Effects of Grazing Pattern on Ecosystem Respiration and Methane Flux in a Sown Pasture in Inner Mongolia, China

Abstract

The establishment of sown pasture is an important agricultural practice in many landscapes. Although both native grassland and sown pasture play a key role in the global carbon cycle, due to lack of data and field experiments, our understanding of grassland CH₄ fluxes and CO₂ emissions remains limited, especially when it comes to sown pasture. We measured ecosystem respiration and CH₄ fluxes in response to a variety of potential drivers (soil temperature, soil moisture, ammonium nitrogen, nitrate nitrogen and dissolved organic carbon) in CG (continuous grazing), RG (rotational grazing) and UG (ungrazed) plots in sown grassland for one year in Inner Mongolia. Fluxes of CH₄ and ecosystem respiration were measured using static opaque chambers and gas chromatography. Grazing significantly reduced ecosystem respiration (p < 0.01), and grazing pattern significantly influenced respiration in CG and RG plots (p < 0.01). We find that the sown grassland is a net sink for atmospheric CH₄. No influence of grazing pattern was observed on CH₄ flux in CG, RG and UG (p > 0.05). Soil temperature is the most important factor influencing ecosystem respiration and CH₄ flux in the sown grassland, with soil moisture playing a secondary role to soil temperature. Variation in levels of ammonium nitrogen, nitrate nitrogen and dissolved organic carbon had little influence on ecosystem respiration or CH₄ flux (except in UG plots). The values obtained for ecosystem respiration of grasslands have a large uncertainty range, which may be due to spatial variability as well as differences in research methods. Mean CH₄ fluxes measured only during the growing season were much higher than the annual mean CH₄ fluxes.

1. Introduction

Methane (CH₄) and carbon dioxide (CO₂), two of the three major greenhouse gases (CH₄, CO₂, N₂O) play a significant role in the radiative balance of the earth’s atmosphere. Atmospheric concentrations of CH₄ and CO₂ have grown by 150% and 40% respectively from about 1750 to 2011 and this increase is the main driver of climate change [1].

Semiarid grasslands, which cover about 11 percent of the global land surface, act as important sources or sinks of greenhouse gases and are widely used as pasture [2,3,4,5]. Although pastures cover a large fraction of Earth’s ice-free land (about 26%) [6], because of increases in human population and overgrazing, degradation and fragmentation of steppe are inevitable [7]. Grazing intensity is currently increasing and is likely to continue rising as demand for animal products increases. However, at some point in the future, because of the degradation and fragmentation of native grassland, reduction of grazing intensity on native grassland is likely to occur and the establishment of sown pasture will become essential, as it can alleviate not only the grazing pressure on native grassland but also the imbalance in grassland use in terms of space and time [8,9]. Though both native grassland and sown pasture play a key role in the global carbon cycle [10,11,12], due to lack of data and field experiments, our understanding of grassland CH₄ fluxes and ecosystem respiration remains limited, especially for sown pasture.

Grasslands in Inner Mongolia are representative of the Eurasian grassland belt [13] and grazing is the dominant land use there [14]. Current studies show that grassland soils are significant sinks for atmospheric CH₄ and sources for CO₂ [15,16]. However, the potential of Inner Mongolian grassland soils to act as sources or sinks may be affected by grazing regime [17,18,19]. The response of CH₄ fluxes and ecosystem respiration when pastures receive increased grazing pressure or changed grazing methods remains uncertain. Nor are the effects of soil moisture, soil temperature and soil disturbance, which have been identified as important factors controlling the soil-atmosphere exchange of CH₄ and CO₂ well documented for grazing lands [20,21,22]. Grazing regime as well as changes of soil moisture, soil temperature and soil disturbance will increase uncertainties in the measurement on the CH₄ flux and ecosystem respiration of sown pasture.

In this study, we measure CH₄ flux and ecosystem respiration (Re) over an entire year (from July 2012 to July 2013) from a sown pasture in Inner Mongolia in order to investigate the effects of grazing pattern as well as soil moisture, soil temperature and other soil properties on soil-atmosphere exchange of CH₄ and ecosystem respiration.

2. Materials and Methods

2.1. Site Description

This study was conducted in a sown pasture belonging to the Chinese Academy of Agricultural Science, located in Shaerqin town, Tumd Left Banner, Huhhot, Inner Mongolia, China (40°34′ N, 111°34′ E, 1055 m). It has a typical temperate semi-arid climate. Mean annual rainfall ranges from 350 to 450 mm and mean annual air temperature ranges from 5.8 to 7.3 °C, with a maximum monthly mean of 23.3 °C in July and a minimum of −11.0 °C in January, with ≥10 °C accumulated temperature of above 2700 °C and frost-free period of around 130 d (The weather bureau of Hohhot, China). Soils are Castanozems in the “Food and Agriculture Organization of the United Nations (FAO) nomenclature”. Soil properties of the upper 300 mm of the soil profile were: pH of 8.5, salt content 0.03%, total N 0.035%, organic matter of 0.8% (Institute of Grassland Research of Chinese Academy of Agricultural Sciences). The experimental site was built and fenced in 2008 and then a mixture of Medicago sativa, Lespedeza floribunda, Leymus chinensis, Elymus dahuricus Turcz and Bromus inermis Leyss was sown in 2009. Grazing by sheep began in 2010.

2.2. Experimental Design

The sown pasture was divided into 3 equal transects, with each transect including 7 experimental plots. Within these 7 plots, there are 5 rotational grazing plots (RG), 1 continuous grazing plot (CG) and 1 ungrazed plot (UG) assigned randomly. Thus, there are in total 15 rotational grazing plots, 3 continuous grazing plots and 3 ungrazed plots in the 3 transects. We randomly chose 6 of the 15 rotational grazing plots as our experimental sites (Figure 1). Each plot covered about 0.67 ha. To ensure the spatial representativeness of the greenhouse gas measurements, there were two sampling points in every UG and CG plot and there was one sampling point in every RG plot that we chose. Thus we had 6 replicate sampling points for each treatment (UG, CG and RG).

Grazing was allowed from June to October since 2010 and the sheep stayed in the winter sheepfold the rest of year. The grazing treatments were set by Institute of Grassland Research of Chinese Academy of Agricultural Sciences, as they represent the traditionally used grazing practices in this region. During the grazing period, the 30 sheep were grazed in every RG plot rotationally (rotational grazing in five plots, one plot for 6 days, so it has a whole rotational grazing every 30 days); the 6 sheep were continuously grazed in every CG plot; and in the meantime grazing was forbidden in all UG plots. All factors considered, the grazing rate in CG, which was consistent with that in RG, was 9 sheep per hectare per 5 grazing months per year. Management details and soil properties were summarized in

Table 1. Experiment management details and soil properties.

Experiment Site	Number of Sampling Points	Duration of Grazing Period (d)	Grazing Rate (sheep ha−1)	pH	SOM 0–20 cm (g kg−1)	SAN 0–20 cm (g kg−1)	SBD (0–10 cm) (g/m3)
RG1.1	1	30	9	8.58 ± 0.11	13.54 ± 1.63	0.07 ± 0.01	1.51
RG1.2	1	30	9	8.28 ± 0.16	12.84 ± 3.20	0.07 ± 0.02	1.51
RG2.1	1	30	9	8.46 ± 0.14	9.92 ± 0.76	0.06 ± 0.02	1.40
RG2.5	1	30	9	8.32 ± 0.17	9.45 ± 1.67	0.06 ± 0.02	1.40
RG3.2	1	30	9	8.33 ± 0.03	14.93 ± 3.11	0.07 ± 0.01	1.55
RG3.3	1	30	9	8.08 ± 0.57	13.36 ± 3.65	0.08 ± 0.03	1.55
CG1	2	150	9	8.33 ± 0.47	16.9 ± 6.45	0.07 ± 0.02	1.65
CG2	2	150	9	8.50 ± 0.27	10.63 ± 3.06	0.05 ± 0.01	1.46
CG3	2	150	9	8.24 ± 0.30	12.03 ± 3.44	0.04 ± 0.04	1.49
UG1	2	-	-	8.53 ± 0.13	12.08 ± 0.67	0.07 ± 0.01	1.39
UG2	2	-	-	8.56 ± 0.11	10.01 ± 1.60	0.05 ± 0.01	1.56
UG3	2	-	-	8.57 ± 0.13	16.20 ± 5.04	0.08 ± 0.02	1.46

SOM: Soil Organic Matter; SAN: Soil Available Nitrogen; SBD: Soil Bulk Density; The given data represent the mean ± standard error.

.

2.3. Gas Flux Measurements

Fluxes of CH₄ and ecosystem respiration (Re, CO₂) were measured using static opaque chambers and gas chromatography [23]. The static chambers were made of stainless steel (thickness = 1 mm) and consisted of two parts. The first part was a pentahedral chamber (length × width × height = 0.4 m × 0.4 m × 0.4 m) with removable bottom, covered with a 3 cm thick layer of foam insulation in order to avoid plant photosynthesis and prevent heat exchange inside and outside the chamber. The chamber was fitted with a gas balance tube, a digital display thermometer and a gas samples interface. The second part of the static chamber was a square base frame without a top and bottom (length × width × height = 0.4 m × 0.4 m × 0.2 m). In the six RG plots one base frame was inserted into soil and in every CG two base frames were installed but at least 10 m apart and were observed simultaneously. Settings in UG were the same way as in CG. Meanwhile, in order to reduce direct destruction and disturbance to the sampling area, all base frames were installed at least one week before sampling. During every sampling time, five air samples were collected from the closed chambers at fifteen-minute intervals using 60 mL plastic syringes only from 9:00 to 13:00. On the basis of variation in concentrations over time of 5 gas samples, we used a nonlinear fitting method to obtain gas flux rates at the outset of the measurement [24,25]. In order to avoid pressure changes in the chamber when sampling, the chambers included gas balance tubes. We collected the air samples once a week in the growing season, twice a week during periods of freeze-thaw (around March) and twice a month during periods when the soil was frozen.

Concentrations of CH₄ and CO₂ were analyzed by a gas chromatograph (Agilent 7890A) equipped with a flame ionization detector (FID) operated at 200 °C. CH₄ was detected directly by FID through a 2 m × 2 mm stainless steel column packed with 13XMS (60/80). CO₂ was separated from other components in a 2 m × 2 mm stainless steel column packed with Porapak Q (60/80) and entered a nickel catalytic converters (375 °C) where it was converted by hydrogen into CH₄ so that it could be detected by FID. The column oven temperature was 55 °C and the carrier gas was N₂ (99.999%) flowing at 30 mL min⁻¹; combustion gas was H₂ (99.99%) flowing at 35 mL min⁻¹; assistant combustion gas was air flowing at 400 mL min⁻¹. All gas samples were analyzed within 24 h of sampling.

2.4. Auxiliary Measurements

Soil temperature (at 5 cm depth) and soil moisture (0–6 cm) were measured by digital thermocouples (JM624, Liwen Electronics LTD, Tianjin, China) and a portable moisture probe meter (MPKit, Ruidisheng Science and Technology LTD, Nanjing, China) during the gas sample collection. Daily precipitation and air temperature were obtained from the local meteorological station. During the winter time, when soil was frozen, samples of the soil layer (0–6 cm) were taken back to laboratory and dried at 105 °C for 24 h to determine the gravimetric water content. Then both volumetric water content and gravimetric water content were converted into water filled pore space (WFPS, %), calculation as follows:

Volumetric water content = gravimetric water content × soil bulk density

Total porosity = 1 − soil bulk density/2.65

WFPS (%) = volumetric water content /total porosity

Soil samples at the depth of 0–15 cm were collected during the gas sampling process and were taken back to the laboratory to measure NH₄⁺-N, NO₃-N (1 mol/L KCL extraction) and DOC (dissolved organic carbon, water-extraction). The bulk density, soil organic matter and soil available nitrogen were measured once a year and soil properties are summarized in Table 1.

2.5. Statistical Analysis

For statistical analysis and figure preparation, we used SPSS 20.0 and Origin 8.5 (Origin Lab Corporation, Northampton, MA, USA). To detect whether different treatments (RG, CG and UG) brought significant differences to CH₄ flux and ecosystem respiration, an ANOVA was employed. Pearson correlation analysis was used to analyze the correlation between CH₄ flux, ecosystem respiration and their influencing factors.

3. Results

3.1. Environmental Factors

Figure 2 shows that precipitation, air temperature, soil temperature and soil moisture all have a distinct seasonal variation. The annual precipitation was 581 mm and the annual mean air temperature was 7.3 °C during the study period (from 5 July 2012 to 9 July 2013). The annual precipitation was above the longer-term average value (400 mm) and 82% of it occurred in June-September. The annual mean air temperature was slightly higher than the historic mean, with the maximum daily mean air temperature (26.5 °C) and minimum daily mean air temperature (−22.6 °C) occurring in July and January, respectively. We obtained the daily mean air temperature and daily precipitation data for the whole year of 2012 and 2013 from the weather bureau of Hohhot, China. We find that there is no significant difference between the two year’s daily mean air temperature (p = 0.47) and daily precipitation (p = 0.75).

There was a similar pattern for the soil temperature in CG, RG and UG plots. There were no significant differences in soil temperature among the three treatments (CG, RG and UG) (p > 0.05) (Figure 2b). Annual mean soil moisture was also not significantly different between CG, RG and UG plots (p > 0.05, WFPS %), with 41.6% in CG, 46.4% in UG and 44.0% in RG (Figure 2c). As can be seen in Figure 2, from November 2012 to February 2013 soil moisture was the highest of the entire period of observation while soil temperature dropped under 0 °C and precipitation barely occurred. In order to gain a better understanding of the seasonal variation of greenhouse gas emissions during the period of observation, we divided the whole year into three periods: the growing season (from July 2012 to October 2012 and from May 2013 to July 2013), the freezing period (when the water was frozen in the soil, from November 2012 to February 2013), and the freezing-thawing period (from March 2013 to April 2013).

3.2. Ecosystem Respiration

The ecosystem respiration of all plots ranged from 1.6 to 617 mg C m⁻² h⁻¹, with the peak value occurring on 2 July 2013 in UG and the lowest value recorded on 2 February 2013 in CG. During the observation period, there were significant differences between CG and UG (p < 0.01), RG and UG (p < 0.01), RG and CG (p < 0.01) in CO₂ emission. Grazing did not change the seasonal pattern of CO₂ emission in CG, RG and UG (Figure 3a). The annual mean CO₂ emissions for CG, UG and RG were 119.9 ± 108.7 mg C m⁻² h⁻¹, 189.8 ± 185.6 mg C m⁻² h⁻¹ and 134.0 ± 124.8 mg C m⁻² h⁻¹, respectively. The mean CO₂ emission in UG was 62.6% and 44.1% higher than that in CG and RG during the growing season and 43.2% and 38.3% higher during the freezing period and 14.6% and 13.1% higher during the freezing-thawing period (

Table 2. Seasonal and annual mean cumulative of ecosystem respiration (Re) and CH₄ flux.

Treatment	Season	Re			CH4 Flux
		Mean a (mg C m−2 h−1)	CV	Cumulative (t C ha−1)	Mean a (μg C m−2 h−1)	CV	Cumulative (kg C ha−1)
CG	GS	182.0 ± 98.1	0.54	8.10	−31.7 ± 8.8	0.28	−1.40
	FP	4.6 ± 2.7	0.59	0.15	−11.3 ± 7.6	0.67	−0.29
	FTP	50.7 ± 37.3	0.74	0.81	−20.36 ± 3.6	0.18	−0.33
	Annual	119.9 ± 108.7	0.90	8.97	−25.4 ± 11.1	0.44	−2.02
UG	GS	295.9 ± 170.6	0.58	12.96	−34.6 ± 10.3	0.30	−1.50
	FP	8.1 ± 4.6	0.57	0.28	−13.8 ± 7.3	0.53	−0.39
	FTP	59.4 ± 54.0	0.91	1.08	−21.9 ± 6.5	0.30	−0.36
	Annual	189.8 ± 185.6	0.98	14.32	−27.9 ± 12.3	0.44	−2.25
RG	GS	205.4 ± 114.0	0.56	9.00	−29.5 ± 8.5	0.29	−1.28
	FP	5.0 ± 2.9	0.57	0.17	−13.9 ± 4.3	0.31	−0.43
	FTP	51.6 ± 36.6	0.91	0.83	−21.3 ± 6.2	0.29	−0.34
	Annual	134.0 ± 124.8	0.93	10.0	−25.1 ± 9.5	0.38	−2.05

GS: growing season (from July 2012 to October 2012 and from May 2013 to July 2013). FP: freezing period (from November 2012 to February 2013). FTP: freezing and thawing period (from March to April 2013). Annual (from July 2012 to July 2013). ^a: mean ± stdev. CV: Coefficient of Variation. Re: ecosystem respiration.

). Ecosystem respiration in the growing season accounted for 88.9% (in CG), 90.3% (in UG) and 90% (in RG) of the annual emission, respectively.

Soil temperature and soil moisture are both important in controlling CO₂ production but their importance varies with different observation periods. According to the regression equations in

Table 3. Stepwise regression equations of Re (ecosystem respiration) and main environmental factors in CG, UG and RG.

Site	Season	Regression Equation	F	Sig.	R2
CG	GS	y = −129.41 + 10.937x1 + 7.604x2	20.847	<0.001	0.654
	FP	y = 16.861 − 0.363x2	51.601	<0.001	0.910
	FTP	y = 6.307 + 6.507x1	55.066	<0.001	0.871
	Annual	y = −78.973 + 10.635x1 + 4.377x2	55.665	<0.001	0.766
UG	GS	y = −132.329 + 18.836x1 + 7.238x2	19.358	<0.001	0.671
	FP	y = 20.954 − 2.231x3	11.806	<0.05	0.730
	FTP	y = −121.547 + 9.7x1 + 22.571x3 + 0.791x5	914.285	<0.001	0.997
	Annual	y = −111.769 + 18.622x1 + 5.723x2	61.013	<0.001	0.795
RG	GS	y = −22.394 + 12.249x1	13.749	<0.001	0.478
	FP	-	-	-	-
	FTP	y = 5.09 + 7.798x1	49.791	<0.001	0.859
	Annual	y = −79.807 + 12.077x1 + 4.061x2	34.881	<0.001	0.708

x₁: soil temperature (°C), x₂: soil moisture (WFPS, %), x₃: NH₄⁺-N (mg N kg⁻¹ dry soil), x₄: NO₃-N (mg N kg⁻¹ dry soil), x₅: DOC (mg C kg⁻¹ dry soil). GS: growing season (from July 2012 to October 2012 and from May 2013 to July 2013). FP: freezing period (from November 2012 to February 2013). FTP: freezing and thawing period (from March to April 2013), Annual (from July 2012 to July 2013). “-”: no fitting regression equation was found.

, soil temperature was the primary environmental factor that determined CO₂ emission in the growing season, in the freezing-thawing period and on a full year scale (CG, UG, RG). Soil moisture became relevant to CO₂ emission on a full year scale (CG, UG, RG) and in the growing season (CG, UG) (Table 3). The data for that time frame are reflected in the Figure 2 and Figure 3. Meanwhile, NH₄⁺-N, NO₃⁻-N and DOC seemed to have little effect on CO₂ emissions except for the UG treatment (Table 3).

As

Table 4. Correlation between Re (ecosystem respiration) and main environmental factors.

Site	Season		SM (WFPS, %)	ST (5 cm, °C)
CG	GS (n = 26)	R2	0.201 * 89.561 + 6.497x	0.485 ** 41.649e0.070x
		f(x)
	FP (n = 8)	R2	0.915 ** 16.644 − 0.355x	0.219 6.371 + 0.386x
		f(x)
	FTP (n = 10)	R2	0.546 * 115.598 − 9.128x + 0.19703x2	0.871 ** 3.4 + 6.712x
		f(x)
	Annual (n = 44)	R2	0.172 ** 56.275 + 11.706x − 0.333x2	0.840 ** 11.197e0.137x
		f(x)
UG	GS (n = 26)	R2	0.093 172.168 + 8.242x	0.602 ** 56.944e0.085x
		f(x)
	FP (n = 8)	R2	0.564 ** 25.714 − 0.502x	0.548 * 12.536 + 0.905x
		f(x)
	FTP (n = 10)	R2	0.432 * 208.147 − 16.069x + 0.336x2	0.83 ** 7.271 + 10.052x
		f(x)
	Annual (n = 44)	R2	0.135 * 113.743 + 15.923x − 0.476x2	0.882 ** 21.889e0.136x
		f(x)
RG	GS (n = 26)	R2	0.187 * 102.403 + 7.022x	0.506 ** 41.811e0.076x
		f(x)
	FP (n = 8)	R2	0.576 * 19.125 − 0.412x	0.861 ** 8.7856 + 0.817x
		f(x)
	FTP (n = 10)	R2	0.271 134.223 − 9.4x + 0.195x2	0.869 ** 7.526 + 7.539x
		f(x)
	Annual (n = 44)	R2	0.163 ** 49.172 + 14.775x − 0.419x2	0.850 ** 11.359e0.144x
		f(x)

f(x): Re (ecosystem respiration), ST: soil temperature, SM: soil moisture, GS: growing season (from July to October 2012 and from May to July 2013). FP: freezing period (from November 2012 to February 2013). FTP: freezing and thawing period (from March 2013 to April 2013), Annual (from July 2012 to July 2013). **: a significance level of 0.01, *: a significance level of 0.05.

shows, CO₂ emissions were positively affected by soil temperature (p < 0.01) in all the treatments except the freezing period in CG. There was a significant positive linear relationship between CO₂ emission and soil moisture in CG and RG during the growing season and there was a negative linear relationship in UG, CG and RG during the freezing period and those two showed a significant nonlinear correlation in UG and CG during the freezing-thawing period (Figure 4). For the whole year, a nonlinear correlation illustrated that the annual CO₂ emission reached a peak while soil moisture was around 42.8%, 40.6% and 42.8% (WFPS) in CG, UG and RG, respectively (Table 4 and Figure 4).

3.3. Methane Fluxes

The CH₄ fluxes from CG, UG and RG were nearly all negative, meaning that the pasture was a net sink for atmospheric CH₄. The annual mean CH₄ fluxes for CG, UG and RG were −25.4 ± 11.0 μg C m⁻² h⁻¹, −27.9 ± 12.2 μg C m⁻² h⁻¹ and −25.1 ± 9.4 μg C m⁻² h⁻¹ (mean ± stdev), respectively (Table 2). The CH₄ fluxes were not significantly affected by the three different grazing patterns (CG, UG and RG) (p > 0.05) during the observation period (Figure 3b). The highest measured CH₄ flux (−81.6 μg C m⁻² h⁻¹) occurred on July 5, 2012 in CG, while the minimum CH₄ flux (−1.3 μg C m⁻² h⁻¹) occurred on 9 January 2013 in CG. Comparing to the freezing period and freezing-thawing period, CH₄ fluxes were highest in the growing season (Table 2). During the growing season, CH₄ uptake accounted for 70% (CG), 65.2% (UG) and 61.9% (RG) of the annual uptake, respectively.

According to stepwise regression analysis, soil temperature was the most important environmental factor driver of temporal variability in CH₄ fluxes (

Table 5. Stepwise regression equations between CH₄ fluxes and environmental factors.

Site	Season	Regression Equation	F	Sig.	R2
CG	GS	y = −17.092 − 0.766x1	9.051	<0.01	0.277
	FP	y = −20.102 − 1.809x1	11.174	<0.05	0.670
	FTP	-	-	-	-
	Annual	y = −15.874 − 0.799x1	50.524	<0.001	0.579
UG	GS	-	-	-	-
	FP	-	-	-	-
	FTP	y = −5.427 − 0.805x1 − 0.213x4	22.313	<0.01	0.859
	Annual	y = −35.084 + 3.189x3 − 0.367x1	13.658	<0.001	0.485
RG	GS	y = −13.105 − 0.987x1	16.377	<0.001	0.490
	FP	-	-	-	-
	FTP	y = −29.6 + 0.567x2	7.814	<0.05	0.460
	Annual	y = −16.149 − 0.825x1	41.391	<0.001	0.591

y: CH₄ fluxes, x₁: soil temperature (°C), x₂: soil moisture (WFPS, %), x₃: NH₄⁺-N (mg N kg⁻¹ dry soil), x₄: NO₃-N (mg N kg⁻¹ dry soil), x₅: DOC (mg C kg⁻¹ dry soil). GS: growing season (from July 2012 to October 2012 and from May 2013 to July 2013). FP: freezing period (from November 2012 to February 2013). FTP: freezing-thawing period (from March to April 2013), Annual (from July 2012 to July 2013). “-”: no fitting regression equation was found.

), explaining 45.2%–63.9% of the annual variation in the three grazing pattern plots (

Table 6. Correlations between CH₄ flux and main environmental factors.

Site	Season		SM (0–6 cm, WFPS %)	ST (5 cm, °C)
CG	GS (n = 26)	R2	0.117 −21.475 − 1.563x + 0.055x2	0.268 ** −18.015 − 0.739x
		f(x)
	FP (n = 8)	R2	0.635 * −40.182 + 0.855x	0.557 * −17.965 − 1.476x
		f(x)
	FTP (n = 10)	R2	0.381 * −29.798 + 0.193x	−0.125 -
		f(x)
	Annual (n = 44)	R2	0.382 ** −25.066 − 0.785x + 0.032x2	0.639 ** −15.61 − 0.836x
		f(x)
UG	GS (n = 26)	R2	0.193 * −20.791 − 2.805x + 0.102x2	0.014 −27.683 − 0.399x
		f(x)
	FP (n = 8)	R2	0.258 −35.149 + 0.61x	0.21 −18.905 − 1.047x
		f(x)
	FTP (n = 10)	R2	−0.123 -	0.255 −16.673 − 0.786x
		f(x)
	Annual (n = 44)	R2	0.374 ** −28.099 − 0.829x + 0.033x2	0.452 ** −19.671 − 0.804x
		f(x)
RG	GS (n = 26)	R2	−0.086 -	0.257 ** −16.383 − 0.709x
		f(x)
	FP (n = 8)	R2	−0.057	−0.157
		f(x)	-	-
	FTP (n = 10)	R2	0.697 ** −30.677 + 0.671x	0.588 ** −16.464 − 1.051x
		f(x)
	Annual (n = 44)	R2	0.169 ** −28.545 − 0.08x + 0.012x2	0.558 ** −16.929 − 0.684x
		f(x)

f(x): CH₄ fluxes, ST: soil temperature, SM: soil moisture. GS: growing season (from July to October 2012 and from May to July 2013). FP: freezing period (from November 2012 to February 2013). FTP: freezing and thawing period (from March 2013 to April 2013). Annual (from July 2012 to July 2013). **: a significance level of 0.01, *: a significance level of 0.05, “-”: no fitting regression equation was found.

). Annual CH₄ fluxes in the three treatments plots were strongly negatively correlated with soil temperature (p < 0.01). However, as Figure 5 shows, instead of a linear relation, annual CH₄ flux showed a concave-shaped relationship with soil moisture, namely, for each peak of CH₄ absorption in CG, RG and UG there is a distinct optimum soil moisture value (30.0% in CG, 32.4% in RG and 30.7% in UG) (Table 6 and Figure 5). In addition, Table 5 reveals that variability in NH₄⁺, NO₃⁻ and DOC had little effect on CH₄ flux (except for the UG treatment).

3.4. The Relationship between CH₄ Flux and Ecosystem Respiration (Re)

There is a significant linear correlation between CH₄ flux and ecosystem respiration in CG (p < 0.05) and RG (p < 0.05) plots but the correlation in UG (0.05 < p < 0.10) is not significant (Figure 6). In all grazing treatments, higher respiration values are associated with more negative CH₄ fluxes.

4. Discussion

4.1. Ecosystem Respiration (Re)

Respiration plays a key role in the global carbon cycle and it can prominently influence soil-atmospheric CO₂ exchange and net soil organic carbon (SOC) storage [26]. Similar to results obtained in other studies, we found a peak in respiration (88.9%–90.3% of annual Re) occurred during the growing season. Highest values of Re also tended to be associated with soil moisture, as observed in other works [16,27]. Re peaked in UG on 2 July 2013 right after four consecutive days of rainfall, which might have affected respiration in two ways: high soil moisture may have enhanced biological activities and the large amount of surface litter fall in UG plots, which resulted from the lack of grazing, supplied substrate for microbial respiration.

The Re observed in grassland has a large uncertainty. In our study, mean annual Re ranged from 119.9 mg C m⁻² h⁻¹ in CG to 189.8 mg C m⁻² h⁻¹ in UG. For comparison, we selected studies that have vegetation and climate similar to our site, as illustrated in

Table 7. Comparison of Ecosystem Respiration among different sites in Inner Mongolia.

Site	Ecosystem Type	Re (mg C m−2 h−1)	Period	Source
Inner Mongolia (43°26′ N, 116°40′ E, 1189 m)	Temperate steppe	47.7	2004	[28]
		20.5	2005
Inner Mongolia (42°27′ N, 116°41′ E, 1350 m)	Temperate steppe	223.3 ± 9.9	2006.6–9	[31]
		138.7 ± 6.9	2007.6–9
Inner Mongolia (40°34′ N, 111°34′ E, 1055 m)	Sown pasture	147.9 ± 21.3	2012.7–2013.7	This study
		227.8 ± 34.7	GS

Re (ecosystem respiration) = mean ± standard error but the standard error in [28] are not found. GS: growing season (from July 2012 to October 2012 and from May 2013 to July 2013).

. Fu et al. [28] reported considerably lower rates of respiration, which may be the result of methodological differences (eddy covariance versus the static opaque chamber—gas chromatography technique that we used). When measuring soil CO₂ efflux, the results from eddy covariance and chamber methods are in agreement at night but they are significantly different during daytime [29,30].The results from Yan et al. [31] are close to ours, which might be because we both used the chamber method. And furthermore, Cheng et al. [32] measured Re in soils along a transect from southern Inner Mongolia to the whole Ningxia province, concluding that Re from soils with plant cover ranged from 44 to 345 mg C m⁻² h⁻¹ with a mean value of 133 mg C m⁻² h⁻¹, which was consistent with our results. But their results also have a large range, which might be due to spatial variability among those experimental sites in drivers such as precipitation and soil temperature. In addition, the general lack of annual observation data in very cold regions also creates considerable uncertainty. For example, there are few long-term continuous measurements of arctic tundra CO₂ fluxes over the full annual cycle [33]. Even if the soil is frozen, CO₂ emissions still occur. Lange et al. reported that once the soil was frozen, CO₂ concentrations increased throughout the frozen period, even during very cold conditions, indicating net CO₂ production [34].

Annual CO₂ emission was best predicted by soil moisture and temperature, which explained 70.8%–79.5% of the variation in annual CO₂ emission. The same results were found in an earlier experiment conducted in a typical steppe in central Mongolia. Using a stepwise multiple regression analysis, the authors found that soil volumetric water content, soil temperature and aboveground green biomass were the three main factors that affect ecosystem respiration, and that aboveground green biomass was the primary factor related to ecosystem respiration [16]. Other studies also found that soil temperature and soil moisture played the major role in driving the temporal CO₂ emission variation [35,36]. For example, Bai et al. think that soil moisture and temperature were positively correlated with CO₂ emissions [35]. Sun et al. obtained similar the same results of Bai et al., they pointed out that temporal variations of CO₂ emission were strongly correlated with air and sediment temperatures [36].These results suggest that higher soil temperatures affect the root systems, enhancing root respiration and microbiological activity along with soil organic matter mineralization [37]. In this study, there is an exponential relationship between ecosystem respiration and soil temperature on a yearly scale and during the growing season (Figure 4). These results agree with many other studies ([38,39]). For example, Wagle and Kakani reported that exponential temperature–respiration functions provided a good fit for soil temperature <30 °C and <23 °C during the 2011 (a) and 2012 (b) growing seasons. Ecosystem respiration declined beyond 30 °C in 2011 and beyond 23 °C in 2012, and the exponential functions were highly significant (p < 0.0001) after excluding data points beyond these ranges [38]. We think that the consistency of the results in Wagle and Kakani’s study and our study may be because of the consistency of the soil temperature in the two studies. Almost all of the soil temperature values (44 data points were observed in a year) in our study are below 30 °C and most of them are below 23 °C (Figure 2). However, when soil temperature is low, such as in freezing period (FP) and freezing-thawing period (FWP), the relationship between ecosystem respiration and soil temperature becomes linear (Figure 4).

4.2. CH₄ Flux

Recent evidence suggests that semiarid grassland is an important sink for atmospheric CH₄ ([40,41,42]) and results from our study). With a mean annual CH₄ uptake of 26.1 µg C m⁻² h⁻¹ (

Table 8. Comparison of CH₄ flux among different sites in Inner Mongolia.

Site	Ecosystem Type	SM (0–6 cm, v/v %)	ST (5 cm, °C)	CH4 Flux (μg C m−2 h−1)	Period	Source
Inner Mongolia (43°33′ N, 116°40′ E, 1268 m)	Temperate steppe	13.5 ± 0.5 a	-	−38.7	2007.8–2008.8	[42]
Inner Mongolia (43°38′ N, 116°42′ E)	Temperate steppe	14.0	12.5	−43.8 ± 2.5 a	2007.10–2008.10	[11]
Inner Mongolia (43°33′ N, 116°40′ E, 1250 m)	Temperate steppe	17.2	-	−66.6 ± 4.2 a	2008.7–9	[40]
Inner Mongolia (42°02′ N, 116°17′ E)	Temperate steppe	8.2	16.9	−60.7	2009.5–9	[43]
		12.5	16.5	−78.7	2010.5–9
Inner Mongolia (43°11′–43°27′ N, 116°22′-117°00′ E)	Temperate steppe	13.7 ± 0.8 a	18.5	−98.6 ± 10.4 a	2010.5–10	[44]
Inner Mongolia (40°34′ N, 111°34′ E, 1055 m)	Sown pasture	18.1 ± 0.6 a	11.1 ± 0.4 a	−26.1 ± 0.9 a	2012.7–2013.7	This study
		14.6 ± 0.2 a	18.0 ± 0.8 a	−31.9 ± 1.5 a	GS

ST: soil temperature, SM: soil moisture. ^a: mean± standard error. GS: growing season (from July 2012 to October 2012 and from May2013 to July 2013).

), mean CH₄ uptake from all three treatments ranged from 25.1 to 27.9 µg C m⁻² h⁻¹ (Table 2). As Table 8 shows, in comparison with rates of annual mean CH₄ uptake [11,42], mean rates of CH₄ uptake measured during the growing season only were much higher [40,43,44]. Moreover, results from our study show that CH₄ uptake in the non-growing season accounts for 25%–33% of the annual CH₄ uptake (Table 2), which is consistent with the result (15%–30%) of an experiment conducted in a short-grass steppe in North America [45]. Current estimates of annual CH₄ exchange rates between steppe soils and the atmosphere will have high uncertainty if CH₄ uptake is only measured during the growing season [46]. From Table 8, we can see that CH₄ uptake rate will be overestimated when it is measured only in the growing season, because the CH₄ uptake rate in the growing season is always higher than average annual rates.

In our study, annual CH₄ fluxes had a significant negative correlation with soil temperature in all plots (p < 0.01, Table 6), similar to results obtained by other researchers [13,21]. Some researchers have concluded that higher temperature and reduced moisture increase net CH₄ uptake in terrestrial ecosystems, as they invariably increase gas diffusion rates and microbial access to oxygen and atmospheric CH₄ [47]. Even though the effect of soil moisture on CH₄ flux is less important than the influence of soil temperature (Table 5), a concave-shaped relationship was found between annual CH₄ fluxes and soil moisture in all three treatments (Figure 5). A similar relationship was also observed in an experiment conducted in the High Plains Grasslands (United States) and the relationship between CH₄ uptake and WFPS was hump-shaped with an optimum WFPS around 24% [48]. This might be due to the fact that the CH₄ uptake rate is limited by low diffusivity of CH₄ into the soil at high soil moisture contents, while very low moisture contents limit the biological activity of methanotrophs.

Some other studies found that soil temperature and soil moisture played the major role in driving temporal variation in CH₄ emissions [49,50,51]. Among them, Zhao et al. found that during the growing season, soil temperature played the dominant role in driving CH₄ emissions [49]. Rong et al. reported that seasonality of CH₄ uptake was related to monthly mean temperature and precipitation, which together explained 56% (range: 40%–83%) of the variability in monthly cumulative soil CH₄ uptake [50]. Roy Chowdhury et al. concluded that temporal dynamics of CO₂ production and methanogenesis at −2 °C showed evidence of fundamentally different mechanisms of substrate limitation and inhibited microbial growth at soil water freezing points compared to warmer temperatures [51].

4.3. Effects of Grazing on Ecosystem Respiration and CH₄ Flux

Alternative and appropriate grazing management can be beneficial to increase plant production and decrease ecosystem respiration, mitigating the negative effects of global climate change on the CO₂ balance in grassland ecosystems [52,53]. We obtained similar results in our study, as ecosystem respiration was 5.4 t C ha⁻¹ year⁻¹ lower in CG and 4.4 t C ha⁻¹ year⁻¹ lower in RG than that in UG. If we allot this ecosystem respiration to the sheep (which ate grass in the CG and RG plots), each sheep reduced CO₂ emission by 1.2 kg C ha⁻¹ year⁻¹ in CG and 1.0 kg C ha⁻¹ year⁻¹ in RG, showing that grazing significantly reduces ecosystem respiration (p < 0.01). The reason for this decline might be the reduction of aboveground biomass caused by sheep grazing in CG and RG. Moreover, different grazing patterns (CG and RG) have significantly different influences on Re. The Re in CG plots is significantly lower than that in RG plots (p < 0.01). It may be because the rotational grazing method gives recovery time for grass, so there is more aboveground biomass in RG plots than that in CG plots and there is more Re emission in RG than that in CG. However, because we have no data of carbon uptake (The absorption of carbon by photosynthesis was not considered), we can only provide limited information on the carbon balance.

Some previous studies have found that grazing exerts a considerable negative impact on CH₄ uptake in semi-arid steppes at regional scales during wintertime [19] but light-to-moderate grazing did not significantly change the annual CH₄ uptake [14]. Another study in China got a similar result, finding that heavy grazing depressed soil CH₄ uptake by 36% but light and moderate grazing had no significant effects in grassland ecosystems. The response of grassland soil CH₄ uptake to grazing also was found to depend upon grazing intensity, grazing duration and climatic types [54]. The CG and RG in our study can be characterized as moderate grazing and did not result in any significant difference in CH₄ uptake among the CG, UG and RG plots. Although the grazing pattern was different in CG and RG, grazing intensity was the same on an annual scale by 9 sheep ha⁻¹ year⁻¹. As shown in Table 2, CH₄ uptake in CG and RG were 0.23 kg C ha⁻¹ year⁻¹ and 0.20 kg C ha⁻¹ year⁻¹ lower than that of UG respectively. This may be attributed to the soil compaction caused by sheep trampling, which leads to an anaerobic environment. However, the influence is not statistically significant. In summary, moderate grazing (CG and RG in our study) did not influence CH₄ uptake significantly.

Not all results from past studies are consistent. For example, a study, in an alpine steppe on the Tibetan Plateau, China, thought that no grazing enhanced CH₄ uptake by 17.8% and 33.8% in 2009 and 2010, respectively, while its effect on CO₂ emission (ecosystem respiration) was not significant [55]. Gao et al. reported that long-term cattle grazing increased soil CO₂ fluxes, while the grazing effect on CH₄ uptake depended on precipitation [56]. The effects of grazing on ecosystem respiration and methane flux are complex. Different results obtained from different locations may be due to different climate and soil characteristics, not just due to grazing.

5. Conclusions

In the sown grassland, grazing significantly reduced ecosystem respiration, with reductions in both continuously grazed and rotationally grazed plots. We find that the sown grassland is a net sink for atmospheric CH₄, but no influence of grazing pattern was observed on CH₄ flux in CG, RG and UG. Soil temperature is the most important factor influencing ecosystem respiration and CH₄ flux in the sown grassland, with soil moisture playing second only to soil temperature. Variation in levels of ammonium nitrogen, nitrate nitrogen and dissolved organic carbon had little influence on ecosystem respiration or CH₄ flux (except UG plots). The values obtained for ecosystem respiration of grasslands have a large uncertainty range, which may be due to spatial variability as well as differences in research methods. Mean CH₄ uptake measured only during the growing season was much higher than the annual mean CH₄ uptake. Methane uptake rate will thus be overestimated when it is measured only in the growing season. In this study we measured ecosystem respiration, but future work should measure carbon uptake and changes in biomass to illustrate CO₂ exchange, not just ecosystem respiration. Availability of NH₄⁺-N, NO₃⁻-N and DOC in soils seemed to have little effect on ecosystem respiration and CH₄ fluxes except for the UG treatment, a finding that will be explored in more detail with collection of additional data.