Distributions and Inﬂuencing Factors of Soil Organic Carbon Fractions under Different Vegetation Restoration Conditions in a Subtropical Mountainous Area, SW China

: Vegetation type is known to affect soil organic carbon (SOC) storage. However, the magnitudes and distributions of SOC sequestration and driving factors for different vegetation types are still largely unknown. Thus, we studied the changes in SOC fractions along soil proﬁles for different vegetation restoration types and their relationships with soil properties. We selected ﬁve vegetation types and collected soil samples from depth intervals of 0–10, 10–30, 30–60, and 60–90 cm. Five soil carbon fractions and the soil properties were tested to evaluate the soil carbon fraction distributions and inﬂuencing factors. Our results demonstrated that the concentrations of total organic carbon (TOC) and ﬁve carbon fractions were strongly affected by vegetation types and soil depths. The concentrations of all ﬁve soil carbon fractions in 0–10 cm depth were higher than those in the other three soil depths and generally increased with vegetation complexity. The Pearson correlations and redundancy analysis showed that the fractions of soil glomalin-related soil protein (GRSP) and Fe oxides as well as the soil bulk densities, were the most signiﬁcant related to soil TOC levels and carbon fractions, which suggests that soil biochemical and physicochemical processes are among the most important mechanisms that contribute to SOC persistence. Considering the sensitive indices of the soil carbon variables and PCA results, soil permanganate oxidizable carbon (POXC) was considered to be the most sensitive index for differentiating the effects of vegetation types. These results provide important information regarding the distributions and driving factors of the carbon fractions that result from different vegetation restoration types and will help to improve our understanding of soil carbon sequestration during vegetation restoration processes. vegetation types on soil organic carbon, carbon fractions, and the sensitivity indices of the soil carbon indices.


Introduction
Soil stores the largest terrestrial organic carbon pool on Earth and it is a major source and sink of atmospheric CO 2 in mitigating the greenhouse effect [1]. Soil organic carbon (SOC) is a key indicator for evaluating soil quality because of its physicochemical properties, biological processes, and ecological functions [2][3][4]. Ecological restoration is widely considered as a way to improve SOC sequestration and ecosystem services for different land-cover types [5]. Through the differences in carbon inputs from biota and losses through decomposition, different types of vegetation restoration have the potential to influence soil carbon storage and carbon dynamics, as reported in previous research [6,7]. the Yunnan Province of China ( Figure 1). The area has a subtropical monsoon climate with an average annual rainfall of 846 mm and a mean annual temperature of 16 • C. The soil in this area is a Cambisol. Most of original natural vegetation was cleared for fuelwood and pastures before the 1960s. Since the 1980s, different vegetation types, mainly including shrub-grass land (SG), coniferous Pinus forest (PF), coniferous and broad-leaved mixed forest (MF), and natural secondary forest (NSF), have been used to recover degraded land. Meanwhile, the undisturbed and mature natural forest (NF) in the Huafoshan Nature Reserve about 6 km apart from the study area was selected as the reference area. The study area and reference area belonged to the same climatic and soil zones. Detailed information about the restoration history and community characteristics of these four ecological measures is provided by Fu et al. [24,25] (different plant community abbreviations) and in Appendix A (Table A1). The study area is located at the Samachang ecological observation station (25°24′09″N; 101°28′18″E) in Muding County mid-Yunnan, about 200 km away from Kunming City in the Yunnan Province of China ( Figure 1). The area has a subtropical monsoon climate with an average annual rainfall of 846 mm and a mean annual temperature of 16°C. The soil in this area is a Cambisol. Most of original natural vegetation was cleared for fuelwood and pastures before the 1960s. Since the 1980s, different vegetation types, mainly including shrub-grass land (SG), coniferous Pinus forest (PF), coniferous and broad-leaved mixed forest (MF), and natural secondary forest (NSF), have been used to recover degraded land. Meanwhile, the undisturbed and mature natural forest (NF) in the Huafoshan Nature Reserve about 6 km apart from the study area was selected as the reference area. The study area and reference area belonged to the same climatic and soil zones. Detailed information about the restoration history and community characteristics of these four ecological measures is provided by Fu et al. [24,25] (different plant community abbreviations) and in Appendix (Table A1).

Soil Sampling
In 2019, 15 comparative study plots were established in three representative sites with SG, PF, MF, and NSF to assess soil C distribution. At least 3 km distance between sites was selected to reduce spatial autocorrelation. Four representative 20 × 20 m plot for every vegetation restoration type were established within each site. In addition, three representative plots of NF in the nearby Huafoshan Nature Reserve were selected as references ( Figure 1). Eight to ten soil cores from four soil depth levels (e.g., 0-10, 10-30, 30-60, and 60-90 cm) were obtained at random in each plot. These samples were sieved (2mm mesh) and air-dried to analyze soil properties and soil C fractions. Additionally, soil bulk density (BD) and soil water content (SWC) were measured gravimetrically with an additional three undisturbed soil samples.

Soil Sampling
In 2019, 15 comparative study plots were established in three representative sites with SG, PF, MF, and NSF to assess soil C distribution. At least 3 km distance between sites was selected to reduce spatial autocorrelation. Four representative 20 × 20 m plot for every vegetation restoration type were established within each site. In addition, three representative plots of NF in the nearby Huafoshan Nature Reserve were selected as references ( Figure 1). Eight to ten soil cores from four soil depth levels (e.g., 0-10, 10-30, 30-60, and 60-90 cm) were obtained at random in each plot. These samples were sieved (2-mm mesh) and air-dried to analyze soil properties and soil C fractions. Additionally, soil bulk density (BD) and soil water content (SWC) were measured gravimetrically with an additional three undisturbed soil samples.

Laboratory Analysis
We measured the soil pH using a soil to water mixture (1:2.5, w:v) with a pH meter and soil clay contents by the Bouyoucos hydrometer method. Soil total nitrogen (TN) was de- termined using the semimicro Kjeldahl method, and total phosphorus (TP) was examined by phosphomolybdate blue methods [26]. The soil free Fe (Fed), poorly crystalline Fe (Feo), and organically complexed Fe (Fep) were extracted using dithionite-citrate-bicarbonate (DCB), ammonium oxalate, and sodium pyrophosphate at pH 10, respectively. The extracted Fe was measured by using atomic absorption spectroscopy [26].
The SOC concentrations were determined with a TOC analyzer. The DOC was extracted with a 1:5 ratio of soil to 0.5 M K 2 SO 4 solution and shaken [27]. The solution was determined with a TOC analyzer after filtration (0.45 µm membrane). The LFOC (<1.7 g cm −3 ) and HFOC (>1.7 g cm −3 ) fractions were determined according to the density fractionation method described by Gregorich and Ellert [26]. The POXC fraction was determined by adding 0.333 M KMnO 4 for oxidation. The suspension was then centrifuged, diluted, and measured spectrophotometrically at 565 nm [28]. The POC fraction was separated using the wet sieving approach. In brief, 20 g air-dried soil with 5 g L −1 (NaPO 3 ) 6 was dispersed for 20 h. The soil suspension was passed through a 53-µm sieve. The fraction retained on the sieve was oven-dried and analyzed with a TOC analyzer.
The soil biochemical properties, including GRSP and EPS, were also determined. The GRSP fractionation was determined using the method described by Singh et al. [29]. Briefly, easily extractable GRSP (EE-GRSP) was extracted from 1.0 g of soil incubated with 8 mL of 20 mM citrate (pH 7.0) by 30 min autoclaving. The T-GRSP was obtained from 1.0 g of soil with 8 mL of 50 mM citric acid (pH 8.0) by six successive autoclaving cycles. The GRSP contents in the extracts were assayed using a Bradford assay [29]. The difficultly extractable GRSP (DE-GRSP) was the difference between T-GRSP and EE-GRSP. The EPS levels were determined on the supernatant by the anthrone method [30].

Calculations of Soil Carbon Indices
The SOC stock was calculated as follows [6]: where BDi is bulk density (g cm −3 ) at soil layer i, SOCi is the soil organic carbon content (g kg −1 ) at soil layer i, and Di is the soil depth (cm). The sensitivity index (SI) was defined as the reductions in soil carbon fractions after different vegetation restorations and was calculated as follows [31]: SI = (carbon variables of restoration vegetation − carbon variables of reference natural forest)/carbon fractions of reference natural forest (2) The carbon management index (CMI), carbon pool index (CPI), and soil carbon sequestration capacity (SCS), which are the systematic, sensitive indicators used to assess SOC changes, were also calculated as follows [31][32][33]: SCS = CPI/C lability (5) where LI (lability index) = C lability of restoration vegetation/C lability of natural forest, C lability = POXC/Non-POXC, and non-POXC = TOC-POXC. Meanwhile, the SIs of these three carbon indices were also calculated using Equation (2).

Statistical Analyses
Normality and homogeneity tests were applied to all variables. Two-way ANOVA was used to compare the effects of vegetation restoration type and soil depth on the concentrations of the basic soil properties and soil carbon variables. The significances of the differences in the tested parameters among the vegetation restoration types were compared using LSD (Least Significant Difference) at the 0.05 level.
Pearson correlations and redundancy analysis (RDA) were applied to elucidate the relationships among the soil carbon variables and soil characteristics among all soil samples from different vegetation restoration types. To summarize the total variance of the data, we performed a principal component analysis (PCA) that included all soil carbon fractions and carbon indices for the combined vegetation types and soil depths. The above analyses were conducted using SPSS 20 (SPSS, Chicago, USA) and CANOCO 5.0.

Basic Soil Characteristics
Two-way ANOVA showed that the vegetation restoration type had a strong effect on all basic soil properties except for the SWC, and soil depth had a significant effect only on the BD and Fep (p < 0.05, Table A2). Throughout the soil profile (0-90 cm), the soil BD and pH values decreased, and the concentrations of soil TN, Fep, Feo, and soil clay contents increased with the vegetation complexity ( Figure 2). The highest values of BD, TP, and pH were found in soils under SG. The TN, Fep, and Feo concentrations were obviously higher in soils under NF than those under the other four vegetation restoration types ( Figure 2 and Table A2).
Forests 2022, 13, x FOR PEER REVIEW 5 of 16 differences in the tested parameters among the vegetation restoration types were compared using LSD (Least Significant Difference) at the 0.05 level. Pearson correlations and redundancy analysis (RDA) were applied to elucidate the relationships among the soil carbon variables and soil characteristics among all soil samples from different vegetation restoration types. To summarize the total variance of the data, we performed a principal component analysis (PCA) that included all soil carbon fractions and carbon indices for the combined vegetation types and soil depths. The above analyses were conducted using SPSS 20 (SPSS, Chicago, USA) and CANOCO 5.0.

Basic Soil Characteristics
Two-way ANOVA showed that the vegetation restoration type had a strong effect on all basic soil properties except for the SWC, and soil depth had a significant effect only on the BD and Fep (p < 0.05, Table A2). Throughout the soil profile (0-90 cm), the soil BD and pH values decreased, and the concentrations of soil TN, Fep, Feo, and soil clay contents increased with the vegetation complexity ( Figure 2). The highest values of BD, TP, and pH were found in soils under SG. The TN, Fep, and Feo concentrations were obviously higher in soils under NF than those under the other four vegetation restoration types ( Figure 2 and Table A2). Vegetation types and soil depths had significant effects on the distributions of soil GRSP and EPS (p < 0.05), except vegetation types had no effect on EPS (p > 0.05) (Table Vegetation types and soil depths had significant effects on the distributions of soil GRSP and EPS (p < 0.05), except vegetation types had no effect on EPS (p > 0.05) (Table A2).

Distributions of Soil Carbon Variables
The concentrations of TOC and those of all five carbon fractions were strongly affected by the community types and soil depths ( Table 1). The TOC, POXC, LFOC, HFOC, and POC concentrations in the surface layer (0-10 cm) were higher than those in the other three soil layers and generally increased with the vegetation complexity ( Figure 4). The DOC concentrations in the surface layer exhibited more significant difference than those in the other soil layers and those of the other carbon indices among the five vegetation types (Figure 4). The CPI and CMI values in the surface layer (0-10 cm) were lower than other vegetation types. The CPI and SCS values exhibited increasing trends with soil depth ( Figure 4, Table A3).

Distributions of Soil Carbon Variables
The concentrations of TOC and those of all five carbon fractions were strongly affected by the community types and soil depths ( Table 1). The TOC, POXC, LFOC, HFOC, and POC concentrations in the surface layer (0-10 cm) were higher than those in the other three soil layers and generally increased with the vegetation complexity ( Figure 4). The DOC concentrations in the surface layer exhibited more significant difference than those in the other soil layers and those of the other carbon indices among the five vegetation types (Figure 4). The CPI and CMI values in the surface layer (0-10 cm) were lower than other vegetation types. The CPI and SCS values exhibited increasing trends with soil depth (Figure 4, Table A3).
The SOC stocks and carbon fractions in different soil layers or throughout the soil profile under NF were higher than those under the other vegetation restoration types ( Figure 5 and Table A3). The SOC stocks did not display significant differences among SG, YF, MF, and NSF, although the SOC stocks exhibited an increasing trend with vegetation complexity ( Figure 5). The SOC stocks and carbon fractions in different soil layers or throughout the soil profile under NF were higher than those under the other vegetation restoration types (Figure 5 and Table A3). The SOC stocks did not display significant differences among SG, YF, MF, and NSF, although the SOC stocks exhibited an increasing trend with vegetation complexity ( Figure 5).

Multivariate Analysis between SOCs and Soil Properties
The Pearson correlations showed that the soil TOCs and all five carbon fractions were significantly negatively correlated with the BD and pH, and positively related to EE-GRSP, DE-GRSP, T-GRSP, EPS, TN, Fep, and Feo ( Figure 6). The RDA results showed that the first axis explained 62.55% of the variances in the carbon fractions. The soil GRSP, Fe oxides, and BD were the most driving factors that were significantly related to the soil TOCs and carbon fractions (Figure 7).

Multivariate Analysis between SOCs and Soil Properties
The Pearson correlations showed that the soil TOCs and all five carbon fractions were significantly negatively correlated with the BD and pH, and positively related to EE-GRSP, DE-GRSP, T-GRSP, EPS, TN, Fep, and Feo ( Figure 6). The RDA results showed that the first axis explained 62.55% of the variances in the carbon fractions. The soil GRSP, Fe oxides, and BD were the most driving factors that were significantly related to the soil TOCs and carbon fractions (Figure 7).
The SOC stocks and carbon fractions in different soil layers or throughout the soil profile under NF were higher than those under the other vegetation restoration types (Figure 5 and Table A3). The SOC stocks did not display significant differences among SG, YF, MF, and NSF, although the SOC stocks exhibited an increasing trend with vegetation complexity ( Figure 5).

Multivariate Analysis between SOCs and Soil Properties
The Pearson correlations showed that the soil TOCs and all five carbon fractions were significantly negatively correlated with the BD and pH, and positively related to EE-GRSP, DE-GRSP, T-GRSP, EPS, TN, Fep, and Feo ( Figure 6). The RDA results showed that the first axis explained 62.55% of the variances in the carbon fractions. The soil GRSP, Fe oxides, and BD were the most driving factors that were significantly related to the soil TOCs and carbon fractions (Figure 7).

Sensitivity of Soil Carbon Variables to Vegetation Restoration Types
The two-way ANOVA results showed that the community type had a significant effect on the SI values for the soil carbon fractions and CPI, CMI, SCS. Soil depth had a significant effect only on the SI value for soil POC ( Table 1). The CMI and CPI values under NSF were highest in all the five vegetation types. The average SI values for the soil carbon indices along the soil profile (0-90 cm) indicated that the SI values for CMI were highest and were followed by the CPI, SCS, and POC, while the DOC, POXC, LFOC, HFOC, and TOC were less sensitive to changes in community types (Figure 8).

Sensitivity of Soil Carbon Variables to Vegetation Restoration Types
The two-way ANOVA results showed that the community type had a significant effect on the SI values for the soil carbon fractions and CPI, CMI, SCS. Soil depth had a significant effect only on the SI value for soil POC ( Table 1). The CMI and CPI values under NSF were highest in all the five vegetation types. The average SI values for the soil carbon indices along the soil profile (0-90 cm) indicated that the SI values for CMI were highest and were followed by the CPI, SCS, and POC, while the DOC, POXC, LFOC, HFOC, and TOC were less sensitive to changes in community types (Figure 8).

Sensitivity of Soil Carbon Variables to Vegetation Restoration Types
The two-way ANOVA results showed that the community type had a significant effect on the SI values for the soil carbon fractions and CPI, CMI, SCS. Soil depth had a significant effect only on the SI value for soil POC ( Table 1). The CMI and CPI values under NSF were highest in all the five vegetation types. The average SI values for the soil carbon indices along the soil profile (0-90 cm) indicated that the SI values for CMI were highest and were followed by the CPI, SCS, and POC, while the DOC, POXC, LFOC, HFOC, and TOC were less sensitive to changes in community types (Figure 8).

Effects of Different Vegetation Types on the Soil Organic Carbon Variables
Our results showed that different vegetation restoration types affected the soil carbon contents and exhibited an increasing pattern with increased vegetation complexity, which was corroborated by previous studies [13]. Generally, the SOC stocks are determined by the balance between plant litter/root inputs and outputs or decomposition. First, the inputs of litter and roots in the relatively complex communities (e.g., MF, NSF, and NF) were higher than those in the SG and YF communities [24,34], so less organic matter was returned to the soil in the SG and YF communities. Second, as the early stages of community succession, the rapid growth of plants in SG and YF communities greatly consumed soil nutrients by roots. In contrast, the higher accumulated biomass levels in the relatively complex communities increased the organic matter levels through litter and roots. Meanwhile, with the improvements in soil structure and functions, the losses of soil carbon and nutrients due to soil erosion and runoff were reduced [24]. Our study also revealed that the concentrations of TOC and most carbon fractions decreased along the soil profile. This possibly occurred because the TOC and its fractions were mainly affected by root residues and secretion at deeper depths and less by litter, which led to carbon decrease along the soil profile for all the different vegetation restoration types [35,36].
Significant positive correlations were found between the TOC and soil carbon fractions (p < 0.001), which suggested that the conversion from SOC to carbon sequestration in the forms of labile and non-labile carbon fractions can increase synchronously with the vegetation complexity. DOC is a key component of soil carbon cycling and plays a vital role in soil processes and functions [11]. The DOC contents in the surface layer exhibited more significant differences, which indicated that DOC was more sensitive to vegetation change than the other carbon fractions and soil depths. The POXC and LFOC fractions have been used as measures of labile carbon under different land cover or vegetation types [28,37]. For example, significant correlations have been reported in many studies. de

Effects of Different Vegetation Types on the Soil Organic Carbon Variables
Our results showed that different vegetation restoration types affected the soil carbon contents and exhibited an increasing pattern with increased vegetation complexity, which was corroborated by previous studies [13]. Generally, the SOC stocks are determined by the balance between plant litter/root inputs and outputs or decomposition. First, the inputs of litter and roots in the relatively complex communities (e.g., MF, NSF, and NF) were higher than those in the SG and YF communities [24,34], so less organic matter was returned to the soil in the SG and YF communities. Second, as the early stages of community succession, the rapid growth of plants in SG and YF communities greatly consumed soil nutrients by roots. In contrast, the higher accumulated biomass levels in the relatively complex communities increased the organic matter levels through litter and roots. Meanwhile, with the improvements in soil structure and functions, the losses of soil carbon and nutrients due to soil erosion and runoff were reduced [24]. Our study also revealed that the concentrations of TOC and most carbon fractions decreased along the soil profile. This possibly occurred because the TOC and its fractions were mainly affected by root residues and secretion at deeper depths and less by litter, which led to carbon decrease along the soil profile for all the different vegetation restoration types [35,36].
Significant positive correlations were found between the TOC and soil carbon fractions (p < 0.001), which suggested that the conversion from SOC to carbon sequestration in the forms of labile and non-labile carbon fractions can increase synchronously with the vegetation complexity. DOC is a key component of soil carbon cycling and plays a vital role in soil processes and functions [11]. The DOC contents in the surface layer exhibited more significant differences, which indicated that DOC was more sensitive to vegetation change than the other carbon fractions and soil depths. The POXC and LFOC fractions have been used as measures of labile carbon under different land cover or vegetation types [28,37]. For example, significant correlations have been reported in many studies.
de Moraes Sá et al. [13] and Sheng et al. [32] found a positive relationship between SOC and POXC and LFOC during ecological succession.

Influencing Factors of Soil Carbon Distributions
The present study showed that the GRSP fractions were the most influencing factors that were significantly correlated with soil TOC and carbon fractions ( Figure 6) and validated the contribution of biochemical properties to increase the carbon pools composed of TOC and carbon fractions. The significant correlation between SOC and GRSP has been reported in several previous studies [17,19,20,38]. By relating the GRSP contents to the SOC contents, the proportions of T-GRSP/SOM were 41%, 40%, 39%, 32%, and 28% for SG, YF, MF, NSF, and NF soils, respectively, which indicate that GRSP contributes highly to the SOC in soils. In addition, high GRSP contents can be found in relatively simple plant communities that are susceptible to changes in abiotic stress. Higher glomalin levels under environmental stresses are produced by arbuscular mycorrhizal fungi [39,40]. On the other hand, GRSP can protect SOC from decomposition by enhancing the formation of soil aggregates and partially sealing the soil pore system to slow down the penetration of water into the aggregates [19,41]. Overall, our results indicated that T-GRSP and its two fractions might be the reliable indicators under vegetation restoration. Similar to GRSP, EPS plays an important role in improving the production of soil aggregates and promoting soil carbon stability [21]. In our study, we found that soil GRSP contributes more to SOC than to EPS. Therefore, soil GRSP are generally considered to be a sensitive indicator of long-term C storage [19,20,42].
In line with previous works that were studied in other tropical and subtropical sites [19,43], our results also showed that Fe and BD have significant correlations with TOC and carbon fractions (Figures 6 and 7). This result indicates an interdependency between SOC, GRSP, and some soil physicochemical properties. The significantly negative relationships between BD and SOC and GRSP indicated that SOC and GRSP increase with decreasing soil structure. In addition, compared to the soil Fed concentrations, significant correlations between the Fep and Feo concentrations and the carbon fractions were also found in our study, which implies that different vegetation restoration types may change the associations of SOC with Fe oxides and eventually influence the SOC contents. Iron oxides in soil play an important role in soil C stability because they are involved in the physical, chemical, and biological protection mechanisms of SOC [44][45][46]. Different forms of Fe oxides have different roles in the process of SOC stability by improving the formation of soil aggregates. Feo and Fep have extensive surface areas and strong binding capacities to stabilize SOC [47]. Compared to Fed, Feo and Fep contributed more to SOC stability because their effects on the formation of soil aggregates were significantly greater than that of Fe d [10].

Indices for Assessing the Vegetation Restoration Effect on the Soil Carbon Pool
The PCA results indicate that all soil carbon fractions and carbon indices had the potential to be used as indicators to assess the vegetation restoration effects on the soil carbon pool (Figure 9). The PCA showed that the first PC mainly distinguished the different restoration vegetation types, and the second PC resulted in the differentiation among soil depths. Considering the SI values and correlations between principal components and carbon variables (Figures 8 and 9), it can be inferred that POXC, POC, and HFOC were the most sensitive indices for differentiating vegetation types, whereas the SCS and CPI were more sensitive for differentiating the effects of soil depth. The POC is more sensitive to both rapid losses and gains in TOC as a result of management or land-use conversions, and the LFOC and HFOC, which are separated by using densitometry techniques, are also more sensitive to land cover changes and management practices [48,49]. The soil POXC is more sensitive to the presence of lignin or lignin-like compounds and therefore to the nature of the vegetation present [50], which may explain the highest sensitivity of POXC to vegetation type. Therefore, POXC is considered to be the most sensitive index for differentiating the effects of vegetation types in our region.

Conclusions
The present study demonstrated that the SOC stocks exhibited an increasing trend with vegetation complexity in this region. However, there were no significant differences among SG, YF, MF, and NSF. Meanwhile, the TOC, POXC, LFOC, HFOC, and POC contents in the surface layer were higher than those in the other three deeper soil layers. These results suggested that the concentrations of TOC and all five carbon fractions were strongly affected by the community types and soil depths. The positive correlations between the SOC fractions and soil BD, Fe, and GRSP suggested that soil physicochemical and microbial processes are among the most important mechanisms that contribute to SOC persistence. Although all soil carbon indices have the potential to be used as indicators to assess vegetation recovery, POXC could be considered as most sensitive indicator of soil carbon changes that are associated with vegetation restoration.