The effects of warming and nitrogen addition on ecosystem respiration in a Tibetan alpine meadow: The significance of winter warming

Abstract In recent decades, global warming has become an indisputable fact on the Tibetan Plateau. Alpine ecosystems are very sensitive to global warming, and the impact may depend on the degree of atmospheric nitrogen (N) deposition. The previous studies have paid more attention to year‐round warming, but the effect of winter warming has been unstudied. In this study, a manipulative experiment was conducted, consisting of warming and N addition. It was carried out since 2010 in an alpine meadow, and three types of warming treatments were set up: no warming (NW), year‐round (YW), and winter warming (WW). Warming significantly increased air and soil temperature, but decreased soil moisture. Under no N addition, YW showed significantly decreased ecosystem respiration (Reco) in 2012, and WW decreased Reco in 2014. Under N addition, neither YW nor WW had significant effects on Reco, indicating that N addition compensated the negative effect of warming on Reco. Annually, YW and WW decreased ecosystem carbon (C) emissions, and the extent of the reduction was even larger under WW. Under no N addition, both YW and WW significantly decreased aboveground biomass. Moreover, especially under no N, YW and WW significantly decreased soil inorganic N. WW also had negative effects on soil microbial biomass C. Structure equation modeling showed that soil moisture was the most important factors controlling Reco, and soil inorganic N content and microbial biomass C could explain 46.6% and 16.8% of the variation of Reco. The findings indicate that soil property changes under warming had substantial effects on ecosystem C efflux. The inhibitory effects of winter warming on ecosystem C efflux were mainly attributed to the decline of soil N and microbial biomass. Thus, the effects of winter warming on ecosystem C emissions in this semiarid alpine meadow are not as serious as expected and largely depend on N deposition.


| INTRODUC TI ON
Global warming and atmospheric nitrogen (N) deposition are important aspects of global change. The IPCC (2007) reported that global temperatures displayed significant seasonal differences, especially in the winter for high-latitude and high-altitude areas.
Under future climate change scenarios, this asymmetric warming trend will be even more pronounced (Kreyling, 2010). N is an important element limiting the productivity of terrestrial ecosystems (Elser et al., 2007;LeBauer & Treseder, 2008;Wedin & Tilman, 1996). The amount of global N deposition increased more than three times in the last century (Gruber & Galloway, 2008;IPCC, 2007) and is projected to increase by two to three times by the end of this century (Lamarque et al., 2005). N deposition increases have seriously affected the structure and function of terrestrial ecosystems (Galloway et al., 2004), but the extent to which the effects on terrestrial ecosystems interact with warming is unclear (Dormann & Woodin, 2002). Although many studies have been conducted on warming and N deposition in terrestrial ecosystems, these studies have mainly been single factor experiments over a short time period. Furthermore, studies on the effects of asymmetric seasonal warming on ecosystems are still lacking (Hutchison & Henry, 2010;Turner & Henry, 2009). Therefore, in order to obtain a deeper understanding of the impacts of global change on terrestrial ecosystems, a comprehensive study on ecosystem carbon (C) emissions, in response to asymmetric seasonal warming and increased N deposition, is urgently needed.
Elevated temperature can affect the ecosystem N cycle. With sufficient soil water content (Sw), warming can stimulate the N mineralization rate (Rustad et al., 2001), and plant productivity increases resulting from warming may increase N demand for plant growth . In addition, warming can potentially increase ecosystem N losses during the winter period, particularly in ecosystems that frequently experience soil freezing and thawing events . In these ecosystems, increased N mineralization rates during winter time, when plants are largely in their dormant period, coupled with soil freezing and thawing changes caused by snowpack decline , can lead to N loss increase from leaching (Joseph & Henry, 2008;Yanai, Toyota, & Okazaki, 2004). In addition, in alpine meadows, winter warming also affected the seasonal partitioning of soil N by plants and soil microorganisms, which can decrease soil nutrient release for plant growth in the early growing season (Edwards & Jefferies, 2013;Jaeger, Monson, Fisk, & Schmidt, 1999). These increased N losses over the winter and the decrease in nutrient release in the early growing season may limit primary productivity increase in response to experimental warming. The impacts of winter warming on ecological processes may be largely different from annual warming, as winter climate may play a critical role in N retention and other important nutrients (Kielland, Olson, Ruess, & Boone, 2006;Schimel, Bilbrough, & Welker, 2004). Thus, studies on the specific effects of winter warming on ecosystems are very important.
Recognition of the controlling factors is critical for accurately estimating C emissions. Illustrating the controlling factors for ecosystem respiration (Reco) is vital for estimating C balance and understanding the mechanisms of ecosystem CO 2 emissions under future global change scenarios. Generally, temperature is one of the important factors which affect Reco. However, in arid or semiarid areas, the relationship between Reco and temperature may be confounded by other environmental factors, such as soil water availability Tang, Baldocchi, & Xu, 2005;Xu & Qi, 2001). The direct impact on Reco is that soil moisture affects the physiological activities of plant roots and soil microorganisms, and the indirect impact is that soil moisture affects the transfer process of the substrates and O 2 for respiration (Luo & Zhou, 2006). Warming and N addition also affects plant production and soil properties, which inevitably causes ecosystem C efflux to change, as plant production and soil microorganisms are important sources of ecosystem C efflux. However, whether or not the controlling factors change under different warming treatments and N addition is still unclear.
Accounting for more than 60% of the area of the Q inghai-Tibet Plateau, alpine meadows are the basis for maintaining forage production and the development of livestock husbandry and are very sensitive to global climate change (Chen et al., 2013). In recent decades, global change has already imposed pronounced effects on ecosystem C and N cycles in alpine grasslands (Chen et al., 2013). Meteorological observation showed that, over the last several decades, asymmetric seasonal warming (with the most significant warming in winter) was very notable on the Tibetan Plateau (Li, Yang, Wang, Zhu, & Tang, 2010;Liu & Chen, 2000).
Relative to other regions, this area is projected to experience a large degree of climate warming in the next several decades (IPCC, 2007). However, studies of soil N dynamics in the winter for the Tibetan Plateau have nevertheless received little attention. Therefore, the recognition of controlling factors on the C cycle under winter warming and increased N deposition can help predict the response as well as the feedback to global change.
In this study, we investigated how warming and N addition regulating ecosystem C efflux in an alpine meadow ecosystem and isolated the specific effect of winter warming from year-round warming. We arranged the experiment in a factorial design with N addition, and we used open-top chamber devices (OTCs) to generate warming effects either for year-round or only winter treatment. We hypothesized that warming and N addition would have interactive effects on Reco. Based on the results that winter warming could increase soil N losses, and that the alpine ecosystem is N-limited, we predicted that warming would increase ecosystem C efflux under N addition treatment. We also predicted that it would restrict ecosystem C efflux under the no N addition treatment. In addition, winter warming may decrease plant production and ecosystem C efflux, as winter warming can increase soil N loss but is not affected by the warmer temperatures over the summer.

| Study area
This study was conducted in an alpine meadow in the Damxung grassland station, approximately 3 km north of Damxung County, Tibet Autonomous Region, China. Damxung County is in the central part of the southern region of the Tibetan Plateau (91°05′E, 30°29′N). The altitude is 4,333 m above sea level, and the climate is a semiarid continental type. The long-term mean annual temperature is 1.3°C, and the precipitation is 477 mm, with 85% of precipitation occurring from June to August (Shi et al., 2006;Zong et al., 2014).
The soil is classified as a meadow soil with sandy loam; the depth is approximately 0.3-0.5 m (Shi et al., 2006), and it is composed by 67.02% sand, 18.24% silt, and 14.74% clay (Zong et al., 2014). The surface soil bulk density is 1.29 g/cm 3 . Detailed soil properties can be found in Zong et al. (2014). The plant community cover is approximately 30%-50%, with Kobresia pygmaea C.B. Clarke var. pygmaea, Carex montis-everestii, and Stipa capillacea Keng as dominant species. In addition, the meadow has also been invaded by Anaphalis xylorhiza due to overgrazing degradation. The total atmospheric N deposition at this site is approximately 10 kg N ha −1 year −1 (Zong et al., 2016). We set up the experiment in July 2010 and synchronously monitored air temperature, soil moisture, and temperature at 5 cm depth by a HOBO weather station (Onset Inc., Bourne, MA, USA) at half-hour frequency. Rainfall data were obtained from the national Damxung weather station (4,288 m a.s.l., 3 km away from study site) and downloaded from the China Meteorological Data Sharing Service System (http://cdc.cma.gov.cn).

| Measurement of ecosystem respiration
Ecosystem respiration (Reco) was measured from June to September in 2012, 2013, and 2014, using a measuring system LI-8100 (LI-COR Biosciences, Lincoln, NE, USA). The LI-8100 system was attached to a chamber, 20 cm in diameter and 4.07 L in volume, and linked to a gas analyzer. At least 1 month before each measurement, one PVC collar (20 cm in diameter and 5 cm in height) was randomly inserted into soil to a depth of approximately 3 cm in each plot for Reco measurement. Plants in the collar were left intact, so that the measured respiration could represent Reco (composed by above and belowground components) Lin et al., 2011).
In each PVC collar, Reco was measured from the linear rate of CO 2 accumulation within the sealed cylindrical headspaces. During the Reco measurement process, PVC collars were covered by a removable lid that contained an opening with a CO 2 sensor. After closing the lid, CO 2 monitoring within the cylindrical headspace lasted for 1.5 min. Ecosystem CO 2 flux rates were calculated as a linear CO 2 increase using the 1-s readings during the 1.5-min closure time, with the initial 15-s mixing time after lid closure discarded in a LI-8100 file viewer application software, (Heinemeyer et al., 2011;Zong et al., 2017). Reco measurement was conducted approximately three times in each month, at an approximately 10-day interval from June to September during every growing season.

| Measurement of plant production and soil properties
Plant aboveground biomass was estimated using a nondestructive method Zong, Chai, Shi, & Yang, 2018). Briefly, for each plot in mid-August of 2012, 2013, and 2014, plant community height and cover were measured using a 50 × 50 cm quadrat divided into twenty-five 5 × 5 cm subquadrates. In 2012, we carried out this process in a nearby alpine meadow by measuring the community height and cover, harvesting, oven-drying, and weighing. The following equation was used to simulate the relationship between aboveground biomass (AGB) and vegetation height (H) and cover (C): AGB = 0.269 + 3.466C + 0.752H (R 2 = 0.658, p < 0.001, N = 80).
Details of this estimation method can be found in Zong et al. (2018).
After plant material collection in mid-August, a soil drill sampler (5 cm in diameter) was used to take 0-to 20-cm soil samples, which were immediately passed through a 2-mm sieve to pick out plant roots. These root samples were washed, separated, oven-dried at 65°C for 48 hr and weighed. The sieved soil was then mixed as a composite sample and refrigerated in the laboratory. NO Soil microbial biomass carbon (SMC) was measured by the chloroform fumigation-extraction method (Vance, Brookes, & Jenkinson, 1987). Briefly, fumigated and unfumigated soil samples were extracted with 0.5 mol/L potassium sulfate (K 2 SO 4 ) and filtered through a 0.45μm membrane. The extractable organic C was determined by a liquiTOC II analyzer (Elementar Co., Hanau, Germany) and converted to SMC using conversion coefficients of 0.45 (Xu et al., 2010).

| Statistical analysis
A repeated-measure ANOVA was applied to assess the effects of warming and N addition on ecosystem CO 2 flux. For monthly average ecosystem CO 2 flux, plant biomass, SIN, and SMC, a two-way ANOVA was used to test the differences between different treatments and followed by Duncan's test for multiple comparisons.
Regression analyses were also used to test the relationships between ecosystem CO 2 flux and Sw, soil temperature, plant aboveground and belowground biomass, SIN, and SMC in different years.
The average growing season Reco was averaged by daily respiration data measured during each growing season. Total C emissions from the entire growing season were the sum of the monthly C emissions.
A previous study in the same ecosystem found that the proportion of C released during the growing season was 97.4% of the total annual amount (Zhang, 2005). All the analyses were performed in SPSS Structure equation modeling (SEM) was also used to evaluate the direct and indirect effects of different environmental variables on Reco. Based on the theoretical knowledge of major environmental factors regulating the variations of ecosystem CO 2 efflux, a path model was developed to evaluate the interactive relationships between Reco, Sw, SIN, SMC, and AGB. The adequacy of this model was evaluated by the chi-square test and Akaike information criterion (AIC). Nonsignificant chi-square tests (p > 0.05) and a low AIC value suggested that the model could be accepted as a potential explanation of the observed covariance structure (Grace, 2006). Based on the AIC values, nonsignificant pathways were removed to improve the model adequacy. Eventually, the final model was relatively strong: χ 2 = 1.044, probability level = 0.307, RMSEA = 0.019, and CFI = 1.00.
Furthermore, in this path model, R-squares for Reco were relatively high. The SEM was performed using Amos 17.0 (SPSS Inc.). devices increased air and soil temperature by 1.6 and 1.4°C, respectively, while reducing soil moisture by 4. 7% (v/v Figure 1d).

| Seasonal variations of ecosystem respiration and annual ecosystem CO 2 efflux
Statistical analysis showed that Reco presented significant seasonal variations (Table 1 Warming tended to decrease Reco, but varied with years (Table 1, By averaging daily Reco in the same month, we calculated total C emissions throughout the growing season. The previous study had indicated that C emissions in the growing season accounted for 97.4% of total annual C emissions. We estimated that the total annual C emissions in ambient plots in 2012, 2013, and 2014 were 4,583, 4,082, and 5,532 kg C ha −1 year −1 , respectively (  (Table 2). Under N treatment, YW and WW had no effect on ecosystem C emissions in 2014 (Table 2).

| Plant aboveground and belowground biomass
Warming significantly affected aboveground biomass (Table 3,

| Soil inorganic N and microbial biomass C
Warming significantly affected SIN (Table 3,  Warming also significantly affected SMC (Table 3,

| Environmental factors regulating ecosystem respiration
Regression analysis showed that the seasonal variation of Reco was marginally and negatively correlated with air temperature, and it can explain only 10.8% of the variations of Reco, with soil temperature only explaining 5.8% of the variation of Reco (Figure 5a,b). These     on Reco was not significant (Figure 7). From the total effects on Reco, Sw was the most important factor affecting Reco (R 2 = 0.494).
SIN and SMC can explain 46.6% and 16.8% of the variations of Reco, while AGB can only explain 6.8% of its variation (Table 4). The results demonstrated that soil properties such as Sw, SIN, and SMC, were key factors regulating the variation of Reco. These findings also indicate that the effects of the changes in warming on soil properties, rather than plant production, affected ecosystem CO 2 efflux.

| D ISCUSS I ON
Our results demonstrate that the warming treatment significantly in- Due to the high latitude, harsh climate, and remote distance of the study site, we mainly collected the data during growing season.
In fact, as mentioned above, during growing season, warming altered the plant production and soil properties, and subsequently the ecosystem C efflux. Therefore, the lack of data during the nongrowing season may have effects on an annual timescale. However, plants only generate production in the growing season, and the effects of nongrowing season warming on plant production are manifested in the growing season. In addition, a previous study showed that ecosystem C emissions in the growing season were 97.4% of total annual C emissions (Zhang, 2005). Therefore, the change in ecosystem C efflux in the growing season could largely account for annual timescale change.

| Effects of warming on ecosystem C efflux and implication of winter warming
The During seasons with rare rainfall events (mostly during the early growing season), warming reduced Reco. Generally, warming promoted N mineralization and provided more N for plant growth, especially in nutrient-limited ecosystems, but these responses only occurred in the case of sufficient water availability for the plants (Sierra, 1997;de Valpine & Harte, 2001). In semiarid alpine regions, rainfall rarely occurs in the early spring, and during this time period, warming can intensively reduce soil moisture. This reduction resulted in the inhibition of plant growth and thus C emissions.
Therefore, the effects of warming on plant growth were more pronounced at the beginning of the growing season, because rare rainfall events reduce soil moisture. The previous study has found that in the semiarid alpine region, soil moisture was an important factor regulating seasonal and large-scale spatial patterns of Reco (Geng et al., 2012;Jiang et al., 2013). The decline of soil moisture resulting from warming will directly limit ecosystem C emissions. In addition, soil microbial activity and substrate supply could also be inhibited due to the reduction in Sw (Niu et al., 2008;Yan, Chen, Huang, & Lin, 2011).
In the nongrowing season, warming generally promotes N mineralization because of temperature increase (Henry & Jefferies, 2003;Rustad et al., 2001), but under low-temperature conditions, dormant plants and soil microbes are not actively retaining nutrients. With an increased frequency of soil freeze-thaw cycles, this may lead to gaseous or leaching soil N loss (Hobbie & Chapin, 1996;Matzner & Borken, 2008;Turner & Henry, 2010). In a previous paper, we also Notes. All the effects were calculated using standardized path coefficients. Sw, soil inorganic N content, aboveground biomass, soil microbial biomass carbon, and Reco represented soil water content, soil inorganic N content, plant aboveground biomass, soil microbial biomass C, and ecosystem respiration, respectively. the early growing season , which would limit plant growth during the following growing season. In addition, another reason for the decrease in Reco in the warming treatment was the inhibition of plant production by warming . This was determined to occur because plant biomass is an important component of Reco Zong et al., 2013 (Vidon et al., 2010;Zhou et al., 2011). Under the background of global warming, air temperature change induced by the decrease in winter snow cover in high-latitude and high-altitude region could increase the frequency of soil freezing and thawing cycles (Henry, 2008), thereby affecting ecosystem soil C and N cycles and storage (Kreyling, 2010). or shortly after the soil thaw event, concurrently, or followed by, a nutrient pulse that can provide an important nutrient resource for plant growth in the early growing season (Edwards & Jefferies, 2013;Jaeger et al., 1999). As seasonal biogeochemical events, the timing and magnitude of nutrient pulses could be affected by winter warming, which has important implications for ecosystem primary productivity and C efflux under future global change scenarios (Edwards & Jefferies, 2013).

| Effects of N addition on ecosystem C efflux
An appropriate quantity of N addition significantly increased ecosystem C emissions, consistent with the effects of N addition on plant production . In general, due to the high altitudes of alpine ecosystems, the low temperature restricts soil N mineralization, and the soil N content is generally very low, so, for an alpine meadow, soil N availability becomes a key factor limiting production (Bowman, Theodose, Schardt, & Conant, 1993;Cao & Zhang, 2001;Jiang et al., 2013).
Exogenous nutrient inputs significantly increased soil nutrient availability, so that leaf N content and photosynthetic capacity increased significantly (Reynold & Thornley, 1982;Lü et al., 2013). A previous study showed that N addition can enhance soil net N mineralization rates , which would stimulate the decomposition of organic matter in soil, which in turn can improve the soil inorganic N, and lead to an increase in plant production ). An improvement of plant production means more respiration for growth and maintenance (Flanagan & Johnson, 2005), and more photosynthetic products delivered to soil microorganisms (Yan et al., 2011). Therefore, improved plant productivity due to N addition is an important factor in increase in Reco.
The N addition offsets the loss of soil N in the warming treatment, which is more pronounced in the late-growing season which has many rain events. This is consistent with the results from a study on an old farmland (Hutchison & Henry, 2010). This study found that there were no treatment effects on plant biomass in dry years, while in wet years, warming (both year-round and winter-only) combined with N addition approximately doubled plant aboveground productivity, and that these effects were additive (Hutchison & Henry, 2010). This finding indicated that the effect of warming may interact very strongly with interannual variation in precipitation.

| Factors regulating ecosystem C efflux on different time scales
Generally, temperature is the most important factor regulating Reco, and the positive correlation between Reco and temperature has been referenced in many ecosystem models (Reichstein et al., 2003;Rey, Petsikos, Jarvis, & Grace, 2005;Zhou, Talley, & Luo, 2009). However, the seasonal dynamic of Reco was less negatively correlated with temperature, but positively correlated with soil moisture (Figure 5), consistent with our previous study . The apparent negative effect of soil temperature on ecosystem and soil respiration could be confounded by the effect of the aboveground biomass, especially under nutrient enrichment . In semiarid areas, soil moisture plays an important role in regulating the activities of plant production (Niu et al., 2008;Xu & Wan, 2008;Yan et al., 2011) and soil microorganisms (Austin et al., 2004;Bi, Zhang, Liang, Yang, & Ma, 2012). Plant production is the source of the substrate for Reco, and the controlling effects have been verified in many previous studies Yan et al., 2011). Therefore, on a seasonal timescale, the relationship between soil temperature and Reco was confounded by soil moisture (Shen, Li, & Fu, 2015) and plant production .
Structure equation modeling analysis demonstrated that soil properties, such as Sw, SIN, and SMC, were key factors regulating the seasonal and interannual variations of Reco. In semiarid areas, soil moisture plays an important role in regulating the activities of plant production (Niu et al., 2008;Xu & Wan, 2008;Yan et al., 2011) and soil microorganisms (Austin et al., 2004;Bi et al., 2012).
Sw not only directly affected Reco but also indirectly affected Reco through soil nutrient availability and microbial biomass. Therefore, the decrease in Sw under warming had significant effects on ecosystem C efflux through the subsequent change in soil properties.
Soil nutrient availability and microbial biomass also directly or indirectly had effects on Reco (Figure 7). The warming treatment, especially winter warming, decreased SIN and SMC (Figure 4), which could account for the decrease in Reco and subsequently the annual C efflux. As the direct effect of winter warming on soil microorganisms, we also infer that winter warming could affect the timing and magnitude of nutrient pulses. This would have important implications for primary productivity and ecosystem C efflux in alpine ecosystems under future global change scenarios (Edwards & Jefferies, 2013;Jefferies, Walker, Edwards, & Dainty, 2010 variations. These findings indicated that in this semiarid alpine meadow ecosystem, rather than plant production, the changes in warming on soil properties affected ecosystem CO 2 efflux. They also indicated that the greater effect of winter warming than yeararound warming on ecosystem C efflux can be interpreted by these mechanisms. Structure equation modeling analysis indicated that in this alpine meadow, the changes in warming on soil property changes, rather than plant production, had greater effects on ecosystem CO 2 efflux. The results can be interpreted as follows. The Turner & Henry, 2010). Therefore, the effects of winter warming on soil properties were direct, while the effects on plant production occurred later and were indirect in growing season. Third, the previous studies also showed that soil microbes were the main source of ecosystem C efflux, and the effects of warming on soil microbes can be directly manifested in Reco. Although SIN was also decreased by YW, SIN only can indirectly affect Reco through its effect on plant production and other ecological processes. Therefore, in this alpine meadow, soil property changes induced by the warming treatment had greater effects on ecosystem CO 2 efflux.

| CON CLUS ION
To our knowledge, this is the first study to evaluate winter warming and separate the effects of warming treatments for ecosystem C efflux and the controlling factors of an alpine meadow on the Qinghai-Tibetan Plateau. Warming can directly reduce ecosystem CO 2 emissions by reducing Sw, while winter warming increased SIN loss and decreased SMC, and indirectly affected ecosystem C emissions. N addition could compensate for the decrease in SIN to some extent. The findings indicated that the effects of warming on soil properties are more important than plant production, to affect ecosystem CO 2 efflux in this semiarid alpine meadow ecosystem. From the aspect of ecosystem C efflux, the effects of winter warming are not as impactful as predicted and largely depend on precipitation pattern and atmospheric N deposition in this semiarid alpine region.

ACK N OWLED G M ENTS
We thank the experts from American Journal Experts (https://www. aje.com/#) for editing the English text of a draft of this manuscript.
This work was jointly supported by the National Natural Science

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
NZ, PLS, and XZZ designed the experiments. NZ, SBG, and CD performed the experiments. NZ, SBG, CD, and XC analyzed the data.
NZ, PLS, and XZZ wrote and revised the manuscript.

DATA ACCE SS I B I LIT Y
Data are available from the Dryad Digital Repository: https://doi.