1 Introduction

Grasslands occupy approximately 40% of the Earth’s terrestrial surface (excluding Antarctica and Greenland; Suttie et al. 2005) and play an important role in providing essential ecosystem services, such as forage production for ruminant livestock (Boval and Dixon 2012), high biological diversity (Habel et al. 2013), water purification and retention, erosion reduction, carbon (C) sequestration as well as landscape and aesthetic values (Isselstein and Kayser 2014; Schirpke et al. 2017). In temperate Europe, managed grasslands are one of the most important land use practices, providing consumed feed sources for efficient ruminant meat and dairy products, and thus are increasingly gaining attention for their benefits to society (Kizeková et al. 2018; Nölke et al. 2021; Schils et al. 2022).

Nitrous oxide (N2O) and methane (CH4) are the two most important greenhouse gases (GHGs) after carbon dioxide (CO2) and are the main contributors to stratospheric ozone depletion (Jackson et al. 2019; Ravishankara et al. 2009), with global warming potentials (GWPs) that are respectively 265 and 28 times greater than CO2 on a 100-year time scale (IPCC 2014). The global atmospheric concentrations of N2O and CH4 have increased steadily at rates of ∼ 0.91 and ∼ 6.33 ppb yr− 1 over the last 20 years, respectively (Lan et al. 2022). Globally, managed grasslands represent huge potential for mitigating GHG emissions and have long been recognized for nitrogen (N) and carbon (C) storage (Wiesmeier et al. 2013). Nevertheless, grassland soils can act as sources or sinks for atmospheric N2O and CH4, depending on changes in environmental factors and management practices (Chang et al. 2021; Dijkstra et al. 2013).

Numerous studies have addressed the responses of N2O and CH4 fluxes to the changes in management intensification, land use change, and N deposition in European temperate grasslands (Drewer et al. 2017; Feigenwinter et al. 2023; Garnier et al. 2019; Ma et al. 2021; Reinsch et al. 2020). Abagandura et al. (2020) reported that intercropping Kura clover (Trifolium ambiguum M. Bieb) with prairie cordgrass (Spartina pectinata link.) not only increased biomass yield but also reduced cumulative N2O emissions and net global GWP compared with the prairie cordgrass monocultures. Most of the previous studies were conducted in controlled laboratory experiments or only during the growing seasons, but it is well documented that excluding the N2O emissions from the non-growing season would result in a reduction of N2O emission factors by up to 30% in agroecosystems globally (Shang et al. 2020), especially during freeze–thaw cycles in winter and early spring (Li et al. 2021; Luo et al. 2013). Furthermore, large episodic N2O emissions emitted from thawing grassland soils have been observed (Byers et al. 2021; Wolf et al. 2010). Wang et al. (2022) reported that CH4 uptake from a cold alpine meadow was increased under warming and the effect was stronger during unfrozen winter. Therefore, with the background of increased global warming during winter seasons, a better understanding of the responses of N2O and CH4 fluxes of grasslands will provide important implications for their management as well as for estimating GHG budgets.

N2O fluxes from soil are the result of a combination of microbial nitrification and denitrification, with N2O being produced as a by-product of nitrification and a gaseous intermediate product, evolved during the reaction processes of denitrification (Butterbach-Bahl et al. 2013). Higher microbial biomass C (MBC) and N (MBN) usually increase soil N2O emissions in terrestrial ecosystems via nitrification (Li et al. 2020, 2022). Whether soil acts as a source or sink of CH4 depends on the balance between the production of CH4 by methanogenic microorganisms under anaerobic conditions and oxidation by methanotrophic microorganisms predominately under well-aerated conditions (Le Mer and Roger 2001). Soil N2O and CH4 fluxes are therefore strongly regulated by environmental parameters that interact with soil water content (Dijkstra et al. 2013). In grassland systems, CH4 fluxes may be closely associated with plant species which can directly regulate soil water content (Fischer et al. 2019), especially during unfrozen winter, as methanotrophs are highly sensitive to temperature (Conrad 2007). Moreover, peak CH4 emissions also occurred during the thaw period, but the magnitude was controlled by soil moisture (Chen et al. 2021; Tagesson et al. 2012).Anaerobic oxidation of CH4 has been recently reported in terrestrial ecosystems (Fan et al. 2022), however its occurrence and relevance in grassland needs further systematic approval. An alternative to N input via fertilization are legume-based grasslands, which provide high-quality forage and decrease production costs by reducing mineral fertilizer demand (Lüscher et al. 2014; Rasmussen et al. 2012). Thus, introducing legumes (e.g., white clover, Trifolium repens) into grasslands may offer the potential to mitigate soil N2O emissions with no reduction in productivity (Barneze et al. 2020; Fuchs et al. 2020). However, knowledge on the different effects of legume and non-leguminous grassland species on soil N2O and CH4 fluxes specifically during winter seasons is still highly uncertain.

In this study, our objectives were to explore soil N2O and CH4 flux dynamics from grasslands of perennial ryegrass (Lollium perenne) and white clover throughout the winter period and to identify the relationships between N2O and CH4 fluxes with soil properties. We established a field study on a five-year-old experiment site in central Germany, with white clover grown as an unfertilized pure stand, N-fertilized and unfertilized perennial ryegrass as the non-leguminous pure stands, and bare soil treatment as no-vegetation reference. Related to these objectives, we hypothesized that (1) white clover and fertilized perennial ryegrass would have comparable soil N2O emissions, but larger than the unfertilized perennial ryegrass due to higher N availability; (2) soil N2O and CH4 fluxes would be mainly regulated by N availability and soil moisture, respectively.

2 Materials and Methods

2.1 Study Site and Experimental Design

This study was conducted at the experimental station of the Georg-August-University Göttingen in Deppoldshausen (51°34’ N, 9°58’ E, 342 m a.s.l), South Lower Saxony, Germany. The mean annual temperature and mean annual precipitation (during 2015–2019) recorded by the nearby weather stations were 9.5 ± 0.2 °C and 593.2 ± 57.8 mm, respectively (Nölke et al. 2022). The soil is classified as a calcareous Luvisol according to the World Reference Base for Soil Resources (WRB 2014) with a silty-clay soil texture (55% clay, 2% sand, and 43% silt in the top 30 cm). The general soil characteristics determined by our previous studies at the same site were as follows: pH (7.4), total N (1.6 g kg− 1), total phosphorus (P) (38 mg kg− 1), and total potassium (K) (133 mg kg− 1) (Nölke et al. 2022).

Prior to the start of the experiment site in 2014, the land was cultivated by conventional farming with crop rotation of winter oilseed rape, winter wheat, and winter barley. In summer 2014, stands were established in a randomized complete block design with four replicates, each plot had a size of 15 m2 (5 m × 3 m). A detailed description of our study site is shown in Nölke et al. (2022). The selected four treatments were: (1) white clover pure stand, (2) perennial ryegrass with and (3) without fertilization, and (4) bare soil. For ryegrass N1, 240 kg N ha− 1 was applied as NH4NO3 in four doses per year (80, 60, 60, and 40 kg N ha− 1), after each biomass harvest of perennial ryegrass.

2.2 Soil N2O and CH4 Flux Measurement

Soil N2O and CH4 fluxes were measured weekly during the winter period from 1 December 2019 to 14 March 2020 using the enclosed static chamber method. In each plot, two weeks before the first sampling day, a transparent chamber with the bottom removed and a resealable airtight lid equipped with a Luer-lock sampling port was inserted approximately 5 cm into the soil. The chambers were made of polyvinyl chloride, with a 22 × 22 cm area and 26 cm height above ground, with a total enclosed air volume of 12.6 L. During the gas sampling, plants growing in the plots of white clover and perennial ryegrass were included in the chamber. On each sampling day, gas samples were collected using a 30-mL polypropylene syringe at 1, 16, 31, and 46 min after the closure of the airtight lid. Each of the gas samples (25 mL) was immediately stored in a pre-evacuated Exetainer with septa (Labco Limited, Lampeter, UK). After each sampling, the collected gas samples were transported to the lab and were later analyzed for gas concentrations using a gas chromatograph (GC, Bruker SCION Model 456, Bremen, Germany), equipped with an auto-sampler, an electron capture detector (ECD; for the determination of N2O), and a flame ionization detector (FID; for the determination of CH4). Soil N2O and CH4 fluxes were calculated from the linear regression of the measured gas concentrations over time (46 min) and adjusted with the meantime measured air temperature and pressure according to the Eq. (1):

$$ Flux= \frac{dC}{dt}\times \frac{VPm}{AR(T+273.2)}$$
(1)

where, Flux is the N2O or CH4 flux (µg N or C m− 1 h− 1), dC/dt is the slope from the linear regression analysis, V is the total enclosed air volume of the chamber (m3), P is the actual atmospheric pressure (KPa), m is the molecular mass of the gas (g mol− 1), A is the soil area covered by the chamber (m2), R is the is the universal gas constant (8.314 Jmol− 1 K− 1), and T is the actual air temperature (°C), respectively.

2.3 Cumulative Fluxes and Global Warming Potential

The cumulative fluxes of soil N2O and CH4 from each plot were estimated using trapezoidal interpolation between gas fluxes and sampling day intervals (Wang et al. 2021). We estimated the combined N2O and CH4 fluxes to non-CO2 GWP for each treatment using the CO2-equivalents (CO2-eq) conversion factors of 265 for N2O and 28 for CH4 on a 100-year time scale (IPCC, 2014).

2.4 Soil Samples Analyses

Following gas sampling, soil samples were collected near each chamber (in the top 20 cm) using a hand auger to determine soil water-filled pore space (WFPS), mineral N (NH4+ and NO3−), and MBC and MBN on the same day. The in-situ soil (in the top 5 cm) and air temperatures were measured using a digital thermometer (GMH 1170, Greisinger electronic GmbH, Regenstauf, Germany). Daily soil temperature and precipitation data were collected from the climate station at Göttingen (station ID: 1691) of the German Meteorological Service. Air pressure was measured using a pressure gauge (GDH 200 − 14, Greisinger electronic GmbH, Regenstauf, Germany). In each replicate plot, three random sub-samples (∼ 50 g each) were taken, pooled, and mixed thoroughly in the field. Each of the mixed soil samples was divided into three parts and transferred to the laboratory for analysis. The first part (∼ 30 g) was oven-dried at 105 °C for one day to determine the gravimetric moisture content, which was calculated for WFPS using a theoretical particle density of 2.65 g cm− 3 for the mineral soil (Shao et al. 2023). The second part (20 g) was put into pre-prepared bottles containing 60 mL 0.05 M K2SO4 for mineral N extraction; the extraction bottles were shaken for 1 h and then filtered through 0.05 M K2SO4 pre-washed filter papers and then immediately frozen until analysis (Zang et al. 2016). The extractable NH4+ and NO3− were analyzed using a San + + continuous flow analyzer (Skalar Analytical, the Netherlands). The third part was used for the determination of soil MBC and MBN, for which about 10 g of mixed soil sample was put into pre-prepared 50 mL glass bottles and fumigated with ethanol-free chloroform for 24 h, followed by extraction with 40 ml 0.05 M K2SO4. The dissolved organic carbon (DOC) and total dissolved nitrogen (TDN) concentrations were measured with a TOC/TIC analyzer (Multi C/N 2100, Analytik Jena, Germany). Then, MBC and MBN were calculated as the difference in the DOC and TDON between the paired fumigated and unfumigated soils, and divided by the correction factors of 0.45 for MBC and 0.54 for MBN, respectively (Brookes et al. 1985).

2.5 Statistical Analysis

The measured soil N2O and CH4 fluxes and soil controlling factors (temperature, WFPS, mineral NH4+ and NO3− concentrations, MBC, and MBN) were tested for normality of distribution using the Shapiro-Wilk test and homogeneity of variance using Levene’s test. Non-normal distributed data were log or square root transformed (adding a constant value if negative values occurred, e.g., soil CH4 fluxes). Linear mixed-effects (LME) models (Crawley, 2007) were used to test the differences in soil N2O and CH4 fluxes and soil controlling factors among treatments. In LME models, treatment was considered as the fixed effect and sampling days and replicate plots were random effects. The significant differences were evaluated using the analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) test for multiple comparisons at P < 0.05. The cumulative fluxes and non-CO2 GWP were not tested for statistical differences among treatments due to the calculation method (trapezoidal extrapolation). The relationships between soil N2O and CH4 fluxes and controlling factors (i.e., temperature, WFPS, mineral NH4+ and NO3−) were determined by the Spearman’s rank correlation test, using the average of four replicates on each sampling day over the measurement period, statistical significances were considered at P < 0.05. All statistical analyses were performed using R version 3.6.2 (R Core Team, 2019).

3 Results

3.1 Soil N2O and CH4 Fluxes During Winter Season

Bare soil acted as a sink of N2O (ranging from − 0.61 to -0.02 µg N m− 2 h− 1; Fig. 1a and d) when WFPS was below 50% or above 79% (Fig. 1d), whereas soils from the three types of grassland were generally N2O sources (1.1–8.11 µg N m− 2 h− 1; Fig. 1a and d). Over the measurement period, fertilized ryegrass and white clover had higher soil N2O emissions than bare soil (P < 0.01; Fig. 2a); N2O emissions from ryegrass without fertilization did not differ significantly from the other treatments (P > 0.12; Fig. 2a).

Fig. 1
figure 1

Dynamics of mean (± SE, n = 4) soil N2O (a) and CH4 (b) fluxes, temperature in the top 5 cm (c), and water-filled pore space (WFPS; d) in the top 20 cm from bare soil, white clover, unfertilized (ryegrass N0) and fertilized (ryegrass N1) perennial ryegrass

Full size image
Fig. 2
figure 2

Season-mean soil N2O (a) and CH4 (b) fluxes from bare soil, white clover, unfertilized (ryegrass N0) and fertilized (ryegrass N1) perennial ryegrass. Lowercase letters above the error bars indicate significant differences among treatments (linear mixed-effects model with Fisher’s LSD test at P < 0.05). The black points represent mean values (± SE, n = 52) over the measurement period

Full size image

Soil CH4 fluxes did not show clear temporal patterns in December 2019 and January 2020, whereas in February and March 2020, significant temporal variations were observed among treatments (P < 0.05; Fig. 1b). Over the measurement period, there were no significant statistical differences in soil CH4 fluxes among treatments (P > 0.05; Fig. 2b). Bare soil and ryegrass without fertilization were net CH4 sources (with mean values were 2.3 and 1.9 µg C m− 2 h− 1, respectively; Fig. 2b), whereas soils from white clover and fertilized ryegrass were net CH4 sinks (with mean values were − 3.81 and − 0.23 µg C m− 2 h− 1, respectively; Fig. 2b).

3.2 Soil Parameters and Their Relationships with N2O and CH4 Fluxes

The daily 5-cm soil temperatures were all above zero and ranged from 0.3 to 8.5 °C (Fig. 1c and Fig. S1). Over the measurement period, the mean soil temperature did not differ among the treatments (P = 0.84; Fig. 3a). The WFPS ranged from 54 to 76%, 57–91%, 42–91%, and 52–84% in the bare soil, white clover, ryegrass with and without fertilization plots, respectively (Fig. 1d). The mean WFPS from white clover was the highest, followed by fertilized and unfertilized ryegrass, and bare soil (P < 0.01; Fig. 3b).

Fig. 3
figure 3

Season-mean (± SE, n = 52) soil temperature (a), water-filled pore space (WFPS; b), NH4+ (c), NO3− (d), microbial biomass carbon (e), and microbial biomass N (f) from bare soil, white clover, unfertilized (ryegrass N0) and fertilized (ryegrass N1) perennial ryegrass. Lowercase letters above the error bars indicate significant differences among treatments (linear mixed-effects model with Fisher’s LSD test at P < 0.05)

Full size image

Fertilized ryegrass had highest soil NH4+ contents of all other treatments (P < 0.03; Fig. 3c). White clover and fertilized ryegrass had higher NO3− contents than unfertilized ryegrass (P < 0.01; Fig. 3d), and had higher MBC and MBN contents than bare soil (P < 0.01; Fig. 3e and f). Unfertilized ryegrass had similar NH4+, MBC, and MBN contents as white clover (P > 0.63; Fig. 3c, e, and f).

Across treatments, soil N2O fluxes showed a weak positive relationship with soil temperature and NH4+ (Fig. 4a and c), whereas they correlated significantly with soil WFPS and NO3− (Fig. 4b and d). Soil CH4 fluxes correlated negatively with WFPS (Fig. 4f), whereas there were no significant relationships detected between soil CH4 fluxes and any other abiotic environmental factors (Fig. 4e, g, and h).

Fig. 4
figure 4

Spearman rank correlations of soil N2O and CH4 fluxes with soil temperature (a, e), water-filled pore space (WFPS; b, f), mineral NH4+ (c, g), and NO3− (d, h). Each data point represents a measurement value from each sampling day, and from bare soil, white clover, unfertilized (ryegrass N0) and fertilized (ryegrass N1) perennial ryegrass. ρ (rho) indicates spearman’s rank correlation coefficient, statistical significances were considered at P < 0.05

Full size image

3.3 Area-Scaled Gas Fluxes and Non-CO2 Global Warming Potential

During the winter season, cumulative N2O emissions indicated soils from all plots acted as N2O sources, with values ranged from 13.9 to 43.5 kg CO2-eq ha− 1. Specifically, fertilized ryegrass had the largest cumulative N2O emissions, which were 56%, 14%, and 212% larger than unfertilized ryegrass, white clover, and bare soil, respectively (Fig. 5a). Cumulative CH4 fluxes varied from − 4.2 to 2.3 kg CO2-eq ha− 1, and soil under white clover acted as a net CH4 sink, whereas soil from the other treatments were all net CH4 sources (Fig. 5b). The non-CO2 GWP across the study period ranged from 16.2 to 44.2 kg CO2-eq ha− 1. The fertilized ryegrass had the largest non-CO2 GWP, followed by white clover, unfertilized ryegrass, and bare soil (Fig. 5c).

Fig. 5
figure 5

Estimated areal mean (± SE, n = 4) cumulative soil N2O emissions (a), CH4 fluxes (b), and combined N2O and CH4 global warming potential (non-CO2 GWP; c) within bare soil, white clover, unfertilized (ryegrass N0) and fertilized (ryegrass N1) perennial ryegrass. Cumulative soil N2O emissions and CH4 fluxes were not tested statistically since these values were calculated using the trapezoidal interpolation

Full size image

4 Discussion

Our results show that soil from all plots were net sources for N2O across the winter period. The mean and cumulative N2O fluxes in bare soil were relatively small, where both N2O production and consumption occurred. Many earlier studies reported that overwinter N2O emissions from managed grasslands contributed a large portion to the annual GHG budget (Byers et al. 2021; Li et al. 2021; Wolf et al. 2010), which resulted from the short-lived peak of N2O fluxes during freeze–thaw cycles. However, we did not detect any N2O pulse emissions in any of our plots, probably due to the above-zero soil temperature limiting freeze–thaw events during the measurement period (Tang et al. 2022). This was in line with a previous study in an alpine grassland that observed no stimulated N2O fluxes, not even during spring thaw, and found average winter emissions accounted for only 16.7% of the annual N2O budget (Li et al. 2012). Contrary to our expectation, soil N2O emissions from unfertilized perennial ryegrass were comparable with those from white clover and fertilized perennial ryegrass. Similar findings were observed by Barneze et al. (2020) in the UK, with no significant differences in soil N2O emissions among legume pure stand, fertilized and unfertilized grass monoculture. In contrast, Byers et al. (2021) reported that red clover in pure stand had significantly higher soil N2O emissions than grass-only plots in southern Norway. Below we provide several reasons explaining why white clover’s N2O emissions may not have exceeded that of the grasses in our study.

(1) Across plots, the significant positive relationships between soil N2O fluxes with NH4+ and NO3− concentrations confirm that N availability strongly regulates N2O emissions during unfrozen winter seasons. Although there was no significant difference in soil N2O emissions from fertilized and unfertilized perennial ryegrass, similar trends in mineral N and N2O emissions were observed in the fertilized plots. Similar to our findings, Röver et al. (1998) reported that N2O emissions did not differ significantly between fertilized and unfertilized arable soils over the winter period, and the source of N2O may be independent from the N-fertilizer applied during growing seasons. Wang et al. 2020 demonstrated that, beside soil mineral N, the availability of organic C was a more limiting factor for denitrification and N2O production in unfrozen winter. This was also in line with our results, that the lowest N2O emissions were recorded on bare soil, which had a much lower organic C input from root exudation and clearly lower microbial biomass than other five-year grass species.

(2) Although soil N2O fluxes were positively related to WFPS, nitrification activities would be limited under wetter conditions. When the WFPS was above 60%, the dominant source of N2O emissions is linked to denitrification, and with further increases of WFPS, more N2O is reduced to N2 (Davidson et al. 2000). During our measurement period, most of the WFPS values ranged from 40 to 65%, except for one measurement conducted immediately after rainfall in February 2020, indicating that nitrification might contribute to most N2O production. The unfertilized white clover had the same level of soil N2O emissions as the fertilized perennial ryegrass, but exceeded that of bare soil, which might be explained by the following two aspects. First, compared with perennial ryegrass, white clover has a taproot and less extensive root system, which makes it possible that less water is required for above- and below-ground biomass production (Karsten and MacAdam 2001). In our study, white clover plots had the highest mean WFPS (64%), which was probably at the optimum moisture conditions favorable to both nitrification and denitrification for N2O production (Rafique et al. 2011), and thus had higher N2O emissions. Second, the high N2O emissions under white clover are probably due to the increased soil mineral N (especially NO3−) from legumes, released from fixed N from root decomposition and limited N uptake from soil (Niklaus et al. 2006). These results could be further supported by the higher MBC and MBN contents in white clover and fertilized perennial ryegrass plots than bare soil, which indicate greater microbial activities, that might be involved in nitrification and denitrification processes for N2O production (Butterbach-Bahl et al. 2013; Li et al. 2020).

We found that only soil from the white clover plots was a net CH4 sink, while the bare soil and unfertilized perennial ryegrass plots were net sources for CH4 throughout the measurement period. Although the fertilized perennial ryegrass plots represented a negative mean soil CH4 flux (-0.2 µg C m− 2 h− 1), the large temporal variation led to a small estimated cumulative CH4 emission. These findings are consistent with the results by Imer et al. (2013), who reported both positive and negative CH4 fluxes occurred and showed highly temporal variations in soils from three managed Swiss grasslands, but concluded that all sites were weak CH4 sinks because the fluxes were dominated by uptake. In this study, negative soil CH4 fluxes from white clover occurred more frequently during the measurement period, and consequently white clover had the highest soil CH4 uptake compared with the other treatments, which supported our second hypothesis.

Previous studies have shown that soil CH4 fluxes could be affected by N availability. As in fertilized soils, the increased NH4+ may compete with CH4 for the same active sites of methane monooxygenase, thus leading to an inhibition of CH4 oxidation (Steudler et al. 1989). Nonetheless, fertilization-induced high NO3− levels can also stimulate CH4 uptake by limiting methanogenesis, which can be inhibited by toxic intermediates produced from denitrification (Roy and Conrad 1999). However, we did not detect any relationships between soil CH4 fluxes with NH4+ and NO3− concentrations, which can possibly be explained by the low and small changes in N availability during the experiment period. Previous studies have shown the positive effect of soil temperature on soil CH4 uptake (Drewer et al. 2017; Wang et al. 2022), since elevated temperature generally contributes to a lower soil water content as well as increased methanotrophic activity, and thus stimulates CH4 consumption in soils (Dobbie and Smith 1996). There was no significant relationship between soil temperature and CH4 fluxes. Given the small temporal variation in soil temperature, and the fact that all treatment plots were established at the same site, current technology may not be able to detect the effect of soil temperature on CH4 emissions. It should be underlined that soil CH4 fluxes were negatively related to WFPS, while it was generally reported that CH4 uptake decreased with increasing soil water content (Liu et al. 2017; van Delden et al. 2018). Other previous studies demonstrate that soil CH4 uptake was parabolically correlated with soil moisture, that the activity of methanotrophs was inhibited under dry condition, and high soil moisture limited air diffusion into the soil (Borken et al. 2006; Dijkstra et al. 2013). Across the measurement period, the white clover plots had the highest WFPS and also the largest CH4 uptake, which may indicate that methanotrophic activity was relatively high and soil moisture in our plots was still under the optimum soil water content for CH4 oxidation in the cold environment (Luo et al. 2013; Wang et al. 2022). Overall, our results suggest that the root system of white clover was able to maintain greater soil water than perennial ryegrass, which probably caused an enhancement of soil CH4 uptake during unfrozen winter seasons. It may also suggest that the extension of white clover in rotation or as an intercrop would have a positive effect on the reduction of atmospheric CH4, at least on a regional scale.

5 Conclusions

We investigated soil N2O and CH4 fluxes from three grassland types and bare soil over the unfrozen winter period on an established five-year-old grassland. Our results suggest that N2O emissions can be maintained at low levels in both legumes and non-legumes during unfrozen winters, mainly due to limitation of mineral N, whereas methane is driven more by soil water-filled pore space, such results suggest that more developed root systems and water-holding capacity may be more advantageous for maintaining soil as a CH4 sink under a comparable climate. Given that anthropogenic climate change (e.g., winter warming) is becoming more frequent globally, our results suggest that more attention should be paid to soil greenhouse gas emissions in grassland systems during wet and non-freezing winter seasons.