Elevated precipitation alters the community structure of spring ephemerals by changing dominant species density in Central Asia

Abstract Global climate change is one of the most pressing conservation challenges; in particular, changes in precipitation regimes have already substantially influenced terrestrial ecosystems. However, the mechanisms influencing precipitation changes on individual plants and the plant communities in desert grasslands have yet to be fully elucidated. We therefore examine the influence of increased precipitation on plant community compositions in the Gurbantunggut Desert, Xinjiang, northwestern China, from 2005 to 2009. We found that growth of all plant species and the community productivities increased markedly with enhanced water input. Cover of ephemeral synusia also significantly increased due to increased precipitation, implying that the role of the ephemeral community for stabilization of sand dunes was strengthened by increased precipitation. The response of plant community compositions to increased precipitation was primarily reflected as changes in plant density, while increased precipitation did not affect plant species richness and the diversity index. Dominant species drove the response of plant density to increasing precipitation during the five‐year study period. However, the relative responses of rare species were stronger than those of the dominant species, thereby potentially driving species turnover with long‐term increased precipitation. This finding improved our understanding of how increased precipitation drives the changes in plant community composition in desert grasslands and will help to better predict changes in the community composition of ephemerals under future global climate change scenarios.


| INTRODUC TI ON
The earth has experienced substantial precipitation changes in the past five decades, and climate change models predict further changes in precipitation regimes, with increased annual precipitation in many regions (IPCC, 2007). Many studies have demonstrated that global climate change has caused a series of ecological consequences, such as those for plant production (De Boeck, Hiltbrunner, Verlinden, Bassin, & Zeiter, 2018;Zhao et al., 2019) or vegetation changes (Jiang, Bao, Guo, & Ndayisaba, 2017;Li, Chen, Li, Deng, & Fang, 2015). However, changes in the precipitation regime are not uniform and are complicated, revealing spatial and temporal variations that are difficult to study (Li, Zhang, et al., 2015;Walther et al., 2002). Precipitation in the mid and high latitudes of the Northern Hemisphere has generally increased by 0.5 ± 1% per decade during the last 1,000 years in some places (IPCC, 2007). Zhang et al. (2007) showing that anthropogenic forcing has had a detectable influence on observed changes in average precipitation during the twentieth century, including increases in precipitation in the mid-latitudes of the Northern Hemisphere. In the past century, Central Asia has experienced a complex pattern of precipitation changes (Lioubimtseva & Henebry, 2009). Overall, the Central Asia dry lands may become moister, and extreme precipitation events may increase as a result of global warming due to a probable intensification of westerly cyclones (Deng & Chen, 2017;Lioubimtseva, Cole, Adams, & Kapustin, 2005;Luo et al., 2018). It is therefore important and necessary to better understand how these precipitation changes affect plant community compositions and ecosystem functions in this changing environment.
If it is possible to identify functional characteristics at an individual level that determines the responses of ephemeral plant communities to precipitation change, the potential for predicting outcomes for a range of different situations will be improved (Morecroft et al., 2004). Observational studies have suggested some likely patterns of response. When drought occurs in perennial grasslands, the dominant grass species tend to die first, reducing community productivity and leaving gaps in the sward for other short-lived plants, followed by increased frequency and abundance of shortlived plants in the following years (Grime, Willis, Hunt, & Dunnett, 1994). It has been proven that deep-rooted species tend to increase after drought events (Dunnett, Willis, Hunt, & Grime, 1998). With an increased frequency of drought events, we may see a shift in grassland composition toward more deep-rooted and short-lived plants. Seed productivity is also an important factor for determining plant community structure through seed sizes and numbers (Leishman, 2001). Changes in precipitation can significantly affect seed productivity and phenology, especially in harsh environments (Chen et al., 2019). Researchers have found that although decreases in precipitation suppress the aboveground biomass of plant communities, net primary production (NPP) is significantly more sensitive to increased precipitation than to decreased precipitation, especially in arid and semiarid environments (Knapp & Smith, 2001;Zhao et al., 2019). Most experimental manipulations have been performed in perennial grasslands (Byrne, Adler, & Lauenroth, 2017;Dunnett et al., 1998), and the Central Asia arid area is one of the largest arid regions at middle latitudes, and includes approximately one-third of the arid lands and almost 90% of the temperate deserts worldwide (Lioubimtseva & Henebry, 2009), but research on the effects of precipitation changes on plant communities in this area are rare (Zhang, Li, Zhang, Zhang, & Chen, 2016).
Water availability is the most important factor determining plant growth and primary productivity in all ecosystems, especially in arid and semiarid grasslands (Dube & Pickup, 2001;Nielsen, Osler, Campbell, Burslem, & Van Der Wal, 2010). Recent studies in Central Asia have shown that increased precipitation is beneficial for the growth of Haloxylon ammodendron (Zhao et al., 2019); significantly improves the growth and seed production of annual desert plants (Chen et al., 2019); and the dynamics of vegetation cover are closely related to precipitation changes (Zhang, Lu, et al., 2016). However, plant community structures may respond more strongly to increased precipitation than to drought, because plants are more drought-tolerant in arid ecosystems (Knapp & Smith, 2001). In addition, the responses of plant communities to changes in precipitation regimes rely greatly on community structures and plant species compositions (Gerten et al., 2008;Suding et al., 2008;Torode et al., 2016). On one hand, precipitation changes, for example, a long-term precipitation increase in grasslands, cause difficulty in obtaining consistent effects for community diversity, because of the high variations across survey years (Collins et al., 2012;Yang et al., 2011). On the other hand, the dominant species in the plant community determine the responses of the plant community to changes in precipitation (Smith, Knapp, & Collins, 2009). If the dominant plant species respond strongly to climate change, then significant changes in the plant community can be expected (Byrne et al., 2017). However, the dominant species in arid lands may be less responsive to increased precipitation because they may already have adapted to the arid environment (Harpole & Tilman, 2007). Therefore, although it is clear that precipitation changes have significant effects on plant communities (NPP), we still have an incomplete understanding how plant communities respond to increased precipitation in arid and semiarid grasslands.
Spring ephemerals are widely distributed in the arid lands of Central Asia, and precipitation is more important than soil chemistry in determining the distribution of ephemerals (Zhang, Liu, Zhang, & Sun, 2016). These spring ephemerals offer an intriguing survey opportunity to observe their responses to increased precipitation. In this area, each year supports a new generation of ephemerals experiencing climates that differ from year to year because ephemerals are very sensitive to precipitation changes (Zhang & Chen, 2002). Thus, only by multiyear sampling can we capture several unique combinations of plant community properties and environmental conditions. Although the lifecycles of spring ephemerals are only 60-70 days and finish before the summer drought arrives (Zhang, 2002), they play important ecological roles in Central Asia. For instance, in early spring, although their biomass may be less than that of shrubs in most arid ecosystems, ephemerals cover the spaces between shrubs and can greatly decrease the frequency and intensity of sandstorms. Moreover, ephemerals are the major components of transitory desert grasslands, and provide forage for livestock in early spring when there are few other resources available (Zhang, 2002). A change in any of these functions is significant for ecosystem stability in Central Asia, but predicting the response of this

| Study site
The Gurbantunggut Desert, a fixed and semifixed desert, is located in the hinterlands of the Dzungaria Basin in Xinjiang, northwestern China. The mean annual temperature, mean growth season precipitation, and mean annual precipitation are 7.1°C, 67.9 mm, and 215.6 mm, respectively (meteorological datasets are from the Fukang station of the Desert Ecology Chinese Academy of Sciences).
During winter, the desert is usually covered by more than 20 cm of snow. Annual evaporation in this region is more than 2,000 mm (Sun & Yang, 2010). Soils are gray desert soils (Chinese classification) with aeolian sands on the surface (0-100 cm; Chen, Wang, Li, & Ruan, 2007). The vegetation consists of a mixture of shrubs and grasses.
Haloxoylon persicum, which grows on the tops of dunes, is the constitutive species in the desert community. Ephedra distachya, a small shrub, is distributed in interdunal valleys. Annual herbs are distributed in the understory of the shrubs (Zhang, 2002).
In spring, ephemeral plants are dominant in the community (Zhang & Chen, 2002). These plants are able to take advantage of the favorable soil moisture and thermal conditions in the spring (around March) to grow and finish their lifecycles within approximately 60-70 days before the summer drought arrives (Zhang, 2002). The ephemerals play a very important role in the stabilization of sand dunes and in reducing the frequency and intensity of sandstorms (Wang, Wang, Jiang, & Zhao, 2005).

| Experimental design
Our experimental site is located at N44°32.407′, E88°16.779′ and is in a valley between dunes with an elevation of 510 m. The site is isolated by a steel fence to prevent grazing by animals. A randomized block field experiment was performed continuously from 2005 to 2009. The properties of the sandy soil in the experimental site are shown in Appendix 1.
The experiments involved several irrigation levels. According to the results of recent publications, precipitation will continuously increase by approximately 50 mm in the next 50-100 years (Takayabu et al., 2007;Zhao, Ding, Xu, & Zhang, 2003), and precipitation fluctuations may become as great as +40% in extreme precipitation years (Liu et al., 2010); therefore, we set 40 mm of water as the minimum water input. In , 2007, the treatments consisted of natural rainfall (as control) and 40 mm of water addition based on natural rainfall. In 2006, an additional treatment of 80 mm water was added to simulate future extreme wet years. Usually, water additions were accomplished with a sprinkler kettle, and irrigation was over the entire plot area. The research area was divided into five zones, and, in each year, one zone was harvested to avoid the effects of destructive harvesting on the results. Each zone consisted of sixteen 1 × 1.5 m 2 plots (in 2006, there were twenty-four plots), and each treatment had eight replicates. The plots were arranged in a randomized block, each plot was separated from its adjacent plots by a 1 m buffer strip. All plots were divided into a central zone of 1 × 1 m 2 to investigate plant community changes and a surrounding buffer zone to reduce edge effects from the surrounding untreated vegetation (Appendix 3).

Simulations of increased precipitation started in late March (ap-
proximately the 25th of March) of each year before plant emergence and harvesting took place in late May (approximately the 25th of May). Water was input every two weeks after the start of the experiment, for a total of four times during the plant growth season for each of 5 years. The volume of water input each time in each plot was 15 L for the 40 mm water treatment; this is equivalent to a one-time increase in precipitation of 10 mm, and 30 L was used for the 80 mm treatment; this is equivalent to a one-time increase in precipitation of 20 mm. Each plot was divided into a central zone of 1 m 2 and a surrounding buffer zone to reduce marginal utility.

| Sampling and analysis
The cover for each plant species and for the entire community, and the plant densities of every species and for entire plots were measured before harvest. Diversity indices, for example, species richness, the Shannon-Weiner index, and the relative abundance of each plant species, were calculated. When most plants were senescing at the end of May (approximately two months after starting), all plants in the plots were dug up. Each plant was taxonomically identified, seed numbers were recorded (this was only done in 2006 and 2007), and each plant was divided into shoots and roots and later dried at 80°C in an oven. The dry weights of each species were recorded. Species richness was represented by the total number of species per plot; the relative abundance of each species was calculated as the percentage of plots containing that species. The biodiversity index and frequency of appearance were calculated by the Shannon-Weiner index as follows: where P i = n i /N and n i = number of individuals for species i, and N = total number of individuals in all species.
Frequency of appearance (%) = Number of plots in which a species emerged during three experimental years × 100/total number of plots in 3 years.
To determine the responses of different plants to experimental water additions, we calculated the differences and response ratios of the plant density responses to water addition for each block as follows: where T m is the plant density for the 40 mm or 80 mm water addition treatment plots in block m, and C m is the plant density for the control plot in block m.
A total of 37 plant species were observed across all plots during the 5 years of the experiment (Appendix 2). Among these plant species, one species, for example, Ephedra distachya, is a small shrub, six are perennials, and six are unidentified, while 24 are ephemeral plants. A species was classified as "dominant" if its frequency was >70% and its relative abundance was >5%. A species was classified as "rare" if its frequency was <50% and its relative abundance was <3% (Ma et al., 2017;Mariotte, Vandenberghe, Kardol, Hagedorn, & Buttler, 2013). This method yielded six ephemeral dominant species, namely Erodium oxyrrhynchum, Ceratocarpus arenarius, Carex physodes, Alyssum linifolium, Hyalea pulchella, and Trigonella arcuata, and 26 rare species (Appendix 2).

| Statistical analysis
An a priori structural equation model (SEM) was applied to test the effects of increased precipitation on alterations of the community structures of spring ephemerals, according to hypothesized causal relationships, using R version 3.2.2 (R Foundation for Statistical Computing, Vienna, Austria, 2013), with the "lavaan" package. We selected unique attributes of the plant community: aboveground biomass (ANPP), species richness, plant density, and community structure in our model. To use community structure as a variable in the model, we used a nonmetric multidimensional scaling (NMDS) analysis of plant density, based on the Bray-Curtis distance measure (McCune, Grace, & Urban, 2002), and selected Axis 1 as the community structure in the model. In our NMDS analysis, Axis 1 and Axis 2 explained 41% and 28% of the total variance in the plant community, respectively (stress = 0.2218). We used the χ 2 test, root mean square error of approximation (RMSEA), and Akaike information criteria (AIC) to evaluate the model fits. Nonmetric multidimensional scaling (NMDS) analyses of plant densities were also performed to examine the effects of increased precipitation on the plant community structures over 5 years.
All data were subjected to a one-way or two-way ANOVA using the SPSS17.0 software package for Windows (version 17.0; SPSS Inc.). Treatments were compared using the least-significant different (LSD) test at p < .05. All figures were created with SigmaPlot 10.0. Water additions promoted the growth and significantly increased the ANPP of the plant community, but with strong interannual variability across the five study years (F 4, 40 = 20.048, p < .0001, Table 1).

| Effects of elevated precipitation on ANPP, cover, and seed production
The greatest increase was in 2006, and the 80mm water treatment significantly increased the biomass of ephemerals by 326%; the smallest increase was in 2008, and the biomass associated with the 40mm water treatment was only significantly higher (22%) than that of the control ( Table 2).
The largest increase was in 2009, and the 40mm water treatment significantly increased the cover by 59% (Figure 1). Elevated precipitation significantly promoted seed production of the ephemeral plants(F 1, 16 = 0.532, p < .001, Table 1). In 2006, the 40 and 80 mm water treatments significantly increased seed production by 298% and 378%, respectively, compared to the control. No differences in seed production were observed between the water addition treatments of 40 and 80 mm. In 2007, seed production associated with the 40mm water treatment was 97% higher than that of the control (Table 2). Elevated precipitation had no significant effects on species richness and the Shannon-Wiener index across the 5 years (Tables 1 and 2).

| Effects of elevated precipitation on the proportion of ANPP and seed production of dominant species
Water addition significantly increased the ANPP of the plant com- TA B L E 1 Results of two-way ANOVA on the effects of year (Y), elevated precipitation (W), and their interactions on the ANPP, coverage, species richness, Shannon-Wiener index, plant density, N uptake, P uptake, K uptake, and seed number In general, compared to the 40 mm water addition, the 80 mm addition had no effect on seed production for the five dominant species. However, both water additions increased the total proportions of seed production for the five dominant species in the 2 years (Figure 3).  Figure 4) clearly reflects the divergence in plant community compositions caused by increased precipitation.

| Effects of elevated precipitation on the community composition of ephemerals
In addition, we combined the plant species to form dominant species group and a rare species group, and these groups

| D ISCUSS I ON
Research regarding the change mechanisms of plant community compositions and ecosystem functions in natural ecosystems is always a major research challenge in ecology (Kane et al., 2017;Xu et al., 2018), especially in response to climate changes in recent decades (Eskelinen & Harrison, 2015). Desert ecosystems are more dependent on water availability than on any other factor (e.g., nutrients) in arid and semiarid regions (Weltzin et al., 2003). Precipitation in the arid regions of Central Asia, such as the Dzungaria Basin, has increased by 20% following global climate changes in recent decades (Li, Zhang, et al., 2015) and seems likely to increase by approximately 50 mm in the future (Takayabu et al., 2007;Zhao et al., 2003), and the precipitation regime has changed (Lioubimtseva & Henebry, 2009).
The effects of changes in precipitation on ecosystem processes, such as vegetation dynamics (Petrie et al., 2018) and plant productivity (Li, Zhang, et al., 2015) in desert ecosystems have been a concern in recent years (Jiang et al., 2017;Muldavin, Moore, Collins, Wetherill, & Lightfoot, 2008;Thomey et al., 2011). Understanding the ecological responses of plants to changes in precipitation at the population and community levels is essential for predicting how vegetation will function in the future. The present study showed that enhanced water inputs strongly changed some important ecological processes in the Gurbantunggut Desert, for example, community regime and seed productivity.

| Effects of elevated precipitation on plant community structures
The stability and resilience of ecosystems are determined by both the richness and diversity of species (Oliver et al., 2015) and by the community structure (Dieleman, Branfireun, McLaughlin, & Lindo, 2015).
Our present study showed that the density, cover, and aboveground biomass of some plant species had a strong positive response to enhanced water addition, while some other species exhibited less-pronounced changes. These impacts caused significant changes in the community structure, although the species richness and biodiversity index were unchanged over the five-year duration of our study.
In the present study, strong interannual variations in the measured variables were observed across entire experimental years (Table 1; Figure 1). Strong year-to-year fluctuations in plant community parameters, including plant cover and ANPP with annual precipitation have been reported for different grassland ecosystems (Heisler-White, Knapp, & Kelly, 2008;Yang et al., 2011). Water availability limits plant growth and primary productivity in all ecosystems, especially in arid and semiarid grasslands (Nielsen et al., 2010). It has been reported that ephemerals are sensitive to precipitation changes, and March-to-May precipitation is mainly responsible for biomass fluctuations in the same region (Zhang & Chen, 2002;Zhang, Liu, et al., 2016). Thus, the strong interannual variability in the measured variables could have been largely attributed to the year-to-year fluctuations in precipitation. In addition, we found that the effects of increasing precipitation on ANPP and cover varied with year and were not in agreement with the fluctuations in growing season precipitation for some years (Figure 1 and Table 2). For example, the growing In our desert plant community, water additions significantly increased plant densities in the community (Table 2), and the dominant species responded most strongly in terms of absolute plant density to water additions over 5 years ( Figure 5). This finding is due to the higher initial densities of the dominant species in every plot. Because the dominant species had higher initial densities, they were better able to respond to precipitation changes in an absolute sense. Studies of other low-productivity grasslands have also demonstrated that the dominant species responded more to fertilization or disturbance than did the subordinate species (Houseman, Mittelbach, Reynolds, & Gross, 2008), and the aboveground biomass of the dominant species in the grasslands of F I G U R E 5 The difference of plant density of two vegetation types (dominant species and rare species) to water addition from 2005 to 2009 (a, 2005; b, 2006; c, 2007; d, 2008; e, 2009 However, in the present study, the rare species exhibited almost the same response ratios in 2007 and even had higher relative response ratios to the experimental water additions than the dominant species in the other 4 years ( Figure 6). This result indicates that rare species may perform better than dominant species relative to their initial densities in the community. These differences between rare and dominant species may be due to individual traits or to the limitation of the dominant species by other resources (e.g., space or nutrient availability) and thus reduced their potential to respond disproportionately to their initial densities (Harpole & Tilman, 2007;La Pierre et al., 2016). Thus, the greater response ratios of the rare species compared to those of the dominant species in our research provided more evidence that a priority effect limits the responses of rare species to resource limitations (La Pierre et al., 2016). The absolute responses of the rare species in the 5 years of study did not drive the plant density of community response to water additions due to of their low densities in the community. However, we should not ignore the important roles of rare species in the plant community. With longterm water additions, rare species may become the dominant species when they overcome the priority effect (Avolio et al., 2014).

F I G U R E 6
The response ratio of plant density of two vegetation types (dominant species and rare species) to water addition from 2005 to 2009 (a, 2005; b, 2006; c, 2007; d, 2008; e, 2009

| Effects of elevated precipitation on the soil seed bank
The soil seed bank represents important propagule pools for plant population dynamics (Maighal, Salem, Kohler, & Rillig, 2016) and plays a crucial role in the ecology of different plants (Bekker et al., 1998). Therefore, the soil seed bank is an important component for understanding plant community compositions and ecosystem functions. The composition of the seed bank can also be altered by changes in the contributions of seed yield by individual plant species due to the different responses of seed productivity to water additions (Walck, Hidayati, Dixon, Thompson, & Poschlod, 2011). Based on our results, the annual total seed amounts of the desert community can be significantly changed by enhancing the water supply (Table 1). The size of the soil seed bank influences plant population dynamics, with a larger soil seed bank causing a greater probability for a larger plant community and later changes in the species compositions and community diversities (Rosbakh et al., 2017;Walck et al., 2011). Such results suggest that changes in precipitation might potentially alter the stability and productivity of the desert ecosystem and then influence ecosystem functions.

| Effects of elevated precipitation on ecosystem functions
Desert ephemerals form a significant community (Zhang, Lu, et al., 2016;Zhang & Chen, 2002) and are widely distributed in both arid woodlands and open desert areas (Robinson, 2004). Ephemerals play a very important role in the stabilization of sand dunes and in reducing the frequency and intensity of sandstorms (Wang et al., 2005).
Moreover, ephemerals are major components of desert grasslands, which provide high-quality forage for livestock in early spring and, during this time, there are few other grass species available (Zhang & Chen, 2002). Zhang (2002) reported that the biomass of ephemerals in the Gurbantonggut Desert accounts for 58 percent of the total biomass in the community of Haloxoylon persicum in precipitationrich years. From the present study, changes in the community structure of ephemerals induced by increased water may induce changes in several ecosystem functions: First, an increase in vegetation cover reduces wind erosion and mitigates the desertification processes. Wind erosion is a leading factor for land desertification and sandstorm disasters and is one of the most serious environmental problems in the world (Guo, Huang, Dong, Van Pelt, & Zobeck, 2014). Many studies have indicated that planting vegetation is an effective measure for control of wind erosion and that vegetation cover was the most important factor (Jiang, Miao, Toshio, & Zhou, 2013). Wind tunnel tests found that wind erosion was greatly reduced when the plant cover was more than 30% (Scott Van Pelt et al., 2013). In our experiment, the cover under 40 and 80 mm treatments was 23% and 32% higher, respectively, than for the control. Such results imply that the role of ephemeral grasslands in the stabilization of sand dunes is strengthened by enhanced water input and would respond more highly in extreme precipitation years. Second, in our research, water additions significantly increased the aboveground biomass of the plant community across the five study years (Tables 1 and 2). Increasing the biomass of individual species and the total net ecosystem primary productivity would supply fodder of higher nutritional quality for livestock production. In addition, aboveground net primary production (ANPP) is a key integrator of carbon uptake and energy flow in many terrestrial ecosystems, especially in desert ecosystems (Li, Zhang, et al., 2015). Dryland ecosystems account for approximately one-third of the vegetation carbon in the world (Smith et al., 2000); vegetation carbon is sensitive to climate change and has been increased by the precipitation changes in Central Asia (Li, Zhang, et al., 2015;Smith et al., 2000). In addition, ephemerals can be divided into two types according to animal favor, unpalatable and palatable grasses (Anderson, Hoffman, & O'Farrell, 2010). Our results showed that both types were affected differently by water additions. The productivity of palatable grass per plot (e.g., E. oxyrrhynchum, T. arcuata, A. linifolium, and H. pulchella) increased, while the unpalatable grass (e.g., C. physodes) showed no change in shoot biomass per plot (Appendix 5). These results support the hypothesis that palatable grasses for animals are more competitive than unpalatable grasses (Moretto & Distel, 1997). The magnitude of the increase in shoot production with enhanced precipitation was more marked for H. pulchella, T. arcuata and A. linifolium than for other plants, suggesting higher water sensitivity for these three palatable grasses.

| CON CLUS IONS
In the face of the global climate change, ephemeral populations and plant communities in desert ecosystems are likely to play critical roles in the stability of these fragile ecosystems. Our study showed that the growth of all plant species and the productivity of the communities increased markedly with enhanced water input.
The response of plant community compositions to the increased precipitation was primarily reflected by changes in plant density, while increased precipitation did not affect the plant species richness and diversity index. Dominant species drove the response of plant density to increased precipitation in the five-year study period.
Moreover, increased precipitation had important effects on ecosystem functions. Increased precipitation increased the soil seed bank and altered its composition, as well as increased the productivity of palatable grasses for animals (e.g., E. oxyrrhynchum, T. arcuata, A. linifolium, and H. pulchella).
These findings have improved our understanding of how increased precipitation drives changes in plant community compositions in desert grasslands and will help us to better predict alterations in the community compositions of ephemerals under future global climate change scenarios. However, whether or how ecosystem functioning would be changed by changes in the precipitation regime in this desert ecosystem are still poorly understood, and more long-term and field experiments need to be performed in the future to answer this question.

ACK N OWLED G M ENTS
This study was supported by the National Program on Key Basic Research Project (2014CB954202). We thank the editor and two anonymous reviewers provided helpful comments on the manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interests.

AUTH O R CO NTR I B UTI O N S
Y.J. and G.F. conceived the ideas and designed methodology; Y.S., Z.S., T.Z., B.M., and C.T. collected the data; Y.J. and Y. S. analyzed the data; Y.J. and Y. S. wrote the article.

O PE N R E S E A RCH BA D G E S
This article has earned an Open Data Badge for making publicly available the digitally-shareable data necessary to reproduce the reported results. The data is available at https ://doi:10.5061/dryad. g1m5k70.

A PPE N D I X 3
General view of the study site and experimental communities. (a) Plot diagram; (b) Field image. The experimental design includes two treatments: control and 40 mm water addition (only in 2006, one more treatment: 80 mm addition).
A PPE N D I X 2 (Continued)