeds are somewhat slower; therefore, no actual down-slope pollen dispersal is found. Accordingly, the upward transportation of spruce pollen from lower elevations interferes with pollen assemblage at the higher elevations.
Because Betula is wind pollinated and produces abundant, well-dispersed pollen, it is often over-represented in the pollen record 38,39 . However, many previous studies have also shown that Betula pollen is equi-represented in the pollen rain 3,40 . High percentages of Betula (up to 40%) undoubtedly indicate the local presence of the   41 . The consistent presence of Betula at >5% in modern pollen spectra might also reflect the presence of birch woodland [42][43][44] . Therefore, Betula pollen showed a strong linear relationship with Betula tree abundance, and the highest Betula values of 11.4% from Zone II in the study area corresponded closely to the distribution of the secondary forest of B. platyphylla after the destruction of the coniferous forest; this accurately illustrated human activities in the modern pollen rain. The mean pollen percentage of Betula was >2% m in vegetation Zones I and II, reflecting upslope wind transport.
In Zone I, the presence of Abies pollen even at low percentages (4.54%) is most likely linked to the existence of A. squamata above the spruce forest zone. Lu et al. 45 also stated that fir forests generally grow above spruce forests. Although J. tibetica was the common tree in the south-facing slope of the study area, its pollen abundance was very low. Cupressaceae pollen may be somewhat over-represented, with low pollen productivity and poor pollen dispersal; therefore, Shen 46 proposed that it is excluded from the major pollen types because it is poorly preserved in fossil pollen spectra.

Conclusion
An analysis of modern pollen rain along two altitudinal transects from alpine vegetation in Eastern Tibet, China, indicated that changes in altitude and precipitation were the primary factors that accounted for distinct surface pollen spectra. Picea pollen was observed in the surface soils of all vertical vegetation belts in the region, implying effective upward transport of spruce pollen from the coniferous forest zone by the upslope wind. The complicated local atmospheric conditions had a large impact on surface pollen dispersion and transportation in the regions. Thereby, the pollen spectra in lower elevations representing a local aspect whereas those in higher elevations recorded the vegetation types from the low altitude. Researchers must therefore pay more attention to vegetation reconstruction based on pollen samples with high spruce pollen concentration in the stratum of the alpine vegetation zone in Eastern Tibet, China. Moreover, the modern pollen data also documented the destruction of the coniferous forest in the study area; this indicated that human activity can change the natural features of pollen assemblage.

Materials and Methods
Study Area. Dingqing County (Fig. 1) (31°1′N to 32°21′N, 94°39′E to 96°17′E) in Eastern Tibet is located in the western section of Changdu District and along the southern slope of the Taniantaweng Mountains. It has a semi-humid monsoon climate with a mean annual temperature of approximately 8-12 °C, mean July temperature of approximately 16 °C, annual accumulated temperature (≥10 °C) of 1 500-4 200 °C, annual theoretic evaporation of 1 538 mm, and annual precipitation of approximately 600-800 mm 47,48 . From top to bottom, it displays a distinct elevational distribution of vegetation 47,49 (sparse alpine cushion, alpine meadow, alpine shrub to montane shrub meadow, montane coniferous forest, and shrub steppe with sparse trees).

Vegetation Investigation and Samples Collection.
We collected modern pollen soil samples (Fig. 1) along two transects in Dingqing County, Changdu District, Tibet. For the first transect (Transect I), 19 samples were connected from 4740 m to 3 693 m. The second transect (Transect II) was located at intervals of 20-100 m from 4700 m to 3510 m.
We treated surface pollen samples in the laboratory with heavy liquid and acetolysis after chemical treatments 50 . Using 400x magnification with an Olympus microscope, we examined two to three slides of each sample and counted roughly 300 terrestrial pollen grains. Unknown and indeterminate pollen grains, such as broken, concealed, or corroded pollen, were excluded. Percentages of pollen were calculated based on the total pollen sum (excluding spores) (Fig. 2), using the Tilia/TiliaGraph software 51 . Statistical Analysis. We selected Pinus, Abies, Picea, Cupressaceae, Betula, Salix, Ericaceae, Gramineae, Chenopodiaceae, Compositae, Artemisia, Leguminosae, Polygonum, Ranunculaceae, Rumex, Solanaceae, Rosaceae, Euphorbiaceae, Urticaceae, Caryophyllaceae, Cruciferae, Geranium, Thalictrum, and Cyperaceae for discriminant analysis and detrended correspondence analysis (DCA). Statistical analysis was processed using the CANOCO software package 52 .