Plant size influences abundance of floral visitors and biomass allocation for the cushion plant Thylacospermum caespitosum under an extreme alpine environment

Abstract Variation in size may influence the abundance of visitors and reproductive allocation for cushion plants in the extreme alpine environments. To assess effects of plant size on the abundance of main visitors and reproductive allocation in Thylacospermum caespitosum populations at two altitudes, the abundance of the visitors, visiting frequency, total number of flowers, number of fruits, number of unseeded flowers, and reproductive allocation were investigated during the period of reproductive growth. Concurrently, the effects of plant size on the visitors' contributions to fruit setting rate were assessed by a bagging experiment. Our results showed that flies (Musca domestica and Dasyphora asiatica) were the main pollinating insects of T. caespitosum, and they could obvious facilitate (p < 0.05) the fruit setting rate of this cushion plant. Seed set and floral visitation were significantly influenced (p < 0.001) by plant size. Moreover, the reproductive allocation and fruit setting rate of T. caespitosum was influenced (p < 0.001) by plant size. More biomass was allocated to reproduction in plants of greater diameter. There is an increase in reproductive success (increases of fruit number with increase in plant size) in relation to plant size. In conclusion, the extent of M. domestica and D. asiatica to facilitate the fruit setting rate mainly depended on the size of T. caespitosum. Size‐dependent reproductive allocation occurred in T. caespitosum and was the chief factor affecting the contribution of flies to fruit setting rate. These traits reflect reproductive fitness of T. caespitosum related to plant size in extreme alpine environments.

Traditionally, bees, hoverflies, and butterflies are frequently studied as pollinators in agricultural and conservation research; wild and managed bees are well documented as effective pollinators in agricultural and other ecosystems (Orford et al., 2015). However, many flies, bumblebees, and Lepidoptera species are known contributors to pollination in alpine ecosystems (Arnold, Savolainen, & Chittka, 2009;Galen, 1996). In the subnival belt, the high altitude, low temperatures, overcast conditions, short growing season, unstable substrate, intense radiation, and relatively unpredictable weather and high winds are all challenging for insect visitors (Cavieres, Quiroz, Molina-Montenegro, Muñoz, & Pauchard, 2005;Körner, 2003). Thus, the levels of diversity and activity of insect visitors are reduced due to the harsh climatic conditions in alpine ecosystems, and pollination rates are accordingly inherently low (Torres-Díaz et al., 2011). These climatic characteristics also progressively reduce pollen availability in alpine plants (Bingham & Orthner, 1998). Alpine plants have developed solutions to deal with low visitor rates and reduction in pollen availability, such as self-compatible (Liu et al., 2017;Sosenski, Ramos, Domínguez, Boege, & Fornoni, 2017) plants are an evolutionary solution for alpine and arctic plants to deal with low visitor numbers; thus, they will face lower risk of extinction due to low supply of pollen from compatible plants (García-Camacho & Totland, 2009;Muñoz & Arroyo, 2006). Alpine plants in the subnival belt demonstrate a high frequency of asexual reproduction (Morgan, Wilson, & Knight, 2005;Reid, Hooper, Molenda, & Lortie, 2014). This reproductive strategy benefits alpine plants in harsh climatic conditions with low diversity and activities of pollinators (Milla, Giménez-Benavides, Escudero, & Reich, 2009). Theoretical and empirical studies have predicted that severe environmental conditions (climatic conditions and insufficient visitors) lead to greater levels of asexual reproduction compared to lower alpine elevations, where high metabolic costs are invested in sexual reproduction relative to asexual reproduction (Chen, Li, Yang, & Sun, 2017;Schuster & Longton, 1983, Stark, Mishler, & McLetchie, 2000. Alpine plants can also compensate for the scarcity of visitors in alpine habitats by increasing the flowering phase and flower longevity (Blionis & Vokou, 2002;Fabbro & Körner, 2004). Although alpine plants benefit from self-compatible, asexual reproduction, longer flowering phase, and flower longevity to compensate for the low diversity and activity of pollinators, many alpine species are strongly dependent on scarce insect visitors to increase their seed set. Alpine flowers (Saxifraga oppositifolia, Dryas integrifolia, and Salix arctica) that are frequently visited by insects (pollinated primarily by flies and bumblebees) are totally or partially dependent on them for seed set (Kevan, 1972;Peeters & Totland, 1999).
Here, we investigated whether and how strongly floral visitors influenced seed set in an alpine plant species, and what plant traits determined visitor abundance. High allocation to the production of reproductive structures also dictates the extent to which floral visitors pollinate flowers (Campbell & Halama, 1993). Producing (more nectar) may increase visitation rates as well as pollination success, and these in turn result in greater fruit and seed production (Mattila & Kuitunen, 2000;Sletvold, Tye, & Ågren, 2017). Moreover, environmental conditions can affect reproductive allocation patterns in plants. Variation in environmental conditions during reproduction may result in differences in ovule number, germination rates, growth rates of pollen tubes, and seed production (Jennersten, 1991;Young & Stanton, 1990). Alpine plants, hereafter cushion plants, tend toward a highly compact growth form that slowly (ca. 0-7 cm per year) grows along the ground, forming dense mats of vegetation. These slow-growing species often have very long-lived leaves (many cushions can grow for centuries or persist for additional centuries ;Chen et al., 2017;Forbis & Doak, 2004;le Roux & McGeoch, 2004;Molau, 1997). Previous studies have shown that some cushion species may need 20 years of vegetative growth in order to begin reproduction for the remainder of their life history (Molau, 1997;Morris & Doak, 1998). These characteristics indirectly reflect that plant size may influence reproductive allocation in cushion plants. Thus, studies examining reproductive allocation need to consider the potential effect of size-related variation in cushion species. After all, study of size dependence of fitness components is usually the only way to apply life history theory predictions to most herbaceous perennials due to the difficulty in evaluating the age of these plants except where long-term demographic studies are conducted (Shabir, Nawchoo, & Wani, 2017).
Size-dependent variation in reproductive allocation is a common phenomenon in many plant species (Samson & Werk, 1986). Plant size is widely used in the prediction of current and future reproductive output, and the relative advantages or disadvantages of reproducing at different sizes (Stearns, 1992;Sletvold, 2002). Ollerton and Lack (1998) indicated that plant size not only directly influences individual plant fecundity but also can indirectly affect reproductive output. The largest plants are commonly the most fecund, and size is closely correlated with total flower production in populations (Weiner, 1988;Herrera, 1993). In fact, the effect of plant size on reproductive output has been broadly discussed in extreme environments (Aarssen, 2015), such as desert and saline environments (Waller, 1988; Aguilar, Ashworth, Galetto, & Aizen, 2006; del Carmen Mandujano, Carrillo-Angeles, Martínez-Peralta, & Golubov, 2010;Benassi et al., 2011;Chacoff et al., 2012;Paasch, Mishler, Nosratinia, Stark, & Fisher, 2015;Salguero-Gómez et al., 2016). Moreover, although alpine plants benefit from asexual reproduction to compensate for the low diversity and activity of pollinators, many cushion species are entirely dependent on sexual reproduction (Chen et al., 2017). Thylacospermum caespitosum (Caryophyllaceae) represents this type of cushion species. But plant size influences the abundance of floral visitors and biomass allocation for the cushion plant under an extreme alpine environment remains unknown. Here, we investigated whether the fruit setting rate promoted by visitors was related to the size of T. caespitosum. We also evaluated the effects of plant size on the abundance of the main visitors, visiting frequency, total number of flowers, number of fruits, number of unseeded flowers, reproductive biomass ratio, stem-leaf biomass ratio, root biomass ratio, and fruit setting rate. We aimed to answer the following ques-

| Study site and study species
The study site was situated in the headwater region of the Urumqi River in the eastern Tianshan Mountains, China. The site is part of the Xinjiang Uygur Autonomous Region. The elevation varies from about 4,100 m to 4,300 m. At the study site, mean annual temperature ranges from ca. 5 during the day and −4°C at night; however, temperature as low as −10 may occur during the vegetative growth season (Liu et al., 2017). As Figure 1 shown, T. caespitosum is a perennial plant that has a woody taproot and forms very dense and solid cushions (Dvorský et al., 2013). Thylacospermum caespitosum is one of the most representative high-alpine cushion plants in Asian high mountain regions. It is distributed along rocky slopes and crevices from 3,600-6,000 m asl. Thylacospermum caespitosum is found in China (Gansu, Qinghai, Sichuan, Xinjiang, and Xizang), India, Kazakhstan, Kyrgyzstan, Nepal, and Sikkim (Flora of China Editorial Committee, 1999). Thylacospermum caespitosum plants have been studied to understand their ameliorating effects on the harsh environments they inhabit (extreme altitude and dry conditions) and their role as nurse plants for other plant species in alpine ecosystems (Bello et al., 2011;Chen et al., 2015;Dvorský et al., 2013;Michalet et al., 2016).
In the study region, T. caespitosum was distributed in the range

| Effects of plant size on reproductive allocation
The selected plants used to determine visitation rate were used to count flowers (during flowering) and fruits (after flowering) of T. caespitosum by using circular PVC (polyvinyl chloride) piles with 10 cm radius (157 cm 2 ). The centers of PVC piles coincided with the cushion plant. With these data, we could measure the total number of flowers, number of fruits, and number of unseeded flowers. At the same time, all of the above plants were collected at the end of the growing season to estimate biomass allocation (g). All samples were dried at 70°C for 48 hr and weighed (Ploschuk, Slafer, & Ravetta, 2005). Plant biomass was divided into reproductive biomass (all reproductive structures), stem-leaf weight biomass, and root biomass.

| Statistical analysis
All experimental results were presented as the mean ± standard error. To test the effect of plant size on abundance of main visitors and reproductive allocation, all parameters (number of flies, visiting frequency, total number of flowers, number of fruits, number of unseeded flowers, flower weight ratio, stem-leaf weight ratio, root weight ratio, fruit setting rate, and RII visitor ) were, respectively, evalu-

Results of field investigations showed that flies (Musca domestica and
Dasyphora asiatica) were the main flower-visiting insects for T. caespitosum (Figures 2 and 3a). Relative abundance of flies was >93% of visitors for all plant sizes of the two populations (Figure 3a). Although  (Figures 2d and 4).
As shown in Table 1 The reproductive biomass ratio of T. caespitosum was significantly affected (p < 0.001) by plant size, but stem-leaf weight ratio and root weight ratio not be affected (p > 0.05, Table 1  was significantly higher (p < 0.05) than V2 except for R1 in the low and high elevation populations (Figure 6a,b). Moreover, plant size has a great effect on RII of T. caespitosum (p < 0.001, Table 1). RII increased with the increase in plant size, with R3 and R4 significantly higher (p < 0.05) than R1 and R2 (Figure 6c,d). These results indicated that flies may facilitate fruit setting in T. caespitosum, which was especially obvious in larger plants.

| D ISCUSS I ON
Insect visitor diversity, abundance, and activity decrease dramatically in extreme alpine environments relative to tropical (or lowland) regions due to harsh climatic conditions (Körner, 2003;Medan et al., 2002). Scarcity of insect visitors adversely affects plant reproduction (Kuijper & Pen, 2014). However, alpine flowers are frequently visited by insects and are totally or partially dependent on them for seed set (Kevan, 1972;Peeters & Totland, 1999). Our results indicated that flies were the major visiting insects for T. caespitosum, and they may promote fruit setting of this cushion plant (especially in larger plants; Figures 2, 3 and 6). But bees, hoverflies, and butterflies (taxa frequently studied as visitors in agricultural and conservation contexts) were not observed in our study. In fact, more researchers are realizing that the contributions of visitors other than bees (such as nonsyrphid Diptera) have been underexplored despite their potential to contribute to plant reproduction and stability in the face of environmental change (Rader et al., 2016;Tiusanen, Hebert, Schmidt, & Roslin, 2016). Lefebvre, Fontaine, Villemant, and Daugeron (2014) indicated that flies (flies represent more than 60% of all visitors, with 54% of them being Empidinae) widely replaced bees as the main flower visitors in a subalpine meadow in the French Alps, and among them the Empidinae might play a key role in pollination. Kevan (1972) and Peeters and Totland (1999) showed that alpine flowers most frequently visited by flies and bumblebees are totally or partially dependent on them for seed set.
We then ask the question, why were floral visitors predominantly flies? On one hand, nonsyrphid dipteran visitors are dominant at high altitudes and latitudes, including alpine and subarctic ecosystems where bees are less abundant, mainly due to the loss of habitat from anthropogenic land use change and intensification (Elberling & Olesen, 1999;Shrestha et al., 2016;Vanbergen et al., 2014;Williams & Osborne, 2009). The proportion of Muscidae species usually increases with altitude, but the proportions of Syrphidae and bumblebees decreases (Pont, 1993). On the other hand, the scarcity of flower visitors with long proboscises in the alpine ecosystem may have opened a niche for flies and put a selection premium (extra charges in selection process) on a longer proboscis and more shallow flowers (Elberling & Olesen, 1999 (Larson, Kevan, & Inouye, 2001;Orford et al., 2015).
Plant size is one of the major biotic factors that determined the amount of energy available for reproduction and seed development.
Generally, large plants usually have large reproductive outputs (for example seeds or flowers) and have been confirmed in alpine plants (Rees & Venable, 2007;Sun et al., 2014;Venable & Rees, 2009).
Large reproductive investment means more materials will be used to attract insects for pollination. Our results showed that the total number of flowers, the number of fruits, number of unseeded flowers, and reproductive allocation of T. caespitosum were clearly influenced by plant size (Figures 3 and 4; Table 1). He et al. (2008) found that the biomass of reproductive organs and allocation of resources to reproduction in three Oxytropis species from the Qinghai-Tibet Plateau increased with increases in plant size. Similar positive relationships between reproductive allocation and plant size have also been found in Polygonum macrophyllum (Meng, Wang, Liu, & Zhu, 2011). Our results similarly showed that the total number of flowers, the number of fruits, number of unseeded flowers, and reproductive allocation were significantly increased (p < 0.05) with increases in plant size in two populations of T. caespitosum (Figures 4 and 5).
This can be simply explained by the fact that larger T. caespitosum produced more reproductive biomass than smaller ones. This may be attributed to plant modular architecture, as relatively large individuals within a population have more vegetative and reproductive modules than smaller individuals (Weiner, 1988;Niklas, 1993).
Underlying size-dependent reproductive allocation is an adaptation of T. caespitosum to extreme environments. Reproductive allometry related to plant size is thought to be due to environmental variability and is interpreted as an adaptive strategy of plant growth and allocation (Bonser & Aarssen, 2009). Allocating more biomass to functions that maximize vegetative growth is more common under harsh environments (Shabir et al., 2017). For small individuals, large resource investment in reproduction may have a negative influence on future reproduction, growth, or survival (Obeso, 2002;Shibata & Kudo, 2016). Inversely, large resource investment in vegetative growth may lead to positive effects on future reproduction, especially in the harsh alpine environment where conditions (high altitude, low temperatures, overcast conditions, short growing season, unstable substrate, intense radiation, and relatively unpredictable weather and high winds) are not advantageous to plant survival. This type of allocation therefore implies trade-offs for T. caespitosum in harsh alpine environments because resources allocated to one function or organ are not available to other functions or organs (Weiner, 2004).
In brief, flies are the main visitors of T. caespitosum, and floral visitors with the capacity to pollinate T. caespitosum over a broader temporal range (primarily nonbee taxa) have a stronger influence on T. caespitosum reproductive fitness than those with a more restricted temporal range. This is significant because it demonstrates how abi-

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
Ruiming Zhao designed and performed the study. Hua Zhang performed statistical analysis. Lizhe An conceived and financed this study. All the authors participated in writing the manuscript.

DATA ACCE SS I B I LIT Y
We agree to make our data publicly available in a relevant repository.