Introduction

Soil organic carbon (SOC) is a complex and diverse amalgamation of organic compounds that experience incremental decomposition and transformation1. In the global soil C pool, which holds 2500 Gt, SOC constitutes approximately 62% of the total2. Any slight alternation in the storage and behavior of SOC can have a substantial effect on atmospheric CO₂ concentrations and consequently exert a significant influence on climate change3,4. SOC can be artificially separated into two fractions of soil labile C and stable C5. The quantity of soil labile C and stable C within the total SOC serves as a crucial determinant of SOC stability6.

Soil aggregates are the primary structural units of soil7. SOC has a close connection with the formation and stability of soil aggregates. Soil aggregates are the primary repository of SOC and play a crucial part in offering physical protection to SOC. Meanwhile, SOC functions as the principal cementing substances that contribute to the formation and stability of soil aggregates8,9. Soil aggregates are typically classified into four fractions: large macroaggregates (> 2 mm), small macroaggregates (0.25–2 mm), microaggregates (0.053–0.25 mm), and silt and clay fraction (< 0.053 mm)10,11. The composition of various soil aggregates affects soil structure and soil aggregate stability (SAS), subsequently influencing soil erosion resistance and soil quality12,13. Enhancing SAS represents an efficient approach to improve soil quality and alleviate environmental concerns related to soil degradation14.

A variety of indicators have been demonstrated to be the prominent factors influencing soil aggregate fluctuations15,16, among which land use17,18 and water level fluctuations19,20 are the two most dominant elements. In terrestrial ecosystems, the impact of land use on SAS and nutrient cycling have been widely documented21,22,23,24. For instance, as discovered by Guo et al.25 in the Chengdu Plain, land use had an impact on C physical sequestration principally due to modifications in surface area, pore development, and the Fe oxides content within macroaggregates. Farahani et al.26 reported that cultivation disrupted soil aggregates, causing soil structure degradation and C accumulation reduction, whereas vegetation restoration increased SAS and C content. Furthermore, adopting appropriate land use patterns tailored to local conditions can minimize harm to soil aggregates, preserve soil structure stability, and enhance nutrient accumulation12,13. However, within the context of a flooding regime, the influence of land use on soil C and nitrogen (N) dynamics might deviate from that observed in terrestrial ecosystems11. The soil nutrient and structure fluctuations in the riparian ecosystems are more susceptible to a variety of external factors compared to typical terrestrial ecosystems27. Zhu et al.28 and Ran et al.11 found that frequent wet–dry alternations triggered aggregate disintegration, expedited organic matter release, and promoted organic N and C mineralization, thus changing the nutrient status and ecological functions of soil. Moreover, drying and wetting cycles alter the conventional land use patterns and land management strategies, affecting the SOC stocks and local C cycle indirectly17. However, despite the acknowledged significance of both land use and hydrological changes, the interaction between these factors and their additive or synergistic effects on SAS and nutrient dynamics in the riparian ecosystems has remained insufficiently investigated10,11.

Situated in Xinyang City, Henan province, China, the Chushandian Reservoir (CR) stands as the only large-scale reservoir in the upper reaches of the main stream of the Huaihe River. With flood control as its main function, it also has comprehensive utilization functions such as water supply, power generation, irrigation, and tourism29. It is of great significance for improving the flood control standards of the Huaihe River, ensuring food security, promoting regional economic development and protecting the ecological environment30. Despite its significant functions and positive impacts, the CR has unfortunately encountered several environmental issues and the ecological function has degraded, which are causing growing anxiety and warranting urgent attention. For example, due to the significant disturbance caused by the dam construction and the altered hydrological regime, the soil erosion has intensified and the ecological degradation has become more severe in the CR area31,32. The “14th Five-Year Plan” for Water Security Guarantee and Water Ecological Environmental Protection in Henan Province has provided planning guidance at the macro level for the construction of the irrigation area project of the CR and its ecological environment protection. Reservoir buffer strips are critical for retaining water and soil, improving water quality, accumulating nutrients, and detaining non-point pollution, which demonstrated significant effects on maintaining environmental benefits and ecosystem functions of the reservoir33,34. However, few studies have made attempts to characterize the aggregate-related N and C levels and SAS in the buffer strips of the CR, considering the impacts of land use and water level fluctuations31. Additionally, the corresponding mechanisms remain to be further explored. Therefore, in order to understand the initial development of soil structure during the early phases of reservoir construction and furnish a scientific foundation for subsequent ecological conservation and land use strategies, it is essential to pinpoint the primary factors influencing SAS and nutrient dynamics within the buffer strips of the CR.

In the present study, we assessed the distribution and stability of soil aggregates along with the N and C content associated with these aggregates, across three land use types (abandoned cropland, grassland and woodland) and at three distances from the watercourse (2, 10, and 20 m). We conducted this evaluation before and after seasonal water level fluctuations (in April and June 2021) within the buffer strips of the CR. We proposed that (1) within the buffer strips of recently established reservoirs, land use significantly affects SAS more than water level changes do; (2) aggregate fractions might regulate the effects of land use and distance from the watercourse on nutrient content and SAS. The specific objectives of this study were to: (1) compare the variations in soil aggregate contents and aggregate stability indices under different land uses and flooding intensities; (2) identify the distribution characteristics of total carbon (TC) and total nitrogen (TN) in aggregate fractions under different land uses and distances from watercourse; (3) evaluate the impact of aggregate fractions on the nutrient content and SAS.

Materials and methods

Experimental reservoir

The CR is located in central China, with its dam positioned in the Huaihe River mainstream of Xinyang City, Henan Province29 (Fig. 1). The CR ceased operations for storage in May 2019 and achieved its standard level by October 202031. Above the dam site, along a stretch of 100 km to the origin of the Huaihe River channel, the reservoir control basin covers an area of 2.9 × 103 km2, with a total reservoir capacity of 12.51 × 108 m3. The climate of the CR basin transitions from subtropical to warm temperate, characterized by warm winters and humid summers, with average annual precipitation and temperature recorded at 1130 mm and 11–16 °C, respectively35. Precipitation across the year varied unevenly, reaching its peak in summer. The dominant soil type in the CR is yellow–brown power clay, categorized as Alfisols in USDA Taxonomy or Luvisols in FAO Taxonomy29. The key plant species include Populus, Camellia sinensis, Castanea mollissima, Imperata cylindrica, Xanthium sibiricum and Conyza canadensis31,36.

Fig. 1
figure 1

Positions of the CR region in China’s Huaihe River Basin (left) and the sampling locations within the buffer strips of the CR (right). Solid green square represents grassland; Solid orange circle represents abandoned cropland; Solid blue diamond represents woodland (Software: ArcGIS v. 10.8 https://www.esri.com/en-us/home).

Full size image

Within the CR buffer strips (N32° 22’ –32° 32’, E113° 89’–113° 96’), three different land uses, covering abandoned cropland, grassland, and woodland were selected as targets in this study (Fig. 1). In each land use, three replicate plots were conducted (nine in total). The plots were predominantly located along the western bank-line of the CR, characterized by more agricultural land and less construction land30. The distance between each plot did not exceed 1.5 km, and the sample plots had similar elevation, geology, slope and historical land utilization, which can be found in our prior research31. The width of sample plots ranged from 25 to 30 m, and we set the sampling strip distance accordingly. To eliminate the boundary effects of adjacent plots, in each land use, sample strips with different distances (2, 10, and 20 m) were carried out along the reservoir bank36. For convenience, the types of land use and scales of distance have been collectively labelled as C2, C10, and C20 for abandoned cropland, G2, G10, and G20 for grassland, and W2, W10, and W20 for woodland. During the discharge period (April 2021), the CR was at the lowest water level, while during the rainstorm period (June 2021), it was at the highest water level. In April and June 2021, within each site before and after seasonal variations in water levels in the CR buffer strips, four soil samples were gathered, totaling 72 samples.

Soil sampling and property analysis

Five intact soil cores were extracted at a depth of 10 cm from each site, merged to generate a uniform composite sample, put in self-sealing bags, and immediately delivered to the laboratory for further analysis. The physicochemical properties of the sample were assessed, including pH, total carbon (TC), total nitrogen (TN), dissolved organic carbon (DOC), ammonium nitrogen (NH4-N), and soil aggregates. An extra intact soil core sample was maintained to quantify soil water content (SWC) and bulk density (BD). All the methodologies were elaborated in Yan et al.31.

Water-stable soil aggregates were analyzed using a modified Yoder apparatus10,37. In summary, a series of 200 mm diameter sieves with mesh sizes of 0.053, 0.25, and 2 mm, holding 20 g of air-dried soil on the uppermost sieve, were submerged in distilled water for 5 min at ambient temperature. Subsequently, the soil samples were filtered underwater by repeatedly lifting and lowering the sieves for a period of 10 min, and the materials retained on the top surfaces of all the sieves were carefully collected. Meanwhile, the material mixture (consisting of water and soil) that had passed through the sieve with a mesh size of < 0.053 mm was gathered in buckets and then centrifuged at 2500 × g for 10 min38. This centrifugation process was intended to effectively separate the aggregate fractions from the mineral particles, thus facilitating further analysis and characterization of each component. Before the measurement of TC and TN levels, all the gathered samples were transferred into an oven and dried at 60 °C for 48 h39.

Data analysis

SAS was evaluated by MWD and GMD, based on formulas established in earlier researches40,41:

$${\text{MWD}} = \sum {\text{x}}_{{\text{i}}} {\text{y}}_{{\text{i}}}$$
(1)
$${\text{GMD = exp}}\left\{ {\frac{{\sum {\text{wi}}\;{\text{lnxi}}}}{{\sum {\text{wi}}}}} \right\}$$
(2)

where \({\text{x}}{\text{i}}\) signifies the mean diameter of the aggregate fraction (mm), \({\text{y}}{\text{i}}\) indicates the fraction’s proportional representation within the whole sample, and \({\text{w}}{\text{i}}\) denotes the weight of the aggregate fraction (g).

Levene’s test and Kolmogorov–Smirnov test were conducted to evaluate the homogeneity and normality of variances of all variables, while those that did not meet the assumptions were transformed with the natural logarithm42. Two-way ANOVA with a LSD test was executed to evaluate the differences in SAS, soil properties, N and C contents, and particle size fractions among sampling sites. The correlations among the above-mentioned indicators were assessed using Pearson correlation analysis (PCA). Variance partitioning analysis (VPA) was utilized to evaluate the combined and individual effects of distance from the watercourse and land use on soil aggregate fractions and their stability.

Furthermore, partial least squares path modelling (PLS-PM) was employed to assess how aggregate fractions, distance from watercourse, soil properties and land uses affect the nutrient content and SAS43. PLS-PM demonstrates excellent proficiency in handling small sample datasets and complex model frameworks, thus making it a significant analytical instrument in diverse research fields44,45. The model functioned on the basis of two essential constituents: the measurement model and the structural model. Acquired variables were presented as rectangles, and composite variables were illustrated as ellipses, which cannot be directly measured11. The combined impact of multiple predictor variables on response variables was determined by load coefficients. The direct correlations between predictor and response variables were quantified through path coefficients. To guarantee the dependability of the outcomes, both the coefficient of determination (R2) and path coefficients were validated via 5000 bootstraps. The comprehensive performance of the model was evaluated using the Goodness of Fit (GOF) statistic46,47. All statistical analysis and calculations were executed in R 4.0.3.

Results

Soil properties

Although the investigated area had consistent soil types, topography, and meteorological characteristics, distinct soil properties were observed due to diverse land uses among the sampling sites (Table 1). The abandoned cropland contained the lowest levels of TC, DOC, and C:N ratios but the highest NH4-N value. The grassland had the lowest NH4-N content and the woodland held the highest DOC content.

Table 1 Physicochemical features of soils from varying land uses and distances from watercourse in the buffer strips of the CR in April and June 2021.
Full size table

Different sampling months also lead to significant changes in soil properties (Table 1). From April to June 2021, the SWC and NH4-N of abandoned cropland, grassland, and woodland increased evidently, whereas the C:N ratios of these three land use types decreased remarkably. The TC content of abandoned cropland decreased, while the TN value of grassland increased. In addition, the DOC content of woodland increased, in contrast to the decrease in TC.

Two-way ANOVA demonstrated that soil DOC and NH4-N were remarkably impacted by land use and distance from the watercourse (Table S1). According to PCA, TC and TN exhibited a positive relationship, and they displayed significantly positive correlations with MWD and GMD at different monitoring sites and sampling months (Fig. 2).

Fig. 2
figure 2

Correlations among soil properties, aggregate fractions, MWD, and GMD in April (a) and June (b) 2021; *p < 0.05, **p < 0.01.

Full size image

Aggregate size distribution and stability

Generally, land uses and distance from the watercourse had significant effects on the proportions of four aggregate fractions, MWD and GMD in the CR buffer strips (Table 2). In contrast to woodland, abandoned cropland and grassland showed a reduction in the > 2 mm and 0.053–0.25 mm fractions, while the 0.25–2 mm and < 0.053 mm fractions saw an increase. Between April and June 2021, the three types of land use decreased the percentage of the > 2 mm fraction, while increasing the percentages of the 0.25–2 mm and < 0.053 mm fractions. Notable variations in MWD and GMD were detected across various land use categories, distances from the watercourse, and different sampling periods (Fig. 3). MWD and GMD were ranked as woodland > grassland > abandoned cropland, and increased with the increase of distance from watercourse. An overall decrease in MWD and GMD was shown from April to June 2021.

Table 2 Percentage of different aggregate fractions (%) and aggregate MWD and GMD in the buffer strips of the CR.
Full size table
Fig. 3
figure 3

Soil aggregate mean weight diameter (MWD) and geometric mean diameter (GMD) in the buffer strips of the CR.

Full size image

The responses of soil aggregate fractions and stability with respect to land uses and distance from the watercourse are presented in Fig. 4. In April 2021, the contributions of land use to soil aggregate fractions and stability were 22.50% (p < 0.01) and 22.38% (p < 0.01), individually, whereas those of distance from the watercourse were 17.68% (p < 0.01) and 15.22% (p < 0.01). In June 2021, the contribution of land uses to soil aggregate fractions and stability increased to 26.79% and 27.73% with p < 0.01, and those of distance from the watercourse decreased to 12.77% and 12.16% with p < 0.01. For both seasons, approximately 60% of the variation in aggregate fractions and stability has remained unexplained. This suggests that the factors influencing soil aggregate fractions and SAS are highly complex and multifaceted.

Fig. 4
figure 4

Variance partitioning analysis of the impacts of land uses (L) and distance from watercourse (D) on soil aggregate fractions (a, b) and stability (c, d) in April (a, c) and June (b, d) 2021.

Full size image

TC and TN contents in aggregate fractions

In soil aggregates, TC and TN levels were profoundly affected by aggregate sizes, land uses and distance from watercourse (Fig. 5, Table S2). Typically, the 0.25–2 mm fraction in various land uses presented the maximum TC and TN levels, followed in order by the > 2 mm, < 0.053 mm and 0.053–0.25 mm fractions. The levels of TC and TN in aggregate fractions varied by land use, ranking as: woodland > grassland > abandoned cropland. As the distance from the watercourse increased, the levels of TC and TN in aggregate fractions raised across the three land use types. In general, the levels of TC and TN in aggregate fractions declined between April and June 2021.

Fig. 5
figure 5

Variations in Total Carbon (TC) and Total Nitrogen (TN) levels within different soil aggregate fractions in the buffer strips of the CR in April (a, c) and June (b, d), 2021. LMA, > 2 mm; SMA, 0.25–2 mm; MI, 0.053–0.25 m; S + C, < 0.053 mm.

Full size image

The ratio of C:N in soil aggregates was significantly influenced by different land uses (Fig. 6, Table S2). Across various aggregate fractions, the C:N ratio in woodland peaked in April 2021, while they reached their lowest levels in June 2021, as opposed to abandoned cropland and grassland. Significant variations in the C:N ratio was observed among aggregate fractions across different land uses. Furthermore, the C:N ratio in aggregate fractions was faintly influenced by distance from the watercourse. From April to June 2021, the C:N ratio in different aggregate fractions increased.

Fig. 6
figure 6

C:N ratios for different soil aggregate fractions in the buffer strips of the CR in April (a) and June (b) 2021. LMA, > 2 mm; SMA, 0.25–2 mm; MI, 0.053–0.25 m; S + C, < 0.053 mm.

Full size image

Generally, the proportional impacts of the four aggregate fractions on TC and TN levels exhibited comparable patterns in both April and June 2021 (Fig. 7). The relative contribution of macroaggregates was more prominent compared to microaggregates. From April to June 2021, the relative contribution of the > 2 mm fraction in grassland to TC and TN levels increased, whereas the relative contribution of the 0.25–2 mm fraction, 0.053–0.25 mm fraction and < 0.053 mm fraction decreased. The opposite trend was observed in woodland.

Fig. 7
figure 7

Proportion of C (a, b) and N (c, d) contents of soil aggregate fractions in the buffer strips of the CR in April (a, c) and June (b, d) 2021. LMA, > 2 mm; SMA, 0.25–2 mm; MI, 0.053–0.25 m; S + C, < 0.053 mm.

Full size image

The influence of aggregate fractions on nutrient content and SAS

PLS-PM analysis identified causal relationships among soil properties, aggregate fractions, aggregate nutrient content, and stability in April and June 2021 (Fig. 8). All variables demonstrated a composite reliability surpassing 0.7, and the average variance extracted (AVE) values were larger than 0.5 (Table S3). These results strongly indicate that the measurement model is highly robust and reliable. The GoF indices of the PLS-PM for April and June 2021 were recorded as 0.52 and 0.67 respectively, which evidently implies a rather satisfactory overall performance of the model. The nutrient content and stability of aggregates were closely associated with land uses and distance from watercourse, and aggregate fractions served as mediators. Specifically, in April 2021, the explained path coefficients of land uses, distance from the watercourse, soil properties, and aggregate fractions on aggregate nutrient content were 0.10 (p < 0.01), 0.05 (p < 0.05), 0.21 (p < 0.001), and 0.27 (p < 0.05), individually (Fig. 8a, Table S4). The explained path coefficients of land uses, distance from the watercourse, and aggregate fractions on SAS were 0.16 (p < 0.01), 0.14 (p < 0.01), and 0.99 (p < 0.001), individually. Being significantly influenced by land uses (0.64, p < 0.001), soil properties had an important impact on aggregate fractions (0.48, p < 0.01).

Fig. 8
figure 8

Partial least squares path model illustrating the relationships among land uses, distance from watercourse, soil properties, aggregate fractions, aggregate nutrient content and stability within buffer strips of the CR in April (a) and June (b) 2021.

Full size image

By June 2021, the impact of land uses on aggregate nutrient content and stability increased to 0.17 (p < 0.01) and 0.22 (p < 0.01), respectively, whereas the influence of distance from the watercourse on these indices decreased to 0.03 (p < 0.05) and 0.10 (p < 0.01) (Fig. 8b, Table S4). From April to June 2021, the impact of soil properties on aggregate nutrient content was enhanced, and the impact of aggregate fractions on nutrient content was weakened.

Discussion

Impacts of land use strategies on SAS

Variations in land use strategies manifest primarily through alterations in species makeup and vegetation patterns17. Shifts in species makeup and plant structure can greatly affect the level of SOC, thereby influencing SAS in either direct or indirect manners10. In this study, the SAS indicators (MWD and GMD) exhibited significant positive correlations with TC and TN in the buffer strips of the CR (Fig. 2). This finding implies a favorable impact of TC and TN on the formation and stabilization of soil aggregates. The underlying mechanism could potentially be attributed to the fact that the organic materials and compounds encompassed within TC and TN function as binding agents. These agents have a vital role in enhancing soil aggregation and also confer a considerable degree of stability to the soil aggregates48.

Land use serves a crucial role regulating aggregate fluctuations23. In the terrestrial ecosystem, it is widely recognized that the stability of aggregates in undisturbed natural land is generally higher than that in agricultural land49,50,51. However, substantial distinctions exist between reservoir ecosystems and terrestrial ecosystems, with the variation characteristics of SAS under diverse land uses also differing. E.g., it has been proven that conservation agriculture is effective in enhancing the stabilization of aggregates in the Three Gorges Reservoir11. Our results indicated that, among the three land uses, the ranking of the SAS indicators in the CR buffer strips during April and June 2021 was woodland > grassland > abandoned cropland (Fig. 3), consistent with the findings of soil samples collected in December 202031. As possible explanations, the woodland boasted the richest species composition and vegetation structure compared with grassland and abandoned cropland. This leads to the maximum levels of SOC and humus, which facilitate the maintenance of SAS49. Second, anthropogenic disturbance, which encompasses activities such as tillage, directly ravages the vegetation cover and severely dismantles the soil structure. This twofold assault not only weakens the natural mechanisms that are crucial for maintaining soil integrity and cohesion but also leads to a reduction in SAS10. The woodland was virtually unaffected by anthropogenic disturbance, which reduces soil erosion, and contributes to the maintenance of SAS52. Furthermore, previous studies have demonstrated that the plentiful root exudates and plant debris present in the woodland serve to enrich the soil with highly polymerized binding substances53,54 and boost soil microbial activity55,56.

Within the CR buffer strips, we observed that the SAS indicators of woodland, grassland and abandoned cropland all decreased from April to June 2021 (Fig. 3). This result was possibly due to wet–dry alternations caused by the seasonal water level fluctuations. When elaborating on the underlying mechanisms from a physical perspective, the alternating wet and dry conditions initiate soil shrinkage and expansion. These processes subsequently disrupt the connections among soil particles, thereby resulting in a reduction in SAS31. At the chemical level, previous research has shown that such wet–dry alternations modify the ionic concentration and redox conditions in the soil, which affects the formation and decomposition of organic and inorganic cementing substances, and ultimately influences the stability of aggregates57. Biologically, it has been reported that wet–dry alternations induce changes in the composition and function of soil microbes, indirectly influencing SAS56.

Impacts of distance from watercourse on SAS

Water level fluctuation constitutes another dominant factor affecting SAS20,31. Comprehensive investigations have revealed the response mechanisms of soil aggregate fractions and their stability within the reservoir riparian zone during seasonal dynamics and long-term wet–dry cycles19,58, and have reached a consensus. For instance, SAS tends to increase in sites with low flooding intensity and decrease in those with strong flooding intensity28. In the current study, at distance scale, the SAS indicators in the CR buffer strips in April and June 2021 followed the order: C20 > C10 > C2 for abandoned cropland, G20 > G10 > G2 for grassland, and W20 > W10 > W2 for woodland (Fig. 3), indicating the SAS increased with the increasing distance from watercourse and the weakening of water level fluctuations. This finding aligns with the prevalent perspective presented in previous studies19,28,58. It is mainly due to the fact that soils near water are more susceptible to direct scouring and erosion induced by water currents, consequently leading to the disruption of the structure of soil aggregates59. In addition, within the CR buffer strips, as the distance from the watercourse increased, a decline in soil moisture and an improvement in soil aeration were observed. As certain scholars have pointed out, under such conditions of moderately low moisture, it is favorable for the bonding and aggregation between soil particles60. This also stimulated soil microbial activity and aids in the augmentation of SAS56.

In this study, variation partitioning and PLS-PM analysis yielded comparable results. The contribution of distance from watercourse to SAS in the CR buffer strips decreased from April to June 2021 (Figs. 4, 8). This finding may be related to soil samples had a relatively higher moisture content in June 2021 in comparison to April 2021, and the distribution of soil moisture did not exhibit significant disparities at the distance scale. As a result, the impact of the distance from the watercourse on SAS was weakened. Our results also suggested that within the buffer strips of the CR, the influence of land uses on SAS was more pronounced compared to the impact of the distance from watercourse (Fig. 4), which was in accordance with our initial hypothesis. Nevertheless, Ran et al.11 reported that, the intensity of dam-triggered flooding played a substantial role in elucidating SAS than land uses did. Therefore, the dominant driving mechanisms underlying SAS within the riparian zone remain a subject of contention. There are different impact mechanisms of land uses and distance from watercourse on the stability of aggregates. The disparities in vegetation cover and SOC content among various land use types have a direct impact on SAS23,41. In contrast, the distance from the watercourse mainly exerts an indirect influence on SAS by modulating the soil moisture conditions31. Moreover, land uses, which frequently encompass human activities, can significantly affect the soil structure and the stability of aggregates26,39.

Distributions of C, N, and C:N ratio in aggregate fractions

Generally, across the study locations within the buffer strips of the CR, the TC and TN levels in aggregate fractions exhibited similar patterns to the SAS, with the woodland showing the highest (Fig. 5, Table S2). These were significantly affected by land uses and distance from watercourse. This implies that varying protection measures should be taken to maintain SAS and aggregate-related N and C contents. Notably, the woodland in western Hubei and Guangdong Province also demonstrated the highest contents of aggregate-related organic carbon and TN47,49. This suggests that such a land use is conductive to retaining the N and C contents within soil aggregates. Besides land use and distance from watercourse, aggregate fractions’ TC and TN contents in the CR buffer strips were also impacted by aggregate sizes (Fig. 5, Table S2), consisting with the previous studies11,23. Furthermore, the coarse aggregate fractions (macroaggregates, > 0.25 mm) contained higher TC and TN levels than the fine aggregate fractions (microaggregates, < 0.25 mm) (Fig. 5, Table S2), compatible with the aggregate hierarchy model reported by Tisdall and Oades61, in which the levels of TC and TN have a positive relationship with aggregate size. It might be attributed to the formation process of larger aggregates from smaller ones through the bonding of polysaccharides derived from plants and microbes7,62. Consequently, macroaggregates contributed a larger proportion to aggregate fractions’ TC and TN than microaggregates (Fig. 7).

Essential factors that greatly impact the decomposition of fresh plant residues include C:N ratios, which are crucial for soil aggregation and C storage63,64. A considerable number of studies have documented soil aggregates’ C:N ratios in the riparian zone of reservoirs under various land use strategies and hydrological stress conditions10,11,65. However, the variation characteristics of aggregate C:N ratios with aggregate fractions are still controversial. Ran et al.11 concluded the C:N ratio of soil aggregates in reservoir riparian zone increased with increasing aggregate size. Zhu et al.10 argued in reservoir riparian zone, a negative correlation was observed between the C:N ratio of soil aggregates and their sizes, indicating that the C:N ratio diminished with larger aggregate sizes. In the current study, soil aggregate C:N ratios in the CR buffer strips generally decreased with increasing aggregate size in most but a few cases (Fig. 6, Table S2), akin to the findings of Zhu et al.10. The high C:N ratio represents a low decomposition and mineralization degree of soil organic matter, whereas the low C:N ratio reveals that the organic matter is readily decomposed in soil21,66. Therefore, coarse aggregate fractions in the CR buffer strips contained more labile organic C than fine aggregate fractions. Traditionally, carbon associated with microaggregates demonstrates a high level of SOC stability and a low turnover rate on account of the significant amount of stable carbon present in microaggregates. In contrast, carbon related to macroaggregates shows a low SOC stability and a high turnover rate as it is abundant in labile carbon content67,68.

Impacts of aggregate fractions on aggregate nutrient content and stability

Soil properties, such as pH, BD, SWC and DOC, directly impacted aggregate fractions (> 2 mm, 0.25–2 mm, 0.053–0.25 mm and < 0.053 mm) within the buffer strips of CR (Fig. 8, Table S4). This highlights that soil properties had a significant and direct effect on the formation and distribution of various-sized aggregates. Previous study has proved that the influence of soil properties on aggregate fractions ultimately affects the nutrient content and stability of aggregates, which in turn exerts an impact on soil quality and ecological functions11. In addition to their direct effects, both land uses and distance from watercourse also affected the nutrient content and stability of aggregates through their effects on aggregate fractions (Fig. 8, Table S4). This indicates that aggregate fractions served as essential mediators not only for nutrient distribution and availability in soils but also for soil structural stability within the buffer strips of CR. In alignment with the research findings in the terrestrial ecosystem, wherein aggregate fractions function as a mediator in the influence of land uses on C stabilization69, this further accentuates the significance and universality of aggregate fractions across diverse ecosystems.

Aggregate fractions could regulate the effects of external environmental variables on aggregate nutrient content and stability, which may be based on several factors. First, the physical structure of aggregates offers storage and protection for soil nutrients70. The complex pore systems constituted by different aggregate fractions decelerate the rate of nutrient loss, enabling nutrients to persist in the soil for an extended period. Secondly, the organic matter within aggregates interacts with soil microorganisms71. The microorganisms consume the organic material, decompose it, and release adhesive substances that enhance aggregate stability. Additionally, the mineral components of aggregates are involved in the adsorption and desorption of nutrients48. When external environmental conditions change, the equilibrium of nutrient adsorption and desorption by mineral shifts, thereby regulating the content and effectiveness of nutrients within aggregates. What’s more, the pore structure and aeration of aggregates have an impact on the redox conditions of the soil. This directly influences the chemical forms and transformation processes of specific nutrient elements such as nitrogen, iron, manganese, etc., consequently regulating their content and stability within aggregates72. In summary, through a variety of mechanisms, including complex physical structures, organic–inorganic interactions, and microbial activities, aggregate fractions modulated the effects of external environmental variables on the nutrient content and stability of aggregates to maintain the function and stability of reservoir ecosystems.

Conclusions

This study presents new evidence regarding the influence of three types of land uses and water level fluctuations on soil aggregates and their corresponding N and C levels in the buffer strip ecosystem of the CR. The findings highlighted the distinct mechanisms of soil aggregate stabilization across various land uses, suggesting that tailored protection strategies are necessary for each land use to minimize soil erosion and promote vegetation recovery in the buffer strips. Land uses and distance from the watercourse were crucial factors in stabilizing soil structure and accumulating soil N and C. The impact of land uses was relatively more prominent. Meanwhile, soil aggregate fractions acted as crucial mediators of the effects of land uses and distance from the watercourse on soil aggregate nutrient content and stability. This research deepens comprehension of SAS mechanisms and offers crucial insights for devising strategies to efficiently control soil erosion and enhance soil quality in reservoir buffer strips. In addition, the buffer strips of the CR are popular destinations for fishing enthusiasts, and disturbance from anthropogenic activity may influence soil structure, leading to soil aggregate instability. Further research should investigate how human actions affect soil stability in the buffer strips of the CR.