Effects of grazing on CO 2 , CH 4 , and N 2 O ﬂ uxes in three temperate steppe ecosystems

. Terrestrial ecosystems play a critical role in regulating the emission and uptake of the most important greenhouse gases (GHGs) such as CO 2 , CH 4 , and N 2 O. However, the effects of grazing on these GHG ﬂ uxes in different steppe types remain unclear. Here, we compared the effects of grazing on seasonal CO 2 , CH 4 , and N 2 O ﬂ uxes in the meadow (MS), typical (TS), and desert (DS) temperate steppe ecosystems in northern China. CO 2 emission rates increased from 311.4 (cid:1) 73.2 to 349.6 (cid:1) 55.4 mg (cid:3) m (cid:4) 2 (cid:3) h (cid:4) 1 in MS, but decreased in TS (from 341.3 (cid:1) 93.0 to 239.5 (cid:1) 81.9 mg (cid:3) m (cid:4) 2 (cid:3) h (cid:4) 1 ) and DS (from 212.1 (cid:1) 53.7 to 163.0 (cid:1) 83.4 mg (cid:3) m (cid:4) 2 (cid:3) h (cid:4) 1 ) in response to summer grazing (SG). N 2 O emission rates increased in MS from 4.7 (cid:1) 2.2 uptake magnitude of coef emission by negative relationship between and simultane-ously in and Our suggest that effects of SG on GHG ﬂ uxes varied in different steppes and the relationship among GHGs was steppe-dependent and SG also changed the relationship by affecting GHG ﬂ uxes induced by varied soil and environmental factors.

INTRODUCTION CO 2 , CH 4 , and N 2 O are the three most important greenhouse gases (GHGs) contributing 64%, 17%, and 6%, respectively, to the total global warming potential of all GHGs (IPCC 2013). Terrestrial ecosystems play a key role in regulating the emission and uptake of GHGs (Han et al. 1999), and fluxes are biogeochemically coupled with the soil and the atmosphere (Han et al. 1999). CO 2 emissions from soil to atmosphere (i.e., soil respiration) are the sum of microbial respiration, root respiration, and bulk soil respiration (Xu et al. 2008a). N 2 O fluxes from soils are derived from microbial-mediated nitrification and denitrification under aerobic and anaerobic conditions, respectively (Holst et al. 2008, Zhong et al. 2014. CH 4 fluxes are the function of a balance between the production by methanogenic microbes and the consumption by methanotrophic microbes (Hou et al. 2012. CO 2 , CH 4 , and N 2 O are the most biogeochemically mobile forms of C and N. Emission and uptake of these gases determine the turnover rate between terrestrial ecosystems and the atmosphere (Schlesinger and Bernhardt 2013). Understanding relationships between GHGs and the underlying mechanisms driving those relationships is very important in improving our understanding of global C and N cycling (Xu et al. 2008a, Zhang et al. 2014a. Few studies have assessed the relationships between multiple GHG fluxes in terrestrial ecosystems (Dong et al. 2000, Xu et al. 2008a, Zhang et al. 2014a. Positive and linear correlations between CO 2 and N 2 O fluxes have been found in temperate grasslands (Dong et al. 2000, Xu et al. 2008a, Zhang et al. 2014a, tropical forests (Garcia-Montiel et al. 2002, and croplands (Zhou et al. 2004). Negative linear relationships between CH 4 and CO 2 fluxes have been found in grasslands (Wu et al. 2010), and a trade-off relationship between N 2 O and CH 4 fluxes was found in croplands (Zhou et al. 2004). However, there is no consensus on the relationships between multiple GHG fluxes and the mechanisms driving these relationships are largely unknown.
Previous studies indicate that soil characteristics such as texture, structure, water-holding capacity, pH, and organic C content strongly influence the metabolic pathways in the C and N cycles (Garcia-Montiel et al. 2004, Xu et al. 2008a. The magnitude, controls, and associations between GHG fluxes differ considerably among different grassland types (Frank et al. 2002, Garcia-Montiel et al. 2004). Grasslands account for approximately 25% of the earth's land surface and have a significant effect on global GHG exchange (Batjes 1998) due to their capacity to produce, store, and cycle C and N substrates . Livestock grazing is the most important human practice in grasslands and has been shown to be an important factor regulating the emission and uptake of GHGs (Sch€ onbach et al. 2012, Skiba et al. 2013, Tang et al. 2013. The biophysical and chemical controls of GHG fluxes in response to grazing vary across steppe types due to site-specific grazing history (Frank et al. 2002, Tang et al. 2013. To assess the effects of grazing on GHG fluxes in meadow (MS), typical (TS), and desert steppes (DS) in Inner Mongolia, China, a three-year field study measuring fluxes (CH 4 , CO 2 , and N 2 O) as well as soil, vegetation, and environmental factors was carried out. We hypothesize that (1) grazing effects changed with different GHGs for special mechanisms of production and consumption, (2) relationships among multiple GHGs fluxes are steppe-dependent for different soil characteristics and environmental conditions, and (3) summer grazing (SG) affects the relationships by changing the soil and environmental factors.

Study site
In 2011, along east-west transect in the Inner Mongolia grasslands, MS, TS, and DS were assigned, respectively. The growing season starts in early May and ends in late September. The mean annual temperature and precipitation

Experimental design and field measurements
Randomly selected paired plots of SG and ungrazed (NG) areas of 100 9 100 m 2 were established at each of the three sites. Summer grazing plots had strictly restricted stocking rates and time (0.5 sheep unit per hm 2 at growth season). The NG areas were enclosed to prohibit grazing. Three flux chamber bases (length 9 width 9 height = 0.5 9 0.5 9 0.1 m) were installed randomly 10 cm into soil 10-15 d prior to measurements in each plot. Canopy biomass in the frame of bases was clipped to ground level one day before gas sampling. Gas samples were collected in the morning (9:00-10:00) and the chamber bases were moved to neighboring locations within each plot after gas sampling was completed. The sampling frequency was once every 10-15 d during the growing season of each year (from May to October 2011-2013).
An in situ static chamber technique was used for measurements of GHG fluxes. Stainless steel chambers, 25 cm high and shaded with a reflective plastic sheet, were placed onto the bases sealed with water. Chamber headspace gas samples were collected at 0-, 10-, 20-, and 30-min time intervals using 100-mL polypropylene syringes. The gases were then injected into 100-mL sealed airbags. The airbags were transported to the laboratory within two days. Gas measurements were done with a Hewlett-Packard 5890 series II gas chromatograph fitted with an electron detector (for N 2 O) and flame ionization detector (for CH 4 and CO 2 ). Certified CH 4 , CO 2 , and N 2 O standards of 1.92, 348, and 0.338 lL/L, respectively, were used for calibration. In situ daily emission/uptake rates of GHGs were determined according to the concentration trend in chamber headspace and recorded as lgÁm À2 Áh À1 for CH 4 and N 2 O, and mgÁm À2 Áh À1 for CO 2 . Mean seasonal fluxes were calculated as the arithmetic mean value.
Soil temperature (ST) was measured concurrently with gas samples using a portable digital thermometer (902C; Shengtong Instrument Factory, Hebei, China). Soil moisture (SM) was also measured concurrently from soil samples (n = 3) collected just outside the bases using a stainless steel corer (3.5 cm in diameter) and oven-dried at 105°C for about 20 h to a constant weight. Aboveground biomass (AGB) was measured monthly by clipping canopy biomass to the ground level in three to six quadrats (1 9 1 m) at each plot each year. Corresponding belowground biomass (BGB) was also sampled in 2012 and 2013 using a stainless steel corer (7.0 cm in diameter) in each quadrat and put in root bags. After being rinsed, BGB and AGB were oven-dried at 70°C to constant weight (about 48 h).
A baseline survey of vegetation and soil characteristics was carried out in May 2012. Soils were sampled using a stainless steel corer (3.5 cm in diameter) with three replications for each plot (SG vs. NG) at each steppe site. The soil samples were sieved through 2-mm mesh and separated into two parts. One part was stored at 4°C for the measurement of NH 4 + , NO 3 À , and microbial biomass C and N (MBC and MBN). MBC and MBN were determined using the chloroform fumigationextraction method (Liu et al. 2007). Soil NH 4 + and NO 3 À concentrations were measured using Continuous-Flowing Analyzer (AutoAnalyzer 3, SEAL Analytical, Norderstedt, Germany). The second part was air-dried for measurements of soil pH, texture, soil total C and N, and available phosphorus and potassium. Soil pH was measured using pH meter in soil water suspension (soil:water = 1:2.5). Soil texture was measured using laser particle size analyzer (Mastersizer 2000, Malvern Instruments Ltd., Worcestershire, UK). Air-dried soil subsamples were ground to fine powder (mesh number 100) with a mortar and pestle, and total C and total N were assayed using an elemental analyzer (vario EL III; Elementar, Hanau, Germany). Available phosphorus and available potassium were assayed using 0.5 mol/L sodium bicarbonate method and flame atomic absorption spectrophotometry. Soil bulk density was determined using the soils sampled by a stainless steel cylinder of 98.2 cm À3 (5 cm in diameter, 5 cm in height).

Statistical analyses
One-way ANOVAs and Duncan's multiple range tests were used to compare differences in soil physicochemical properties, environmental factors, and vegetation between no gazing and SG and steppe types. A t test was used to examine SG effects on the mean seasonal flux of CH 4 , CO 2 , and N 2 O in each year. The path analysis of structural equation model (SEM) was performed to analyze pathways that may explain environmental and vegetation effects on CH 4 , CO 2 , N 2 O fluxes and their relations under NG and SG in each steppe. Amos version 17.0.2 (Amos Development Corporation, Chicago, Illinois, USA) was used to parameterize the model. SigmaPlot 10.0 (Systat Software Inc., San Jose, California, USA) was employed to create figures.

Soil and vegetation characteristics
The DS and TS soils were significantly sandier than the meadow (Table 1). Soil bulk density was greatest in the desert site and lowest in the meadow site. Grazing did not affect soil bulk density.
Soil pH was close to neutral at the TS and slightly alkaline in the meadow and desert sites (Table 1). Grazing significantly decreased soil pH in the MS and significantly increased soil pH in the TS (Table 1). Soil total C and N were significantly higher in the MS (Table 1). Significant differences in the soil C/N ratio were detected between sites, and values were highest in the MS and lowest in the DS (Table 1). Again, no significant grazing effects on C/N ratio were found. Soil available N was significantly lower in the DS than in the MS and the TS (Table 1). Grazing significantly decreased available N in the TS and DS, but significantly increased available N in the MS. Grazing significantly decreased available potassium in the MS and DS (Table 1).
The MS and TS showed comparable MBC and MBN, while the DS had significantly lower values (Table 1). Grazing only affected soil MBN in the TS (Table 1).
Aboveground biomass in the MS was significantly greater than in other two sites. Grazing significantly reduced AGB in all sites (Table 1). Summer grazing increased BGB in the MS and significantly decreased BGB in the other two sites (Table 1).
Soil temperature was significantly lower in the TS compared to the MS and DS due to its higher altitude (Table 1). Grazing had no significant effects on ST. The TS and DS had similar SM values, which were significantly (P < 0.05) lower than that in the MS (Table 1). Grazing significantly decreased mean SM in the MS, but did not affect the other sites (Table 1).

Effects of grazing on CH 4 , CO 2 , N 2 O fluxes
Under NG, mean seasonal CH 4 uptake of the three years was greatest in the DS (75.8 AE 7.2 lgÁm À2 Áh À1 ) followed by the typical (64.4 AE 7.6 lgÁm À2 Áh À1 ) and finally the MS (47.1 AE 10.4 lgÁm À2 Áh À1 ; Fig. 1A-C). Summer grazing significantly decreased the mean seasonal CH 4 uptake rate by 43% in the MS, significantly increased it by 13% in the TS, and had little effects in the DS ( Fig. 1A-C).
Under NG, mean seasonal CO 2 emissions of the three years in the MS (311.4 AE 73.2 mgÁm À2 Áh À1 ) and TS (341.3 AE 93.0 mgÁm À2 Áh À1 ) were significantly higher than in the DS (212.1 AE 53.7 mgÁm À2 Áh À1 ; Fig. 1D-F). Summer grazing Notes: NG, no grazing; SG, summer grazing; BD, bulk density; TC, total carbon; TN, total nitrogen; C/N, total carbon/total nitrogen ratio; AP, available phosphorus; AK, available potassium; MBC, microbial biomass carbon; MBN, microbial biomass nitrogen; ST, soil temperature; SM, soil moisture; AGB, aboveground biomass; BGB, belowground biomass. Different lowercase letters in the same row indicate significant differences (P < 0.05) based on Duncan's multiple range tests. significantly enhanced the mean seasonal CO 2 emission rate by 12% in the MS and significantly reduced it by 30% in the TS and by 23% in DS (Fig. 1D-F).
Under NG, mean seasonal N 2 O emissions in the TS (9.2 AE 4.2 lgÁm À2 Áh À1 ) were slightly higher than those in the MS (4.7 AE 2.2 lgÁm À2 Áh À1 ) and the DS (6.3 AE 1.5 lgÁm À2 Áh À1 ; Fig. 1H-J). Summer grazing significantly enhanced the mean seasonal N 2 O emission rate by 79% in the MS, whereas no significant effects were detected in the TS and DS (Fig. 1H-J).
Relations among CH 4 , CO 2 , and N 2 O fluxes and their changes by grazing Under NG, the main driving factors of seasonal CH 4 uptake were ST, AGB, and SM in the MS, TS, and DS, respectively ( Fig. 2A-C). Summer grazing changed the driving factor from AGB to ST in the TS (Fig. 2E). No change was ❖ www.esajournals.org observed in the other two steppes (Fig. 2D, F). Soil temperature, SM, and AGB had significant effects on seasonal CO 2 emissions in the three steppes under NG and their total contribution to seasonal CO 2 emission decreased from the MS (R 2 = 0.84), to the TS (R 2 = 0.77), to the DS (R 2 = 0.31; Fig. 2A-C). Summer grazing altered the driving environmental factors on CO 2 flux in all three sites (Fig. 2D-F). Seasonal N 2 O emission was significantly related to SM and AGB in the MS, ST and AGB in the TS, and SM in the DS ( Fig. 2A-C). Summer grazing also altered the driving environmental factors on N 2 O flux in all three sites (Fig. 2D-F).
A significant positive correlation under NG was found between seasonal CO 2 and N 2 O emissions in the MS (r = 0.59, P < 0.01) and the DS (r = 0.44, P < 0.01), but not in the TS ( Fig. 2A-C). Summer grazing significantly decreased the magnitude of the correlation coefficients between seasonal CO 2 and N 2 O emissions in the DS, not in the MS and TS ( Fig. 2A-C). A significant negative relationship between seasonal CH 4 uptakes and N 2 O emissions was found in the MS (r = 0.32, P < 0.05) and the DS (r = 0.46, P < 0.01); SG significantly reduced the magnitude of the correlation coefficients in the both steppes ( Fig. 2A, C, D, F). In contrast, no significant relationship was detected between seasonal CH 4 uptake and N 2 O emission in the TS, regardless of grazing (Fig. 2B,  E). No significant relationship between seasonal CH 4 uptake and CO 2 emission was detected in all three steppes under NG and SG plots (Fig. 2).

Effects of grazing on soil and vegetation traits
Grazing significantly reduced soil water content in the MS likely as a result of the combined effects of enhanced soil surface evaporation induced by decreased vegetation cover and reduced infiltration caused by soil compaction (Wang et al. 2002, Zhong et al. 2014. Increases in soil pH in response to grazing at the typical and desert sites could be related to decreased litter accumulation and root biomass as well as urine and dung input (Table 1; Cui et al. 2005). In contrast, decreased soil pH value at the MS site could be caused by an increased root biomass and leguminous species leading to the formation of organic acids (Table 1; Brady and Weil 2002). Increased available soil N in the MS likely resulted from urine and dung inputs, an apparent increase in leguminous species, and significantly decreased SM which inhibits N absorption by plants Evans 1997, Hamilton andFrank 2001). Decreases in soil available N in the TS and DS could be caused by grazing-enhanced N absorption by plants and decreased leguminous species. N absorption by plants has been shown to increase in response to herbivory (Xu et al. 2008b). In addition, N in urine and dung at these steppes is more prone to volatilization under a dry and hot climate (Zhang et al. 2013). Significant decreases in soil inorganic N under heavy livestock grazing are well documented at TS sites (Wu et al. 2011) and DS (Su et al. 2005), whereas significant increases have been reported at MS (Han et al. 2008). Decreases in available soil potassium by grazing in all three steppe types could be attributable to increased K output via canopy biomass consumption by livestock.

Effects of grazing on CH 4 , CO 2 , and N 2 O flux
The mean seasonal CH 4 uptake rates (47.1-75.8 lgÁm À2 Áh À1 ) of the temperate steppe presented here fall within the range of 2-105 lgÁm À2 Áh À1 and keep in pace with a previous summary by Wang et al. (2014) who calculated a mean CH 4 uptake rate (standard deviation) of 59.62 AE 56.96 lgÁm À2 Áh À1 by obtained 465 sets from 50 articles in grasslands. For each steppe type in our study, the mean seasonal CH 4 uptake rate in the MS (47.1 lgÁm À2 Áh À1 ) was slightly higher than CH 4 mean value (28.45 AE 24.32 lgÁm À2 Áh À1 ) of 128 published data in meadow in China , while in the TS, it (64.4 lgÁm À2 Áh À1 ) was slightly lower than that mean value (71.45 AE 61.21 lgÁm À2 Áh À1 ) of published 337 data in China , and in the DS, it (75.8 lgÁm À2 Áh À1 ) was largely higher than that mean value (41.19 AE 26.01 lgÁm À2 Áh À1 ) of published 21 data in semiarid DS in China .
The mean seasonal CH 4 uptake was greatest in the DS followed by typical and finally the meadow site, which is consistent with previous studies (Tang et al. 2013). This can be largely attributed to the differences in soil water content and organic matter and partly to sand content, topsoil NH 4 + , and pH value (Wu et al. 2010, Fang et al. 2014. Higher SM is favorable for CH 4 production in soil but not for CH 4 oxidization and diffusion of atmospheric CH 4 into the soil (Koschorreck and Conrad 1993). High availability of NH 4 + can enhance ammonium-oxidizing bacteria at the expense of the CH 4 -oxidizing bacteria (Brady and Weil 2002); thus, higher SM and NH 4 + lead to lower CH 4 uptake in the MS (Table 1). Lower organic matter restricts CH 4 production and higher sand content is more conducive to CH 4 oxidation, which leads to a relatively higher magnitude of CH 4 uptake in the TS and DS soils (Wang and Han 2005).
The observed decrease in mean seasonal CH 4 uptake in the grazed MS resulted from decreased O 2 and CH 4 diffusion due to soil compaction by grazing and trampling. The significantly increased soil NH 4 + concentration, and significantly lowered pH value, suppressed CH 4 oxidation and uptake (Table 1; Wang andHan 2005, Fang et al. 2014). Significant increases in CH 4 uptake in the other two grazed steppes were more likely related to increased pH value and decreased soil NH 4 + concentration that favored CH 4 oxidation and uptake (Table 1; Fang et al. 2014).
The mean seasonal CO 2 emission rate (163-349 mgÁm À2 Áh À1 ) was similar to previous studies in Inner Mongolia grassland (Dong et al. 2005, Jia et al. 2007). For each steppe type, the mean seasonal fluxes of the three years in the MS (349 mgÁm À2 Áh À1 ) were similar to previous studies (352.3 mgÁm À2 Áh À1 ) by Dong et al. (2005). The mean seasonal CO 2 emission rate (341.3 mgÁm À2 Áh À1 ) in the TS was greater than those found in previous studies in other adjacent experiment sites with range of 206.7-267.7 mgÁm À2 Áh À1 (Li et al. 2000, Jia et al. 2007). In the DS, the mean seasonal CO 2 emission rate (163 mgÁm À2 Áh À1 ) fall within the range of 150-253 mgÁm À2 Áh À1 in the similar experiment sites in Inner Mongolia grassland (Zhang et al. 2014a, b).
Differences in mean seasonal CO 2 emissions between steppe types were largely attributable to the differences in soil organic matter, soil water content, and live root biomass (Table 1; Li et al. 2000, Zhang et al. 2003. Significantly reduced mean seasonal CO 2 emissions by SG in the TS and DS could be the result of significant decreases in litter accumulation and root biomass as previous studies have suggested (Wan and Luo 2003, Bahn et al. 2008. In contrast, the mechanism for grazing-enhanced CO 2 emission in the MS involves increased root biomass, N availability, and lowered pH values (Table 1).
The mean seasonal N 2 O emission rates (4.7-9.2 lgÁm À2 Áh À1 ) were in accordance with previous studies in Inner Mongolia grassland (Du and Chen 1997, Liu et al. 2015. In the MS, the mean seasonal value of N 2 O flux was similar to previous studies by Lu et al. (2015). In the TS, previous studies indicated that the mean seasonal values in L. chinensis were 6.9 and 7.9 lgÁm À2 Áh À1 Chen 1997, Chen et al. 2000), slightly lower than that in our studies (9.2 AE 4.2 lgÁm À2 Áh À1 ).
The significantly higher mean seasonal N 2 O emission in the TS coincided with much higher contents of MBC, MBN, and available ammonium and nitrate (Table 1). The significantly lower soil pH and the mean seasonal temperature in the TS also played an important role, as supported by the significant negative relationship between seasonal N 2 O emission and ST observed (Table 1, Fig. 2B, E).
The significantly enhanced mean seasonal N 2 O emissions with SG in the MS can be explained by the following three soil water status-dependent mechanisms (Schrama et al. 2013). First, due to the wetter soil at the meadow site, input of dung and urine by livestock may provide more substrates for nitrification and denitrification Oenema 1995, Oenema et al. 1997). Second, soil compaction by livestock can enhance N 2 O emission due to decreased aeration of the soil (Clayton et al. 1997). Third, grazing significantly reduced SM leading to decreased N absorption by plants in favor of microbial denitrification (Brady and Weil 2002). This mechanism could also explain the lack of grazing effects on N 2 O emissions in the TS and DS. As a rule, N in dung and urine is more prone to ammonia volatilization in hot and semiarid climates and compaction has minor effects on aeration in sandy soils. In addition, competitive N absorption between plants and microbes at these N-limited sites may have constrained N 2 O-producing processes such as nitrification and/or denitrification (Brady and Weil 2002).
Our study indicated that GHG fluxes varied among steppe types, which were due to not only large differences in annual precipitation, annual temperature, but also differences in vegetation composition and soil type. Some previous studies also indicated that vegetation and soil type can play important roles in GHG fluxes for certain types of steppe (Du and Chen 1997, Wang et al. 1997, Dong et al. 2000. Vegetation and soil type can affect soil microbial composition, diffusion of O 2 and GHGs, and ST and SM Chen 1997, Dong et al. 2005). The effects of vegetation and soil type on GHGs will be considered in next studies.
Correlations among GHG fluxes by steppe type and grazing A significant positive correlation was found between seasonal CO 2 and N 2 O emissions in the MS and DS ( Fig. 2A, C), which is previously reported trend in several MS communities ❖ www.esajournals.org (Dong et al. 2000, Holst et al. 2008. This can be explained by SM and AGB, which are driving factors for the fluxes of both gases ( Fig. 2A, C;Yao et al. 2010). Mechanistically, CO 2 production derives from root respiration and microbial respiration via the aerobic microbial decomposition of the soil organic matter (Xu et al. 2008a, Zhang et al. 2014a. N 2 O fluxes are produced directly by the anaerobic denitrification of nitrates derived from mineralization, whereas aerobic organisms decompose organic N (95-99% of total soil N) largely in protein form (Brady andWeil 2002, Zhang et al. 2014a, b). Plant residues, soil organic matter, and microbial biomass stoichiometrically regulate both processes forming the basis for the relationship between CO 2 and N 2 O fluxes in terrestrial ecosystems (Xu et al. 2008a). CO 2 and N 2 O production were derived mainly from the decomposition of organic matter in NG plots, thus explaining positive linear relationship between CO 2 and N 2 O fluxes. Summer grazing significantly decreased the magnitude of the correlation coefficients between CO 2 and N 2 O emissions in the DS (Fig. 2C, E), but did not in the MS ( Fig. 2A, D). According to results of the SEM (Fig. 2), SG did not change the driving factors affecting CO 2 and N 2 O emissions in MS, and the increase in root biomass, total C, NH 4 + , and NO 3 À by SG increased CO 2 and N 2 O emissions (Table 1, Fig. 1). It is well documented that the denitrification of excess soil nitrate can greatly enhance the production of CO 2 from organic C (Brady andWeil 2002, Zhang et al. 2014a, b). In addition, high soil N has been widely reported to shorten the root life span of grassland species, enhancing C turnover (Bai et al. 2008, Zhang et al. 2014a. In the DS, SG decreased SM effects on N 2 O emissions and increased AGB effects on CO 2 emissions, both of which lead to a decrease in the magnitude of the correlation coefficients between CO 2 and N 2 O emissions. In addition, the decrease in total C, total N, NH 4 + , NO 3 À , and root biomass by SG increased the competition for these nutrients (Zhang et al. 2014b). No significant relationship was found between CO 2 and N 2 O emissions in the TS likely because N 2 O and CO 2 emissions had different driving factors (Fig. 2B). The variation in water condition between years may confuse the relationship between CO 2 and N 2 O emissions in the TS. A significant correlation between ST and AGB was the main reason for the significant negative relationship between CH 4 uptake and N 2 O emissions. Summer grazing decreased the magnitude of the correlation coefficients by decreasing the relationship between ST and AGB. However, SG improved the effect of SM on N 2 O emission in the MS ( Fig. 2A, D). Mechanistically, CH 4 flux is the function of balance between the consumption by methanotrophic microbes and the production by methanogenic microbes (Hou et al. 2012. N 2 O is emitted from soils by microbialmediated nitrification and denitrification under aerobic and anaerobic conditions, respectively (Zhang et al. 2014b; ST had significant effects on microbial activities and metabolism processes (Dunfield et al. 1993). In addition, ST affects root activity and their ability to absorb nutrients, which provides nutrients to the shoot biomass. The increase in NH 4 + and NO 3 À contents by SG (Table 1), increase in N 2 O emissions and decrease in CH 4 uptake (Fig. 1A, D), via inhibition of CH 4 oxidation by CH 4 -consuming bacteria, and increase in N 2 O production were observed (Zhang et al. 2014a, b). In the DS, SM was the main factor regulating relationship between CH 4 uptake and N 2 O emissions and SG decreased the magnitude of the correlation coefficients by strengthening the effect of AGB on N 2 O emissions in DS (Fig. 2C, F). In drought conditions, SM has a significant effect on microbial activities and nutrition movements related to CH 4 oxidation and N 2 O production . Thus, the mechanisms that regulate the relationship between CH 4 uptake and N 2 O emission were different among steppes and SG effects were also different.
The lack of significant seasonal correlations between CO 2 emission and CH 4 uptake in all steppes is consistent with previous study (Zhang et al. 2014b). The driving factors were different for both gas fluxes among steppes and grazing (Fig. 2). All ST, SM, and AGB had significant effects on CO 2 emission, but only ST had a significant effect on CH 4 uptake in the MS. Limited SM, a major controlling factor for the seasonal dynamics of soil respiration but not of CH 4 uptake, can explain this observation (Zhang et al. 2014b). Grazing had minor impacts on this correlation and the seasonal pattern of SM and relevant relations to CO 2 emission and/or CH 4 uptake.
We found a significant positive relationship between seasonal CO 2 and N 2 O emissions, and a significant negative relationship between CH 4 uptake and N 2 O emission in the MS and DS. No significant relationships were found between CH 4 uptake and CO 2 emission. Summer grazing did not change the relationship between CO 2 and N 2 O emissions in the MS, but decreased the magnitude of the correlation coefficients in the DS. The negative relationship between CH 4 uptake and N 2 O emission was due to the significant relationship between ST and AGB in the MS, and to the significant effects of SM on both GHG fluxes in the DS. Summer grazing decreased the magnitude of the correlation coefficients between CH 4 uptake and N 2 O emission in the MS by decreasing the significant relationship between ST and AGB. Grazing enhanced SM effects on N 2 O emissions and enhanced AGB effects on N 2 O emission in the DS.

ACKNOWLEDGMENTS
We thank JB Gu, LH Li, and X Li for their substantial help in building field facilities and conducting field surveys, and Duolun Restoration Ecology Research Station of Institute of Botany, the Chinese Academy of Sciences, for providing access to the historical data and laboratories. We would also like to thank Alison Beamish at the University of British Columbia for her assistance with English language and grammatical editing of the manuscript. Finally, we thank the editor and anonymous reviewers for their thoughtful comments. Huiqiu Shi, Longyu Hou, and Liuyi Yang conducted all field work, carried out laboratory work, participated in data analysis, and drafted the manuscript; Dongxiu Wu, Lihua Zhang, and Linghao Li participated in the design of study, coordinated the study, and helped draft the manuscript. All authors gave final approval for publication. This work was funded by the National Natural Science Foundation of China (41273090, 31130008, and 31229001) and the National Key Research and Development Plan (2016YFC0500600).