2.1 Study area

The study area is in Duolun County, Inner Mongolia, China (42.02°N, 116.17°E; 1341 m). The mean annual temperature in this region is 2.1 °C, and the mean annual precipitation is 385.5 mm, most of which falls between May and September (Niu et al., 2009). The typical soil in this region is chestnut (Soil Taxonomy Research Group, 2001), corresponding to Calcic Petrocalcids based on the U.S. Soil Taxonomy (Soil Survey Staff, 2014). The soil bulk density is 1.31 g cm^–3. The area is dominated by perennial bunch grass Stipa krylovii, perennial rhizome grass Leymus chinensis, perennial semi-shrub Artemisia frigida, and perennial forbs such as Agropyron cristatum and Allium bidentatum (Niu et al., 2009).

The mean soil temperatures for the three growing seasons of 2013, 2014 and 2015 were 19.4 °C, 19.8 °C, and 18.8 °C, respectively. The mean soil moisture levels (soil volumetric water content, VWC) for the three growing seasons were 13.4%, 10.4%, and 11.9%, respectively. Soil volumetric water content was closely associated with precipitation events (Fig. 1).

2.2 Experimental design

Atmospheric N deposition plus fertilizer application in this area are about 5 g N m^–2 yr^–1 (Wang et al., 2015). Eighteen 15 m × 10 m plots were established in the site, separated from one another by 4-m-wide buffer zones. Six N addition treatments, of 0 (N0), 2 (N2), 4 (N4), 8 (N8), 16 (N16), and 32 (N32) g N m^–2 yr^–1 with three replicates for each level were randomly assigned to the eighteen plots. Since July 2003, a dry urea N fertilizer (CO(NH₂)₂, 46.67% N) has been manually spread on the surface of the plots. Based on the local weather forecast, N addition was conducted just hours before the rain in mid-July (during the rainy season) each year.

2.3 Soil respiration

A PVC collar (20 cm diameter, 14 cm height) was inserted 5–6 cm deep at random locations in each plot to measure soil respiration (Rs, µmol of CO₂ m^–2 s^–1). Aboveground biomass in the collar was clipped and removed 24 h before the measurement. All collars were removed after the measurement, and new ones were randomly insert into the soil during the next year. Respiration was measured during approximately 14-day periods in the mid-growing period, from 23 July to 4 August in 2013, from 27 July to 8 August in 2014, and from 26 July to 5 August in 2015. The Rs measurements would be postponed for 1–2 days after precipitation to avoid CO₂ flux pulses. During each period, the 18 plots were each observed only once. The Rs values in three plots with six collars were measured on each date. The respiration rate was measured with a soil C-flux automatic measurement system (LI-COR, NE, USA), and data were recorded once every 1 h for each 24-h measurement period. The annual mean of Rs was used for subsequent analysis. In the meantime, soil temperature (°C) and moisture (%) in the top 0–10 cm layer of the soil near the collar were measured with an auxiliary sensor attached to the LI-COR 8150.

A soil sensor attached to a data logger (EM50; ICT International, Australia) in the experimental site recorded the soil temperature and volumetric water content at a 10-cm depth. Data were recorded every hour during the growing season (June to September). Precipitation events at the site were recorded by the staff working in the Duolun Restoration Ecology Research Station.

Fig. 1

Dynamics of soil temperature and soil moisture for 0–20 cm surface soil and precipitation during the growing seasons of 2013 (a), 2014 (b), and 2015 (c).

2.4 Soil properties and aboveground biomass

Soil sampling was conducted during the soil respiration measurements in each year. Three 0–20 cm surface soil samples were randomly collected from each plot to form a composite sample. The soil pH, dissolved organic carbon (DOC), ammonium nitrogen (-N), and nitrate nitrogen (-N) were measured. Fresh soil was sieved through 2-mm-pore filters and stored at 4 °C for PLFAs, pH, DOC, -N, and -N measurements. To measure the DOC, a 0.5 M K₂SO₄ solution was used to extract soil organic carbon. The extracts were measured with a TOC analyzer (Liqui TOC II, Elementar, Germany). Ion chromatograph analyzers were used to measure inorganic ions such as -N and -N (DX-600 and ICS-2100, Dionex, CA, USA), and inorganic nitrogen (IN) was calculated as the sum of -N and -N. The other portion of each soil sample was air dried and then ground for soil total organic carbon (TOC) and total nitrogen (TN) measurements. TOC was measured by the potassium dichromate-oxidation method, and TN was analyzed with an elemental analyzer (Perkin-Elmer, MA, USA). Aboveground biomass (ABM) in one quadrat (1 m × 1 m) of each plot was clipped and determined as described by Fang et al. (2012).

2.5 Soil phospholipid fatty acids

The microbial composition was measured in fresh soil using PLFAs as described by Bossio and Scow (1998). For comparisons to be valid among years, 18 biomarkers (i14:0, i15:0, a15:0, i16:0, a16:0, 16:1w5c, 16:0, i17:0, a17:0, 17:1w8c, cy17:0, 10Me17:0, 18:2w6c, 18:1w9c, 18:1w7c, 18:1w5c, 18:0, and 10Me18:0) that existed in all three years were selected, and the proportions of the individual PLFA biomarkers (i.e., the ratios of the selected biomarker to the total PLFAs that existed in a growing season) were used for further analysis. According to previous studies, i14:0, i15:0, a15:0, i16:0, a16:0, 16:0, 16:0, i17:0, a17:0, cy17:0, 17:1w8c, 18:1w7c, 18:1w5c, and 18:0 indicate bacteria; 18:2w6c and 18:1w9c indicate fungi; 10Me17:0 and 10Me18:0 indicate actinomycetes; i14:0, i15:0, a15:0, i16:0, i17:0, and a17:0 indicate gram-positive bacteria; a16:0, 17:1w8c, cy17:0, 18:1w7c, and 18:1w5c indicate gram-negative bacteria; and 16:1w5c indicates arbuscular mycorrhizal fungi (AMF) (Bi et al., 2012; Bossio and Scow, 1998; Wei et al., 2013). The ratios of fungi: bacteria (F: B ratio) and gram-positive bacteria: gram-negative bacteria (G⁺ : G^– ratio) were considered as potential tools for assessing soil quality and environmental stress (Kaur et al., 2005; Canarini et al., 2017). Thus, these ratios were calculated according to the PLFA biomarkers mentioned above.

2.6 Statistical analysis

Temperature sensitivity of soil respiration (Q₁₀) was calculated by the following exponential functions:

where Rs is the 24-h dynamic soil CO₂ flux rate (µmol CO₂ m^–2 s^–1) in the PVC collar, and T is the 24-h dynamic soil temperature (°C) near the collar. Parameter a indicates the basal respiration rate, and b is the exponent used to calculate Q₁₀.

Two-way ANOVA was used to evaluate the effects of N addition and interannual variability on Rs, Q₁₀, soil properties and microbial composition. Pairs of mean values were compared with the least significant difference (LSD) test. Pearson correlations were used to analyze the correlations of soil properties, microbial composition, Rs and Q_10.

SPSS 20.0 was used to perform a principal component analysis (PCA) on microbial composition (measured by the proportions of individual PLFA biomarkers) in soil samples in the six N addition levels and three years. Graphs were created using SigmaPlot 12.5 (Systat Software, CA, USA).

A structural equation model (SEM) was used to assess the direct and indirect effects of N addition on microbial community, soil respiration, and Q₁₀. The model included nine categories of variables, which were “N addition”, “soil temperature”, “soil moisture”, “soil pH”, “C and N availability”, “aboveground biomass”, “Rs”, “Q₁₀”, and “microbial community”. Soil DOC, TOC, and TN were considered as potential indicators for “C and N availability”. In light of the complexity of defining the microbial community, we quantified such variables in each model as follows: “PLFAs” and “microbial composition”. Prior to the path analysis, PCA was performed for the indicators representing “microbial composition” and “C and N availability”, respectively, then the loading of PC1 was used in the model (Table 1). A bootstrapping with 200 resamples was performed, and the 95% bootstrap confidence interval was used to judge whether the estimated path coefficients were significant in SEM. The criteria for evaluating model fit, such as chi-square, P values, and root mean square errors of approximation (RMSEA) were adopted according to Schermelleh-Engel et al. (2003). The SEM was carried out using Amos 21.0 (Smallwater Corporation, Chicago, IL, USA).

3.1 Soil properties and aboveground biomass

The DOC and IN show significant differences along the N gradient (Table S1, P < 0.001). The DOC was promoted by N addition in 2014 and 2015 but did not change in 2013 (Fig. 2a). The DOC in 2013 was significantly higher than in 2014 and 2015 (Table S2, P < 0.001). The IN was significantly positively correlated with N addition for the three years (Fig. 2b), with the highest content in 2014 (Table S2). TOC and TN did not respond significantly to N addition (P > 0.05), but they had significant differences along the interannual gradient (P < 0.001) (Fig. 2d and Fig. 2e). In contrast, the pH was significantly inhibited by N addition for the three years (P < 0.001), but there was no significant difference between years (P > 0.05) (Fig. 2c). Aboveground biomass did not significantly respond to N addition for the three years (P > 0.05) but was significantly different between years (P < 0.01) (Fig. 2f).

3.2 Soil microbial phospholipid fatty acids

In the three growing seasons, the PLFA biomarkers selected in this study accounted for 64%–69% of the total nitrogen gradients, with the sum of the leading five markers (16:0, 18:1w7c, 18:1w9c, i15:0 and i16:0, all of which are bacterial biomarkers) accounting for 36%–40% of the total. Soil PLFAs had significant differences between N gradients (Table S1, P < 0.001). Soil PLFAs were inhibited by N addition in 2013 and 2015, but did not change in 2014 (Fig. 2g). F:B was significantly inhibited by N addition for the three years (Table S1, P < 0.001), and the highest ratio was in 2014 (Fig. 2h). In contrast, G+:G– significantly increased by N addition (P < 0.001), and the highest ratio was in 2015 (Fig. 2i).

Table 1

Loadings of microbial compositions and C and N availability with PC1 in principal component analysis on soil samples with six N addition levels and three years

Fig. 2

Soil variables with six N gradients and three years (n=3) Note: (a) Dissolved organic carbon; (b) Inorganic nitrogen; (c) Ph; (d) Total organic carbon; (e) Total nitrogen; (f) Aboveground biomass; (g) PLFAs; (h) Fungi to bacteria ratio; (i) Gram-positive bacteria to gram-negative bacteria ratio; (j) Soil respiration; (k) Q₁₀.

Principal component analysis was used to determine the relationships between soil microbial communities. Soil samples with decreasing N gradients were generally distributed from left to right along PC1, indicating that PC1 was related to N addition rate. In addition, samples in 2015 were allocated in the ordination plot where PC2 > 0, but the opposite was found for samples in 2013 and 2014. These patterns implied that interannual variability was associated with PC2. The two axes explained 51.56% of the variation of the microbial community composition in the soil samples (Fig. 3).

Fig. 3

Principal component analysis of microbial community composition in soil samples with six N gradients and three years (mean±SE) (n=3). Soil samples in the three years are presented by different colors, as black indicates 2013, white indicates 2014, and grey indicates 2015.

3.3 Rs and Q ₁₀

Rs shows significant differences between N gradients (Fig. 2j, Table S1, P < 0.001). Compared to the N0 treatment, Rs under the N2, N4, N8, N16 and N32 treatments decreased by –2.58%, 14.86%, 22.62%, 23.97% and 19.87% for the three growing seasons, respectively. On the other hand, Q₁₀ did not significantly respond to N addition (Fig. 2k, Table S1, P > 0.05), with values of 1.71, 1.65, 1.76, 1.67, 1.73 and 1.58 for the six N gradients.

3.4 Relationships between soil respiration and the various impact factors

Pearson correlation analysis showed that Rs was positively correlated with soil moisture, soil pH and PLFAs, and negatively correlated with N addition, DOC and IN. Q₁₀ was significantly positively correlated with soil moisture and Rs, and negatively correlated with soil temperature, DOC and TN (Table 2).

Table 2

Pearson correlation analysis of soil respiration and the impact factors

Two SEM models were constructed using the two microbial community indicators of soil PLFAs and microbial community composition. The models explained the variation between 62% and 65% for Rs and 52% for Q₁₀, respectively. The path coefficients in the models show that N addition significantly reduces soil pH and changes soil microbial community composition. The aboveground biomass was not significantly affected by various factors of the soil, but had a significant direct effect on soil PLFAs and soil microbial community composition. Soil pH, soil moisture and C and N availability have significant positive effects on soil PLFAs, and aboveground biomass inhibits soil PLFAs. N addition, soil moisture, and C and N availability have significant direct negative effects on soil microbial community composition, and aboveground biomass has a significant positive effect on it (Fig. 4).

From the standard total effect coefficient, N addition inhibited Rs, but had less effect on Q₁₀. The main factors affecting soil PLFAs and soil microbial community composition were soil pH and N addition, respectively. Rs was more affected by soil moisture, soil pH, soil PLFAs and microbial community composition. Soil moisture was the most important factor in Q₁₀ change, and other factors had less influence on Q₁₀ (Table 3).

4.1 The responses of soil properties and soil microbial community to N addition

After urea fertilizer is hydrolyzed to -N, the soil pH decreases because -N is transformed into -N and H⁺ is released into the soil (Kopáček et al., 2013). This process explains the significant correlations between N addition and soil pH. In this study, soil pH had less relative impact on microbial composition than C and N availability. This finding concurs with the review of Zhou et al. (2017), in which N addition was found to regulate the F: B ratio and G⁺: G^– ratio by increased N availability rather than by decreased soil pH. However, there are studies which considered soil acidification as the primary factor shaping the microbial composition (Chen et al., 2015) and soil respiration components (Li et al., 2018). Since the fertilizer used in these two studies was NH₄NO₃, one explanation for this contradiction is the difference in the form of N fertilizer. Although both forms induced soil acidification (Tian and Niu, 2015), the addition of urea and NH₄NO₃ were found to have inconsistent effects on other key factors, including soil respiration (Zhou et al., 2014), DOC (Yue et al., 2016), and SOC (Cheng et al., 2017). Therefore, the mechanisms of altering soil C fractions, including the soil microorganisms, might differ between two forms of N addition. The significant inhibition of total PLFAs by N addition echoed other studies (Riggs and Hobbie, 2016; Yue et al., 2016), which might be caused by soil acidification in this study, the release of toxic minerals such as Al³⁺, Fe³⁺ and Mn³⁺ (Tian et al., 2016), and depressed enzyme activities.

Fig. 4

Structural equation model for the effects of nitrogen addition, soil temperature and soil moisture on soil pH, C and N availability, aboveground biomass, total PLFAs, microbial composition, Rs and Q10. Note: Dashed, red, and blue paths indicate the effect is insignificant, positive, and negative, respectively. The width of the arrow indicates the magnitude of the standardized path coefficient. Bold and italic-typed values indicate the proportions of the variance explained for each variable. χ²—chi-square, df—degrees of freedom, P— P value, RMSEA—root mean square error of approximation, CFI—comparative fit index.

Table 3

Standardized total effects in structural equation models

As commonly occurrs in N saturation conditions, the DOC is expected to decrease when C becomes the limiting factor (Kopáček et al., 2013; Tahovská et al., 2013), but the DOC increased with N addition in this study. The sophisticated in- and outflows of labile C in the soil processes can make it difficult to untangle the lines linking N fertilization and DOC, let alone considering the effects of the different forms of N added and the changes in soil acidity (Evans et al., 2008). In their meta-analysis, Yue et al. (2016) found that DOC in the O-horizon of terrestrial ecosystems was enhanced by N addition; and the decreased microbial biomass and increased litter decomposition (Liu et al., 2010) in this study site might explain such a phenomenon (Rodriguez et al., 2014).

The groupings of the samples as explained by the first ordination axis in the PCA revealed the impact of N fertilization on the soil microbial composition in this semiarid grassland soil. In agreement with Zhou et al. (2017), when N was added to the soil, bacterial communities outcompeted saprophytic fungal communities, and gram-positive bacteria outcompeted gram-negative bacteria. Consistent with Wei et al. (2013), the decrease of the F: B ratio with N addition was mainly caused by a decrease in the fungi biomarker. The soil pH preferences among microbial communities have been inconsistent among studies (Zhalnina et al., 2014; Chen et al., 2015), and soil acidity had a more negative impact on fungi than on bacteria in this study. The inhibition of expression of genes coding for lignin decomposition enzymes, the lower C investment in mycorrhizal fungi, and decomposition prevention and toxicities due to the formation of complex-structured humus with N addition were all found to severely suppress the growth of fungi (Treseder, 2008; Zak et al., 2011). For the bacterial communities, there was a trade-off between the ability to decompose recalcitrant carbon compounds and the high N demand (Treseder et al., 2011). Gram-positive bacteria relied more on nutrients for building complex cell walls and enzymes (Schimel et al., 2007; Treseder et al., 2011), thus they thrived under the conditions of high N availability.

4.2 The responses of Rs and Q ₁₀ to N addition

Rs includes the respiration of soil microorganisms, plant roots, soil animals, etc., and was affected by many factors such as soil temperature and moisture, soil physical and chemical properties, aboveground and underground biomass of vegetation, and soil microbial community. N addition had a negative effect on Rs in this grassland soil. However, this is not necessarily inconsistent with the reviews in which Rs increased by N addition (Zhang et al., 2014; Zhong et al., 2016). Studies have shown that organic matter formed by aboveground biomass through photosynthesis will increase the substrate concentrations for root respiration and microbial respiration, and significantly increase Rs (Wen et al., 2014). However, the aboveground biomass in this study did not change significantly under N addition. The reviews concluded that the increased Rs was closely correlated with enhanced root biomass. However, root biomass showed no significant response to N fertilization in this study site (Song et al., 2014). Moreover, decreased soil heterotrophic respiration by N addition in grasslands was caused by reduced microbial biomass (Riggs and Hobbie, 2016). Therefore, the lack of a response of aboveground biomass and root biomass compounded by the negative response of total PLFAs to N addition might explain the decrease in Rs. N addition changed the microbial community composition, and because different microorganisms have different preferences and efficiencies for C decomposition and utilization, it also had a significant impact on Rs.

Q₁₀ is the proportional rate of increase in soil respiration when soil temperature increases by 10 °C, which largely determines the feedback of the terrestrial soil carbon cycle to global climate change (Yang et al., 2011). In this study, N addition had no significant effect on Q₁₀. The three-year result of Q₁₀ was consistent with what we primarily found in 2013 (Li et al., 2015), thus N addition in this grassland might not be able to regulate Rs in a warmer temperature. However, the inhibition and stimulation of Q₁₀ were both found by N addition in other grasslands (Zhong et al., 2016; Fang et al., 2018). Q₁₀ is affected by many factors such as soil temperature and moisture, plant community composition, the number and composition of soil microbial communities, and the quality and quantity of soil substrates (Moyano et al., 2013), which also explains the varying results in different studies. Studies have shown that N addition promotes the formation of difficult-to-decompose organic carbon in the soil, so a higher activation energy is required during the decomposition process, leading to an increase of Q₁₀ (Wang et al., 2018). Studies have also shown that a decrease in soil microbial biomass will reduce Q₁₀, and at the same time, the amount of litter and root biomass will also have a greater effect on Q₁₀ (Liu et al., 2016.). Therefore, the non-response of Q₁₀ to N addition in this study may be caused by a combination of interacting factors. Multi-factor experiments found soil moisture was the primary factor in the semi-arid ecosystems (Wang et al., 2017; Zhang et al. 2015). The response of Q₁₀ to N addition exhibits a parabolic relationship with soil temperature and soil moisture (Zhong et al., 2016), suggesting that N addition could increase Q₁₀ when soil temperature and moisture were unsuitable, such as under an extreme climate. In this study, increased soil moisture had positive effects on Q_10. Due to the importance and complexity of Q₁₀, the mechanisms involved in its changes need to be further explored.