The study area received 370 mm and 185 mm of rain for 2008 and 2009, respectively, and showed great variations in temporal distribution (Figure 1). Most of the rain fell over the summer months (from June to August), which accounted for 90% of the annual total precipitation in the two years. Soil temperature and moisture varied seasonally at this site (see Ref. [20] ). Briefly, soil temperature at 5 cm depth ranged from 13.1˚C to 31.1˚C with an average of 22.3˚C in 2008, 8.5˚C to 27.9˚C with an average of 18.1˚C in 2009. And the temperature generally negatively correlated with mowing height. Soil moisture peaked in July and had similar temporal fluctuation patterns over the two growing seasons, although the amount of rainfall was substantially different during the same period.

Above-ground biomass and litter measured are presented in Table 1. The above-ground biomass and total litter decreased progressively as cutting height decreased from 15 cm to 2 cm, and significant lower litter mass was found for the M₂ treatments (P < 0.01). A similar pattern was also observed for ANPP, which indicates that a cutting height greater than 10 cm has minimal effects on net primary productivity. These results indicate that the

effects of greater cutting heights (M₁₅ and M₁₀) on net primary production were limited in this grassland ecosystem. In addition, the biomass of Artemisia frigida, a species with low height, contributed to most of the total biomass of M₂ and M₅ plots relative to that of the control, M15 and M10 plots (data not shown).

Mean N₂O emission rates from the different mowing treatments varied in a range of −31.7 to 67.2 μg N₂O m⁻²∙h⁻¹ with means of 12.7, −4.2, 4.8, −5.2 and 4.1 µg N₂O m⁻²∙h⁻¹, for treatments of M_ck, M₁₅, M₁₀, M₅, and M₂, respectively, over the growing season of 2008, and means of 13.6, −9.0, 3.7, −2.9 and 5.5 µg N₂O m⁻²∙h⁻¹ for M_ck, M₁₅, M₁₀, M₅, and M₂, respectively, in 2009. The variation of N₂O fluxes was relatively narrow in the dry year (2009) compared to the wet year (2008) (Figure 1).

The higher variation in N₂O emission rate resulted in high variation of the monthly cumulative N₂O flux with a range of 0.48 to −0.35 kg N₂O ha⁻¹ over the two growing seasons (Figure 2). Measurements showed that grassland could be a N₂O source (positive value)/sink (negative value) (Figure 2). Regardless of the mowing heights, the monthly cumulative N₂O flux was negative in July for most of the treatments, except for the controls both in 2008 and 2009, and the M₁₀ and M₂ in 2008 (Figure 2).

The seasonal cumulative N₂O emissions based on a linear interpolation of spatially averaged daily or monthly mean fluxes. Over the two growing seasons, an estimated amount of 0.22 kg N₂O ha⁻¹ per season was emitted in the control treatment, of which more than 88% occurred in May and June. Occasionally, negative values were observed in the control and mowing treatments (Figure 1). The estimated seasonal cumulative emissions in the different treatments ranged from −0.27 kg N₂O ha⁻¹ to 0.33 kg N₂O ha⁻¹ (Table 1). For the total fluxes over the two growing seasons, the value for M₁₅ was the only mowing treatment significantly different from that of the control (P < 0.05) (Table 1). Compared with the control, the M₁₅, M₁₀, M₅ and M₂ mowing treatments decreased N₂O emission (Table 1).

The results of regression analyses showed that the closest relationship existed between the N₂O fluxes and soil

moisture (r² = 0.8, P < 0.0001), followed by above-ground plant biomass (r² = 0.30, P = 0.02) in this experiment (Figure 3(a), Figure 3(b)). Our stepwise regression analyses showed that a combination of soil moisture (partial R² = 0.42, P = 0.0005), soil temperature (partial R² = 0.13, P = 0.02) and total phosphorus (partial R² = 0.08, P = 0.047) explained 62.5% of the variation in the N₂O fluxes (P < 0.005) in May. Microbial biomass nitrogen explained 21.2% of the variation of the N₂O fluxes (P = 0.02) in June, and soil moisture explained 21.4% of the variation (P = 0.02) in July. Other factors did not contribute significantly to the explanation of the variation in N₂O fluxes in these summer months. In addition, we did not find a significant correlation between N₂O fluxes and any other variables measured at P < 0.05 in August and September.

Using the pooled data of the different treatments in the whole growing season, multi-linear regression analyses showed that N₂O fluxes was correlated with soil temperature, moisture (ST, SM) and soil organic carbon (SOC) in the control treatment and correlated with ST and SM in the M₅ treatment, with SOC in the M₂ treatment (Table 2). Combined ST, SM and SOC can explaining 31% of the variation in the N₂O fluxes in the control treatment and combined ST and SM can explaining 27% in M₅ treatment. Across the 25 sites all of the treatments, combined ST, SM and SOC can explaining 12% of the variation in N₂O in spatial (Table 2).

Our observations, showing a non-linear decrease in N₂O emission with decreases in mowing height, generally supported the hypothesis that a long term mowing treatment would decrease grassland N₂O emissions to the atmosphere. Our observations are in line with the results of another similar investigation also conducted in a grazed grassland in Inner Mongolia [21] , but opposite to the findings in a heavily grazed alpine grassland by Gao [22] . Mowing not only led to lower N₂O emission rates, but also led to lower above-ground plant and litter biomass (Table 1), and most likely to a removal of nitrogen from the system. Nitrification is commonly repressed by nitrogen limitation, which explains the decline in N₂O with mowing height. Indeed we observed a linear relation between N₂O fluxes and total soil nitrogen though this relationship was not robust because of low measurement frequency and was only found for soil temperatures ranging from 16˚C to 18˚C. This was in line with findings that N mineralization and nitrification were repressed during the growing season but increased after that period [23] .

In July of 2008 and 2009, a number of weak N₂O uptake peaks occurred in some of the mowing treatments (Figure 2(a), Figure 2(b)), leading to significant uptake of N₂O on monthly basis (Figure 2(b)). Though temperature was favorable in July, microbial activity may have decreased due to the low precipitation. Additionally, plants strongly compete for the mineralized N, leaving less N for nitrifying and denitrifying bacteria [8] [24] . This competition may be the reason for the N₂O uptake observed in grasslands during July in our study. This result confirms that grassland soils may act as net sinks of atmospheric N₂O under certain conditions due to high N₂O reduction activity by denitrifying microorganisms [16] [25] . N₂O reduction by denitrifying microorganism is associated with reduced levels of oxygen, which is usually related to increased moisture contents [26] . Despite the different patterns of N₂O fluxes in July 2008 and 2009, soil moisture contents were not significantly different between these months [20] . Hence, unknown soil factors might have been responsible for the observed differences in the fluxes in July.

In August, however, there was a net N₂O emission except for the control in 2008 and M₅ in both years (Figure 2(b)). At this time of the year, the temperature is still high enough for soil microbial activity (data not shown). Therefore soil microbes will decompose plant material, resulting in increased mineralization of N. However, most of the plants are getting senescent and require fewer nutrients from soil for growth at that time, consequently the competition between plants and microorganisms for mineralized N will be reduced. The higher N availability likely promotes microbial nitrification leading to a high N₂O production.

Using the pooled data of the whole growing season and all treatments, stepwise regression analyses showed

One asterisk, P < 0.1; two asterisks, P < 0.05; three asterisks, P < 0.01; ST-SM refers to soil temperature-soil moisture; ST-SM-SOC refers to soil temperature-soil moisture-soil organic carbon.

that soil moisture was the most important controlling factor for N₂O fluxes, explaining 80% of the observed variations in seasonal average N₂O fluxes (P < 0.0001) at our study site (Figure 3(a)). Similar results were reported for grasslands in Ireland [27] and New Zealand [28] . Therefore, our results confirm soil moisture is a major control of seasonal N₂O fluxes and determines microbial N₂O emissions in the semi-arid grassland of Inner Mongolia, China. In some agricultural regions, soil moisture was also found to be the controlling factor of N₂O fluxes [29] . Because soil moisture is the key determinant of the microbial processes that consume or produce N₂O, soil moisture shifts in arid and semiarid regions will likely affect N₂O fluxes.

The relationship between N availability and N₂O flux is commonly examined for predicting N₂O fluxes on large scales, and Millar [30] reported that N availability is the only factor affecting N₂O fluxes at meter-scales. However, in our study, we find the significant positive relationships between total nitrogen and N₂O fluxes only in a narrow temperature range (Figure 3(c)).

Our other hypothesis was that both soil biotic and abiotic factors underlying mowing effects of decreasing plant biomass play important roles in regulating N₂O fluxes. Since the cumulative seasonal N₂O fluxes were only slightly different among different mowing treatments, and no consistent changes were found among all of the mowing treatments (Table 1, Table 3), we tested the effects of mowing on N₂O fluxes by comparing the pooled data of all the mowing treatments including that of the control. The result suggests that mowing might decrease N₂O emissions by reducing plant litter and above-ground biomass (Table 3), since the decreased plant litter and biomass might result in decreases in the substrates supplied to microbes, which are involved in nitrification and denitrification processes. The above of our finding agrees with that of some previous studies. Zou [31] established a positive linear relationship between above-ground plant biomass and N₂O emissions. Kammann [32] found that increasing numbers of cuts reduced N₂O emissions. However, in other cases, for instance, Beck and Christensen [33] found that N₂O emissions increased when all above-ground grass was removed, and Klumpp [34] only observed some small peaks of N₂O emission in response to cutting events. There was a significant effect of the differences in soil moisture on the N₂O fluxes after mowing (Figure 3(a)), which explains 80% of the variation in the N₂O emissions induced by mowing (P < 0.001). Similar results were found in a study performed in grazed grasslands by Wolf [21] .

This study was carried out at the Duolun Restoration Ecology Research Station (116˚17'E, 42˚02'N), Institute of Botany, Chinese Academy of Sciences (IBCAS), which is located at Duolun County in Inner Mongolia, China. The area is situated in a semiarid, middle temperate zone and characterized by a continental monsoon climate. The mean annual air temperature is around 2.1˚C, with monthly mean temperatures ranging from −17.5˚C in January to 18.9˚C in July. The mean annual precipitation is approximately 385 mm, with 80% precipitation occurring from mid-June to late September. The topography is featured by low foothills at elevations of 1150 - 1800 m. The soil at the study site is classified as Chestnut soil in the Chinese soil classification, which equals to Calcic Kastanozem in IUSS system. And it contains 62.75% ± 0.04% sand, 20.30% ± 0.01% silt and 16.95% ± 0.01% clay, with mean soil (0 - 10 cm) bulk density of 1.31 g∙cm⁻³ and pH value of 7.12. The dominant plant species in the temperate grassland are Stipa krylovii Roshev., Cleistogenes squarrosa (Trin.) Keng., Artemisia frigida Willda, Potentilla acaulis L., Allium bidentatum Fisch. Ex Prokh. and Agropyron cristatum (L.) Gaertn.

The mowing experiment was set up in 2003 and consists one control (non-mowing, M_ck) and four mowing height treatments, i.e., mowing heights at 2 cm (M₂), 5 cm (M₅), 10 cm (M₁₀) and 15 cm (M₁₅) above soil surface, respectively, in a complete randomized block design, with five replicates. The five treatments were randomly assigned to the five plots (10 m by 20 m in size) in each of five blocks. Any adjacent blocks and plots within a block were 4 m apart from each other. Mowing with complete removal of the plant cuttings was carried out in late August each year starting in 2003. The precipitation data from 2008 to 2009 were provided by a meteorological station in an open field, which is approximately 2 km away from the experimental site and run by the Duolun Restoration Ecology Research Station, IBCAS. Prior to the setup of the experiment, the site had been kept free from disturbance and large animal grazing by fencing since 2001.

N₂O fluxes were measured by a static chamber technique following Zhang [35] -[37] . Briefly, the chamber consists of a stainless steel permanent base (50 × 50 × 12 cm) and a stainless steel top (50 × 50 × 50 cm). The base, with a 3-cm-deep groove on the upper side for water sealing, was driven into soil down to 12 cm each year approximately a month before the measurement started in each plot [38] . The chamber top was covered by heat-isolating and light-impenetrable materials outside and equipped with one rubber septum for gas sampling and two electric fans inside for mixing the air in the chamber headspace continuously and thoroughly [39] . During the measurements, the top was installed on the base, with its downside placed in the groove. The grooves were filled with water to seal the chamber. Gas samples were collected at 10-min intervals for 30 min (i.e., 0, 10, 20 and 30 min) through the septa using 60-ml syringes with airtight stopcocks. All gas samples were brought to a laboratory at the research station for N₂O analyses within 12 h after sampling. N₂O concentration was analyzed using a gas chromatograph (HP 5860, Agilent Technologies). N₂O flux was calculated from the linear slope of the mixing ratio changes in the four samples taken at 0, 10, 20 and 30 min after the chamber was closed. Negative flux values indicate gas uptake from the atmosphere, and positive flux values indicate gas emissions to the atmosphere (A threshold value was used for selecting “good” from “bad” measurements, e.g., r² > 0.90). The detection limit for our approach was 3 μg∙N∙m⁻²∙h⁻¹ for N₂O based on the chamber dimensions, the sampling time, and the reproducibility of repeated measurements of the N₂O standard gases (0.6%). The N₂O flux was measured weekly from June to September in 2008 and once every two weeks from May to September in 2009 on 9:00 am-12:00 am, local time. Concurrent with the N₂O flux sampling, air temperature, soil temperature (ST) and soil moisture (SM, v/v%) next to the chambers were measured. Air temperature was measured at 40 cm height above the soil surface, and soil temperature was measured at 5 cm depth using a portable digital thermometer. Soil moisture was measured using a portable soil moisture measuring kit ML2x (ThetaKit, Delta-T Devices, Cambridge, UK).

Prior to experiments in 2008 and 2009, vegetation variables were measured once at peak biomass (i.e., 10-20 August). A 1 × 1 m frame with a 10 cm × 10 cm grid was put above the canopy adjacent to the flux chambers in each 10 m × 20 m plot. The soil cover of each species was visually estimated following Yang [40] in all of the 100 grid cells, and then summed as the total coverage for the quadrat. After the measurement of the coverage, all plants were clipped at ground level within the quadrat of 1 m² and separated into living vegetation as above ground biomass (ANPP; see also reference 20) and dead plant material as standing litter. The total litter was the sum of the standing litter and litter collected from the surface within the quadrat. All plant material sampled was oven-dried at 65˚C for 48 hours and weighed.

Soil samples (0 - 10 cm layer) were collected monthly during the growing season 2009 using a soil corer (5 cm diameter) at the time of the fourth gas sampling. At each soil sampling, three soil cores were taken randomly at each plot and mixed evenly. These samples were separated into two sets of sub-samples: one set was stored at 4˚C for microbial analysis, and the other was air-dried for soil organic carbon (SOC), total nitrogen (TN) and total phosphorus (TP) analyses (for details see Zhang et al., 2012).

A repeated measures analysis of variance (ANOVA) was performed to examine inter-annual variability in SOC, TN and TP with the pooled data of all treatments during the growing season. Between-subject effects were evaluated as mowing effects and within-subject effects were interpreted as time-of-season effects. Regression analyses were made between N₂O fluxes and the measured variables (e.g., ST, SM, SOC, TN, TP, MBC, MBN and ANPP, total litter, total coverage). Because some parameters were measured at different times, we used the means of the whole growing season. Therefore, every variable had five treatments with five replicates. Differences in seasonal cumulative N₂O flux, total litter and ANPP among treatments were determined by simple one- way ANOVA. To examine which variable had the strongest effect on N₂O fluxes, a stepwise multiple regression analysis was applied between the mean N₂O fluxes over the whole growing season (as independent variable) and the measured variables (as dependent variables). All statistical analyses were conducted with the SAS 8.0 software package (SAS Institute Inc., Cary, NC, USA).