The study was carried out in a wetland at the Naqu Ecological Environment Observation Station (E91°12ʹ∼93°02ʹ, N30°31ʹ∼31°55ʹ) of the Tibetan Plateau Research Institute, Chinese Academy of Science. The type of the wetland is an alpine swamp meadow, with an altitude of 4,480 m and a soil type of inceptisol (Lin et al., 2020). Hollows comprise approximately 10∼20% of the surface area with 5–15 cm of standing water or ice and are dominated by the Carex community. Hummocks, elevated above the water table throughout the snow-free seasons, comprise approximately 80∼90% of the surface area and are dominated by the Tibetan Kobresia community. The above-ground biomass in hollows and in hummocks was estimated by using the empirical formula: biomass (g dry weight/m²) = 24.038 * height (cm) * coverage +0.1261 deduced from a bunch of height and weight measurements in 1 m * 1 m patches in the Qinghai–Tibet Plateau (Wang, 2011).

The site has a cold, alpine mountain climate with an average annual precipitation of 406.2 mm and a mean annual temperature of −2.1°C. The growing season is between the end of May and early October with more than 80% of the annual precipitation. The freezing season is between the end of October and early May, but the specific dates vary from year to year. July–August and January feature the highest and lowest mean temperatures, respectively. Observations were carried out through a vegetation period from May to October in 2019.

IPCC suggested a global warming threshold of 1.5°C above pre-industrial levels. In order to evaluate the effect of global warming of 1.5°C on the GHG emission from the Tibetan wetlands, two warming treatments, designed to warm 1°C (T1) and 2°C (T2) for soil temperature according to our pilot study, as well as a warming control (with no warming device), were conducted for hollows and hummocks in the study area (Figure 1). The warming treatments were permanently set up with passive solar hexagon open-top chamber (OTC) warming devices. The OTC devices were constructed with polymeric methyl methacrylate beads with a light transmittance of 95% according to previous publications (Godfree et al., 2011; Li et al., 2020). Specifically, the bottom edges of the T1 and T2 devices were 75.1 and 86.7 cm, respectively. The top edges of both devices were the same, 57.7 cm. The heights of the T1 and T2 devices were 60 and 100 cm, respectively. The bottom edges were tightly embedded in the soil. Two small fans were installed in the chambers to mix the air. All treatments were repeated at four different sites for both hollows and hummocks. In order to reduce the effects of different patches on the GHG emission, warming devices were adjacently distributed for hollows and hummocks. At some sites, PVC pipes were set up to observe the dynamic changes of the water table. A portable soil temperature and soil-moisture recorder (RHD-13, Handan Ruihua Electronics Co., Ltd., Handan, China) was used to observe the soil temperature and moisture at 5-cm depth when measuring GHGs.

During the growing season (May to October) in 2019, CO₂ and CH₄ fluxes were measured with a portable greenhouse gas analyzer (LGR-UGGA 915-0011, ABB Inc., Quebec, Canada) every 1–2 weeks, when the weather was sunny with few clouds. The measurements were conducted between 09:00 a.m. and 11:00 a.m. to better represent a daily mean value (Xu and Qi, 2001). The measuring duration lasted for 2 min each time, and the instrument automatically recorded data every 1 s. The CO₂ and CH₄ concentrations, as well as the air chamber temperature and atmospheric pressure, were recorded as air passed through the analyzer. The flux of CO₂ under light conditions represented the net ecosystem carbon exchange (NEE). The flux of CO₂ in darkness, as measured with a black cloth covering the OTC device, represented ecosystem respiration (ER), which was measured directly after the NEE measurements for each treatment. To avoid the influence of operation, the data recorded in the first 25 s and the last 25 s were excluded. NEE and ER were calculated with CO₂ concentrations under light and darkness, respectively, with the following formula. F c = V × P a v × ( 1000 − W a v ) R × A × ( T a v + 273 ) × d C d t .

Fc: flux of CH₄, NEE, or ER (μmol·m⁻²·s⁻¹); V: volume of the assimilation chamber (m³); W_av: partial pressure of water in the chamber during the measurement (mmol·mol⁻¹); P_av: the average atmospheric pressure (KPa) in the chamber during the measurement; R: the atmospheric constant (8.314 J mol⁻¹·K⁻¹); A: the area of the base (m²); T_av: the average temperature in the chamber during the measurement (°C); and dC/dt: the rate of change of dry CO₂ and CH₄ concentrations in the chamber with time.

A positive value indicates that the ecosystem is in carbon release (carbon source), and a negative value indicates that the ecosystem is in carbon absorption (carbon sink). Gross primary production (GPP) was calculated by subtracting ER with NEE with a black cloth covering the outsides of the chambers. CH₄ fluxes were also measured simultaneously.

The experimental data were statistically analyzed by SPSS 16.0 statistical software. The paired sample t-test was used to analyze the significance of carbon flux differences among various warming treatments. Pearson correlation was used to analyze the correlation between the carbon flux and environmental parameters. Correlations and differences were considered to be significant at a level of p < 0.05.

During the growing season, soil temperatures at the 5-cm depth in hollows averaged 7.32°C in control treatments and 7.94 and 8.40°C in T1 and T2, which were 0.62 and 1.08°C higher than the control treatments, respectively (Figure 2A). In hummocks, soil temperatures at the 5-cm depth averaged 8.54°C in control treatments and 9.38°C and 10.85°C over the growing season for T1 and T2, which were 0.84 and 2.31°C higher than the control treatments, respectively (Figure 2B). Naturally, the soil temperature of the hummocks was warmer than that of the hollows. In contrast, the changes of the temperature in warming chambers recorded from 10:30 a.m. to 1:00 p.m. were largely dependent on the measuring time, other than the different treatments (Supplementary Figure S1).

For most cases, the hummocks elevated above the water table for 0–25 cm, and the hollows were logged in water (Supplementary Figure S2). The water table obviously declined in October. Warming treatments did not change the water table under the chambers significantly. However, soil moisture at the 5-cm depth decreased significantly with warming treatment (Figures 2C,D). During the peak growing season (from mid-June to mid-August), the hollows were flooded, and the soil moisture was saturated (Figure 2C). For hummocks, the warming treatments significantly reduced the soil moisture throughout the growing season (p < 0.05). The average soil moisture under the control treatment was 58.7%, but it decreased to 54.1 and 50.8% under warming treatments in T1 and T2, respectively (Figure 2D). On the whole, the average values of the above-ground biomass from hummocks were higher than those from the hollows (Supplementary Figure S3). The average values of the above-ground biomass under warming treatments (H1 and H2) were higher than those under control conditions. However, there were no significant differences among the treatments for both hummocks and hollows across the growing season.

NEE refers to the net flux of CO₂ between the ecosystem and atmosphere. Since we follow the atmospheric sign convention, negative NEE fluxes denote an uptake by the ecosystem whereas positive fluxes denote emissions. The average values of NEE in the daytime were all negative and significantly different from each other among the different warming treatments (including control, T1, and T2 treatments) for both hollows and hummocks (Figure 3A). Among them, the respective average NEE values of hollows under the control, T1, and T2 treatments were −1.82, −3.11, and −6.05 μmol m⁻²·s⁻¹, respectively. The respective average NEE values of hummocks under the control, T1, and T2 treatments were −18.92, −26.26, and −28.61 μmol m⁻²·s⁻¹, respectively (Figure 3A). CO₂ fluxes in darkness represent ecosystem respiration (ER). Within the same microtopographic features, ER values under warming treatments were significantly higher than those from control treatments, but no significant differences were observed between the two warming treatments (Figure 3B). Specifically, the average ER values of the hollows under the control, T1, and T2 treatments were 5.42, 6.39, and 7.13 μmol m⁻²·s⁻¹, respectively, while for hummocks, they were 12.80, 13.98, and 14.73 μmol m⁻²·s⁻¹, respectively (Figure 3C). Like NEE, ER values from hummocks were also significantly higher than those from hollows. The differences between NEE and ER fluxes represent the uptake by the ecosystem, i.e., the gross primary production (GPP). The average GPP values of hollows under the control, T1, and T2 treatments were 7.24, 9.51, and 13.18 μmol m⁻²·s⁻¹, respectively. The respective average GPP values of hummocks under the control, T1, and T2 treatments were 31.72, 40.23, and 43.34 μmol m⁻²·s⁻¹, respectively. Generally, NEE, ER, and GPP values from hummocks were all significantly higher than those from hollows.

Over the course of time, the varying patterns of NEE, ER, and GPP differed greatly (Figure 4). NEE reached their largest fluxes in late July for both hummocks and hollows (Figures 4A,B). Positive values of NEE from hollows during May and June suggested the sites were CO₂ sources in this period. ER values in hummocks and hollows were the greatest during the first half of the growing season from May to July (Figures 4C,D). The fluxes of GPP reached their greatest values during late June in the hollows (Figure 4A) and during late July in the hummocks (Figure 4B). According to the average time course of CO₂ fluxes (Figures 3, 4), the warming treatments greatly increased CO₂ and its component fluxes of both hummocks and hollows.

Warming also significantly increased the average CH₄ emissions throughout the growing season relative to the control treatment for both hollows and hummocks under both light and dark conditions (Figures 3D,E). Generally, the fluxes of CH₄ under the light condition were higher than those under the dark condition. The fluxes of CH₄ from hummocks were higher than those of hollows. With warming treatments, the average fluxes during the growing season increased from 0.76 μmol m⁻²·s⁻¹ (control treatment) to 1.33 μmol m⁻²·s⁻¹ (T2 treatment) for the hollows with light and from 0.11 μmol m⁻²·s⁻¹ (control treatment) to 0.62 μmol m⁻²·s⁻¹ (T2 treatment) for the hollows in darkness. For the hummocks, it increased from 1.38 μmol m⁻²·s⁻¹ (control treatment) to 3.65 μmol m⁻²·s⁻¹ (T2 treatment) with light and from 0.20 μmol m⁻²·s⁻¹ (control treatment) to 0.85 μmol m⁻²·s⁻¹ (T2 treatment) in darkness.

Interestingly, the CH₄ emissions exhibited different temporal variation patterns (Figure 5). At the first stage, before the end of June, CH₄ emissions from hollows were all around 0 μmol m⁻²·s⁻¹ for the different warming treatments, regardless of the lighting condition. Then, fluxes increased rapidly for the warming treatments with light but increased only moderately over the same time period for the warming treatments when measuring in darkness (Figures 5A,C). The fluxes reached the highest for the hollows at the end of August then decreased rapidly for the measurements with light and decreased more moderately for the dark measurements. For the hummocks, the variation patterns of the CH₄ fluxes in darkness were similar to those for the hollows in darkness. The CH₄ fluxes for the hummocks in light increased with fluctuations from early May to the end of August and then decreased gradually (Figures 5B,D).

The correlations of fluxes of CH₄ and CO₂ and environmental factors were evaluated (Table 1). During the growing season, GPP, ER, and NEE of hollows were all significantly correlated with the soil temperature and soil moisture. GPP, ER, and NEE of hummocks were all significantly correlated with the soil temperature and above-ground biomass. Additionally, the ER of hummocks was also significantly correlated with soil moisture. CH₄ emissions from both hollows and hummocks were significantly correlated with soil temperature and above-ground biomass, regardless of the lighting condition. However, the CH₄ emissions from hummocks in light were also significantly correlated with soil moisture (Table 1).

Our observations of CO₂ fluxes demonstrate that all the components of these CO₂ fluxes in the wetland were highly sensitive to changes in temperature. GPP is realized by plants. It is well-known that global warming has a positive effect on GPP, especially for alpine environments where temperature is a limiting factor for GPP (Xia et al., 2009; Peng et al., 2014). ER is mainly contributed by both plants and microbes. Higher temperature also has a positive effect on ER from plants and microbes (Atkin et al., 2000; Wang et al., 2014). The size of NEE is jointly determined by the size of GPP and ER. In our study, NEE increased with global warming which suggested that the increase in GPP was greater than that in ER under global warming. A previous study in the alpine meadows of the Qinghai–Tibet Plateau also found the similar results, i.e., NEE increased under warming conditions (Peng et al., 2014).

Our study also showed increases in the soil temperature accelerated the loss of soil moisture (Figure 2). The decrease in soil moisture will inevitably affect root respiration, microbial community composition, and plant physiology and thus affect the global carbon cycle (Selsted et al., 2012; Xu et al., 2014). Specifically, moisture determines the soil’s redox potential, i.e., decrease in soil moisture will increase O₂ concentrations and redox potential of the soil (Dinsmore et al., 2009; Rubol et al., 2012; Esch et al., 2016). In wetlands where water is not a limiting factor for microbes and plants, an increase in O₂ concentration induced by a decrease in soil moisture might increase both GPP and ER. A previous study showed that insignificant changes in GPP and ER resulted in a consequently small NEE response under global warming in a semi-arid temperate steppe in northern China (Xia et al., 2009). In the arid environment, warming reduced soil moisture to a certain extent, resulting in weakening of the plant and soil microbial respiration, which inhibited plant stomatal conductance, reduced plant photosynthesis, and affected NEE (Guo et al., 2011). These contrasting findings with the results from this study highlight the importance of soil moisture in mediating the response of NEE to global warming.

Distinct temporal variations were also obvious in NEE, ER, and GPP fluxes along with the soil moisture and soil temperature. During the early growing season, the snow did not totally melt and soil temperature was a little above 0^oC (Figure 2). Low temperature inhibits the activity of microorganisms (Liu et al., 2009) and affects vegetation respiration (Hicks Pries et al., 2015). During this period, soil temperature was the main factor affecting GPP and ER. Consequently, warming promoted GPP and ER, as well as NEE. From July to early September, the weather of the Tibetan Plateau is controlled by the Indian Ocean monsoon with optimal temperature and precipitation (Hou et al., 2014). The amount of precipitation and cloudy days increase in comparison with those of the former period. Vigorous vegetation growth contributed to high GPP and NEE under warm air temperature (Supplementary Figure S1). During this period, soil temperature still tends to be the most important environmental factor that affects ER. In September and October, plants enter the wilting period (Supplementary Figure S3), and photosynthesis weakens, resulting in a decrease of NEE for both hollows and hummocks.

Our study showed that warming significantly increased CH₄ emission from the wetlands for both hollows and hummocks, regardless of the lighting condition. Increase in CH₄ emission was also observed from microcosms with soil from the Zoige wetland in the Tibetan Plateau under warming conditions (Cui et al., 2015). CH₄ emission is determined by the balance between methanogenesis and methanotrophy, which is affected by various environmental conditions (Serrano-Silva et al., 2014). A previous study showed that high temperature increased methanogenesis and decreased CH₄ oxidation in rice soils (Das and Adhya, 2012). Rise in temperature potentially leads to decrease in the soil redox potential, which favors methanogenesis but inhibits the activity of CH₄-oxidizing bacteria (Das and Adhya, 2012). Such contrasting effects on methanogenesis and methanotrophy necessarily resulted in the fact that warming increases CH₄ emission, which was consistent with our observations. The significant correlation between the CH₄ flux and soil temperature (Table 1) suggested that temperature was one of the controlling factors affecting the CH₄ flux in the alpine wetlands on the Tibetan Plateau.

The higher flux of CH₄ emissions with light than those in darkness may be caused by vegetable internal convective gas transport which was reported to quickly respond to changes in irradiation (Günther et al., 2014). Many plants, such as Typha, Phragmites, Kobresia, and Carex dominating the wetland in this study, have been found to transport CH₄ from the pedosphere to the air through aerenchyma (Morrissey et al., 1993; Yavitt and Knapp, 1998; Beckett et al., 2001; Cao et al., 2008). By this way, CH₄ escapes from oxidation in aerobic peat layers (Günther et al., 2014). Notably, the internal convective gas transport is sensitive to irradiation through stomatal control (Morrissey et al., 1993; Günther et al., 2014). So, our findings about the differences of CH4 emission in darkness and light were largely due to the stomatal control under different light conditions.

Our study also showed that the fluxes of CH₄ exhibited a temporal pattern. The peak CH₄ emission in the two microtopographic features of the wetland in this study appeared in summer, consistent with previous research results in other wetlands (Dise, 1993; Chen et al., 2008; He et al., 2014; Wei et al., 2015). Beyond the fact that the temperature is optimal for CH₄ emission in summer as discussed previously, summer is also the optimal period for vegetation growth. A high input of fresh organic carbon to the soil resulting from vigorous vegetation growth will fuel methanogenesis (Serrano-Silva et al., 2014). The significant correlation between the CH₄ flux and above-ground biomass also explained such seasonal changes in the flux of CH₄ emission.

Microtopography also influenced CH₄ fluxes, i.e., the CH₄ fluxes of hummocks were higher than those of hollows, which may be attributed to two reasons. First, hummocks grow greater amounts of above-ground biomass than hollows, which will allocate more fresh organic carbon to the pedosphere and facilitate methanogenesis (Serrano-Silva et al., 2014). Second, high moisture or the water table affects CH₄ fluxes. The wetland of this study was recharged with melt water, which was dynamically oxic. As a result, oxic water may inhibit methanogenesis and promote CH₄ oxidation.

According to our study, global warming strengthens the CO₂-sinking effect in the daytime and also increases CH₄ emission from the Tibetan wetlands. Warming for 1–2 ^oC increased NEE by an average of 7.3–9.7 μmol m⁻²·s⁻¹ under sunlight during the growing season for hummocks contrary to the control conditions and 1.3–4.2 μmol m⁻²·s⁻¹ for hollows. The values for CH₄ emission were lower than these NEE values, which were 1.3–2.3 μmol m⁻²·s⁻¹ and 0.35–0.57 μmol m⁻²·s⁻¹, respectively. However, CH₄ possesses a higher global warming potential (GWP), which is about 28 times higher than that of CO₂ (IPCC, 2014). From this aspect, an increase in CH₄ emissions would bring 3.8 to 7.5 times the warming effect contrast to NEE in the daytime during the growing season.