Freezing and thawing cycles affect nitrous oxide emissions in rain-fed lucerne (Medicago sativa) grasslands of different ages

Site description

The study was undertaken at the Qingyang Loess Plateau Research Station of Lanzhou University, China (lat. 35°39′N, long. 107°51′E, elevation above sea level 1,297 m) during 2013–2014. The climate is a typical cool continental climate, with mean annual precipitation of 548 mm, with the highest abundance of rainfall from late summer (July) to autumn (September). The average annual temperature ranges from 8 °C to 10 °C. The mean temperatures in the warmest (July) and coldest (January) months are 21.3 °C and −5.3 °C, respectively. The growing season extends from spring (March) to mid autumn (October) for about 255 d with 110 frost-free days on average. The soil type is Heilu soil, which is defined as silty loam soil based on the U.S. classification system (Entisol of FAO classification). The properties of sampled soils at the start of the measurement periods are presented in Table 1.

Experimental design and N₂O emissions measurements

The experiment was a randomized block design comprising two treatments with 4-year-old lucerne and 11-year-old lucerne (Medicago sativa L. cv Longdong) cultivation, which were planted following maize crops (Zea mays L.). Afterwards, neither of these two lucerne grasslands was further reseeded. A fallow plot (F) with no retention of previous crop residue and no soil disturbances was used as control. Plot dimensions were 3 × 4 m and three replicates were used. Both lucerne grasslands were drilled sown, with 0.3 m row spacing and a sowing rate of 15 kg ha⁻¹. Lucerne grasslands were rain-fed and ammonium phosphate was applied at a rate of 302 kg ha⁻¹ at the recovering (or regreening) stage each year. Lucerne was cut twice each year as forage when the grassland reached the flowering stage, which is the typical cultivation mode in this region (Shen et al., 2009).

N₂O emissions were measured in the autumn-winter period (28 November 2013), the winter-spring period (6 March 2014), and during a mild period in winter (7 January 2014). These 3 days were selected as they represent typical temperature fluctuations during a freeze/thaw for across the soil profile, cycle amplitude, and minimum temperature of sub-surface soil (Henry, 2007; Kariyapperuma et al., 2011). Emissions were measured using a N₂O/CO near-infrared gas analyzer (Model DLT-100; Los Gatos Research, Inc., San Jose, CA, USA) connected to automated sampling chambers (diameter 350 mm, height 600 mm, model SC-03, LI-CA, China) placed in each plot. Sampling chambers were fitted with a small fan mounted inside the lid to maintain well-mixed conditions. The chambers were distributed randomly in each plot and were embedded with chamber collars, which were pressed into the soil to a depth of 10 cm two weeks prior to the first gas measurement (Fig. S1). For one replicate of N₂O measurement, e.g., the chamber under F was closed while chambers under 4-year-old and 11-year-old grasslands remained open. Gas samples of each plot were taken every 1 min in sequential order; hence, one replicate of N₂O measurement from F, 4-year-old, and 11-year-old grasslands required a total of 20 min. The sampling sequence was repeated a further two times, yielding a total of three replicates over a 60 min period. Automated chambers were sealed airtight during the sampling procedure by two lids that closed and opened via an air pump.

Data were logged by a soil flux system (MCC-1-8, LI-CA, China). The main problem observed during these experiments was related to chamber sealing in the field condition; thus, the despike function in the ‘oce’ package (Kelley, 2018) was used to identify N₂O spikes with respect to a “reference” time-series. Spikes with N₂O values of 2 min before and after the time point when a spike occurred were replaced with a reference value. Eventually, out of 648 emissions, 69% were real observations, and 21% were replaced by a reference value. The remaining 10% had no values. Daily N₂O emissions were calculated and integrated over time to obtain cumulative emissions over the study period.

Soil measurements and analysis

During each sampling period, soil temperature and volumetric water contents were measured hourly by probes placed close to each chamber, installed at depths of 50, 150, and 250 mm. Soil samples, at depths of 0–100, 100–200, and 200–300 mm, were taken at five sites near each chamber using a 50 mm diameter gouge auger at around 10:00 am. Soil samples were placed in airtight plastic bags, transported to the laboratory, and stored temporarily at 4 °C in the dark (for less than 1 week). Fresh soil was analyzed for MBC, MBN, and mineral N (NH₄⁺-N and NO₃⁻-N). Total organic carbon (SOC), total N (TN), and soil pH were analyzed using air-dried soil.

SOC was measured using dichromate oxidation and total N according to the Kjeldahl method (Kjeltec™ 8400; FOSS, Sweden). Soil mineral nitrogen (NO₃⁻ and NH₄⁺) were extracted with 1 M KCl and measured with a colorimetric continuous flow analyzer (Model FIAstar 5000; FOSS, Sweden). Soil pH (1:2.5 soil-water paste) was measured using electrometry with a pH electrode (Model PSH-25C; JTONE, Shantou, China).

Soil MBC and MBN were determined using the fumigation-extraction method (Brookes et al., 1985; Vance, Brookes & Jenkinson, 1987). One portion of fresh soil (20 g oven-dried equivalent) was shaken with 80 mL of 0.5 M K₂SO₄ for 30 min, filtered through quantitative filter paper (medium-speed), and frozen at −20 °C prior to analysis. Simultaneously, another soil sample was fumigated with ethanol-free chloroform for 24 h at 25 °C and then, the extraction was undertaken as above. Dichromate oxidation was used to determine total organic extracts. The concentration of N in the extracted sample was measured colorimetrically with a continuous flow analyzer (Model FIAstar 5000; FOSS, Sweden). MBC and MBN were calculated as the difference between SOC and TN in the fumigated and non-fumigated extracts using K_EC = 0.38 and K_EN = 0.54, respectively (Brookes & Joergensen, 2005).

Detection of microbial numbers

To calculate the most probable numbers (MPN), the technique of dilution and incubation of replicated cultures across several serial dilution steps was used (Oblinger & Koburger, 1975). In brief, an aliquot of 10 g (fresh weight) of mixed material from a soil depth of 0–100 mm was suspended in 90 mL of autoclave-sterilized distilled water in a 250 mL flask (a dilution of 10⁻¹). Soil suspensions were shaken for 30 min on a shaker (130 rpm) at room temperature (20 °C). The diluted soil samples were allowed to settle for 5 min; then, 1 mL was decanted by a sterile pipette into 10 mL dilution tubes containing 9 mL sterilized distilled water, yielding a 10⁻² dilution. This was repeated down to a dilution of 10⁻⁵, to obtain a series of 10-fold soil dilutions from 10⁻¹ to 10⁻⁶. One mL of each soil dilution was mixed with modified Stephenson media (Sylvia et al., 2005). The tubes were then closed with a sterile permeability sealing membrane (for denitrifying microbes, sealing film was used) and incubated for a period of 14 days at 25 °C and 65% relative humidity in the dark.

In summary, in this technique, replicate portions of the original sample were cultured to determine the presence or absence of microorganisms in each portion. After subdividing the sample, each portion was incubated in a specific nutrient milieu that selects certain organisms or groups of organisms. At the end of the appropriate incubation period, each portion was checked for presence or absence of growth. The MPN of target organisms present in the original sample is either calculated or determined by consulting a table (Williams & Busta, 1999). Measurements of MPN were used to provide an indication of the ability of soil microbes to utilize various nitrogen sources, following a procedure adapted from Sylvia et al. (2005).

Statistical analyses

Soil N₂O emissions at each hour, soil water content, soil temperature, MBC, MBN, mineral-N (both NH₄⁺-N and NO₃⁻-N), microbial numbers, and cumulative N₂O emissions were analyzed statistically using Genstat (version 17.0, VSN International Ltd., Hemel Hempstead, UK) (Cambareri et al., 2017). ANOVA was applied for testing significant differences between the ages of lucerne grassland (4-year-old and 11-year-old grasslands) during the same period. Comparisons between F, 4-year-old, and 11-year-old grasslands were made using the least significant difference (LSD) test at 5%. For MPN comparison, treatment was set as fixed factors and the sampling date was set as a random factor.

Principal component analysis (PCA) was performed to reduce the number of variables in the correlation matrix obtained from Pearson correlation analysis, and to select the most discriminating variables affecting N₂O emissions. Normality and homogeneity of variances were assessed with the Shapiro–Wilk and Levene and Barlett tests.

N₂O emissions

Mean (± standard deviation) N₂O emissions from the F, 4-year-old, and 11-year-old grasslands for each freeze/thaw cycle were 0.0287 ± 0.0009, 0.0230 ± 0.0019, and 0.3522 ± 0.0029 mg m⁻² h⁻¹, respectively (Figs. 1A, 1B and 1C). On 28 November, the mean N₂O emissions ranked 11-year-old grassland > F > 4-year-old grassland, and emissions of the 11-year-old grassland ranged from −0.0014 to 0.0879 mg m⁻² h⁻¹ (Fig. 1A, P < 0.05). On 7 January, the variability was highest for F plots. The average emissions ranked 11-year-old grassland > F > 4-year-old grassland with values of 0.0477, 0.0439, and 0.0173 mg m⁻² h⁻¹, respectively (Fig. 1B, P < 0.05). On 6 March, the daily variability in N₂O emissions was highest for the 11-year-old grassland (P < 0.05), which ranged from −0.0007 to 0.1519 mg m⁻² h⁻¹ (Fig. 1C). Variability of N₂O emissions was small for the three treatments during the time from 12:00 to 16:00 when mean emissions ranked F > 4-year-old grassland > 11-year-old grassland with values of 0.0086, 0.0066, and 0.0005 mg m⁻² h⁻¹, respectively (P < 0.05). Compared with the daily range of N₂O emissions in January and March, the observed range was much lower in November, especially during the hours from 12:00 to 14:00.

Cumulative N₂O emissions were 7.56, −16.98, and 9.61 mg N m⁻² for F, 4-year-old grassland, and 11-year-old grassland, respectively. Cumulative N₂O emissions from F and 11-year-old grasslands showed no difference, and were significantly higher than that of 4-year-old grasslands.

Soil temperatures and soil water content

Regarding soil temperature for the three treatments at a depth of 0–100 mm (Figs. 1D, 1E and 1F), the F plot showed the highest variation on 28 November, with values ranging from −1.2 °C to 8 °C, and values for the three treatments were below zero from 01:00 to 11:00 (Fig. 1D). On 7 January, the soil temperature of the three treatments was below zero and showed low variability (Fig. 1E). On 6 March, soil temperatures were below zero during the period from 00:00 to 11:00, and the variation was highest for the 11-year-old grassland, where it ranged from −1.6 °C to 5.5 °C (Fig. 1D).

On 28 November, the values of soil water content at a depth of 0–100 mm ranged from 12.20% to 12.74%, 11.51% to 11.74%, and 10.99 to 11.08% for the F, 4-year-old, and 11-year-old grasslands, respectively (Fig. 1G). On 7 January, soil water contents ranged from 7.01% to 7.84%, 6.85% to 7.75%, and 7.28% to 8.22% for the F, 4-year-old, and 11-year-old grasslands, respectively (Fig. 1H). On 6 March, soil water content followed the order of 11-year-old > 4-year-old > F (P < 0.05), with values of 12.20% to 12.74%, 11.61% to 12.01%, and 12.62% to 12.91%, respectively (Fig. 1I).

MBN and MBC concentrations

The values of MBN and MBC decreased with increasing soil depth (Table 2). MBN and MBC values in the topsoil (0–100 mm) were higher than at a depth of 100–200 mm (P < 0.05), and both were higher than those at a depth of 200–300 mm (P < 0.05). Values of MBN and MBC were higher for the 11-year-old grassland than those for the 4-year-old grassland (P < 0.05). Values of MBN in the topsoil (0–100 mm depth) for the lucerne grassland were higher compared with values for F (P < 0.05). MBN decreased from the 28 November to the 7 January, and then increased to 6 March, while values of MBC gradually decreased from November to March.

Soil NH₄⁺-N and NO₃⁻-N concentrations

On 28 November soil NH₄⁺-N concentrations at a depth of 0–300 mm remained below five mg g⁻¹ for all three treatments (Fig. 2A). Higher concentrations of soil NH₄⁺-N were measured at a depth of 0–100 mm, which followed the ranking of 4-year-old grassland > 11-year-old grassland > F (P < 0.05). Soil NH₄⁺-N concentrations at a depth of 0–300 mm for the three treatments were also <5 mg g⁻¹ on 7 January (Fig. 2B). Higher soil NH₄⁺-N concentrations were measured at a depth of 100–200 mm, which followed the rankings of 11-year-old grassland > 4-year-old grassland > F with values of 4.77, 4.27, and 1.31 mg g⁻¹, respectively (P < 0.05). On 6 March, soil NH₄⁺-N concentrations at a depth of 0–300 mm were only <5 mg g⁻¹ for F (Fig. 2F). Concentrations measured at a depth of 0–100 mm were all higher at 1.70, 7.99, and 8.84 for F, 4-year-old and 11-year-old grasslands, respectively.

On 28 November, soil NO₃⁻-N concentrations for the three treatments at a depth of 0–100 mm followed the order of 11-year-old grassland > 4-year-old grassland > F with values of 10.77, 10.34, and 6.46 mg g⁻¹, respectively (Fig. 2D, P < 0.05). Soil NO₃⁻-N contents at a depth of 0–300 mm for the three treatments remained <10 mg g⁻¹ on January 7 and higher values were measured at a depth of 0–100 mm for the 11-year-old grassland and the F treatment (Fig. 2E). Soil NO₃⁻-N concentrations at a depth of 0–100 mm followed the order of F > 4-year-old grassland > 11-year-old grassland with values of 9.92, 5.50, and 4.86 mg g⁻¹, respectively (P < 0.05). On 6 March, soil NO₃⁻-N concentrations at a depth of 0–300 mm remained <10 mg g^–1 for all three treatments (Fig. 2F). In the 0–100 mm layer, values were 2.98, 9.24, and 8.93 for the F, 4-year-old, and 11-year-old grassland, respectively.

Numbers of nitrification and denitrification microbes

The MPN method was used to estimate changes in the numbers of nitrifying and denitrifying organisms throughout the freeze/thaw cycles (Table 3). Highest MPN values of nitrosobacteria of approximately 9,380 were found in the F treatment on 28 November (P < 0.05). The highest MPN value for nitrifying bacteria of 15,970 was observed in the 4-year-old grassland on 28 November (P < 0.05). The maximum value of 31,470 denitrifiers per unit of soil dry mass was found in the 4-year-old grassland on 6 March (P < 0.05). The number of MPN denitrifiers in the 11-year-old grassland was 15,550 on 28 November (P < 0.05).

Factors influencing N₂O emissions

Data from the different treatments were analyzed to estimate the impact of freeze-thaw cycles on N₂O emissions, and to remove additional factors such as the age of lucerne grasslands and nitrogen fixing capacity. The correlation matrix at a depth of 0–100 mm (Table 4) indicated that N₂O emissions were significantly correlated with soil water content, soil temperature, NH₄⁺-N concentrations, and the number of nitrosobacteria and denitrifying bacteria. MBN was positively correlated with MBC (P < 0.01). The NH₄⁺-N concentration was also correlated with soil water content (P < 0.05). Further, the numbers of nitrosobacteria and denitrifying bacteria were related to soil temperature (P < 0.05). The number of denitrifying bacteria was significantly correlated with the soil NH₄⁺-N concentration (P < 0.05). The correlations between N₂O emissions and these variables at depths of 100–200 and 200–300 mm were similar to those for the 0–100 mm depth.

Autocorrelated relationships in the matrix were analyzed to identify the main factors that contribute to N₂O emissions throughout freeze/thaw cycles. The results of PCA for soil variables at depths of 0–100 mm are shown in Fig. S2. PCA showed that N₂O emissions were dominantly attributable to the number of denitrifying bacteria during the freeze/thaw period. The number of nitrosobacteria, soil temperature, and soil NH₄⁺-N concentration were dominant variables to explain differences in N₂O emissions.

Diurnal dynamics of N₂O emissions during the freeze/thaw cycle

The highest daily average N₂O emission was 0.3522 mg m⁻² h⁻¹ for the 11-year-old grassland on 28 November (Fig. 1A). Increased emissions with increasing cultivation age are likely attributable to different N and C concentrations in both plant and soil (Wang et al., 2014b). The better developed root system in older grasslands produces more residue and litter and secrets more SOC, TN, and organic acids (Hong, 2009), thus increasing the potential for N₂O production (Rochette & Janzen, 2005). Further, the mean N₂O emissions for the 11-year-old grassland on 7 January were higher than average emissions during the growing season (Rochette & Janzen, 2005). The maximum N₂O emission for the 11-year-old grassland (0.0879 mg m⁻² h⁻¹) in the freeze/thaw cycle was significantly higher than the maximum value (0.0323 mg m⁻² h⁻¹) during the rest of the growing season.

Increased soil temperature during the daytime enhanced soil water content, thus limiting soil gas diffusion (Müller et al., 2002) and leading to reduced N₂O emissions (Burton & Beauchamp, 1994). The low variability in N₂O emissions between 11:00 and 14:00 (Fig. 1) may be explained by the diurnal dynamics of soil temperature and soil water content. High values of soil water content for the 11-year-old grassland on 6 March were associated with low emissions (Fig. 1). High emissions between 18:00 and 22:00 could be explained by changes in the soil volume following freezing and the release of trapped N₂O into the atmosphere, which had accumulated in the frozen soil (Burton & Beauchamp, 1994).

Negative N₂O emissions were measured on a number of ocassions (Fig. 1), which is consistent with other findings (Chapuis-lardy et al., 2007). Negative fluxes usually indicate that N₂O is reduced and emitted as N₂, representing the final step of the microbial denitrification process. Anaerobic conditions could stimulate N₂O consumption by soils resulting in the conversion of N₂O to N₂, with little mineral N and readily available organic matter as energy sources for denitrifying activity of microbes (Bremner, 1997; Chapuis-lardy et al., 2007).

Influence of factors on N₂O emissions during freeze/thaw cycles

Changes in soil temperature and soil water content affect N₂O emissions by influencing metabolic activities of microorganisms, soil aeration, substrate availability, and nutrient redistribution (Wagner-Riddle et al., 2017). The N₂O/N₂ emissions ratio has been shown to decrease with increasing temperature (Holtan-Hartwig, Dörsch & Bakken, 2002). At the tested site, the correlation between soil temperature and N₂O emissions was negative (Table 4, Fig. S2). Other findings also identified soil temperature, rather than soil water content, as more important for regulating N₂O emissions (Koponen & Martikainen, 2004; Luo et al., 2013). This is likely attributable to the effects of soil temperature on enzymatic activity and substrate limitation (Stark & Firestone, 1995). Soil temperature has also been shown to affect the osmotic status of microbial cells, the rates of gas diffusion, and soil water content (Wu et al., 2013). Although the findings of the present study showed that soil water content was a significant factor influencing N₂O emissions in freeze/thaw cycles, soil temperature could better explain N₂O emissions and microbial numbers (Table 4, Fig. S2).

Soil freeze-thaw cycles exert profound impacts on soil physical structure, solute distribution, and nutrient availability. N₂O reductase activity is important for regulating N₂O emissions during freeze/thaw cycles, which is strongly influenced by the availability of decomposable organic matter (Wang et al., 2014a) as well as C and N sources (Christensen & Christensen, 1991; Müller et al., 2002). The present study showed that MBC, MBN, and the MBC/SOC ratio were positively correlated with N₂O emissions (Table 4). Although other research has found increases in the availability of carbon from fragmented organic matter or dead microorganisms during freeze/thaw cycles (Christensen & Christensen, 1991; Wang et al., 2014a), the results obtained in the present study showed that MBC concentration did not significantly change during the freeze/thaw cycles. A probable explanation is that MCB was consumed by microbes that survived the freeze/thaw cycles. A significantly positive correlation was found between MBN and MNC (Table 4), which agrees with previous findings (Gou et al., 2015).

The soil NH₄⁺-N concentration increased during freeze/thaw cycles, whereas the soil NO₃⁻-N concentration decreased significantly (Fig. 2). This increase in NH₄⁺-N concentration is likely attributable to the disturbance of the topsoil structure because of freeze/thaw cycles and the accelerated soil nitrogen mineralization, influenced by soil water content and temperature (Song et al., 2017). Soil water concentration was positively correlated with soil NH₄⁺-N concentration (Table 4). The reduction in soil NO₃⁻-N content during freeze/thaw cycles can be explained by low N uptake, leading to high rates of denitrification and N₂O production (Pu, Saffigna & Strong, 1999). Soil NO₃-⁻N as the direct source of N₂O decreased during the period when N₂O emissions are highly variable (Table 4, Figs. 1 and 2).

The number of nitrosobacteria was linked to soil temperature but not to soil water content (Table 4). Potential explanations are that nitrification depends on soil water content (Stark & Firestone, 1995) or that soil temperature is the limiting factor during freeze/thaw cycles. Ammonia-oxidizing bacteria are essential for nitrification and freeze/thaw cycles as these influence the number of the nitrifying organisms (Tzanakakis, Taylor & Bottomley, 2020). However, the correlations between N₂O emissions and the numbers of nitrifying bacteria were not significant (Table 4). Similarly, although a positive correlation was found between the numbers of nitrosobacteria and nitrification substrates (NH₄⁺-N), as well as a negative correlation between the numbers of nitrosobacteria and nitrification products (NO₃⁻-N), these correlations were not significant. A possible explanation is that the activities of nitrifying microorganisms or nitrification might be diminished during freeze/thaw cycles (Papen & von Berg, 1998).

Nitrification is known to be the largest source of N₂O emissions in semi-arid regions where soils are rarely sufficiently anaerobic to induce denitrification (Wagner-Riddle et al., 2017). However, freeze/thaw cycles induce N₂O emission because of the limitation of the supply of oxygen needed for nitrification (Goreau, Kaplan & Wofsy, 1980). Thawing of the top soil layer and the creation of anaerobic conditions resulting from water saturation would induce denitrification (Müller et al., 2002; Dusenbury et al., 2008; Wu et al., 2013). In the present study, N₂O emissions were significantly linked to the number of denitrifying bacteria (Table 4). Also, the number of denitrifying bacteria in the tested soils was significantly greater than the number of nitrifying microorganisms (Table 3). This is consistent with the findings of Holtan-Hartwig, Dörsch & Bakken (2002) who suggested that freeze-thaw cycles enhance denitrification. Peterjohn (1991) found that denitrifying enzymes can survive under drought and can be reactivated when soil water content increased. Also, soil water saturation stimulated soil microbial activity, restricted the oxygen supply, and increased CO₂ concentrations in soil (Chamindu et al., 2019), all of which could stimulate denitrification.

Large diurnal amplitudes in surface soil temperature are characteristic for the Loess Plateau, resulting from thawing during the daytime and refreeze at night (Figs. 1D–1F). This favors the coupling of nitrification and denitrification (Müller et al., 2002), which could have both contributed to N₂O production in the present study. Here, both the number of denitrifying bacteria and nitrosobacteria bacteria were found to correlate significantly with N₂O emissions, which is in general agreement with prior findings.

Interpretations of this study are restricted by data availability. In particular, the authors are aware of the importance of the soil microclimate (i.e., temperature and water content) and microbial activities throughout the year; however, the present study is limited because of a lack of such data. Thus, broader interpretations should be made with caution. Moreover, further studies of freeze/thaw cycles in the field with higher frequency of N₂O measurement are warranted.