The samples were gathered in Zhenxiong County and Weixin County, Yunnan Province, an area at the junction of Yunnan, Guizhou, and Sichuan provinces. The area has typical karst mountains, dominated by mountainous, semi-mountainous, and frigid alpine areas. The area has an elevation of 480–2,416 m. The area belongs to a warm monsoon climate, characterized by cool and rainy weather. The annual temperature varies between 11 and 20°C, and the annual precipitation is in the range of 600–1,200 mm (Yang and He, 1999).

Zhenxiong County and Weixin County have a total area of 5,088.7 km², including 1,562.4 km² farmland, 43.2 km² garden land, 2,373.1 km² forest land, 38 km² grass land, 34 km² water area, 202.5 km² construction land, and 835.5 km² unused land. The main vegetation types in this area include coniferous forest, broad-leaved forest, shrub, and grass (Yang and He, 1999).

The strata in this area mainly include the early and later Paleozoic, early Mesozoic, and Quaternary strata. The widely exposed strata from old to new are the Maokou Formation of early-Permian; the Emeishan, Longtan, and Changxin Formations of mid-Permian; and the Feixianguan and Yongningzheng Formations of early-Trias and the Quaternary. The lithology of the Maokou Formation is bioclastic limestone and chert nodule limestone. The host rock of the Emeishan Formation is vesicular amygdaloidal basalt. The lithology of the Longtan Formation mainly includes shale and sandstone interbedded with coal seam. The Changxin Formation consists of limestone interbedded with shale and fine sandstone. The Feixianguan Formation is characterized by siltite, muddy siltstone, calcareous siltstone, and mudstone. The Yongningzheng Formation is dominated by argillaceous limestone, limestone, and sandy shale. Besides, other strata sporadically occur (Figure 1).

The samples were gathered for the unit of stratum and bedrock. Totally, 10 types of bedrocks in different stratum were involved (Figure 1). These profiles are silty limestone in the Baota Formation of mid-Ordovician (O₂b), argillaceous limestone in the Wufeng Formation of late-Ordovician (O₃w), limestone in the Qixia Formation of early-Permian (P₁q), bioclastic limestone in the Maokou Formation of early-Permian (P₁m), basalt in the Emeishan Formation of mid-Permian (P₂β), shale in the Longtan Formation of mid-Permian (P₂l), limestone interbedded with shale in the Changxin Formation of mid-Permian (P₂c), siltite in the Feixianguan Formation of early-Trias (T₁f), dolomite in the Guanling Formation of mid-Trias (T₂g), and siltite and sandstone in the Xujiahe Formation of late-Trias (T₃x). These profiles cover the main lithology and strata in this area.

SOC is governed by multiple factors, and some other factors affecting SOC were tried to be excluded when sample locations were selected. The soil profiles were away from pollution sources such as industry or agriculture. All the samples were from the shrub land because the shrub land is widely distributed in this area. Moreover, a successive relationship between the bedrock and soil was roughly estimated in the field through the grain size, mineral composition, residual texture, slope, and topography. For every bedrock unit, soils in the surface and mineral horizons, and bedrocks were, respectively, sampled (Figure 2); that is, 3 samples for each bedrock unit were taken. The surface soil was taken in the upper 10 cm layer, and the soil was taken near the interface with the bedrock (within 10 cm) in the mineral horizon. The bedrock should be fresh rock. The samples were taken from 3 to 5 locations in the same horizon and mixed into one. The samples were kept in canvas sampling bags.

The samples were sent to the laboratory and air-dried at room temperature. The samples were ground to less than 100 mesh for analysis. SOC was determined by the wet oxidation method with K₂Cr₂O₇ following the study proposed by Lu et al. (1999). The samples were dissolved by HCl–HNO₃, and then the analysis of elements was performed by inductively coupled plasma–atomic emission spectroscopy (ICP-AES). For quality control, standard samples and parallel samples were also measured, and the relative errors were less than 5%. The analytical data are provided in Supplementary Table S1.

For correlation analysis, software SPSS 13.0 and Excel 2019 were used. p < 0.01 indicated significant correlation at 0.01 level, and p < 0.05 indicated correlation at 0.05 level.

The successive relationship and soil origins are key and important to reveal the SOC characteristics on different bedrocks. Geochemistry acts as an effective tool to trace soil origin, and a series of geochemical indexes was discussed. Fe, Al, and Ti are relatively immobile or weakly mobile during weathering (Hill et al., 2000). Al/Ti, Fe/Ti, and Al/Fe remain roughly equivalent between the bedrock and the in situ soil, and thus can be used to ascertain their succession (Young and Nesbitt, 1998).

The ratios of Al/Ti, Fe/Ti, and Al/Fe are listed in Table 1. The ratios for limestone in the Qixia Formation of early-Permian show obvious deviations between the bedrock and the soil, which indicates that the soil does not originate from the underlying bedrock although the residual texture, slope, and topography had been considered in the field. So, this profile was excluded for this research. For other profiles, the ratios of Al/Ti, Fe/Ti, and Al/Fe between the soil and the bedrock are not absolutely equivalent because of the different active coefficients, illuviation, and soil-forming conditions, but they show roughly equivalent values in the mineral horizon. Additionally, the reason for some greater deviations between the surface soil and the bedrock may possibly be because the surface soil is more affected by the external environments (Sharma and Rajamani, 2000). Significantly, positive correlations between these ratios in the bedrock and in the mineral horizon were observed, with R² _Al/Ti = 0.82, R² _Fe/Ti = 0.69, and R² _Al/Fe = 0.61. The ratios of soils in mineral horizons and in surface horizons on the same bedrock were also significantly positively correlated (R² _Al/Ti = 0.66, R² _Fe/Ti = 0.83, and R² _Al/Fe = 0.56).

Many researchers have discussed the soil origin in this studied area or its neighborhood using mineralogy, trace element geochemistry, particle-size characteristics (Sun et al., 2002; Liu et al., 2004), acid-dissolved extraction experiments (Wang et al., 1999), REE (Wang et al., 1999; Sun et al., 2002), grain size of quartz (Feng et al., 2009), and grain size of soil profiles (Feng et al., 2002). Generally, most researchers confirmed that the soil was the accumulation of insoluble residue from the underlying rock (Isphording, 1978; Danin et al., 1982), while only a few considered that the soil originated from other sources such as clastic materials transported by water or air, volcanic ash, and desert dust (Yaalon, 1997; Wang et al., 2019). Our results also support that the soil is the weathering product of the underlying rock, and the soil geochemistry is succeeded from the bedrock.

The average SOC contents on different bedrocks are summarized in Table 2. The soils in the mineral horizon had SOC contents of 0.19–2.74% (an average of 1.24%), with the highest of 2.74% on basalt in the Emeishan Formation of mid-Permian and the lowest of 0.19% on dolomite in the Guanling Formation of mid-Trias. The SOC for surface soil showed higher contents of 1.26–4.01% (a mean of 2.63%), with the highest of 4.01% on shale in the Longtan Formation of mid-Permian and the lowest on silty limestone on the Baota Formation of mid-Ordovician. The bedrock organic contents only range between 0.01 and 1.63%.

SOC in the mineral horizon on different bedrocks differs greatly, indicating the effect of underlying rocks on SOC in the mineral horizon. SOC in the mineral horizon is not correlated with organic carbon in rocks (Figure 3), implying the SOC contents in the bedrock itself cannot explain the contents in the mineral horizon. There exists a positive correlation between SOC contents in the mineral horizon and in the surface horizon, indicating SOC in the surface and mineral horizons is affected by similar soil geochemistry since they originate from the same bedrock. However, the correlation coefficient (R ² = 0.58) is low (Figure 3), possibly because the external inputs contribute more to the surface soil than to the soil in the mineral horizon.

The correlations between SOC and soil geochemistry in the mineral horizon are shown in Figure 4. SOC in the mineral horizon is significantly positively correlated (p < 0.01) with Co (R ² = 0.674), Cu (R ² = 0.746), Ti (R ² = 0.81), V (R ² = 0.688), and Zn (R ² = 0.722), and positively correlated (p < 0.05) with Ni (R ² = 0.595). The elements of Co, Cu, Ti, V, Zn, and Ni strongly bind to SOC by forming organic metal complexes because they belong to the first transition metal ions (Goodman and Cheshire, 1973; Cheshire et al., 1977; Xing and Zhu, 2003). Moreover, organic metal complexes have been proven to maintain SOC stability and slow down the SOC decomposition rate (Clough and Skjemstad, 2000; Aran et al., 2001; Tonneijck et al., 2010). Therefore, these elements in the mineral horizon seem to play the most important role in governing SOC. Ca is also documented to enhance SOC stability by Ca–SOC complexing actions (Clough and Skjemstad, 2000; Schmeide and Bernhard, 2009; Saito et al., 2010). The positive Ca–SOC correlation (p < 0.05) was also observed in the mineral horizon, with a low correlation coefficient (R ² = 0.457). Additionally, the significantly positive phosphorus–SOC (p < 0.01) correlation (R ² = 0.73) may result from the fact that SOC and phosphorus are both soil nutrient elements.

The correlations between surface SOC and soil geochemistry are shown in Figure 5. Only Co (R ² = 0.698), Cu (R ² = 0.685), V (R ² = 0.687), and Al (R ² = 0.613) have positive correlations (p < 0.05) with SOC. Compared with the soils in the mineral horizon, SOC is not positively correlated (p < 0.05) with Ti, Zn, and Ni in the surface horizon. Moreover, no significantly positive correlations (p < 0.01) between surface SOC and soil geochemistry were observed. Surface SOC is partly affected by Co, Cu, and V in this study area. Importantly, surface SOC is more affected by the external inputs and external environments, resulting in the difference of its correlations with the first transition metal ions from the SOC in the mineral horizon.

The SOC pool is derived from the by-products of microbial processes and the decomposition of organic matters, and is estimated to be 1,505 Pg to 1 m depth of the soil. The atmosphere only has a carbon pool of 820 Pg (Batjes, 1996). CO₂ emission from the soil to the atmosphere is a substantial constituent of greenhouse gas emissions and is a vital part of the carbon cycle. A small emission from the SOC pool may largely affect atmospheric CO₂ and aggravate global warming. Also, a small increase in the SOC pool results in a strong drawdown impact on atmospheric CO₂. Thus, thoroughly understanding and managing the SOC pool is of a critical importance to limit global warming, and SOC evolution is an important process for the global carbon cycle.

However, the SOC content and its influencing mechanisms on different rocks are difficult to characterize, although a series of factors related to SOC has been widely discussed. Surface soils have wide sources of organic carbon and cannot accurately exemplify the effect of bedrock on SOC. Contrarily, soils in the mineral horizon, which are less affected by the external environments, were focused on in this research.

The variation of SOC content affects the emission amount of CO₂ into atmosphere, and carbon fixation in soils is significant for global carbon cycles (Schlesinger and Andrews, 2000; Lal, 2010). The SOC content is the result of the equilibrium between gains from organic supplies and decomposition (Lal, 2010). SOC in the mineral horizon seems to be largely affected by its decomposition since the gained organics are markedly less than those in the surface soil. SOC is rich in oxygen-bearing functional groups such as COOH, OH, and C=O, which can form metal–organic complex. Nitrogen-containing groups such as -NH₂ and -N=N- are fewer in SOC, but have a stronger complexing ability with metal ions than oxygen-bearing functional groups (Stevenson, 1991). The metal–organic complexation is also one of the important processes controlling SOC contents. Transition metal ions, especially the first transition metal ions, have priority, high affinity, and strong ligand ability to complex with SOC because of their free electron orbits, preferentially forming stable complexes (Xing and Zhu, 2003).

The mechanisms protecting SOC from decomposition are still the subject of debate (Baldock and Skjemstad, 2000; Tonneijck et al., 2010; Kida and Fujitake, 2020). The metal–organic complex has been proven to maintain SOC stability by restraining against biological decomposition. The formation of metal–organic complexes, which accounted for 40% of the carbon accumulation, was observed to be related to SOC stabilization in Andosols (Tonneijck et al., 2010). Clough and Skjemstad (2000) found organic complexes could not be effectively decomposed even though the soil microbial activity was improved. Complexation with SOC resulted in functional groups becoming more condensed and less susceptible to biological attack (Baldock and Nelson, 2000), and Sollins et al. (2006) stated the average residence time of SOC increased with increasing density. Aran et al. (2001) performed disaggregation tests and incubation experiments, and concluded the complex had strong resistance against disaggregation and biodegradation. Wang et al. (2013) concluded that Fe-o affected by sorption and Fe-p by complexation on SOC preservation, and Fe-p had higher contributions to fix organic carbon than Fe-o. Olatunji and Osibanjo (2014) found the stability of metal–organic carbon complexes depended on the organic carbon type, nature of the complex formed, and the degree of stearic effect. Organo-metallic complexes were widely documented to be responsible for the stabilization of SOM and the high SOC stocks (Blaser et al., 1997; Jiang et al., 2021). Consequently, the SOC decomposition rate is slowed down, and more organic carbon is accumulated in soils with a higher content of transition metal ions.

Also, the relationships between SOC and transition elements have been widely documented. Goodman and Cheshire (1973) added Cu into peaty humic acid and gained the Cu–porphyrin complex compound. Lakato et al (1997) deemed that Cu ligands with heterocyclic nitrogen atoms, not with porphyrin. Senesi et al. (1977) detailed humic acid’s complexation with Fe²⁺. Goodman and Cheshire (1976) described how VO³⁻ converts into VO²⁺ when it complexes with SOC. Also, many studies have revealed that SOC is positively correlated with these transition metal ion contents, such as with Cu²⁺ (Ni et al., 2000; Zhu et al., 2006), Ti²⁺ (Gao et al., 1994), V²⁺ (Goodman and Cheshire, 1976), Fe²⁺ (Wang et al., 2005; He et al., 2006), and Co²⁺ (Fan et al., 2006).

A series of simulated experiments about the interactions between metal ions and SOC was also recorded. Georg et al. (1994) confirmed that the SOC respiration, as well as SOC solubilization, was reduced by metal–organic complexes. Boudot (1992) concluded that organo-metallic complexes account for the low SOC mineralization rates. The addition of Ca²⁺ to soil has also been shown to reduce SOC mineralization (Muneer and Oades, 1989). Boudot et al. (1989) observed that the protective effect exerted by metallic ions arose from the formation of metallic hydroxide trapping organic molecules. Cai (2018) found that organic–metal complexes greatly increased under the exogenous addition of Cu and Zn. Zhou (2013) observed the high metal ions decreased the SOC turnover and increased stable organic carbon because of the low mineralization rates.

The rock–soil geochemistry of the first transition metal ions was analyzed to discuss their successive relationship. Ca (Mg) CO₃ leaching has been the most important process of soil forming in the carbonate area because carbonate is largely composed of Ca (Mg) CO₃, which differs greatly from the non-carbonate area. The trace transition metal ions highly accumulate in soil when Ca (Mg) CO₃ leaches. Moreover, Ca (Mg) CO₃ leaching is related to the conditions such as pH, Eh, climate, and clay content. Thus, different carbonates have different concentration coefficients of these transition metal ions since they have different Ca (Mg) CO₃ contents and Ca (Mg) CO₃ leaching abilities. Therefore, no significant correlations of transition metal ions between the bedrock and soil were observed in the carbonate areas. But in the non-carbonate area, Co, Cu, Ti, and V elements have significant positive correlations (p < 0.01), and Zn has positive correlation (p < 0.05) between the bedrock and soil in the mineral horizon (Figure 6). Co, Cu, Ti, and V elements have significant positive correlations (p < 0.01), and Ni has a positive correlation (p < 0.05) between the bedrock and the surface soil (Figure 7). The content of the first transition metal ion in the non-carbonate area largely governs its content in soil.

To summarize, the higher first transition metal ions in the bedrock result in their higher contents in soil during weathering. The SOC stability increases, and its decomposition rate decreases with the increasing of SOC–the first transition metal ion complexes. Consequently, more SOC accumulates in soil with higher first transition metal ions, resulting in higher SOC contents (Figure 8).