Responses of soil respiration to soil management changes in an agropastoral ecotone in Inner Mongolia, China

Abstract Studying the responses of soil respiration (R s) to soil management changes is critical for enhancing our understanding of the global carbon cycle and has practical implications for grassland management. Therefore, the objectives of this study were (1) quantify daily and seasonal patterns of R s, (2) evaluate the influence of abiotic factors on R s, and (3) detect the effects of soil management changes on R s. We hypothesized that (1) most of daily and seasonal variation in R s could be explained by soil temperature (T s) and soil water content (S w), (2) soil management changes could significantly affect R s, and (3) soil management changes affected R s via the significant change in abiotic and biotic factors. In situ R s values were monitored in an agropastoral ecotone in Inner Mongolia, China, during the growing seasons in 2009 (August to October) and 2010 (May to October). The soil management changes sequences included free grazing grassland (FG), cropland (CL), grazing enclosure grassland (GE), and abandoned cultivated grassland (AC). During the growing season in 2010, cumulative R s for FG, CL, GE, and AC averaged 265.97, 344.74, 236.70, and 226.42 gC m−2 year−1, respectively. The T s and S w significantly influenced R s and explained 66%–86% of the variability in daily R s. Monthly mean temperature and precipitation explained 78%–96% of the variability in monthly R s. The results clearly showed that R s was increased by 29% with the conversion of FG to CL and decreased by 35% and 11% with the conversion of CL to AC and FG to GE. The factors impacting the change in R s under different soil management changes sequences varied. Our results confirm the tested hypotheses. The increase in Q 10 and litter biomass induced by conversion of FG to GE could lead to increased R s if the climate warming. We suggest that after proper natural restoration period, grasslands should be utilized properly to decrease R s.

The T s and S w significantly influenced R s and explained 66%-86% of the variability in daily R s . Monthly mean temperature and precipitation explained 78%-96% of the variability in monthly R s . The results clearly showed that R s was increased by 29% with the conversion of FG to CL and decreased by 35% and 11% with the conversion of CL to AC and FG to GE. The factors impacting the change in R s under different soil management changes sequences varied. Our results confirm the tested hypotheses. The increase in Q 10 and litter biomass induced by conversion of FG to GE could lead to increased R s if the climate warming. We suggest that after proper natural restoration period, grasslands should be utilized properly to decrease R s .

K E Y W O R D S
cropland, grazing grassland, restoration grassland, soil temperature, soil water content

| INTRODUCTION
Soil respiration (R s ) is a crucial process in the global carbon cycle (Bahn, Janssens, Reichstein, Smith, & Trumbore, 2010). Minor changes in R s have the potential to significantly affect atmospheric CO 2 concentrations (IPCC, 2007). Large-scale soil management changes have been affecting R s , with considerable impacts on the terrestrial ecosystem carbon cycle. It has been estimated that global net flux due to land use change during the period of 1,850-2,000 is 148.6 Pg C (Kaul, Dadhwal, & Mohren, 2009); however, the mechanisms of this effect remain subject of debate (Nazaries et al., 2015). In recent years, considerable efforts have been made to understand the effects of soil management changes, that is, conversion of cropland to woodland or forest (Kellman, Beltrami, & Risk, 2007;Saurette, Chang, & Thomas, 2007;Zhang et al., 2015), forest to cropland, or grassland (Sheng et al., 2010). However, few studies have focused on the conversion of grazed grassland to cropland, and not many studies focus on conversion of cropland to grassland in degraded ecosystems (Shi, Yan, Zhang, Guan, & Du, 2014;Zhang et al., 2015).
Soil management changes can potentially alter soil temperature (T s ) and soil water content (S w ) (Chen et al., 2016;Rong, Ma, Johnson, & Yuan, 2015;Wang, Gong, et al., 2015), which are the main abiotic factors affecting R s (Fang & Moncrieff, 2001;Gomez-Casanovas, Matamala, Cook, & Gonzalez-Meler, 2012), and these two factors affect the productivity and the decomposition rate of soil organic matter . Temperature sensitivity (Q 10 ) describes the relationship between R s and temperature and can therefore also be changed with soil management changes (Gong et al., 2014;Rong et al., 2015).
Furthermore, soil management changes can impact biotic factors, such as net primary production, belowground biomass (BGB), soil organic carbon (SOC) (Deng, Liu, & Shangguan, 2014;Frank, Liebig, & Tanaka, 2006;Sheng et al., 2010;Zhang et al., 2015), all of which greatly affect R s . However, the effects of soil management changes from cropland to grassland on R s have not been consistent among studies, some studies indicate that it increases R s (Frank et al., 2006;, while other studies show that it reduces R s (Iqbal et al., 2008;Zhang et al., 2015). Moreover, the effects of grazing on R s also have no consistent conclusion (Rong et al., 2015). Therefore, additional studies are needed to clarify the effects of the soil management changes on R s .
The northern agropastoral ecotone of China, which is a transition zone between agricultural and pastoral regions and encompasses various ecosystems, occupies an area of 8 × 10 5 km 2 . Soil management changes from grassland to cropland or from cropland to grassland are frequently occurring in this region, making it the most sensitive eco-environmental area in China . In this area, the typical soil management changes sequences have occurred, including the conversion of free grazing grassland (FG) to cropland (CL), both of which are under human disturbance, and the conversion of FG and CL to restoration grassland-grazing enclosure grassland (GE) and abandoned cultivated grassland (AC). This variety of different soil management changes provides a unique opportunity to study the response of R s to soil management changes. Previous studies of R s in the temperate grassland in China primarily focused on the influence of grassland management practices on R s (Li & Sun, 2011;Lu, Liao, & Liao, 2016;Su, Li, Cui, & Zhao, 2005), further studies are needed to study the impacts of conversions from FG to CL and GE, CL to AC on R s , biotic (aboveground biomass (AGB), BGB, SOC, etc.), and abiotic factors (T s , S w ).
In this study, we measured R s , AGB, BGB, SOC, and soil microclimate in degraded areas of the agropastoral ecotone (soil management types: FG, CL, GE, and AC) in Inner Mongolia from 2009 to 2010. The objectives of our study were to (1) quantify daily and seasonal patterns of R s in four soil management types, (2) evaluate the influence of abiotic (T s , S w ) factors on R s in these soil management types, and (3) detect the effects of soil management changes on R s . We hypothesized that (1) most of daily and seasonal variation in R s could be explained by T s and S w , (2) soil management changes could significantly affect R s and (3) soil management changes affected R s via the significant change in abiotic and biotic factors.

| Site descriptions
The study was conducted in Duolun County in Inner Mongolia

| Measurement of R s , T s , and S w
In each site, five 2 × 2 m sample plots were randomly selected with the constraint that the plots were located at least 10 m from the edge of the site to avoid edge effects. One day before the measurements, PVC collars were inserted (20 cm in diameter by 5 cm in height) 2 cm into the soil at each plot. To exclude respiration from aboveground vegetation, all visible living plants were removed from the collars before measurements.
The measurements were taken eight times per day at 3-hr intervals from 6:00 a.m. to 03:00 p.m., and instantaneous R s in each plot was measured three times with an LI-8100 Automated Soil CO 2 Flux System (LI-COR Environmental, Lincoln, NE, USA) with a 90-s enclosure period and a 30-s delay between measurements. The final instantaneous R s for a given plot consisted of the average of the three measurements, and if necessary, one or more additional measurements were taken until the coefficient of flux variation was below than 2%. The measurements were conducted from August to October in 2009 and May to October in 2010. When measuring R s , T s was determined at a depth of 5 cm adjacent to each PVC collar with a thermocouple. Simultaneously, S w at a depth of 5 cm was measured using a Theta Probe Soil Moisture Sensor. Field meteorological data were obtained from the local meteorological station.

| Biomass and soil characteristics
AGB and BGB were collected and measured at the end of every month (May to September) in 2010, five representative 1 × 1 m quadrats were established at each site. For determination of AGB, all plants from five quadrats were clipped above the soil surface at each site, and litter was collected using rake. After removal of AGB and litter biomass, five belowground core samples at the depth of 0-30 cm were collected from each quadrant using a soil auger with a diameter of 8 cm. The root was separated from the soil by washing over a 0.2mm mesh to determine BGB. All plant samples were dried at 65°C to constant weight for biomass determination.
On 15 August 2010, three soil samples were collected in each plot using a soil auger with a diameter of 4 cm at a depth of 0-30 cm at 10-cm intervals, the samples were mixed to obtain one composite sample per plot. Subsequently, the samples were sieved (<2 mm) and any roots were removed. Then, samples were ground in a ball-bearing mill and sieved (<0.9 mm) prior to analysis of SOC. We determined SOC using the K 2 Cr 2 O 7 -H 2 SO 4 digestion method (Nelson & Sommers, 1982). Soil bulk density was determined using soil cores (volume of 100 cm 3 ) obtained from depths of 0-10 cm.

| Data analysis
Soil respiration, T s , and S w were calculated by averaging the five replicates on each sampling day. We conducted all statistical analyses using SPSS 19.0 (SPSS Inc., Chicago, IL, USA). Spatial variation of R s in all sites was analyzed using one-way ANOVA. There was no temporal autocorrelation of soil respiration with soil temperature and water content according to the autocorrelation function (ACF) test; consequently, regression analysis was conducted between R s and T s , S w . Multiple regression analysis was used to examine the relationships between R s and T s , S w . Least significant difference (LSD) was used to evaluate differences between sites, and analyses were performed at significance levels of p < .05, p < .01, and p < .001. Monthly cumulative R s (gC/m 2 ) were calculated by linear interpolation of daily R s from 5 August to 15 October in 2009 and from 5 May to 31 October in 2010.
To simulate the relationship between R s and T s , we applied the exponential regression model: where R s is the soil respiration rate (μmol CO 2 m −2 s −1 ), T s is the soil temperature (°C) at a depth of 5 cm, and β is constants fitted by the models.
The Q 10 values were calculated with the Van't Hoff model:

| Abiotic and biotic factors among four soil management types
Changes in T s coincided with air temperature, and the maximum and minimum T s values were observed in July and October (Figures 1 and   2a); levels of S w at the depth of 0-5 cm coincided with that of irregular rainfall, with the highest values in September due to high precipitation (Figures 1 and 2b). Levels of T s significantly decreased from 21.9 ± 0.3 to 20.7 ± 0.2°C and 21.3 ± 0.1 to 20.4 ± 0.3°C with the conversion of CL to AC and FG to GE, respectively (p < .05). S w significantly decreased from 13.6 ± 0.5% to 10.2 ± 0.7% and 11.5 ± 0.4% with the conversions of FG to CL and GE (p < .05).
The biotic parameters also varied significantly with soil management changes (p < .05). In particular, SOC decreased from 33.07 ± 0.05 g/ kg to 11.78 ± 0.28 and 17.98 ± 0.87 g/kg with conversions of FG to  (Table 1).

| Dynamic change in soil respiration
The R s showed similar daily variations in the four soil management types.
The extremely low R s values coincided with a drought on July 5, and after the onset of rain on July 11, R s sharply increased to its highest values of the

| Correlation between R s and T s , S w
For the four soil management types, daily mean R s significantly increased exponentially with T s (p < .001). T s explained 26%-70% of the variation in R s (Figure 4, Table 3). Although the relationship between R s and T s was similar among the four soil management types, the coefficient of determination (R 2 ) was greater in CL than FG and AC (R 2 : 70% vs. 26% and 49%), and lower in FG than that in GE (R 2 : 26% vs. 50%) ( Table 3). Values of Q 10 were 1.55, 2.66, 2.10, and 2.01 in FG, CL, GE, and AC, respectively (Table 3).
Daily mean R s was significantly correlated with S w (p < .001) and followed parabolic pattern for four soil management types. The S w explained 26%-40% of the variation in R s (Figure 4, Table 3). When S w values were above about 2%, R s increased with the increasing of S w in CL, whereas R s will decrease when S w reached the threshold for the other sites. Multiple polynomial regression analysis showed that T s and S w explained 66%-84% of the variation in daily mean R s ( Figure 5, Table 4), while total monthly precipitation (MTP) and monthly mean temperature (MMT) explained 78%-96% of the variation in monthly cumulative R s (Figure 6, Table 4).

| Soil respiration among soil management changes
Values of the daily mean R s during growing season varied with soil management changes, and it significantly increased by 29% with the conversion of FG to CL and decreased by 35% and 11% with the conversion of CL to AC and FG to GE during the growing season in 2010 (Table 2). From June to August, R s in CL was significantly higher than that in the other site types (p < .001), while at the end of the growing seasons, we observed no significant difference of R s between the four soil management types (Figure 4).

| Soil respiration under four different soil management types
In this study, daily mean R s values of grassland during the growing seasons ranged from 1.21 to 1.43 μmol CO 2 m −2 s −1 , which were lower than that in the northern grassland in China (1.87 μmol CO 2 m −2 s −1 ) (Zhang et al., 2015). Again, daily mean R s during the growing seasons in CL (1.85 μmol CO 2 m −2 s −1 ), was also lower than that in the agricultural ecosystem in the northern China plain (3.95 μmol CO 2 m −2 s −1 ) (Zhang et al., 2013). The lower R s observed in our study is possibly due to lower precipitation (382 vs. 532 mm and 560 mm for cropland and grassland, respectively). Cumulative R s ranged from 226.42 to 344.74 gC/m 2 during the growing seasons, and it fell right into the range reported for temperate grassland (range: 132-830 gC/m 2 ) (Raich & Schlesinger, 1992), and close to the study in the same region (262-309 gC/m 2 ) (Gong et al., 2014). Different ecosystem types, in regard to different plant community patterns within an ecosystem, exhibit different Q 10 values. In our study, Q 10 values ranged between 1.55 and 2.66 for the four soil management types, which fell right into the range of the Q 10 in China T A B L E 1 Vegetation and soil characteristics for free grazing grassland (FG), cropland (CL), grazing enclosure grassland (GE), and abandoned cultivated grassland (AC) (values represent mean ± standard error, n = 5). Different letters in each column indicate significant differences at p < .05

Soil management
Bulk density (g/cm 3 , 0-10 cm) T A B L E 2 Daily mean soil respiration (R, μmol CO 2 m −2 s −1 ) and cumulative soil respiration (CR, gC/m 2 ) during the growing seasons from 2009 to 2010 for free grazing grassland (FG), cropland (CL), grazing enclosure grassland (GE) and abandoned cultivated grassland (AC) (values represent mean ± standard error, n = 5). Different letters in each column indicate significant differences at p < .05

SOC (g/kg, 0-30 cm) AGB (g/m 2 ) BGB (g/m 2 , 0-30 cm)
(range: 1.28-4.75) (Zheng et al., 2009). Values of Q 10 values far greater in CL than that in the other sites, which confirmed previous conclusion that cropland may be more sensitive to climate warming, whereas grassland may have adapted to climatic warming (Tungate, Israel, Watson, & Rufty, 2007). In FG, Q 10 was 1.55, which was close to the lowest value of the range, study has indicated that small quantities of fungi and bacteria may result in lower Q 10 values in grazed grasslands in China (Cao et al., 2004).

| Effects of abiotic factors on R s in different soil management types
Generally, T s and S w are considered two of the most important abiotic factors controlling temporal variations of R s (Raich & Potter, 1995;Rong et al., 2015). In fact, all biogeochemical processes associated with R s inevitably relate to T s and S w (Risch, Haynes, Busse, Filli, & Schuetz, 2013), and the dependence of R s on T s and S w could be explained by the influence of T s and S w on availability of carbon substrate (Campbell & Law, 2005). Although the relationship between R s and T s and S w is similar, the coefficient of determination (R 2 ) was different (Table 3), suggesting that soil management changes changed the R s through influencing the relationship between R s and abiotic factors.
The effect of S w on R s is complex, as soil water affects not only enzymatic activities and physiological processes, but also gas diffusion (Balogh et al., 2011;Unger, Maguas, Pereira, David, & Werner, 2010). Low values of S w slow down solute diffusion and limit the supply of organic substrate for microorganisms (Moyano, Manzoni, & Chenu, 2013). In our study, S w had the threshold in grassland ecosystem, above or below the threshold, R s would decrease, but values of R s for CL were almost positively related to S w , and the result agreed with finding of Rong et al. (2015).
Although the result suggested that T s was more important than S w in determining soil respiration during growing seasons (Table 3), the dramatic change in S w has a significant influence on R s . Conant, Dalla-Betta, Klopatek, and Klopatek (2004) indicated that S w had an overriding influence on R s , particularly during the dry season in semi-arid environments.
In a similar study, Rey et al. (2011) found that S w was the main driver of R s for most of the year when soil temperatures were above 20°C in semi-arid steppe ecosystems of Spain. In our study, dramatic increase or decrease in R s were observed after rainfall or dry events for all soil management types (Figure 2), and the result was in agreement with findings from the other studies (Rey et al.,2011;Rong et al., 2015), suggesting that the dramatic change in S w have a pronounced influence on R s .
Abiotic factors explained less daily variation of R s in FG than that in the other sites (R 2 : 66% vs. 79%-86%, Figure 5), and it indicated that other factors also play an important role in R s in FG. Grazing significantly changes soil physical and chemical characteristics, including SOC, soil bulk density, and soil texture (Bremer, Ham, Owensby, & Knapp, 1998;Gong et al., 2014;Wilsey, Parent, Roulet, Moore, & Potvin, 2002), so the terms above should be considered to improve the explanation of daily mean R s in FG.

| Effects of soil management changes on soil respiration
Soil management changes result in changes in the soil microclimate and the biotic factors such as vegetation structure, primary productivity, and soil organic matter, thus indirectly affecting R s (Chen et al., 2006;Gong et al., 2014;Rong et al., 2015). Values of R s increased by 29% due to the conversion of FG to CL, and the result is similar to previous studies that conversion of natural grassland to cropland can F I G U R E 4 The relationship between daily mean R s (μmol CO 2 m −2 s −1 ) and soil temperature (T s , °C, 0-5 cm), soil water content (S w , %, 0-5 cm) for free grazing grassland (FG), cropland (CL), grazing enclosure grassland (GE), and abandoned cultivated grassland (AC). (a) relationship between R s and T s ; (b) relationship between R s and S w increase R s (Frank et al., 2006;. Changes in soil microclimate due to soil management changes could not explain the increased R s . S w decreased, while there was no significant change in T s , whereas R s increased with the conversion of FG to CL. This significant change may be explained as follows: First, Q 10 was higher for CL compared to FG, which suggests that soil management change from Soil management R s = a*e b*Ts R s = a*S w 2 + b*S w + c T A B L E 3 Regression equations for daily mean R s (μmol CO 2 m −2 s −1 ) with soil temperature (°C, 0-5 cm) and soil water content (%, 0-5 cm) for free grazing grassland (FG), cropland (CL), grazing enclosure grassland (GE), and abandoned cultivated grassland (AC) F I G U R E 5 Correlation of daily mean R s (μmol CO 2 m −2 s −1 ) with soil temperature (T s , °C, 0-5 cm), soil water content (S w , %, 0-5 cm) for free grazing grassland (FG), cropland (CL), grazing enclosure grassland (GE), and abandoned cultivated grassland (AC). The regression equations for each curve are listed in Table 4 T A B L E 4 Regression equations for monthly cumulative R s (gC m −2 month −1 ) and monthly total precipitation (MTP), monthly mean temperature (MMT) for free grazing grassland (FG), cropland (CL), grazing enclosure grassland (GE), and abandoned cultivated grassland (AC) F I G U R E 6 The relationship between monthly cumulative R s (gC m −2 month −1 ) and monthly total precipitation (MTP), monthly mean temperature (MMT) for free grazing grassland (FG), cropland (CL), grazing enclosure grassland (GE), and abandoned cultivated grassland (AC). The regression equations for each curve are listed in Table 4 FG to CL could increase the temperature sensitivity of soil respiration and lead to a concomitant loss of soil carbon storage. Second, soil management change from FG to CL increased R s through decreasing soil carbon levels (Xie et al., 2007) and soil bulk density. In this study, SOC and soil bulk density in cropland were both significantly lower than that in the FG (Table 1), suggesting that annual tillage enhanced substrate availability and soil aeration, which in turn may have led to increased soil microbial activity and decomposition of soil organic carbon, resulting in a rapid increase in R s . Third, the higher R s in CL may relate to higher AGB, plant photosynthesis has a driving effect on R s (Tang, Baldocchi, & Xu, 2005), and R s increase with increase in AGB (Gong et al., 2014). In this study, AGB was significantly higher in CL than in FG and AC (  (Lai et al., 2017). In our study, manure fertilization of cattle slurry was used in CL site, which may promote the emission of soil CO 2 .

Soil management R s = a•T s 2 + b•S w 2 + c•T s •S w + d•T s + e•S w + f R s = a•MMT 2 + b•MTP 2 + c•MMT•MTP + d•MMT + e •MTP + f
Soil management change from CL to AC remarkably decreased R s by 35%, and it may relate to lower T s , Q 10 and higher soil bulk density in AC. When the cropland converted to abandoned cultivated grassland, SOC increased and AGB decreased (Table 1), which was conducive to the accumulation of soil organic matter and decrease the release of CO 2 . The result differed from the finding of , which showed that R s was increased with the conversion of CL to AC. This may be due to shorter restoration year (10 vs. 15 years), which result in lower accumulation of litter biomass (38 vs. 103 g/m 2 ). Litter is the main source of soil organic carbon and provides substrate for soil microbial activity, resulting in heterotrophic respiration (Ngao, Epron, Brechet, & Granier, 2005), and it is one of the main factors affecting R s along the restoration chronosequence . Our results suggest that after a proper natural restoration period, restoration grasslands should be utilized properly to decrease R s .
The values of R s significantly decreased by 11% with the conversion of FG to GE. On the one hand, T s and S w both were significantly decreased with soil management change from FG to GE (Figure 2a, b), which lead to the decrease in R s . On the other hand, grazing animals deposit large amounts of manure that could increase SOC and BGB in FG (Table 1) and, consequently, increased R s . The active carbon pool in the soil directly provides respiration substrate for decomposition and R s increase with the increase of SOC (Smith 2003;Chen, Huang, & Zou, 2010), while root respiration accounts for 13%-55% of total R s in temperate grasslands (Gong et al., 2014). The conversion of FG-to GE-induced changes in soil microclimate also contributed to the relatively high Q 10 values (Table 3); the same result also has been reported for the Yellowstone National Park and Tibetan Plateau (Chen et al., 2016;Chuckran & Frank, 2013), and it implies that the carbon stored in the soils of the GE may be particularly vulnerable to the climate warming. Moreover, litter biomass was significantly higher in GE than that in FG (Table 1), and it will continuously increase with restoration years. Given these results, we predict that the rate of CO 2 release is faster in GE than that in FG if the climate warming.

| CONCLUSIONS
Our findings support the hypotheses that most of the daily and seasonal variation in soil respiration could be explained by soil temperature and soil water content, and soil respiration is significantly affected by soil management changes. This study monitored the effects of soil management changes from free grazing grassland to cropland and grazing enclosure grassland, cropland to abandoned cultivated grassland on soil respiration in Inner Mongolia, China, and it is critical for enhancing our understanding of the global carbon cycle and has practical implications for grassland management. Soil temperature and soil water content significantly influenced soil respiration for all soil management types and explained 66%-86% of the variability in daily soil respiration. Monthly mean temperature and precipitation explained 78%-96% of the variability in monthly cumulative soil respiration. The results showed that soil respiration increased by 29% with the conversion of free grazing grassland to cropland and decreased by 35% and 11% with the conversion of cropland to abandoned cultivated grassland and free grazing grassland to grazing enclosure grassland. The increase in Q 10 and litter biomass induced by the conversion of free grazing grassland to grazing enclosure grassland could lead to increased CO 2 emissions if the climate warming.
Given the limitations of biotic factors data of this study, soil organic carbon and bulk density were only determined once a year, and biomass was only measured once a month from May to September in 2010, further studies are worthwhile to evaluate the influence of biotic factors (soil organic carbon, aboveground biomass, root biomass, and litter biomass) on soil respiration, and to detect relative contribution of different factors to soil respiration in different soil management types. Furthermore, the studies about the contribution ratio of root and heterotrophic respiration to soil respiration under soil management changes are also needed to fully explain the effect of soil management changes on soil respiration.

This work was supported by the National Key Research and Development
Program of China (2016YFC0500608-3). We are grateful to the students and teachers who have ever helped us and also thank the anonymous reviewers for presenting valuable suggestions to improve this paper.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
Haili Xue analyzed the data and wrote the paper. Haiping Tang designed the research and revised the manuscript. All authors have read and approved the final manuscript.