1 Introduction

Soil organic carbon (SOC) stocks greatly exceed the amount of carbon stored in both the atmosphere and terrestrial vegetation (Scharlemann et al. 2014). It has been suggested that CO2 released from the decomposition of SOC makes an important contribution to climate change. Furthermore, global warming is likely going to increase carbon release from soil to the atmosphere by accelerating decomposition of SOC (Lin et al. 2011; Sanderman et al. 2003). Understanding of SOC dynamics has attracted much attention as it is central to the evaluation of ecosystem carbon balance, and the feedbacks of this to climate change (Kirschbaum 2000; Campbell et al. 2009; Wang et al. 2013; Averill 2014).

It is difficult to detect statistically significant changes of the SOC stock over short-time periods, due to its large pool size, slow turnover rate, and huge spatial heterogeneity (Van Groenigen et al. 2014). Moreover, studies of the microbial processes controlling SOC turnover are necessary to understand the mechanisms governing SOC dynamics (Schimel and Schaeffer 2012); however, these process-based studies can be expensive and time consuming in the field (Campbell et al. 2009). Consequently, carbon stable isotope composition (δ13C) measurements of more easily collected plant biomass and soil samples offer an alternative and relatively cost-effective method to comprehend SOC dynamics (Garten et al. 2000; Acton et al. 2013; Wang et al. 2013; Yang et al. 2015).

Some studies have estimated SOC turnover and degree of SOC decomposition and explored the mechanisms governing SOC dynamics by investigating the pattern and variation of soil δ13C along environmental gradients (Garten et al. 2000; Bai et al. 2012; Peri et al. 2012; Wang et al. 2013; Acton et al. 2013; Angelo and Pau 2015). In general, soil δ13C decreases with mean annual precipitation (MAP) (Lee et al. 2005; Feng et al. 2008; Peri et al. 2012), and increases with mean annual temperature (MAT) (Cheng et al. 2011). The effects of climatic factors on soil δ13C are mainly attributed to their effects on plant δ13C, relative abundance of C4 plants (PC4) or both (Lee et al. 2005; Feng et al. 2008; Cheng et al. 2011; Peri et al. 2012). Mostly, soil δ13C decreased with soil C:N, SOC, and soil slit content (Wang et al. 2015; Yang et al. 2015), and increased with soil sand content (Wang et al. 2013). Edaphic factors may affect soil δ13C by changing microbial activities, substrate quality and availability and stabilizing effect (Wynn et al. 2006; Xu et al. 2016).

Previous studies have demonstrated that climatic and edaphic variables have important roles in shaping patterns of soil δ13C and consequently soil carbon dynamics (Peri et al. 2012; Wang et al. 2013; Yang et al. 2015). However, the controlling factors of soil δ13C differ for various ecosystems and regions. For instance, Peri et al. (2012) investigated relationships of soil δ13C in Southern Patagonia’s native forests with 28 climatic and soil variables, and found that MAP, but not soil properties, was the controlling factor of soil δ13C. At a continental-scale in Australia, Wynn and Bird (2008) found that climate and soil texture (percent of mineral particles < 63 μm) were the dominant controllers of soil δ13C. Wang et al. (2013) found that MAT and SOC were the controlling factors of soil δ13C in alpine meadow of the Tibetan Plateau.

Compared with soil δ13C, the difference of δ13C between surface soils and source plants (Δδ13C) may be a more direct and profound indicator of soil carbon dynamics along environmental gradients (Yang et al. 2015). This is because soil δ13C is not only determined by the rate of SOC decomposition but also by the δ13C value of source vegetation (Ehleringer et al. 2000; Peri et al. 2012). Thus, Δδ13C can eliminate the effect of δ13C value of source vegetation on soil δ13C, and so provide a more accurate indicator of soil carbon dynamics, i.e., large ∆δ13C translates to faster SOC decomposition rates (Yang et al. 2015). However, previous studies have mostly concentrated on soil δ13C and rarely on Δδ13C (Yang et al. 2015; Wang et al. 2018). Although several studies have investigated the relationships of Δδ13C with climatic and edaphic factors in alpine grassland and temperate forest ecosystems (Garten et al. 2000; Yang et al. 2015), the results indicated that controlling factors of Δδ13C differed among ecosystems. MAT was the dominant controlling factor of Δδ13C in temperate forest ecosystems (Garten et al. 2000). Soil silt content and C:N were the dominant controlling factors of Δδ13C in alpine meadow, while in alpine steppe it was SOC (Yang et al. 2015). Moreover, the soil properties, climatic conditions, and vegetation types of these studies are very different from those of temperate grassland ecosystems. Thus, it is important to conduct related research on a temperate grassland ecosystem to determine if its relationships of Δδ13C with climate and soil factors differ from those of other ecosystems.

Xilingele League, Inner Mongolia, China, is a region with typical temperate grassland ecosystems. Thus, we conducted research in this region with the following primary objectives: (1) to verify the hypothesis that in temperate grasslands soil δ13C is positively correlated to vegetation δ13C and is higher than vegetation δ13C, as soil organic matter is mainly derived from plant inputs and SOC decomposition discriminates 13C; and (2) to quantify the individual and combined effects of climatic and edaphic variables on vegetation, soil δ13C and Δδ13C, in order to determine the controlling factors of soil carbon dynamics in temperate grasslands.

2 Materials and methods

2.1 Study area

Plant and soil samples were obtained from 29 grassland sites along northeast–southwest transects located at longitude 111° 25′~ 119° 58′, latitude 41° 3′~46° 45′ in the Xilingele League, Inner Mongolia, China (Fig. 1).

Fig. 1
figure 1

Locations of the sampling sites on the Xilingele League, Inner Mongolia, China

Full size image

The Xilingele League has an average elevation of about 1200 m. It has an arid and semi-arid continental climate with water availability being the major abiotic factor limiting plant growth. MAP ranges from 100 to 450 mm, MAT ranges from − 3 to 6 °C. There is a significant negative correlation between MAP and MAT (r2 = 0.74, P < 0.001). The natural vegetation cover is temperate grassland. The main grassland types, in transition from east to west, are temperate meadow steppe, temperate typical steppe, and temperate desert steppe (DAHV and CISNR 1996). Dominant plant species are as follows: Stipa grandis, Stipa baicalensis, Stipa krylovii, Stipa tianschanica, Leymus chinensis, Agropyron cristatum, Cleistogenes squarrosa, Cleistogenes songorica, Artemisia frigida, Salsola passerine, Salsola arbuscula, Chenopodium glaucum, Allium bidentatum, and Allium mongolicum (DAHV and CISNR 1996). For more detailed information, see Table A1 - ESM. The soils are mostly chernozems, kastanozems, calcisols, and cambisols (Fu et al. 2011).

2.2 Sampling

Plant and soil samples were obtained in August 2015. To minimize influences of short-term human disturbance, such as grazing, sampling sites were at least 500 m away from human habitation, fenced, and not grazed in the sampling year. Sites were separated from each other by about 50 km and their geographic positions were recorded using GPS (Garmin, Montana 650, USA). At each site, an approximately 10 m × 10 m plot was established, and within that plot, three 1 m × 1 m quadrats were randomly located. All aboveground standing plant biomass in each quadrat was cut to ground level and put in a large envelope. Three replicate soil samples from the top 10 cm of each quadrat were collected using a 70-mm inner diameter soil auger.

2.3 Sample analysis

All aboveground plant biomass samples were washed with deionized water to remove dust particles, de-enzymed at 105 °C for 15 min, oven-dried at 80 °C to constant weight, and then were ground and thoroughly mixed using a plant grinding machine (FW100, China). Finally, in preparation for analysis, subsamples of about 20 g were finely ground using a ball mill (Retsch MM400, Germany). After air-drying, soil samples were sieved to 2 mm, and plant residues were removed. Before isotope measurements, soil samples were ground to fine powder as for plant samples, and treated with 0.1 N HCl solution for 24 h at room temperature to remove any carbonates, then washed with distilled water and finally oven-dried at 65 °C to constant weight.

The carbon and nitrogen concentrations and carbon isotope compositions (δ13C, ‰) of plant and soil samples were determined by a coupled elemental analyzer (FlashEA1112; Thermo Electron) and isotope ratio mass spectrometer (Finnigan MAT-253; Thermo Electron). δ13C values of samples were calculated as:

$$ {\delta}^{13}C\left({\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}} \right)=\left[\left({R}_{\mathrm{samples}}-{R}_{\mathrm{standard}}\right)/{R}_{\mathrm{standard}}\right]\times 1000 $$
(1)

where Rsample and Rstandard, respectively, are the 13C/12C ratios in the sample and the standard. The universally accepted standard of Vienna Pee Dee Belemnite (VPDB) was used. The precision for isotope composition measurement was 0.1‰.

Percentages of clay, silt, and sand contents in the soil samples were measured using a particle size analyzer (Malvern Mastersizer 2000; Malvern Instruments Ltd., Worcestershire, UK) after removal of organic matter and calcium carbonates by adding 30% H2O2 and heating at 100 °C for 45 min. Soil pH was determined using a pH electrode in a mixture of soil and water, with a water:soil ratio of 2.5:1 (Bao 2005).

2.4 Meteorological data

Mean monthly temperature (MMT), MAT, mean monthly precipitation (MMP), and MAP for each sampling site were extracted from a regional climate database with 1-km spatial resolution. This database was spatially interpolated based on the records of 107 climatic stations located in Inner Mongolia using the software package Anuspline (ver. 4.2). Using MMT and MMP, we then calculated the growing season temperature (GST) and growing season precipitation (GSP) of each sample site.

2.5 Data analysis

Paired-sample T tests were conducted to compare the mean values of vegetation and soil δ13C. Simple regression analysis was applied to investigate the relationship between vegetation and soil δ13C and also the relationships of vegetation δ13C, soil δ13C, and Δδ13C with climatic and edaphic variables (that is, GSP, MAP, GST, MAT, soil pH, clay content, silt content, sand content, soil organic carbon, and soil C:N). Then, stepwise regression was used to evaluate integrative effects of climatic and edaphic variables on vegetation δ13C, soil δ13C, and Δδ13C. Further, redundancy analysis (RDA) was conducted and T value biplots were drawn to explore primary controlling factors for vegetation δ13C, soil δ13C, and Δδ13C. Data analyses were conducted using SPSS (ver.18.0; SPSS Inc., USA) and Canoco (ver. 5.0; Microcomputer Power, Ithaca, NY, USA).

3 Results

3.1 Variation of vegetation and soil δ13C among sampling sites

Soil δ13C values ranged from − 24.7 to − 19.5‰, with a mean of − 22.4 ± 0.3‰, and with 55% of these values within the range of − 24 to − 22‰ (Fig. A1a, Electronic Supplementary Material - ESM). Vegetation δ13C values ranged from − 26.4 to − 21.6 ‰, with a mean of − 24.5 ± 0.2‰, and with 69% of these values within the range of − 26 to − 24‰ (Fig. A1b - ESM).

3.2 Relationship between vegetation and soil δ13C

Soil δ13C increased linearly with vegetation δ13C (Fig. 2). There was a clear pattern of soil being 13C enriched relative to the source vegetation, and this was more pronounced at sites with lower vegetation δ13C (Fig. 2). On average, △δ13C was 2.1‰ and higher than 1.3‰ for 59% of the sample sites.

Fig. 2
figure 2

Relationship between soil δ13C and vegetation δ13C in temperate grasslands on the Xilingele League, Inner Mongolia, China

Full size image

3.3 Variation of vegetation δ13C, soil δ13C, and Δδ13C along climatic gradients

All soil δ13C, vegetation δ13C, and △δ13C showed linear increasing trends with GST and MAT, and decreasing trends with GSP and MAP (Fig. 3 and Fig. A2 - ESM). Moreover, the relationships of soil δ13C with climate factors were stronger than those of vegetation δ13C (Fig. 3).

Fig. 3
figure 3

Relationships of soil δ13C, vegetation δ13C, and △δ13C with GST and GSP in temperate grasslands on the Xilingele League, Inner Mongolia, China. GST growing season temperature, GSP growing season precipitation

Full size image

3.4 Variation of vegetation δ13C, soil δ13C, and Δδ13C along edaphic gradients

Soil δ13C was positively related to soil pH (Fig. 4a), and negatively related to SOC content and soil C:N (Fig. 4b, c). Vegetation δ13C was not related to soil pH, SOC content, and soil C:N (Fig. 4d~f). △δ13C was negatively related to SOC content and soil C:N (Fig. 4h, i), but was not related to soil pH (Fig. 4g). None of vegetation δ13C, soil δ13C, and △δ13C were related to soil texture characteristics (clay, slit, and sand content; Fig. A3 - ESM).

Fig. 4
figure 4

Relationships of soil δ13C, vegetation δ13C, and △δ13C with soil pH, SOC, and soil C:N in temperate grasslands on the Xilingele League, Inner Mongolia, China

Full size image

3.5 Integrative effects of climatic and edaphic variables on vegetation δ13C, soil δ13C, and Δδ13C

When the integrative effects of climatic and edaphic variables were evaluated in stepwise regression analyses, only GST entered the model for soil δ13C and explained 74% of the variation in soil δ13C (soil δ13C = − 38.38 + 0.96 × GST, r2 = 0.74). GST and soil C:N entered the model for vegetation δ13C and together explained 45% of the variation in vegetation δ13C (vegetation δ13C = − 41.59 + 0.74 × GST + 0.61 × C:N, r2 = 0.45). Only soil C:N entered the model for △δ13C and explained 51% of the variation in △δ13C (△δ13C = 9.71–0.97 × C:N, r2 = 0.51).

The ordination diagram of RDA showed that most variation of δ13C was explained by climatic factors, soil C:N and SOC, but not by soil texture (Fig. 5a). Further, T value biplots of RDA showed that GST was related positively to soil and vegetation δ13C, but was not related significantly to △δ13C (Fig. 5b), soil C:N was related negatively to △δ13C and positively to vegetation δ13C, but was not related significantly to soil δ13C (Fig. 5e). None of GSP, SOC, clay content, silt content, and soil pH were related significantly to soil δ13C, vegetation δ13C or △δ13C (Fig. 5 c, d, and f~h). There was no T value biplot for sand content as the relationships between sand content and δ13C was too weak. Thus, the results of RDA confirm the results of stepwise regression analyses.

Fig. 5
figure 5

Ordination diagram and T value biplots of redundancy analysis (RDA) plotted with the first axis shown in horizontal (X) direction and second axis shown in vertical (Y) direction. Response variables (soilδ13C, vegetation δ13C, and △δ13C) were displayed as solid arrows, explanatory variables (GST, growing season temperature; GSP, growing season precipitation; SOC, soil organic carbon content; C:N, soil carbon:nitrogen ratio; clay, soil clay content; silt, soil slit content; sand, soil sand content; soil pH) were displayed as hollow arrows. Ordination diagram (a); T value biplots with Van Dobben Circle drawn for GST (b), GSP (c), SOC (d), C:N (e), clay (f), silt (g), and soil pH (h). Response variables with arrows that end in Van Dobben Circles had significant relationships with particular explanatory variables

Full size image

4 Discussion

4.1 Potential mechanisms for soil 13C enrichment relative to vegetation

As expected, soil δ13C was higher than vegetation δ13C and this is consistent with other findings (Peri et al. 2012; Yang et al. 2015). An explanation for soil 13C enrichment relative to vegetation is the Suess effect (Revelle and Suess 1957). This arises from a shift to lighter C in the atmosphere due to the burning of fossil fuels (Garten et al. 2000). The maximum of the Suess effect on 13C-CO2 is about 1.3% (Graven et al. 2017). But, the mean △δ13C of our study was 2.1‰, suggesting that the Suess effect cannot be the only mechanism for soil 13C enrichment in our study. Thus, other mechanisms, such as isotopic fractionation during SOC decomposition, could also play an important role (Yang et al. 2015).

Soil δ13C increased linearly with vegetation δ13C (Fig. 2), a result in accord with previous observations in alpine grasslands and temperate forests (Peri et al. 2012; Yang et al. 2015). This indicates that the positive relationship between soil δ13C and vegetation δ13C is common to various ecosystems and that soil δ13C reflects the δ13C of source plants due to soil organic matter being derived mainly from plant inputs.

4.2 Effects of climate and edaphic factors on vegetation δ13C

Although GST, MAT, GSP, and MAP were all related significantly to vegetation δ13C (Fig. 3 and Fig. A2 - ESM), when they were considered together, only GST was related to vegetation δ13C. Firstly, MAT and GST are both expressions of temperature, but GST can more accurately reflect the effect of temperature on vegetation δ13C than MAT, because MAT largely depends on winter temperatures which always remain below biological zero in the region of the study and have no effect on growing vegetation. Secondly, this result indicated that temperature is the dominant controlling factor for vegetation δ13C, a result inconsistent with those showing that precipitation is the dominant controlling factor for δ13C of C3 species in Inner Mongolia (Wittmer et al. 2008; Zhou et al. 2016). The reason for this inconsistency is that the previous research was conducted at the species-level and only included C3 species, whereas the present study was conducted at the community-level and included both C3 and C4 species. There is a significant difference of δ13C between C3 and C4 species, with mean δ13C values of − 12‰ for C4 species, and − 26‰ for C3 species (O'Leary 1981). Thus, when there are only C3 species present, precipitation is the major determinant of plant δ13C in the Inner Mongolia region (Zhou et al. 2016). However, when both C3 and C4 species are present, the relative abundance of C4 plants (PC4) is the major determinant of vegetation δ13C in the region (Chen et al. 2007). Moreover, there is abundant evidence that PC4 is primarily controlled by temperature and it has been found that PC4 increased with GST in Inner Mongolia grassland ecosystems (Wittmer et al. 2010; Auerswald et al. 2012). Thus, at the community-level, GST is the dominant climatic factor determining vegetation δ13C in the temperate grasslands of Inner Mongolia.

Although there was no correlation between vegetation δ13C and soil C:N in simple regression (Fig. 4f), when the climate and edaphic factors were considered together, vegetation δ13C was positively related to soil C:N (Fig. 5e). The mechanisms underlying the positive correlation between vegetation δ13C and soil C:N are not clear. Possibly it is because soil C:N is positively related to PC4. To test this hypothesis, data on PC4 across soil C:N gradients in our study area is required. However, there was significantly positive correlation between plant and soil C:N (P = 0.007) in our study, and leaf C:N of C4 species was significantly higher than that of C3 species (Zheng and Shangguan 2007), which supports our hypothesis above.

4.3 Effects of climate and edaphic factors on △δ13C

Although there were significant positive relationships of △δ13C with GST, MAT, GSP. and MAP in simple regression (Fig. 3 and Fig. A2 - ESM), when the climate and edaphic factors were considered together, the climate factors were not significantly related to △δ13C (Fig. 5b, c). This indicates that compared with edaphic factors, climate factors are not the predominant controlling factors of SOC decomposition. This result is inconsistent with the finding of Peri et al. (2012), but is in accord with the finding of Yang et al. (2015) that edaphic rather than climatic factors controlled over △δ13C in alpine grasslands on the Tibetan Plateau, i.e., △δ13C regulated by soil C:N and silt content. We found that △δ13C was regulated only by soil C:N as shown by a negative association between C:N and △δ13C (Fig. 5). This negative association could be largely driven by the higher rate of SOC decomposition in soils with lower C:N ratios (Yang et al. 2015).

It is known that in many ecosystems, soil texture can play an important role in regulating SOC dynamics. Generally, fine textured soils can increase physical and hydrological protection of SOC by inhibiting decomposition, and stabilizing SOC (Schimel et al. 1994; Wynn et al. 2006; Xu et al. 2016). This would reduce gaseous carbon losses to the atmosphere and the leaching of dissolved organic carbon leading to lower △δ13C. Indeed, previous studies have found that △δ13C was negatively related to soil silt content (Powers and Schlesinger 2002; Yang et al. 2015). However, △δ13C was not related to soil silt content or soil clay or sand content in our study (Fig. A3 - ESM). Wynn et al. (2006) found that SOC decomposition was regulated by soil texture in some climatic regions, but not in others. Furthermore, previous studies have demonstrated that the importance of clay content on SOC decomposition could vary across soils with different clay mineralogy (Spain 1990; Percival et al. 2000). Thus, the absence of association between △δ13C and soil texture might be attributed to the climate and the physical and chemical properties of clay and silt minerals in our study area. Clearly, these hypotheses need to be further considered and tested in future studies.

Based on the results and analysis above, we suggest that climate factors might indirectly affect SOC decomposition through their effects on species composition and plant C:N, and subsequently soil C:N. The significant positive relationships between plant and soil C:N (P = 0.007), and significant relationships between climate factors and plant and soil C:N (GST vs. plant C:N, P < 0.001; GSP vs. plant C:N, P < 0.001; GST vs. soil C:N, P < 0.001; GSP vs. soil C:N, P < 0.001) support this suggestion.

It is relevant to note that nitrogen deposition may affect △δ13C by changing plant species composition and microbial activity (Berendse et al. 2001; Waldrop et al. 2004). Thus, it is necessary to investigate the relationship between nitrogen deposition and △δ13C in future. Moreover, another factor requiring attention is that: as it is difficult to obtain source vegetation δ13C information, present vegetation δ13C is mostly used as proxy for source vegetation δ13C. In this context, using Δδ13C to accurately assess soil carbon dynamic is based on the assumption that the present vegetation δ13C reflects the source vegetation δ13C of soil. Thus, in certain situations, i.e., where there has recently been large plant composition (especially C4/C3 abundance) transformation, this approach is not entirely satisfactory for understanding SOC dynamics. In regard to consideration of interannual variation of vegetation δ13C, the average value of vegetation δ13C for several years should more accurately reflect the source vegetation δ13C. Unfortunately, the present study only considered vegetation δ13C for 1 year.

4.4 Effects of climate and edaphic factors on soil δ13C

We found that GST had a positive effect on soil δ13C. An explanation for this is that as SOC is derived mainly from plant inputs, SOC largely preserves the isotopic signal of the plants it derives carbon from, and that GST has a positive effect on vegetation δ13C by its effect on PC4 (Figs. 3c and 5b). Thus, GST can indirectly affect soil δ13C through its effect on vegetation δ13C. However, unexpectedly the relationship between soil δ13C and GST was stronger than that between vegetation δ13C and GST. This indicates that GST controls soil δ13C not only through its effects on vegetation δ13C but also through other mechanisms, e.g., SOC decomposition which can, to some extent, lead to 13C enrichment, resulting in higher soil δ13C (Garten et al. 2000). Based on the relationships of △δ13C with climatic and edaphic factors, we suggest that climate factors might indirectly affect SOC decomposition through their effects on plant species composition and plant C:N, and then soil C:N. This suggestion is supported by the negative relationship between GST and soil C:N (P < 0.001).

As there was a significant relationship between soil C:N and △δ13C (Figs. 4c and 5e), soil C:N should have an effect on SOC decomposition, and then to have an effect on soil δ13C. However, when the climatic and edaphic factors were considered together, there was no relationship between soil δ13C and soil C:N (Fig. 5e). This may be because soil C:N had a positive effect on vegetation δ13C and a negative effect on SOC decomposition (Fig. 5e), i.e., the effect of soil C:N on soil δ13C was eliminated, due to its offsetting effects on δ13C value of source vegetation and SOC decomposition.

5 Conclusions

Firstly, this study finds a positive association between soil and plant δ13C and soil 13C enrichment relative to vegetation in temperate grassland in the Inner Mongolia, that is similar to previously observed isotopic patterns in alpine grasslands. Secondly, the study demonstrates that GST is the dominant controlling factor of vegetation δ13C, most likely through its effect on PC4. More importantly, by exhibiting that △δ13C was regulated by soil C:N, our results indicates that edaphic rather than climatic factors are the dominant controlling factor of SOC decomposition. Further, the effects of climatic factors on △δ13C are indirect and induced by their effects on soil C:N through influencing species composition and plant C:N. Finally, we suggest that compared with soil δ13C, △δ13C may provide more accurate information about soil carbon dynamics, i.e., larger △δ13C means faster SOC decomposition rate, as soil δ13C may either underestimate or overestimate the effects of some factors on SOC decomposition. This emphasizes that measuring plant δ13C, soil δ13C, and △δ13C simultaneously are necessary to comprehensively understand how environmental change influences the carbon cycles of ecosystems.