Introduction

Soil is the largest carbon pool in terrestrial ecosystems, with soil organic carbon (SOC) constituting a vital component of this carbon pool1. Globally, the SOC stock has been estimated to be up to 2400 Gt (1 Gt = 109) at a depth of 2 meters2. The dynamics and turnover of SOC are profoundly influenced by climate and vegetation3, and fluctuations in SOC stocks are linked to environmental change through a negative feedback mechanism4. An annual increase in global SOC stocks of 4 ‰ could offset the annual CO2 emissions resulting from the combustion of fossil fuels5. Consequently, understanding the mechanisms of SOC sequestration, maintaining stable soil carbon sink functions, and ensuring a high state of stability are crucial processes for soil to adapt to global environmental changes.

Active metal oxides are one of the primary entities that provide protection for SOC. Metal (hydr)oxides and metal cations can form metal-organic carbon complexes with SOC through hydrogen bonding, cation bridging, ligand exchange and complexation, thereby preserving SOC by limiting microbial metabolism6,7. It is estimated that the major reactive metal minerals sequester approximately 600 Gt of SOC globally8. Metal (hydr)oxides of iron (Fe), aluminum (Al), or manganese (Mn) bind SOC through adsorption, co-precipitation and containment in small pore spaces9,10,11. Since Lalonde et al.12 quantified Fe-bound organic carbon (Fe-OC) in marine sediments, research on Fe-OC has been widely reported across various terrestrial ecosystems, such as forests, agricultural lands, and wetlands13,14,15,16,17. Moreover, metal-bound organic carbon (metal-bound OC) can be preserved in a stable form over the long term, with Fe-OC persisting for thousands of years in marine sediments18. In contrast, calcium (Ca) typically facilitates SOC association with clay mineral surfaces through cation bridging19. In recent years, the protective role of Ca on SOC has also garnered attention, with its content recognized as the strongest explanatory variable for grassland soils20. Ca enhances the adsorption capacity of goethite for organic carbon21 and promotes the formation of mineral-bound organic carbon through mediating bio-abiotic coupling mechanisms22. It is noteworthy that the interactions between Fe, Ca, and SOC have predominantly been studied in surface soils (<30 cm) or at smaller scales. It is widely acknowledged that complex processes at the terrestrial surface are subject to extensive stress due to changes in temperature and precipitation, while the relatively stable environment of deeper soils may induce distinct driving mechanisms. However, there is a limited understanding of the response patterns of Fe-OC and calcium-bound organic carbon (Ca-OC) in deeper soil layers.

The quality and stability of SOC directly influence its accumulation. The resistance of SOC to acid hydrolysis is considered one of the important properties for quantifying the quality and stability of SOC, and the percentage of acid-resistant carbon to SOC has been shown to be a strong predictor of SOC mineralization23,24. Acid hydrolysis is used to simulate the oxidative degradation of enzymes and microorganisms in soil. On the one hand, acid hydrolysis can break the weak chemical bonds of easily decomposable components (such as the glycosidic bonds in carbohydrates23). On the other hand, it can preserve the components resistant to microbial enzymatic degradation (such as lignin25). SOC bound to minerals through strong chemical bonds is also protected by the matrix stabilization mechanism26,27. Radiocarbon dating studies have shown that the age of carbon that is not hydrolyzed after 13 M H2SO4 hydrolysis is much greater than that of hydrolyzable carbon28,29. Based on the difficulty of acid hydrolysis of SOC, it can be divided into labile organic carbon (LOC) and recalcitrant organic carbon (ROC)23. Early studies have shown that the resistance of SOC to chemical oxidation is positively correlated with specific active minerals30. Metal oxides have been repeatedly proven to be closely related to the reserves of SOC and the resistance of SOC to microbial degradation. For example, the enhancement of SOC sequestration capacity in paddy field surface soils mainly depends on the accumulation of ROC, and Fe (hydr) oxides mediate the accumulation process of SOC by increasing the content of ROC24. Compared with Fe-OC, Ca-OC has a higher content of aromatic/olefinic and phenolic C functional groups31, and exchangeable Ca has been widely confirmed to be positively correlated with SOC concentration and its resistance to oxidation31,32. However, the role of Ca and Ca-OC as another important carrier in protecting SOC stability is less known.

Moreover, alterations in soil moisture balance (i.e., aridity) and soil depth can lead to variations in soil pH and other related properties. Soil pH reflects the overall chemical status of the soil system, including the types of dissolved metals and the predominant organic-mineral bond types33. As soil pH shifts from acidic to alkaline conditions, the importance of different multivalent cations and their mineral forms for SOC stability (specifically resistance to acid hydrolysis) transitions from Al3+ or Fe3+ to Ca2+ 8,34,35,36,37. In water-limited alkaline soils, exchangeable calcium strongly predicts SOC content, whereas with increasing water availability and acidity, oxides become better predictors35. However, how the natural shifting dominance of Fe-OC and Ca-OC affects SOC stability lacks direct evidence from in-situ soil within natural processes.

To address the aforementioned concerns and fill the gaps in our understanding, we conducted a large-scale transect survey spanning 1500 km across a typical semi-arid to semi-humid region in China. The purpose of this survey was to collect natural evidence demonstrating the role of metal-bound OC in enhancing the stability of SOC. Based on existing knowledge, aridity, soil pH, and metal-bound OC exhibit a degree of covariability. Therefore, we hypothesize that SOC stability exhibits an explicit spatial covariation trend with aridity and shows regular variations along the vertical scale. Consequently, we propose that the strength of control exerted by different metal-bound OC (Fe and Ca) on SOC stability is symmetrically correlated with their relative abundance. As soil pH varies with aridity, the dominance of different metal-bound OC also shifts. To test our hypothesis, we collected 16 representative soil profiles distributed along an aridity gradient to analyze the spatial and vertical distribution of Fe-OC, Ca-OC, and ROC. Systematic sensitivity and pathway analyses were utilized to identify key factors contributing to the stabilization of SOC, and to investigate the cascading effects of soil pH on these processes.

Results

The distribution patterns of metal-bound OC and ROC across spatial and profile gradients

The spatial distribution characteristics of Fe-OC% and Ca-OC% along the climatic gradient were distinct (Fig. 1). The two forms of metal-bound OC exhibited opposing spatial patterns. Fe-OC% primarily showed a decreasing trend from northeast to southwest, whereas Ca-OC% gradually increased along this gradient (Fig. 1). Overall, the spatial pattern of ROC closely resembled that of Fe-OC, decreasing from northeast to southwest (Supplementary Fig. 1), although no distinct spatial distribution features of ROC% were detected (Fig. 1).

Fig. 1: The spatial distribution patterns of Fe-OC %, Ca-OC % and ROC % in different soil layers.
figure 1

The points of different colors represent the Fe-OC %, Ca-OC % and ROC % values of different soil sites. The larger the size of the point, the larger the value of the variable. The colored map represents the aridity. Fe-OC%, percentage of iron (Fe)-bound organic carbon in soil organic carbon; Ca-OC%, percentage of calcium (Ca)-bound organic carbon in soil organic carbon; ROC%, percentage of recalcitrant organic carbon in soil organic carbon.

Full size image

Statistical analysis from 10 cm-increment soil layers indicated that metal-bound OC and ROC exhibited distinct vertical distribution patterns (Fig. 2 and Supplementary Table 3). With depth, the metal-bound OC shifted from Fe-OC to Ca-OC, coincident with an increase in the chemical stability of SOC (ROC%) (Fig. 2). The concentration of Caex2+ generally showed a trend of gradual increase with depth along the profile, but in some soil profiles (e.g., AD and CF), it exhibited an initial increase followed by a decrease with depth (Fig. 2h, p). For the six soil profiles containing soil inorganic carbon (SIC), SIC was predominantly concentrated in the 100–140 cm soil layer (Supplementary Fig. 2).

Fig. 2: The profile distribution patterns of Fe-OC%, Ca-OC%, and ROC% across various soil sites.
figure 2

Soil sites are reorganized along the aridity gradient sequence. The gradient keys on the X and Y axes represent the progressive increase in aridity along subfigures (a–p). The numbers in the upper right corner denote specific soil sites. Red and blue rectangles represent Fe-OC% and Ca-OC% of each soil layer, respectively. Rectangles in different shades of green represent ROC% and LOC% for each soil layer. Black dots indicate the vertical dynamics of Caex2+. The areas shaded in dark gray correspond to the range of the Mollic layer. LOC% percentage of labile organic carbon in soil organic carbon, Caex2+ exchangeable calcium ions.

Full size image

We further analyzed the differences in soil properties across depth, pooling measurements from 10 cm-increment samples into 0–30 (n = 48), 30–100 (n = 97), and 100–200 cm depth increment categories (n = 107) (Fig. 3). Results from one-way ANOVA indicated that SOC, ROC, Fe-OC and Ca-OC contents decreased significantly with increasing depth increment (p < 0.05) (Fig. 3). Ca-OC % and ROC % increased significantly with soil layer (p < 0.001), while Fe-OC% showed the opposite trend (p < 0.05) (Fig. 3d, f, h). The results of simple linear regression indicated that Fe-OC and Ca-OC were significant factors affecting ROC (Supplementary Fig. 3). In all of the three depth increments, ROC exhibited a significant positive linear relationship with Fe-OC (p < 0.001) (Supplementary Fig. 3). In contrast, in the upper 100 cm of the two soil layers, we did not detect a significant linear relationship between ROC and Ca-OC, but in the lower 100 cm, this relationship shifted to a significant positive correlation (p < 0.001) (Supplementary Fig. 3). Furthermore, ROC% was also significantly positively correlated with SIC, Fed, Ald, and Caex2+ (p < 0.05) (Supplementary Figs. 4 and 5).

Fig. 3: Significant difference of relevant soil properties between different soil layers.
figure 3

a Soil organic carbon content. b Soil pH. c Recalcitrant organic carbon content. d Percentage of recalcitrant organic carbon in soil organic carbon. e Fe-bound organic carbon content. f Percentage of Fe-bound organic carbon in soil organic carbon. g Ca-bound organic carbon content. h Percentage of Ca-bound organic carbon in soil organic carbon. Lowercase letters indicate differences in relevant soil properties among different soil layers at the p < 0.05 level (one-way ANOVA test). The contours of the violin plots delineate the distribution patterns of the data. Within each violin plot, the upper and lower extremities of the embedded box correspond to the 75th and 25th percentiles, respectively. The solid line and black dot within the box signify the median and mean of the data, respectively.

Full size image

The distribution of metal-bound OC along natural soil pH gradients

Fe-OC% (significantly positively correlated) and Ca-OC% (significantly negatively correlated) both exhibited strong linear relationships with soil pH (Fig. 4a). Across the three soil layers, we observed the regulatory effect of soil pH on metal-bound OC (Fig. 4). In other words, within three distinct soil pH ranges (pH ≤ 6.5, 6.5 < pH ≤ 7.5, pH > 7.5), the dominant metal-bound OC in the soil layers shifted from Fe-OC to Ca-OC with increasing soil pH, and the content of metal-bound OC within the same soil layer across different soil pH gradients was significantly different (p < 0.05) (Fig. 4b). The regulatory effect of soil pH on metal-bound OC exhibited a significant exponential increase (Fig. 4c). Specifically, the ratio of Ca-OC to Fe-OC remained relatively low at pH values below 7.5. As the pH exceeded 7.5, this ratio gradually increased, and it begun to rise sharply when the pH was greater than 8.5 (Fig. 4c).

Fig. 4: Variations of Fe-OC and Ca-OC along the soil pH gradient.
figure 4

a Linear relationship between metal-bound OC and soil pH. b Transitional patterns of Fe-OC% and Ca-OC% across different soil layers and pH intervals; c Relationship between the ratio of Ca-OC to Fe-OC and soil pH in different soil layers (fitted with an exponential function). Rectangles of different colors in b represent the average values of Fe-OC% or Ca-OC% in the respective soil layers and pH intervals, with whiskers indicating one standard deviation. Lowercase and uppercase letters show differences in Ca-OC% and Fe-OC% among different pH levels within the same soil layer at p < 0.05 (one-way ANOVA). Points of different colors in c denote various aridity sequences.

Full size image

Sensitivity analysis of ROC to metal-bound OC

The sensitivity of ROC to Fe-OC and Ca-OC exhibited asymmetrical observations in the vertical direction (Fig. 5). In soil layers above approximately 60 cm, ROC was more sensitive to variations in Fe-OC. In contrast, below 80 cm, ROC was more sensitive to variations in Ca-OC (Fig. 5a). Further statistical analysis indicated that Fe-OC, Ca-OC, and SIC contents have distinct vertical distribution characteristics (Fig. 5b). Along the vertical profile to a depth of 200 cm, Fe-OC content gradually decreased, reaching its minimum value at approximately 190 cm with a large data variability (from 4.11 to 0.34 g C kg−1). In contrast, the vertical variation in Ca-OC content was not pronounced (from 1.45 to 1.16 g C kg−1). SIC content exhibited a distinct distribution pattern, uniformly increasing from 60.17 g C kg−1 at 20 cm to 143.16 g C kg−1 at 100 cm, and then uniformly decreasing to 87.30 g C kg−1. Notably, both Ca-OC and SIC content reached their peak values within the soil profile at the 100 cm layer. We observed that the sensitivity of ROC to Fe-OC and Ca-OC was not consistent with their absolute contents but was closely related to their relative magnitudes (Fig. 5c). We identified a significant synergistic pattern between the relative content differences of Fe-OC and Ca-OC and the differential sensitivity of ROC to these two forms of metal-bound OC, which conform to a strict S-shaped mode. Specifically, when Ca-OC predominated in the soil layer (Fig. 5c, Step 1, corresponding to soil pH greater than 7.6), ROC was primarily governed by the prevalent Ca-OC. When Fe-OC content in the soil layer initially exceeded Ca-OC content (Fig. 5c, Step 2, corresponding to soil pH ranging from 7.2 to 7.6), the metal-bound OC that predominantly influenced ROC shifted rapidly to Fe-OC. Soil surface charge is low at pH 7.2–7.6 and higher at the extremes of this range, indicating a shift in soil control from Fe to Ca mineral phases within the pH range of Step 2 (Supplementary Fig. 6a). As Fe-OC content in the soil layer continued to increase (Fig. 5c, Step 3, corresponding to soil pH less than 7.2), the sensitivity of ROC to both forms of metal-bound OC remained stable, indicating that it no longer exhibited a persistently high sensitivity with the further increase of Fe-OC content. Figure 5c indicated that the sensitivity of ROC to the contents of Fe-OC and Ca-OC did not strictly adhere to the symmetrical assumption that higher content equated to higher sensitivity. Specifically, in stage 2 of Fig. 5c, when Fe-OC content in the soil layer slightly exceeded Ca-OC, Ca-OC remained the dominant factor influencing ROC changes in that soil layer (Fe-OCmean/Ca-OCmean > 1 and Fe-OCk/Ca-OCk < 1). However, this phenomenon was ephemeral and rapidly transitions to being dominated by Fe-OC. In the left part of stage 1 and the right part of stage 3 in Fig. 5c, when the contents of Fe-OC and Ca-OC were extremely high, the sensitivity of ROC to them ceased to change. This also demonstrated that the assumption of higher content equated to higher sensitivity did not always hold. Therefore, our results revealed multidimensional asymmetric phenomena, both in the vertical dimension and from the perspective of metal-bound OC contents.

Fig. 5: Sensitivity differences of ROC to Fe-OC and Ca-OC.
figure 5

a Distribution patterns of ROC sensitivity to Fe-OC (Fe-OCk) and Ca-OC (Ca-OCk) across the profile scale; b Distribution patterns of Fe-OC, Ca-OC, and SIC contents across the profile scale in different soil layers; c Relationship between the relative sensitivity of ROC to Fe-OC and Ca-OC and the content of Fe-OC and Ca-OC. Points of different colors and shapes in b represent the average values of Fe-OC, Ca-OC, and SIC in specific soil layers; the solid line in c represents the fitting results based on an S-shaped distribution regression (R2 = 0.97, p < 0.001). Backgrounds of different colors indicate the three stages of ROC regulation by Fe-OC and Ca-OC. The inset illustrates the relationship between the relative sensitivity of ROC to Fe-OC and Ca-OC and soil pH.

Full size image

The driving patterns of metal-bound OC on ROC and SOC content

The SEM resulted in good model fits for the 0–30 cm, 30–100 cm, and 100–200 cm soil layers (χ2/df < 5, p > 0.05, Fig. 6). According to the SEM, Fe-OC exerted a dominant direct effect on ROC across all three soil layers (with standardized path coefficients of 0.97, 0.91, and 0.46 for Fig. 6a–c, respectively). In contrast, Ca-OC has a significant direct effect on ROC only in the 100–200 cm soil layer (with a standardized path coefficient of 0.46 in Fig. 6c). Soil pH did not have a direct effect on ROC in the 30–100 cm and 100–200 cm layers but can exert a significant indirect effect on ROC by influencing Fe-OC and Ca-OC. ROC directly affected SOC in all soil layers (with standardized path coefficients of 0.83, 0.91, and 0.91 for Fig. 6a–c, respectively).

Fig. 6: Drivers of SOC and ROC contents across different soil layers.
figure 6

a–c represent the driving processes in the 0–30 cm, 30–100 cm, and 100–200 cm soil layers, respectively. Arrows of different colors indicate the varying effects of the driving pathways. Red arrows signify positive effects, while blue arrows denote negative effects. Dashed lines represent non-significant driving effects. The thickness of the arrows corresponds to the absolute magnitude of the driving effect. Numbers above the paths indicate the standardized path regression coefficients. *, **, and *** represent significant rejection of the null hypothesis at the 0.05, 0.01, and 0.001 levels, respectively. The explained variance (R2) for each response variable is displayed alongside the model.

Full size image

Discussion

Along the climatic gradient from northeast to southwest, aridity increases progressively, leading to a gradual decrease in the content of SOC, ROC, Fe-OC, and Ca-OC (Supplementary Fig. 1). Habitats with high SOC content typically exhibit higher ROC, Fe-OC, and Ca-OC contents. For instance, the higher SOM content commonly found in humid environments also promotes the organic complexation of Fe and the co-precipitation of organic-metal complexes38. Therefore, the increase in aridity contributes to the reduction of SOC through pathways such as decreased productivity and impacts on microbial activity39, which in turn leads to lower contents of ROC, Fe-OC, and Ca-OC. In contrast, Fe-OC% and Ca-OC% display opposing spatial patterns (Fig. 1). That is, as aridity increases, Fe-OC% decreases while Ca-OC% increases. This indicates that the relative importance of Fe and Ca for SOC stability varies with changes in aridity and acidity35. Under arid conditions, evaporation and soil pH increase, reducing the activity of iron oxides in the soil, thereby affecting the binding of iron oxides with organic carbon, which results in a decrease in Fe-OC%. In arid conditions with low precipitation and high evaporation, the Ca2+ produced by calcium carbonate persists or accumulates in arid and semi-arid regions19. In humid regions, due to the leaching effect of precipitation, the content of Ca2+ in the soil is low (e.g., exchangeable Ca2+ in YL soil is approximately 12–18 cmol kg−1), and as the climate gradient shifts towards arid regions, the content of Ca2+ in the soil gradually increases (e.g., exchangeable Ca2+ in BC soil is approximately 75–90 cmol kg−1) (Fig. 2n). The increase in Ca2+ content promotes the bridging of calcium with organic carbon19 (Supplementary Fig. 7), hence, as aridity increases, although the content of Ca-OC decreases, Ca-OC% increases progressively.

Variations in soil depth are accompanied by soil physicochemical properties change. For instance, soil pH gradually increases with increasing soil depth (Fig. 3). Soil pH can reflect soil moisture conditions, and determines numerous geochemical gradients, including the dissolved metals types, minerals and the charge of organic molecules, and the forms of interaction between minerals and organic carbon to a certain extent33,40,41,42. Our results indicate that soil pH is a key factor in regulating the dynamic distribution of metal-bound OC within the soil profile (Figs. 4 and 6). With the increase in soil pH (and soil layer depth), the dominant form of metal-bound OC shifts from Fe-OC to Ca-OC (Fig. 4b). Consequently, the ratio of Ca-OC to Fe-OC increases exponentially with soil pH (Fig. 4c). Different pH environments are associated with distinct charge environments, leading to the binding of organic carbon with various metal oxides. A decrease in soil pH increases Fe3+ and Al3+ content in the soil. Thus, lower soil pH enhances the adsorption of SOC by iron oxides, subsequently promoting the formation of Fe-OC. That is, when soils are acidic (pH <6.5), the interactions between Fe3+, Al3+, and organic carbon dominate, while the role of Ca2+ is diminished19,43. When the pH is between 6.5 and 7.5, the ratio of Ca-OC to Fe-OC begins to rise gradually. At this pH level, the surface charge of most Fe (III) (hydr) oxides is slightly negative or close to zero, thus limiting the adsorption and complexation of SOC by Fe (Supplementary Fig. 6b, the point of zero charge within the soil pH range of 6.5–7.5 is 6.97–7.64). When soils are alkaline (pH > 7.5), the increasing pH enhances the aging and crystallization of Fe (III) (hydr) oxides44, a process that reduces the reactivity of Fe (III) (hydr) oxides, thereby weakening their interaction with organic carbon45. Concurrently, Ca²⁺ begins to link negatively charged organic groups (e.g., carboxyl groups of SOC) through cation bridging43, leading to increase rapidly in the Ca-OC to Fe-OC ratio. In the 100–200 cm soil layer, some samples with soil pH greater than 9 maintain a lower ratio (Fig. 3b), which may be due to Ca2+ combining with carbonate ions to form calcium carbonate precipitates. This reduces the availability of Ca2+ to a certain extent, causing fluctuations in the ratio, but this does not necessarily indicate a weakening of the interaction between Ca and organic carbon. In alkaline soil environments, Ca exerts a stronger influence on SOC stability than other divalent cations. This is attributed to its higher relative abundance and larger ionic radius, which enable the formation of stronger binding energies46,47.

Fe-OC dominates the ROC in the upper soil layers (approximately <80 cm) (Fig. 5a). Iron oxides can bind with organic carbon through adsorption (ligand exchange and cation bridging) and co-precipitation (forming occluded organic carbon), thereby enhancing the stabilization of SOC10. In contrast, Ca-OC dominates the ROC in deeper soil layers (approximately >80 cm) (Fig. 5a). Ca2+ can directly form cation bridges with carboxylic acids and to a lesser extent with phenols or other -OH functional groups, thus enhancing the stability of SOC48. It is noteworthy that a special interval exists during stage 2, where despite Fe-OC content being greater than Ca-OC, the metal-bound OC that primarily influences ROC has shifted from Fe-OC to Ca-OC (Fig. 5c). This indicates that the metal-bound OC dominating SOC stability does not strictly follow the pattern of greater content equating to stronger effect. This result is inconsistent with our hypothesis that the dominance of Fe-OC and Ca-OC on SOC stability is symmetrical with their relative content. This may be related to the forms in which iron oxides combine with organic carbon at different levels. The adsorption of organic carbon by iron oxides has a saturation point, with the maximum carbon-to-iron ratio (OC:Fe ratio) that can be stabilized being 110. The OC:Fe ratio in co-precipitation with iron oxides can reach 10, which can protect more SOC than adsorption alone49,50. The average OC:Fe ratio in all soil layers below 80 cm in this study is less than 1 (Supplementary Fig. 8a), suggesting that iron oxides in deeper soils primarily bind with organic carbon through adsorption, leading to a weakened effect of Fe-OC on ROC (Figs. 5a and 6c). Additionally, Ca2+ may preferentially adsorb more stable organic compounds in Ca-rich environments. For example, certain active iron mineral surfaces may participate in the adsorption of specific categories of organic matter, such as proteins, lignin, and phenolic compounds51,52. However, in Ca-rich environments, there is scant research attention given to the potential preferential adsorption of organic compounds by Ca, and further investigation into this mechanism is warranted in future studies.

Climate variability leads to differences in precipitation and evapotranspiration, influencing the movement of calcium carbonate (CaCO3) as bicarbonate with percolating water. When evapotranspiration exceeds precipitation, caliche layer forms within the soil profile. CaCO3 is a significant acid-buffering substance in soils; in soils with pH > 7.5, the dissolution of carbonate or bicarbonate minerals is the primary mechanism of soil acid buffering. The buffering capacity within this interval is determined by the content of soil carbonates, primarily CaCO3, which buffers soil pH through its hydrolysis53. This study demonstrates that within the pH range of 7.2 to 7.6, the dominant metal-bound OC influencing ROC shifts rapidly from Fe-OC to Ca-OC (Fig. 5c), which coincides with the buffering process of CaCO3 hydrolysis releasing a large amount of Ca2+ and HCO3−. This suggests that CaCO3 is crucial for the formation of Ca-OC, likely due to its role as a rich source of calcium in soils, providing more Ca2+ to bind with organic carbon. Moreover, during pedogenesis, CaCO3 may undergo dissolution and re-precipitation, thereby stabilizing SOC through co-precipitation or encapsulation with CaCO354,55. Deeper soil layers, constrained by energy limitations such as oxygen and lower microbial activity, and less affected by new carbon inputs, are more stable than surface layers56 (Fig. 2). Furthermore, our study indicates a highly significant positive correlation between CaCO3 and ROC% (Supplementary Fig. 5), implying that CaCO3 is one of the key factors contributing to the higher stability of deep soil layers. Six profiles containing CaCO3 do not entirely conform to the pattern of increasing SOC stability with depth; instead, they exhibit higher SOC stability at the caliche layer (Fig. 2). This may be attributed to CaCO3 acting as a glue that immobilizes SOC. On one hand, CaCO3 can directly adsorb various organic compounds, including carboxylic acids, alcohols, and amino acids57. On the other hand, secondary CaCO3 crystals can reduce the porosity of soil micro-aggregates, thereby decreasing the accessibility of SOC within the micro-aggregates to decomposers, preserving SOC from decomposition58.

Thus, we propose the concept of the dynamic changes in SOC stability within the soil profile along the climatic gradient (Fig. 7): In humid areas (Aridity < 0.5) with high rainfall and high vegetation cover, the soil surface exhibits an acidic environment (pH <6.5). Iron oxides form Fe-OC predominantly through co-precipitation, which dominates the stability of SOC. In such acidic surfaces, although Ca2+ is leached to deeper layers by precipitation, local effects of calcium may still occur due to biological cycling19. Soil pH gradually increases with depth. When pH > 6.5, Ca2+ within the soil profile begins to increase, and the contribution of Ca-OC formed by calcium bridging with organic carbon to SOC stability begins to gradually appear. Under conditions of high precipitation and low evapotranspiration, caliche layer forms in deeper soils (> 100 cm), and soil pH further increases to > 7.5. The dominant metal-bound OC influencing SOC stability shifts rapidly from Fe-OC to Ca-OC. In semi-humid to semi-arid areas (0.5 < Aridity < 0.7), the surface soil pH is relatively higher compared to humid regions. The content of Ca2+ in the surface soil has increased relative to humid areas, but Fe-OC remains the dominant metal-bound OC for surface SOC stability. In these areas, as evapotranspiration increases, the layer of CaCO3 deposition moves upward, and the layer which Ca-OC dominates SOC stability also moves upward. In arid areas (Aridity > 0.7), due to low rainfall and high evapotranspiration, iron oxides and organic carbon primarily bind through adsorption (Supplementary Fig. 8c). The soil is in a Ca-rich environment, and caliche layer appears in the upper layers (<100 cm) of the soil. At this point, Ca-OC may dominate the SOC stability throughout the soil profile.

Fig. 7: A schematic diagram of dynamic changes of Fe-OC and Ca-OC dominating ROC within the soil profile at regional scale.
figure 7

The three schematic soil profiles from right to left correspond to increasing aridity and soil pH. With the increase in aridity, which is accompanied by a gradual decrease in precipitation and a progressive thinning of surface cover, the depth of calcium carbonate accumulation gradually shifts upward. The response of ROC to Ca-OC and Fe-OC exhibited variations both horizontally and vertically across different soil layers.

Full size image

Our study demonstrates the important role of Ca-OC and Ca2+ in maintaining SOC stability, which is often neglected. CaCO3 forms through chemical weathering involving water, CO2, and Ca2+. The bicarbonate ions (HCO3−) produced by the dissolution of CO2 in water react with Ca2+ and precipitate as CaCO3, also releasing water and CO2 in the process59. Under conditions of high soil water flux or acids, the dissolution of CaCO3 can promote the leaching of Ca2+, thus combining with SOC to form Ca-OC to enhance SOC stability. This study proposes that carbonate leaching switches the role of metals (Fe and Ca) in stabilizing SOC. This process modulates the relative importance of Fe-OC and Ca-OC in the sequestration of SOC. The widespread occurrence of carbonate leaching suggests that this is not an isolated regional phenomenon but rather a ubiquitous process that may exist in ecosystems.

Methods

Study region and soil sampling

We selected a semi-humid to semi-arid climate transect spanning 1500 km in northern China (38–53°N; 120–135°E; Supplementary Fig. 9) as the study site, transitioning from temperate continental monsoon climate to subarctic climate. The annual mean precipitation, ranging from 350 to 700 mm, decreases spatially from southeast to northwest, forming a natural aridity gradient that supports the research questions. Phaeozems and Chernozems are the predominant soil types in this region44. In July and October 2023, we carried out extensive field surveys and collected 16 natural profiles along the climatic gradient (profile depth of 60–200 cm) (Supplementary Fig. 10 and Supplementary Table 1). Soil samples were collected at 10 cm intervals from each soil layer, resulting in a total of 252 soil samples included in this study. In the statistical analysis, to clearly present our results while preserving the originality of the data, we consolidated all soil layers within the top 30 cm of 16 soil profiles into 0–30 cm. Specifically, soil samples from the 0–10 cm, 10–20 cm, and 20–30 cm layers of the 16 profiles were grouped into the 0–30 cm, resulting in a total of 48 soil samples. Similarly, we categorized the soil layers into 30–100 cm (n = 107) and 100–200 cm (n = 97). The Aridity index (AI) is derived from the Global Aridity Index and Potential Evapotranspiration Climate database (https://cgiarcsi.community/)60. We calculated the Aridity as 1-AI, so a higher aridity value indicates drier climate conditions39. Mean annual temperature (MAT) and mean annual precipitation (MAP) were sourced from the WorldClim 2.0 database61.

Laboratory analysis

Soil organic carbon (SOC), soil inorganic carbon (SIC), soil exchangeable Ca2+ (Caex2+), soil pH were measured according to the methods by Zhang and Gong62. The soil surface charge was determined by potentiometric titration (Supplementary Note 1)63,64,65. Dithionite-extractable iron (Fed) and dithionite-extractable aluminum (Ald) were analyzed by dithionite-citrate-bicarbonate (DCB) method (Supplementary Note 2). Another equal soil sample was extracted with NaCl as a control. Organic carbon associated with Fe mineral phases was calculated as the organic carbon difference between the control and DCB-treated soil residues12. The adsorption (<1.0) and complexation (> 6) between organic carbon and active Fe were distinguished by calculating the molar ratio of OC:Fed. Using 0.5 M Na2SO4 to extract organic carbon associated with Ca bridges (Supplementary Note 3)66,67,68. We used acid hydrolysis method to evaluate the chemical stability of SOC (Supplementary Note 4)23. Soil samples were hydrolyzed with both 2.5 M and 13 M H2SO4 at 105 °C, followed by centrifugation. The labile organic carbon was hydrolyzed by acid, and the recalcitrant organic carbon was not hydrolyzed.

Sensitivity analysis

We established a framework for measuring the sensitivity between factors to identify the differential sensitivity of ROC to different metal-bound OC (Fe-OC and Ca-OC). Initially, we used linear regression equations to quantify the relationship between ROC and metal-bound OC within fixed soil layers, with the slope of the linear regression serving as a measure of the sensitivity of ROC to a specific metal-bound OC in that layer. Following hierarchical regression, we obtained the sensitivity sequence of ROC to Fe-OC and Ca-OC (Fe-OCk and Ca-OCk) within the 200 cm soil profile. The correlation model parameters between different soil layer regression factor groups were documented in the supplementary materials (Supplementary Table 2). Subsequently, we calculated the ratio of the average contents of Fe-OC to Ca-OC as an indicator of the relative content of metal-bound OC in fixed soil layers (Fe-OCmean/Ca-OCmean). The ratio of Fe-OCk to Ca-OCk, indicating the relative difference in sensitivity, was used to delineate the dominant metal-bound OC component that stabilizes ROC in a given soil layer (Fe-OCk/Ca-OCk). The dominant metal-bound OC acting on the vertical scale was defined as a function of Fe-OCk/Ca-OCk relative to Fe-OCmean/Ca-OCmean. This approach allows us to systematically evaluate the influence of metal-bound OC on the stability of SOC within the soil profile.

Structural equation model

The piecewise structural equation model was used to further elucidate the direct and indirect effects of metal-bound OC on SOC content and its stability. In the R 4.4.1 environment, the “piecewiseSEM” package was used to fit the functional relationships from different variables. p value (Chi-square) > 0.05, normed chi-square (χ2/df) <5, comparative fit index (CFI) > 0.9, root mean square error of approximation (RMSEA) <0.08, standardized root mean square residual (SRMR) <0.08 indicated that the final model had an acceptable fit with the data69.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.