Temperature‐mediated responses of carbon fluxes to precipitation variabilities in an alpine meadow ecosystem on the Tibetan Plateau

Abstract Effects of climate warming and changing precipitation on ecosystem carbon fluxes have been intensively studied. However, how they co‐regulate carbon fluxes is still elusive for some understudied ecosystems. To fill the gap, we examined net ecosystem productivity (NEP), gross ecosystem productivity (GEP,) and ecosystem respiration (ER) responses to multilevel of temperature increments (control, warming 1, warming 2, warming 3, warming 4) in three contrasting hydrological growing seasons in a typical semiarid alpine meadow. We found that carbon fluxes responded to precipitation variations more strongly in low‐level warming treatments than in high‐level ones. The distinct responses were attributable to different soil water conditions and community composition under low‐level and high‐level warming during the three growing seasons. In addition, carbon fluxes were much more sensitive to decreased than to increased precipitation in low‐level warming treatments, but not in high‐level ones. At a regional scale, this negative asymmetry was further corroborated. This study reveals that future precipitation changes, particularly decreased precipitation would induce significant change in carbon fluxes, and the effect magnitude is regulated by climate warming size.

patterns of ecosystem structure and function in response to precipitation changes is critical for predicting their provisioning of ecosystem services (Nimmo, Mac Nally, Cunningham, Haslem, & Bennett, 2015).
Precipitation variability is a key driver of ecosystem structure and function for arid and semiarid ecosystems (Liu et al., 2012;Suttle, Thomsen, & Power, 2007). Ecosystems respond to precipitation regime changes through shifts in species composition, distribution, and abundance (Scott, Hamerlynck, Jenerette, Moran, & Barron-Gafford, 2015), as well as water and carbon balances (Kulmatiski & Beard, 2013). For example, six percent expansion of vegetation cover bring about fourfold strengthened sensitivity of net carbon uptake to precipitation change in Australia (Poulter et al., 2014). Hence, in evaluating carbon fluxes responses to climate change, precipitation variations should be fully considered (Xia, Niu, & Wan, 2009), particularly alternated dry or wet seasons (Bonal et al., 2008).
Strong precipitation variability leads to asymmetrical responses of carbon fluxes to increased and decreased precipitation. Gross primary productivity (GPP) or net primary productivity (NPP) is reported to be much more sensitive to increased precipitation (Unger & Jongen, 2015;Wilcox et al., 2017;Wu, Dijkstra, Koch, Penuelas, & Hungate, 2011) or decreased precipitation (Luo et al., 2008;Zscheischler, Michalak, et al., 2014). Currently, much of our knowledge is centered around ecosystem structure and function in response to naturally occurring climatic variations , extreme precipitation experiments (Knapp et al., 2015), or synthesis analysis (Wilcox et al., 2017), while significant knowledge gap exists for some typical ecosystems. In a related synthesis analysis, 83 studies of experimental precipitation manipulations in grasslands were incorporated worldwide, but no single case on the TP (Wilcox et al., 2017). Based on the optimized model, a recent study had reported that TP ecosystem was more sensitive to drying than to wetting (Liu et al., 2018). In addition, plenty of studies evidenced influences of precipitation of both nongrowing season and growing season on ecosystem structure and function on the TP (Cong et al., 2017;Shen, Piao, Cong, Zhang, & Jassens, 2015;Shen, Tang, Chen, Zhu, & Zheng, 2011). Despite our growing awareness and concern, a vital knowledge gap exists about whether the sensitivity of carbon fluxes differs under precipitation increases versus decreases on the TP.
A growing body of evidences demonstrated that warming would export strong direct and indirect effects on carbon fluxes. Climate warming can alter plant community structure and composition (Botkin et al., 2007;Gedan & Bertness, 2009). The warming effects on carbon fluxes vary with plant species (Chen, Luo, Xia, Shi, et al., 2016;Chen, Luo, Xia, Wilcox, et al., 2016), functional groups (Niu, Sherry, Zhou, & Luo, 2013), and root depth (Zhu, Zhang, & Jiang, 2017). Warming also can indirectly regulate carbon fluxes through stimulating evapotranspiration, reducing soil moisture, and exacerbating water stress (Niu et al., 2008). Weakened soil water availability related to warming will exacerbate water limitations on arid and semiarid ecosystems, offsetting part of positive warming effects (Niu et al., 2008). This phenomenon is more likely to be associated with precipitation changes Dermody, Weltzin, Engel, Allen, & Norby, 2007). Studies on the TP also revealed that the interactions between changes in temperature and precipitation would regulate ecosystem structure and function (Ganjurjav et al., 2018;Shen et al., 2014). Both direct and indirect effects of warming on carbon fluxes were related to their magnitudes, because lowlevel warming and high-level warming induce different changes in soil water availability , water use efficiency (Quan et al., 2018), and community composition (Li, Wang, Yang, Gao, & Liu, 2011). Therefore, warming could regulate the precipitation effects on carbon fluxes through modulating the water availability and community composition. To date, few studies have reported this phenomenon on the TP, and multilevel warming was even more uncommon.
Here, we conducted a field experiment to investigate effects of multilevel warming on carbon fluxes in an alpine meadow ecosystem across the TP. This study was conducted for three hydrologically contrasting growing seasons (dry in 2015, wet in 2016, and normal in 2017), which presents a unique opportunity to reveal how precipitation change affects carbon fluxes under multilevel temperature increasing scenario. Specifically, two main questions were set to be addressed: (a) How carbon fluxes respond to interannual variability of precipitation under multilevel temperature increases? and (b) Whether there exists an asymmetric response of carbon fluxes to increased and decreased precipitation under multilevel of temperature increases?

| Study site
The study area was conducted at the Tibet Alpine Grassland Ecosystem Research Station, which is operated by Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences (31°38.513′N, 92°0.921′E, 4,585 m). The study area represents a typical alpine meadow ecosystem. The longterm mean annual temperature and precipitation are −1.05°C and

| Experimental design
Open top chambers (OTCs) were used as passive warming devices based on the International Tundra Experiment design standard (Marion et al., 1997). The OTCs used in the current study were similar to those described in other studies (Chen, Luo, Xia, Shi, et al., 2016;Chen, Luo, Xia, Wilcox, et al., 2016;Dabros & Fyles, 2010).
The 15 plots were separated by a 3.5-m buffer and arranged following a randomized block design.

| Measurements of carbon fluxes
Ecosystem carbon fluxes were measured by an infrared gas analyzer (IRGA; LI-6400, LiCor Inc.) attached to a transparent chamber (0.3 × 0.3 × 0.3 m 3 ). When the conditions within the chamber achieved a steady state, 30 consecutive CO 2 concentration recordings were obtained on each base at 2-s intervals. In order to ensure the air uniformity in the static chamber, two small electric fans were installed inside the chamber and ran continuously to mix the air inside. During measurements, CO 2 concentration was allowed to build up or draw down over time, from which flux rates were determined and net ecosystem productivity (NEP) was calculated. Positive and negative NEP indicated net carbon uptake and net carbon release, respectively. Following NEP measurement, the chamber was vented and replaced on each frame. The chamber was covered by an opaque cloth, and CO 2 exchange measurements were repeated. Under the second set of measurements, light was eliminated (and hence photosynthesis) and the obtained values represent ecosystem respiration (ER). The sum between NEP and ER was treated as instantaneous gross ecosystem productivity (GEP). Ecosystem gas exchange was measured every 5-10 days at 9:00a.m.-12:00p.m. from May to September in 2015-2017. These measuring processes followed the same standards of a previous study in our experimental site (Zhu et al., 2017).

| Measurements of community coverage
A 1 × 1 m frame with 100 equally distributed grids (0.1 × 0.1 m) was placed above the vegetation canopy to measure vegetation coverage (1 × 1 m). Grids with plants appearing over 1/2 of the grid were marked as 1, otherwise marked as 0. The coverage was mainly measured in middle and late growing season .

| Soil temperature and water content
Soil temperature and moisture at 5 cm belowground were measured in the centre of the plots using Campbell CS655 sensors (Campbell Scientific, Logan, UT). Measurements of soil temperature and moisture were taken with 30-min intervals, and averages of the fortyeight measurements were stored as the daily averages. In each warming treatment (three plots), we installed soil sensors in two of them and took their average .

| Regional precipitation and primary productivity products
The annual precipitation data with a spatial resolution of 1 km × 1 km from 2000 to 2015 were provided by the Cold and Arid Regions Science Data Center (http://westdc.westg is.ac.cn). The NPP data with a spatial resolution of 1 km × 1 km were calculated by the Carnegie-Ames-Stanford approach (CASA) from 2000 to 2015.

| Quantifying sensitivity to precipitation
Sensitivity was calculated as the response range relative to the amount of precipitation variability (Knapp, Ciais, & Smith, 2017;Wilcox et al., 2017). The advantage of this method is that ecosystem responses are comparable after they are standardized by the range of precipitation variability: where, X dry and X wet represent the productivity in dry and wet years, respectively. X t represents the productivity means across 2015-2017 and 2000-2015 for in situ measurement and remote sensing products at the regional scale. PPT dry and PPT wet represent the precipitation amounts in dry and wet years, respectively. PPT t is the precipitation means across 1955-2017 and 2000-2015 for in situ measurements and remote sensing products at the regional scale. In this study, the absolute value for the calculated dry minus mean and wet minus mean is required to be roughly equal (the error being <10%). The Sens wet > Sens dry indicates positive asymmetry, with the opposite indicating negative asymmetry.

| Statistical analysis
The one-way ANOVA was applied to compare the sensitivity of

| Microclimate
The three growing season mean soil temperature was, on average,   Figure 1h-i). In July of 2015, it was 63.2% (SPI = −2.2) lower than the long-term mean (Figure 1g, Table 1). In growing season of 2016, precipitation exhibited a unimodal pattern, and July precipitation (SPI = 1.6) was 39.1% higher than the long-term mean (Figure 1h; Table 1). In 2017, June precipitation (SPI = 1.6) was higher than the long-term mean, while July precipitation (SPI = −1.4) was 47.5% lower than the long-term mean (Figure 1i,
Based on the min-max normalization, we calculated the standard deviations (SDs) of carbon fluxes among the three growing seasons under each warming treatment. The average SD of NEP, ER, and GEP under low-level warming (control, W1 and W2) was 0.09 μmol m −2 s −1 , 0.02 μmol m −2 s −1 , and 0.08 μmol m −2 s −1 higher than that of high-level warming (W3 and W4; Figure 3d). In addition, K-means clustering results showed that SD of carbon fluxes among the three growing seasons under the five experimental treatments can be divided into two categories, with control and W1 being assigned to one class and W2-W4 being assigned to the other class ( Figure 4).  Table 4). While there was marginal difference in precipitation sensitivity of NEP (p = 0.149) and ER (p = 0.809) between dry and wet growing seasons under the highlevel warming, but not GEP (p = 0.041; Figure 5; Table 4), the sensitivity of carbon fluxes to dry or wet in low-level warming was higher than that of high-level warming ( Figure 5). At the regional scale, ecosystem NPP exhibited higher sensitivity to decreased precipitation for 53.5% of alpine meadow region ( Figure 6). Both the in situ measurements and ecosystem modeling results at the regional scale pointed to the negative and asymmetric responses of carbon fluxes to precipitation variations.

| Impacts of biotic and abiotic factors on carbon fluxes
The seasonal variabilities of GEP and ER were mainly regulated by soil moisture and K. pygmaea coverage under control, W1, and W2 (Table 5; p < 0.05). In contrast, soil temperature and Potentilla coverage mostly had insignificant effects on GEP and ER under W3 and W4 (Table 5; p > 0.05). We further found that the slopes between biotic, abiotic factors, and GEP were steeper than that of ER in all the warming treatments (Table 5), which indicates the stronger responses of GEP to biotic and abiotic factors than that of ER.

| D ISCUSS I ON
For alpine grasslands, carbon fluxes responses to precipitation variability differed with warming magnitude. At a situ scale, carbon fluxes were much more sensitive to precipitation variability in low-level warming than that of high-level warming. The contrasting response F I G U R E 5 NEP, ER, and GEP sensitivity to precipitation variations (µmol m −2 s −1 /10 mm) in 2015 (dry) and 2016 (wet). Different letters in insets indicate significant differences (p < 0.05). High-level warming: W3 and W4; low-level warming: control, W1, and W2 F I G U R E 6 NPP sensitivity to precipitation variations (µmol m −2 s −1 /10 mm) across the alpine meadow. W (gray region): positive asymmetry in NPP responses to precipitation; D (black region): negative asymmetry in NPP responses to precipitation pattern can be attributable to soil water availability, as well as biotic features of each species.
In this study, the variabilities of GEP and ER were positively related to precipitation. Consistent with previous study (Aires, Pio, & Pereira, 2008)
Conventionally, shallow-rooted plants mostly utilize shallow soil water and being highly sensitive to precipitation variations (Liu et al., 2012). Modified plant community cover would cause a series of changes in soil evaporations, autotrophic respiration, and canopy photosynthesis (Liu, Cieraad, Li, & Ma, 2016;Verburg et al., 2004).
Ecosystem GEP is mainly controlled by its photosynthesis capability (Xia et al., 2015). For the alpine ecosystem, ER variations are dominated by those of autotrophic plant respiration (Chen, Luo, Xia, Shi, et al., 2016;Chen, Luo, Xia, Wilcox, et al., 2016). Considering all these interactions, precipitation plays a key role in regulating carbon fluxes.
For the alpine meadow ecosystem, K. pygmaea, as a dominant species, is shallow-rooted species (Dorji et al., 2013) and relies strongly upon soil surface water (Liu et al., 2012).  (Xu & Li, 2006;Xu, Li, Xu, & Zou, 2007 So their contribution to precipitation-driven variability in carbon fluxes was weaker relative to the shallow deep-rooted species (Liu et al., 2016).

| Asymmetric responses of carbon fluxes to decreased and increased precipitation
This study revealed that NEP and its two components were much more sensitive to decreased than to increased precipitation under low-level warming treatments, but exhibited marginal differences in high-level warming treatments. Extreme climates cause differential survivorship among species and modify community structure and species distributions (Engelbrecht et al., 2007;Miriti, Rodriguez-Buritica, Wright, & Howe, 2007). Drought restricts leaf emergence and canopy development, leading to decreased plant cover and increased plant mortality (Dong et al., 2011). In addition, dry air and/ or soil conditions downgrade leaf stomatal conductance (Jia, Zha, Gong, Wang, et al., 2016;. Declined plant cover related to stomatal closure could suppress canopy photosynthetic capacity (Chen, Luo, Xia, Shi, et al., 2016;Chen, Luo, Xia, Wilcox, et al., 2016) and further restrain GEP. Suppressed root and microbial activities under drought conditions, together with reduced plant cover, lead to lowered ER (Liu et al., 2016;Niu et al., 2008).  Table 6). In contrast, Potentilla possesses stronger drought resistance under low-level warming (p = 0.546) and high-level warming (p = 0.394; Figure 7; Table 6), especially for Potentilla bifurca Linn (Wei & Li, 2003). The positively linear correlation between K. pygmaea coverage and carbon fluxes, together with the asymmetrical response of K. pygmaea to hydrologically contrasting conditions, resulted in negative asymmetry in carbon fluxes responses to precipitation in low-level warming treatments.
Climate influences ecosystem productivity by adjusting vegetation phenology (Pau et al., 2011). Although plentiful precipitation fell in August of a dry growing season in 2015, its effects on carbon fluxes are marginal (Craine et al., 2012). The ecosystem productivity variation in August is highly related to late July precipitation (Craine et al., 2012). In the late growing season, solar radiation weakens and photosynthetic capacity of old leaves subsides (Gunderson et al., 2012). An advanced leaf emergence in spring exerts greater influences on seasonal carbon uptake than an equivalent delay duration of fall senescence (Marchin, Salk, Hoffmann, & Dunn, 2015).
This phenomenon can also be explained by the "slow in, rapid out" principle of net ecosystem exchange. The negative anomalies in precipitation may induce drought-stress mortality. Recovery from the drought through regeneration may be slow, even under adequate precipitation (Korner, 2003). Therefore, negative asymmetry response is also explained by the seasonal distribution of precipitation in a dry growing season.

| CON CLUS IONS
Carbon fluxes in low-level warming treatments were more sensitive to precipitation variability than that in high-level warming treatments for the alpine meadow ecosystem. Such distinctions were ascribable to different in soil water availability and dominances of shallow-rooted plant under low-and high-level warming treatments.
Furthermore, carbon fluxes respond more strongly to decreased than to increased precipitation, leading to their negative and asymmetric responses. This study highlights that the interannual variations of precipitation play a critical role in modulating ecosystem carbon cycling, whereas this effect varies with warming magnitude.

ACK N OWLED G M ENTS
This study was financially supported by the International Partnership Program of Chinese Academy of Sciences (131C11KYSB20160061) and National Natural Science Foundation of China (41725003;41571195;41801083;41501103).

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
Ning Chen and Yangjian Zhang conceived and designed the study.

DATA ACCE SS I B I LIT Y
I agree to deposit the main data of the article in Dryad. https ://doi. org/10.5061/dryad.506p9k0.