The Xilin River is situated in the southeastern part of the Inner Mongolia Autonomous Region in China (43∘26′–44∘39′ N, 115∘00′–117∘30′ E). It is a typical inland river of the Inner Mongolia grasslands. The river basin area is 10 542 km2, total length is 268.1 km, and average altitude is 988.5 m. According to the meteorological data provided by the Xilinhot Meteorological Station (Xi et al., 2017; Tong et al., 2004), the long-term annual mean air temperature is 1.7 ∘C, and the maximum and minimum monthly means are 20.8 ∘C in July and -19.8 ∘C in January, respectively. The average annual precipitation was 278.9 mm for the period of 1968–2015. Precipitation is distributed unevenly among the seasons, with 87.41 % of the total precipitation occurring between May and September.

Soil types in the Xilin River basin are predominantly chernozems (86.4 %), showing a significant zonal distribution as light chestnut soil, dark chestnut soil, and chernozems from the northwest to southeast. Soil types in this basin also present a vertical distribution with elevation. Soluble chernozems and carbonate chernozems are primarily observed at altitudes above 1350 m, with a relatively fertile and deep soil layer. Dark chestnut soil, boggy soil, and dark meadow with high humus content are distributed between the altitudes of 1150 and 1350 m. Meanwhile, light chestnut soil, saline meadow soil, and meadow solonchak with low soil humus, a thin soil layer, and coarse soil texture are distributed between the altitudes of 902 and 1150 m above sea level (Xi et al., 2017).

In this study, five representative transects were selected as the primary measurement sites in the entire Xilin River. Each transect cuts through the riparian wetlands near the river and the hillslope grasslands further away (Fig. 1).

The layout of the sampling points of each transect is shown in Fig. 2. Each sampling point, from T1–T5, was extended from either side of the river to the grassland on the slopes by using five to seven sampling points for each transect, resulting in 24 points in total. The sampling sites on the left and right banks were defined as L1–L3 and R1–R4 from the riparian wetlands to the hillslope grasslands. As transect T3 was located on a much wider flood plain, none of its sampling points were located on the hillslope grassland. The last transect (T5) was located downstream in the dry lake and contained seven sampling points. They were defined as S1–S7, where S1, S2, and S7 were located along the lakeshore (the lakeside zone), and S3–S6 were located in the dry lake bed (S3 and S4 in the mudbank, S5 in saline–alkali soil, and S6 in sand–gravel geology). Moreover, characterizations for the T1, T2, and T3 transects were located along the continuous river flow, and the T4 and T5 transects were located along the intermittent river flow.

The CO2, CH4, and N2O emissions from each site were measured in August (wet season) and October (dry season) in 2018 using a static dark chamber and the gas chromatography method. The static chambers were made of a cube-shaped polyvinyl chloride (PVC) pipe (dimensions: 0.4m×0.2m×0.2m). A battery-driven fan was installed horizontally inside the top wall of the chamber to ensure proper air mixing during measurements. To minimize heating from solar radiation, white adiabatic aluminum foil was used to cover the entire aboveground portion of the chamber. During measurements, the chambers were driven into the soil to ensure airtightness and connected with a differential gas analyzer (Li-7000 CO2/H2O analyzer, LI-COR, USA) to measure the changes in the soil CO2 concentration. The air in the chamber was sampled using a 60 mL syringe at 0, 7, 14, 21, and 28 min. The gas samples were stored in a reservoir bag and taken to the laboratory for CH4 and N2O measurements using gas chromatography (GC-2030, Japan). The measurements were scheduled for 09:00–11:00 or 15:00–17:00 UTC+8.

Soil temperature (ST) was measured at depths of 0–10 and 10–20 cm with a geothermometer (DTM-461, Hengshui, China). Plant samples were collected in a static chamber and oven-dried in the laboratory to obtain aboveground biomass (BIO). A 100 cm3 ring cutter was used to collect surface soil samples at each site, which were placed in aluminum boxes and immediately brought back to the laboratory to measure soil mass moisture content (SMC) and soil bulk density (ρb) using national standard methods (NATESC, 2006). Topsoil samples were collected, sealed in plastic bags, and brought back to the laboratory to measure soil pH, electrical conductivity (EC), total soil organic carbon (TOC) content, and soil C:N ratio.

The CO2, CH4, and N2O emissions were calculated using Eq. (1) (Qin et al., 2016): 1F=VA×dcdt×ρ=H×dcdt×MV×273.15273.15+t, where F denotes the flux of CO2, CH4, and N2O emissions (mgm-2h-1); H is the height of the static chamber (0.18 m); M is the relative molecular weight (44 for CO2 and N2O, and 16 for CH4); V is the volume of gas in the standard state (22.4 Lmol-1); dc/dt is the rate of change of the gas concentration (10-6 h-1); and T is the temperature in the black chamber (∘C).

The annual cumulative emissions were calculated using Eq. (2) (Whiting and Chanton, 2001): 2M=∑Fi+1+F12×(ti+1-ti)×24, where M denotes the total cumulative emission amounts of CO2, CH4, or N2O (kghm2); F is the emission flux of CO2, CH4, or N2O; i is the sampling frequency; and ti+1-ti represents the interval between two adjacent measurement dates.

In this study, a 100-year timescale was selected to calculate the global warming potential (GWP) of soil CH4 and N2O emissions (Whiting and Chanton, 2001): 3GWP=1×[CO2]+25×[CH4]+298×[N2O], where 25 and 298 are the GWP multiples of CH4 and N2O relative to CO2 on a 100-year timescale, respectively.

All statistical analyses were performed using SPSS for Windows version 18.0 (SPSS Inc., Chicago, IL, USA). Statistical significance was set at P<0.05. Pearson correlation analysis was conducted to estimate the relationships between GHG fluxes and environmental variables. A Wilcoxon test was used to determine the differences in the GHG fluxes between the two seasons.

The temporal and spatial variations in SMC10 occurred in the following order: wet season > dry season and riparian wetlands > hillslope grasslands (Fig. 3a, c, and e). Similar variations were observed in SMC20 (Fig. 3b, d, and f). The average SMC10 and SMC20 in the continuous river transects in the riparian zones (SMC10 values were 37.44 % in the wet season and 19.40 % in the dry season, while SMC20 values were 25.96 % in the wet season and 17.39 % in the dry season) were higher than those in the hillslope grasslands (SMC10 values were 9.12 % in the wet season and 4.15 % in the dry season; SMC20 values were 6.51 % in the wet season and 5.96 % in the dry season). During the study period, both SMC10 and SMC20 changed as the distance from the river increased, and the highest value was observed at the near-stream sites (L1 and R1). SMC10 fluctuations were low in the intermittent transect compared with those in the upstream transects, with mean values being 11.79 % in the wet season and 3.72 % in the dry season in the riparian areas. The mean SMC10 in the hillslopes was 6.58 % in the wet season and 2.86 % in the dry season. SMC20 showed similar fluctuation; it was 7.22 % in the wet season and 2.98 % in the dry season in the riparian areas and 7.56 % in the wet season and 4.4 % in the dry season in the hillslopes. In transect T5, average SMC10 and SMC20 at the center of the lake (SMC10 values were 29.00 % in the wet season and 13.36 % in the dry season; SMC20 values were 29.30 % in the wet season and 9.69 % in the dry season) were higher than those along the lakeshore (SMC10 values were 4.90 % in the wet season and 3.13 % in the dry season; SMC20 values were 3.34 % in the wet season and 5.22 % in the dry season).

Spatiotemporal differences in ST during the entire observation period are displayed in Fig. 4. ST variations in the wet season (mean was 27.4 ∘C) were noticeably higher than those in the dry season (mean was 8.97 ∘C). Moreover, ST at riparian sites (mean, 26.0 ∘C in the wet season and 8.41 ∘C in the dry season) was slightly lower than that at the hillslope grasslands (mean, 30.9 ∘C in the wet season and 10.3 ∘C in the dry season) for the 0–10 cm soil depth, with the exception of transect T5. Similar results were observed for the 10–20 cm soil depth.

Figure 5 shows the spatiotemporal variations in GHG emissions in the wet and dry seasons in each transect. CO2 emissions in each transect were higher in the wet season than in the dry season. The average emissions in the riparian wetland transects T1–T4 (1582.09±679.34 mgm-2h-1 in the wet season and 163.24±84.98 mgm-2h-1 in the dry season) were higher than the transects in the hillslope grasslands (1071.54±225.39 mgm-2h-1 in the wet season and 77.68±25.32 mgm-2h-1 in the dry season). High CO2 fluxes occurred in the riparian zones, while lower CO2 fluxes were observed in the hillslope grasslands in continuous river transects (T1, T2, and T3). Transect T4 exhibited lower CO2 emission in the riparian wetlands near the channel than at sites away from the channel. CO2 emissions in transect T5 in the wet and dry seasons decreased from the lakeshore to the lake center.

CH4 emissions at the continuous river flow transects (T1, T2, and T3) varied between the wet and dry seasons, except for those at T4 (characterized by intermittent river flow) and T5 (the dry lake). In the wet season, the near-stream sites (L1 and R1) in T1, T2, and T3 were characterized as high-CH4 sources (average, 3.74±3.81 mgm-2h-1), but the sites located away from the river gradually turned into CH4 sinks. Moreover, all the sites in transects T4 and T5 were sinks. CH4 emissions (mean value: 0.2±0.45 mgm-2h-1) at the wetland sites were always lower in the dry season than those in the wet season. However, the sites on the hillslope grasslands served as CH4 sinks (mean value: -0.05±0.03 mgm-2h-1). In transect T5, CH4 emissions showed the opposite trend; a CH4 sink was observed in the wet season, but it was transformed into a CH4 source in the dry season.

Similar to the CO2 and CH4 emissions, N2O emissions showed a distinct spatiotemporal pattern in all the transects. N2O emissions in the wet season were higher than those in the dry season. These emissions were higher in the riparian wetlands than in the hillslope grasslands. Moreover, almost all sites with continuous river flow were N2O sources, while more than half of the sites with intermittent river flow were sinks.

Table 3 shows that CO2 fluxes were significantly correlated between the wet and dry seasons, while CH4 and N2O fluxes were not correlated between the two seasons.

Figure 6 shows the detailed spatial and seasonal patterns of GHG emission in the wet and dry seasons in the longitudinal direction from the upstream (T1, T2, and T3) to the downstream areas (T4 and T5). The CO2, CH4, and N2O emissions were calculated using the average values of the respective emissions in the wetlands and hillslope grasslands in each transect.

CO2 emissions at the riparian wetlands (Fig. 6a) in the wet season decreased from 2444.69±228.58 mgm-2h-1 in the upstream area to 665.08±347.57 mgm-2h-1 downstream, and the corresponding values for the dry season were 238.12±48.20 and 94.14±7.67 mgm-2h-1, respectively. However, in the hillslope grasslands (Fig. 6b), CO2 emissions exhibited no significant seasonality between the upstream and downstream areas, with mean values being 1103.40±190.44 mgm-2h-1 in the wet season and 79.18±24.52 mgm-2h-1 in the dry season. In addition, the CO2 emissions in transect T5 were low for both months, with averages of 162.83±149.15 mgm-2h-1 and 63.26±12.40 mgm-2h-1 in the wet and dry seasons, respectively. The upstream riparian zones exhibited higher CO2 emissions (894.32±868.47 mgm-2h-1) than their downstream counterparts (621.14±704.10 mgm-2h-1). Mean CO2 emissions showed no significant differences in the grasslands, averaging 524.16±450.10 mgm-2h-1 upstream and 508.06±534.77 mgm-2h-1 downstream.

CH4 emissions showed a marked spatial pattern in the riparian zones from upstream to downstream (Fig. 6c). The transects with continuous river flow were CH4 sources in the wet and dry seasons, with average emissions of 1.42±3.41 and 0.27±0.49 mgm-2h-1, respectively; while those with intermittent river flow served as CH4 sinks, with the corresponding means of -0.21±0.45 and -0.02±0.05 mgm-2h-1, respectively. Moreover, the hillslope grassland sites in all transects were CH4 sinks (Fig. 6d).

N2O emissions in riparian wetlands (Fig. 7e) showed spatial patterns similar to those of CH4 emissions. In the wet season, the transects with continuous river flow served as N2O sources, with a mean emission of 0.031±0.031 mgm-2h-1; meanwhile, transects with intermittent river flow acted as N2O sinks with an average emission of -0.037±0.05 mgm-2h-1. In the dry season, N2O emissions occurred as weak sources in the longitudinal transects, exhibiting an average emission of 0.002±0.007 mgm-2h-1. However, the N2O emission in the hillslope grasslands did not show any spatial patterns (Fig. 7f).

The results of this study emphasized that the rate of CO2 emission in the riparian wetlands was higher than that in the hillslope grasslands, owing to a variety of factors. ST is an important factor affecting GHG emission. Mclain and Martens (2006) showed that seasonal fluctuations in ST and SMC in semi-arid regions have important effects on CO2, CH4, and N2O emissions in riparian soils. Poblador et al. (2017) studied the GHG emission in forest riparian zones and suggested that the difference in the CO2 and N2O emissions in these zones is caused by the spatial gradient of the regional SMC. In this study, the upstream riparian wetlands were characterized by higher TOC content, lower soil C:N ratio, and more abundant BIO than those in the hillslope grasslands (Table 1). These soil conditions benefited the soil microbial activity, ultimately enhancing respiration as well as CO2 emissions. However, the CO2 emission in the downstream areas was nearly identical to that in the grasslands, likely because the wetlands gradually evolved into grasslands after their degradation. N2O emission showed spatial patterns similar to that of CO2 emission, likely because the CO2 concentration was closely related to the nitrification and denitrification processes. High CO2 concentrations can promote the carbon and nitrogen cycles in soil (Azam et al., 2005), increasing belowground C allocation, which is associated with increased root biomass, root turnover, and root exudation. Elevated pCO2 in plants provides the energy for denitrification in the presence of high available N, and there is increased O2 consumption under elevated pCO2 (Baggs et al., 2003). Moreover, soil respiration increases during soil denitrification (Liu et al., 2010; Christensen et al., 1990). In this study, a weak correlation was observed between the CO2 and CH4 emissions in the riparian zones (r=0.228), but CO2 emission was significantly correlated with N2O emission (r=0.322, P<0.05). The soil became anaerobic in the riparian areas as the SMC increased, and this was conducive to the survival of CH4-producing bacteria and to denitrification reactions, eventually leading to an increase in CH4 and N2O emissions. Jacinthe et al. (2015) reported that inundated grassland-dominated riparian wetlands were CH4 sinks (-1.08±0.22 kgCH4-Cha-1yr-1), and Lu et al. (2015) also indicated that grasslands were CH4 sinks. In our study, a marked water gradient across the transects led to the transformation of the soil from anaerobic to aerobic soil, which changed the wetland to either a CH4 source or sink. Therefore, during the transition from the riparian wetlands to the hillslope grasslands, CH4 sources only appeared in the near-stream sites, while sinks appeared at other sites.

Further, we compared the GHG emissions in the riparian wetlands and the hillslope grasslands around the Xilin River basin with those in various types of grasslands (meadow grassland, typical grassland, and desert grassland) in the Xilingol League in Inner Mongolia (Table 5). CO2 emission in the wet season decreased in the following order: upstream riparian wetlands > downstream riparian wetlands > hillslope grasslands > meadow grassland > typical grassland > desert grassland. Moreover, the upper riparian wetlands acted as sources of CH4 emission, while the downstream transects and grasslands served as CH4 sinks. Similarly, except in the downstream transects, N2O emissions occurred as weak sources in different types of grasslands and upstream riparian wetlands. The GHG emissions showed similar spatial patterns in October. Although these estimates were made only in the growing season in August and the non-growing season in October, our results suggest that the riparian wetlands are the potential hotspots of GHG emission. Thus, it is important to study GHG emission to obtain a comprehensive picture of the role of the riparian wetlands in climate change.

We roughly estimated the annual cumulative emission amounts of CO2, CH4, and N2O from the riparian wetlands and hillslope grasslands around the Xilin River basin and further calculated their global warming potential. As Table 6 indicates, annual cumulative emissions of CO2 and CH4 decreased in the following order: upstream riparian wetlands > downstream riparian wetlands > hillslope grasslands. N2O decreased in the following order: upstream riparian wetlands > hillslope grasslands > downstream riparian wetlands. In this study, we used the static dark-box method to measure CO2 emissions, which does not consider the absorption and fixation of CO2 by plant photosynthesis. Therefore, the total annual cumulative CO2 emissions are high. This result clearly showed the more significant impact of CO2 emission than that of CH4 and N2O emissions on global warming. The GWP depends on the cumulative emissions of the GHGs. The GWPs, shown in Table 6, were in the following order: upstream riparian wetlands (13 474.91 kghm-2) > downstream riparian wetlands (8974.12 kghm-2) > hillslope grasslands (8351.24 kghm-2). Therefore, both the riparian wetlands and the grasslands are the “sources” of GHGs on a 100-year timescale. The source strength of the wetlands is higher than that of the grasslands, further indicating that the riparian wetlands are hotspots of GHG emission.

The hydrology and soil properties showed evident differences between transects because the downstream zone was dry all year due to the presence of the Xilinhot Dam (Fig. 1). The dam caused the degradation of the riparian wetlands, resulting in reduced GHG emission. The average CO2 emission amounted to 1663 mgm-2h-1 in the upstream transects (T1, T2, and T3) at the riparian wetlands, while the downstream transects (T4 and T5) recorded an average emission of 1084 mgm-2h-1, 35 % lower than that in the upstream transects. The N2O emission from the riparian wetlands was lower in the downstream transects.

Wetland degradation first resulted in the continuous reduction of SMC, which led to the deepening of the wetland's aerobic layer thickness. Besides, SMC may affect ST and thus transport the CH4 emissions from a source to a sink by affecting methanogen activity (Yan et al., 2018). Second, the reduction of SMC impeded physiological activities of aboveground plants and inhabited related enzyme activities in the respiration process. Meanwhile, various enzyme reactions of underground microorganisms under water stress influence and reduced CO2 emissions (Zhang et al., 2017). Finally, after wetland degradation, long-term drought led to an extremely low SMC, which is not conducive to the growth of nitrifying and denitrifying bacteria and causes the transport of N2O emissions from source to sink (Zhu et al., 2013). As Table 1 shows, soil TOC content in the upstream transects (average: 25.1 gkg-1) was higher than that in the downstream transects (average: 8.41 gkg-1). The relatively low SMC and the aerobic environment were conducive to the mineralization and decomposition of the TOC. The degradation of plants in the wetlands led to the gradual reduction of BIO. Ultimately, the plant carbon source input of the degraded wetlands decreased, and the bare land temperature increased due to the reduced plant shelter. This accelerated the decomposition of TOC, leading to its decrease. This result indicates that wetland degradation caused the soil carbon pool's loss and weakened the wetland carbon source–sink function. These results are in agreement with those of Xia et al. (2017).

The degraded wetlands also caused soil desertification and salinization, leading to a decline in the physical protection afforded by organic carbon and a reduction in soil aggregates. Thus, the preservative effect provided by organic carbon declined. The TOC content and SMC in the dry lake bed in transect T5 were relatively high; however, the GHG emission was very low along this transect because soil pH values increased after the degradation of the lake soil, exceeding the optimum range required for microorganism activity. The soil C:N ratio was very high, resulting in severe mineralization and a low microbial decomposition rate, thus affecting the GHG emissions.