While global climate change has received a great deal of attention in recent years, effective predictions of future climate change depend on relevant information about climate in the geological past. Over recent decades, stable carbon isotopes in sediments such as loess and paleosol, as well as in lacustrine and marine sediments, have been widely used to reconstruct paleovegetation and paleoenvironments, and have provided important insights into patterns of past climate and environmental changes. For example, numerous researchers have used organic carbon isotopes of loess to reconstruct paleovegetation and paleoprecipitation. Vidic and Montañez (2004) conducted a reconstruction of paleovegetation of the central Chinese Loess Plateau during the Last Glaciation (LG) and Holocene using organic carbon isotopes in loess from Jiaodao, Shanxi Province. Hatté and Guiot (2005) carried out a paleoprecipitation reconstruction by inverse modeling using the organic carbon isotopic signal of the Nußloch loess sequence (Rhine Valley, Germany). Rao et al. (2013) reconstructed high-resolution summer precipitation variations on the western Chinese Loess Plateau during the LG using a well-dated organic carbon isotopic dataset. Yang et al. (2015) reconstructed a minimum 300 km northwestward migration of the monsoon rain belt from the Last Glacial Maximum to the Mid-Holocene using organic carbon isotope data from 21 loess sections across the Loess Plateau. However, to our knowledge, there are no paleotemperature reconstructions using organic carbon isotope records of loess and paleosol because it has been argued that temperature exerts only a slight or even no influence on soil organic carbon isotope values (δ13CSOM). While this may be likely, it needs to be investigated because few studies have addressed the influence of temperature on organic carbon isotopes of modern surface soil. Lee et al. (2005) and Feng et al. (2008) both reported no relationship between temperature and surface soil δ13C in central–East Asia. However, Lu et al. (2004) discovered a nonlinear relationship between mean annual temperature (MAT) and δ13CSOM for the Qinghai–Tibetan Plateau. Sage et al. (1999) compiled the data from Bird and Pousai (1997) and also found a nonlinear trend for the variation in δ13CSOM along a temperature gradient in Australian grasslands and savannas.

Plant residues are the most important source of soil organic matter. Values for δ13CSOM are generally close to plant δ13C values, despite isotopic fractionation during decomposition of organic matter (Nadelhoffer and Fry, 1988; Balesdent et al., 1993; Ågren et al., 1996; Fernandez et al., 2003; Wynn, 2007). Thus, the factors influencing plant δ13C might also influence δ13CSOM. Plant δ13C values, especially those of C3 plants, are tightly associated with precipitation, suggesting that precipitation may also affect soil δ13C (Diefendorf et al., 2010; Kohn, 2010). In addition to the effect of precipitation, numerous other factors such as temperature, air pressure, atmospheric CO2 concentration, altitude, latitude, and longitude may also influence δ13C in plants (Körner et al., 1991; Hultine and Marshall, 2000; Zhu et al., 2010; Xu et al., 2015). Although variation patterns of plant δ13C with respect to temperature are so far unresolved (e.g., Schleser et al., 1999; McCarroll and Loader, 2004; Treydte et al., 2007; Wang et al., 2013), it is widely accepted that temperature has a slight effect on plant δ13C. Therefore, if the 13C enrichment during soil organic matter decomposition is a constant value, we expect only a slight or no influence of temperature on soil δ13C. However, 13C-enrichment is affected by environmental and biotic factors (Wang et al., 2015). Thus, it is difficult to determine whether or how temperature affects soil δ13C, and there should be specific investigations focusing on this issue. Although the relationship between temperature and δ13CSOM has been investigated in the studies mentioned above, these studies were unable to effectively separate the influence of temperature from the effect of precipitation. In addition, there are no meteorological stations near most of the sampling sites in the aforementioned studies, suggesting that meteorological data had to be interpolated, which can lead to unrealistic precipitation data in regions with strong topographical variability. This interpolation could have introduced errors in the relationships between temperature and δ13CSOM that were established in these studies.

The present study includes a detailed investigation of the variation in δ13CSOM with respect to temperature across a temperature gradient along the 400 mm isohyet (isopleth of mean annual precipitation; MAP) in China. We sampled surface soil along a specific isohyet to minimize the effect of precipitation changes on δ13CSOM.

In addition, we collected samples only at sites with meteorological stations. Thus, the climatic data we obtained from these stations are likely more reliable than interpolated values.

Except for one δ13CSOM value (-18.8 ‰), all other data vary between -20.4 and -27.1 ‰, with a mean value of -23.3 ‰ (n= 80, SD = 1.45). Multiple regressions with MAT, MAP, altitude, latitude, and longitude as independent variables and δ13CSOM as the dependent variable show that only 9 % of the variability in soil δ13C can be explained by a linear combination of all five environmental factors (p= 0.205; Table 2). Considering the possibility of correlations among the five explanatory variables, stepwise regression was used to eliminate the potential influence of collinearity among them. Variables with P values < 0.05 were incorporated into the model and variables with P values > 0.1 were excluded. Statistical analysis shows that only latitude is included in the stepwise regression model (R2= 0.077, p= 0.012). In order to better constrain the relationship between soil δ13C and each environmental factor, bivariate correlation analyses of soil δ13C against some of the environmental factors were conducted. The bivariate correlation analyses show that δ13CSOM is not related to MAT (p= 0.114) or SMT (p= 0.697) along the isohyet (Fig. 2a, b). In addition, in order to further determine the response of δ13CSOM to temperature, we considered three subsets of our soil samples defined according to the climate, topography, or vegetation type of the Qinghai–Tibetan Plateau (mainly alpine meadow, 10 sites), steppe or grassland (11 sites), and coniferous forest (6 sites; Table 1). Bivariate correlation analyses within these subsets also show no relationship between δ13CSOM and MAT for all categories. The correlation analysis of δ13CSOM with respect to altitude is shown in Fig. 3, which displays no relationship (p= 0.132). Although longitude is not found to influence δ13CSOM in the above stepwise regression, bivariate correlation analyses show that both latitude and longitude are negatively correlated with δ13CSOM (p= 0.012 and 0.034, respectively; Fig. 4a, b).

In addition to the effects of quantifiable environmental factors, qualitative factors such as soil type and vegetation type may influence δ13CSOM. Various concepts have been introduced in soil taxonomy, leaving varied soil nomenclatures in use. In this study, we adopted the Chinese Soil Taxonomy and the World Reference Base (WRB) to describe the observed soils. The soil samples can be divided into eight or six types based on the Chinese Soil Taxonomy or the WRB, respectively (Table 1). One-way ANOVA analyses suggest that both soil and vegetation type play a significant role for δ13CSOM (p= 0.002 for soil type based on the Chinese Soil Taxonomy, p= 0.003 for soil type based on the WRB, and p= 0.001 for vegetation type; Fig. 5).

To further constrain the effects of soil and vegetation type on δ13CSOM, multiple regressions with soil and vegetation type as dummy variables were conducted. Considering the tight relationship between soil type and vegetation type, especially in the Chinese Soil Taxonomy, soil variables and vegetation variables were separately introduced into the statistical analyses. Multiple regression, in which the five aforementioned explanatory environmental factors were taken as quantitative variables and the eight soil types of the Chinese nomenclature as values of a dummy variable, shows that environmental factors and soil types account for 37.4 % of the soil δ13C variance (p < 0.001; Table 2). Using the six soil types based on the WRB rather than the Chinese nomenclature, 29.7 % (p= 0.003) of the variability is explained (Table 2). Similarly, multiple regression with vegetation types as dummy variables shows that the five environmental factors and vegetation types together can explain 36.2 % of the variability in soil δ13C (p= 0.001; Table 2). Compared to the multiple regressions in which only quantitative environmental variables were introduced, the multiple regressions in which soil and vegetation were also introduced explain more of the variance, suggesting that soil type and vegetation type play a significant role in δ13CSOM variability.

Soil δ13C depends on the δ13C of plants and on carbon isotopic fractionation during organic matter decomposition. δ13C values of C3 plants vary between -22 and -34 ‰ with a mean of -27 ‰, and C4 plants range from -9 to -19 ‰ with a mean of -13 ‰ (Dienes, 1980). Carbon isotope fractionation occurs during the process of plant litter decomposition to soil organic matter in most environments, especially in non-arid environments, causing 13C-enrichment in soil organic matter compared to the plant sources (Nadelhoffer, 1988; Balesdent et al., 1993; Ågren et al., 1996; Fernandezet al., 2003; Wynn et al., 2005; Wynn, 2007). A detailed investigation of isotope fractionation during organic matter decomposition, which was conducted on Mount Gongga, an area of the Qinghai–Tibetan Plateau dominated by C3 vegetation with herbs, shrubs, and trees, showed that the mean 13C-enrichment in surface soil (0–5 cm depth) relative to the vegetation was 2.87 ‰ (Chen et al., 2010). Another investigation of 13 soil profiles from the Tibetan Plateau and north China showed that the δ13C difference between surface soil and the original biomass varied from 0.6 to 3.5 ‰ with a mean of 1.8 ‰ (Wang et al., 2008). Thus, the δ13CSOM dataset from this study (δ13CSOM ranges from -20.4 to -27.1 ‰) indicates that the modern terrestrial ecosystem along the isohyet is dominated by C3 plants. This result is consistent with the observations of vegetation along the isohyet completed in our previous study (Wang et al., 2013). The relatively heavy soil δ13C values (mean: -20.4 ‰) at Rikaze (Site 2; Fig. 3 and Table 1) are surprising because only four species of C3 plants grow there, and C4 species are absent. This observation suggests that very large carbon isotope fractionation during SOM degradation has occurred in the local ecosystem. Previous studies have observed a similar phenomenon, although the mechanism responsible for the unusually large isotopic fractionation remains unclear. For example, Wynn (2007) reported that isotopic fractionation enriched soil organic carbon by up to ∼ 6 ‰ with respect to the original biomass.

The MAT, MAP, altitude, latitude, and longitude combined are responsible for only 9 % of the variability in soil δ13C in the multiple regression model, suggesting that the contribution of these five environmental factors to soil δ13C variance is very small. Our previous study conducted along the same isohyet indicated a strong positive relationship between the δ13C of plants and MAT, with a coefficient of 0.104 ‰ ∘C-1 (Wang et al., 2013). The difference between the maximum and minimum temperature along the isohyet is 15 ∘C, so the greatest possible effect of temperature on plant δ13C along the temperature gradient is 1.56 ‰, which is not very substantial. Because the main source of soil organic matter along the isohyet is C3 plants, the induced variance in soil δ13C by plant δ13C can also not be very high. On the other hand, although the 13C-enrichment during SOM degradation follows a Rayleigh distillation process (Wynn, 2007), our recent study shows that temperature does not influence carbon isotopic fractionation during decomposition of organic matter (Wang et al., 2015), which also explains the lack of a relationship between soil δ13C and temperature. Feng et al. (2008) and Lee et al. (2005) reported no relationship between soil δ13C and MAT and SMT, respectively, which is consistent with our results. Their field campaigns were conducted in central Asia, which is also dominated by C3 plants, similar to the area along the 400 mm isohyet. This is the reason why the same pattern exists in both central Asia and in the area along the 400 mm isohyet.

Observations by Bird and Pousai (1997) and Sage et al. (1999) appear to be inconsistent with our findings; the authors found a nonlinear relationship between soil δ13C and MAT in Australian grasslands. However, if they considered only soil with pure C3 plants (MAT is below 16 ∘C), soil δ13C and temperature were not related in Australian grasslands, which agrees with our results. Below 15 ∘C, the C4 contribution to productivity in Australian grasslands is negligible, whereas above 23 ∘C, C3 contribution is negligible. Between 14 and 23 ∘C, soil δ13C is positively correlated with MAT, indicating an increase in C4 representation with increasing MAT (Sage et al., 1999). Lu et al. (2004) also reported a nonlinear relationship between soil δ13C and MAT. Similarly, if their soil data with C4 plants are excluded from the nonlinear correlation, soil δ13C is also not related to MAT (see Fig. 5b in Lu et al., 2004). Thus, the present study and the previous observations are consistent in showing that in a terrestrial ecosystem in which the vegetation is dominated by C3 plants, temperature does not influence soil δ13C variance.

Because all soil samples were taken along the 400 mm isohyet, this study shows that the contribution of precipitation to the variability in soil δ13C is negligible. Although stepwise regression and correlation analysis both show a significant influence of latitude on soil δ13C (p= 0.012; Fig. 4a), which was also described by Bird and Pausai (1997) and Tieszen et al. (1979), the five environmental variables, including latitude, are responsible for only 9 % of the variability in soil δ13C in a multiple regression model (Table 2), suggesting that the contribution of latitude to soil δ13C is also limited. Latitude is a comprehensive environmental factor, and change in latitude can bring about changes in other environmental factors, such as temperature, irradiation, cloud amount, and moisture. Among those, temperature and irradiation should be most strongly related to latitude. The observed relationship between latitude and soil δ13C suggests that environmental factors other than temperature might also contribute to the variance in soil δ13C.

Control of soil δ13C by vegetation type mainly reflects the effect of plant life forms on plant δ13C, which in turn influences isotope fractionation during organic matter decomposition. Communities in which life forms of dominant plants are similar are generally treated as the same vegetation type. Plant δ13C is closely related to plant form (Diefendorf et al., 2010; Ehleringer and Cooper, 1988), which causes δ13C differences among varying vegetation types, resulting in the observed effect of vegetation type on soil δ13C.

Substrate quality partly quantifies how easily organic carbon is used by soil microbes (Poage and Feng, 2004). It can be related to plant type and is often defined using the C / N ratio, lignin content, cellulose content, and/or lignin content / N ratio (Melilloet al., 1989; Gartern et al., 2000). Our study of Mount Gongga, China, showed that litter quality plays a significant role in isotope fractionation during organic matter decomposition, and that the carbon isotope fractionation factor α increases with litter quality (Wang et al., 2015). Thus, the isotope fractionation factor should differ among sites because litter quality is dependent on vegetation, which causes changes in soil δ13C with vegetation type.

The effect of soil type on soil δ13C may be associated with the effect of soil type on isotope fractionation during organic matter decomposition, which involves at least two mechanisms (see Wang et al., 2008, for a detailed discussion). First, properties and compositions of microbial decomposer communities are dependent on soil type (Gelsomino et al., 1999). Different microbes can have different metabolic pathways, even when they decompose the same organic compound (Macko and Estep, 1984), and the extent of isotope fractionation during decomposition may be closely related to the metabolic pathways of microbes (Macko and Estep, 1984). Second, physical and chemical properties such as pH, particle size fraction, and water-holding capacity are considerably different among soil types, which causes organic compounds to decay at different rates in different soil environments. The magnitude of isotope fractionation during decomposition is linked to the degree of organic matter decomposition (Feng, 2002). Thus, soil type plays a significant role in soil carbon isotopic fractionation.

The present study analyzed organic carbon isotopes in surface soil along a 400 mm isohyet of mean annual precipitation in China. Our results indicate that both soil type and vegetation type significantly influence soil organic carbon isotopes. However, temperature is found to have no observable impact on δ13CSOM, suggesting that δ13C signals in sediments cannot be used for temperature reconstructions and that the effect of temperature on δ13CSOM should be neglected in reconstructions of paleoclimate and paleovegetation that use carbon isotopes of soil organic matter.

There is no underlying material and related items in this paper. All data will be provided in the Supplement.