Soil Carbon–Water Trade-Off Relationships and Driving Mechanisms in Different Forest Types on the Yunnan Plateau, China

Abstract

Semi-humid subtropical montane regions face the dual pressures of climate change and water scarcity, making it essential to understand how soil carbon–water coupling varies among forest types. Focusing on seven representative forest types in the central Yunnan Plateau, this study analyzes the spatial distribution, trade-offs, and drivers of soil organic carbon storage (SOCS) and soil water storage (SWS) within the 0–60 cm soil layer, using sloping rainfed farmland (SRF) as a reference. We hypothesize that, relative to SRF, both SOCS and SWS increase across forest types; however, the direction and strength of the SOCS–SWS trade-off differ among plant communities and are regulated by litter traits and soil structural properties. The results show that SOCS in all forest types exceeded that in SRF, whereas a significant increase in SWS occurred only in ACF. Broadleaf stands were particularly prominent: SOCS rose most in the 23 yr SF and the 20 yr ACF (274.44% and 256.48%, respectively), far exceeding the 9–60 yr P. yunnanensis stands (44.01%–105.32%). Carbon–water trade-offs varied by forest type and depth. In conifer stands, SWS gains outweighed SOCS and trade-off intensity increased with stand age (RMSD from 0.48 to 0.53). In broadleaf stands, SOCS gains were larger, with RMSD ranging from 0.21 to 0.45 and the weakest trade-off in SF. Across depths, SOCS gains exceeded SWS in 0–20 cm, whereas SWS gains dominated in 40–60 cm. Regression analyses indicated a significant negative SOCS–SWS relationship in conifer stands and a significant positive relationship in 0–20 cm soils (both p < 0.05), with no significant correlations in other forest types or depths (p > 0.05). Correlation results further suggest that organic matter inputs, N availability, and soil physical structure jointly regulate carbon–water trade-off intensity across forest types and soil depths. We therefore recommend prioritizing native zonal broadleaf species, as well as protecting SF and establishing mixed conifer–broadleaf stands, to achieve synergistic improvements in SOCS and SWS.

1. Introduction

Soil organic carbon (SOC) constitutes the largest carbon pool in terrestrial ecosystems [1], storing roughly 3.3 times as much carbon as the atmosphere [2]. It plays a pivotal role in mitigating global climate change and in enhancing ecosystem resilience to extreme events such as heatwaves and droughts [3]. Even minor changes in soil carbon stocks can substantially affect atmospheric CO₂ concentrations and climate processes. Meanwhile, soil moisture is a fundamental regulator of plant growth and microbial activity. By controlling the availability of carbon, nitrogen, and other nutrients, it governs nutrient cycling and the dynamics of ecosystem productivity [4]. Critically, soil carbon and water cycles are not independent. Accumulation of SOC alters soil structure and porosity and thereby influences soil water content, whereas soil moisture directly constrains the inputs, turnover, and transport of organic carbon [5]. This carbon–water coupling underpins key ecosystem services, including climate regulation and carbon sequestration, drought buffering, and biomass production [6,7]. Recent work in arid and semi-arid regions indicates that carbon-oriented vegetation restoration can deplete soil water to unsustainable levels, ultimately jeopardizing long-term carbon sinks and ecosystem functioning [8]. Clarifying carbon–water coupling in soils is therefore essential for balancing soil water conservation with carbon sequestration in forest ecosystems and, under climate change, is central to improving system resilience and advancing sustainable forest management.

However, existing research indicates substantial uncertainty and context dependence in how soil carbon and water respond to vegetation restoration or changes in land use. On the one hand, some studies have observed synergistic improvements during the early stages of restoration or in surface soils, whereby higher moisture promotes organic matter accumulation and microbial processes, thereby increasing topsoil soil organic carbon storage (SOCS) [9,10,11], whereas losses of soil carbon can also reduce soil moisture [12]. On the other hand, deep-rooted plantations or orchards (e.g., Robinia pseudoacacia, apple) often enhance carbon accumulation while markedly depleting deep soil water and forming persistent dry layers, exhibiting a characteristic carbon–water trade-off [13,14,15]. Long-term natural succession may increase SOC but is often accompanied by declines in soil moisture, thereby constraining the long-term gains in SOC [16]. Similarly, converting sloping cropland to orchards or forests often causes substantial reductions in soil moisture, resulting in only limited increases in SOC [17,18]. Some studies have also reported that, in certain contexts, land-use change has no significant effect on either SOCS or soil water storage (SWS) [11]. From a spatiotemporal perspective, soil carbon–water relations exhibit marked spatial heterogeneity and temporal dynamics. Topography redistributes both soil water and organic matter; footslope positions typically accumulate more moisture and organic matter and thus show weaker trade-offs, whereas upslope positions are more prone to water deficits [19]. Along the soil profile, land-use change often exerts stronger effects on carbon and water in deeper horizons [20]. Over time, as restored stands mature, both soil moisture and carbon accumulation can shift in ways that alleviate early-stage carbon–water trade-offs. For example, in Robinia pseudoacacia plantations on the Loess Plateau, severe deep-soil water deficits were observed in the early restoration phase, but deep-soil moisture recovered after approximately 30 years [21]. Zhang and Shangguan likewise reported that the correlation between SWS and SOCS weakened with increasing restoration age [16]. These divergences indicate that soil carbon–water relations are context dependent, with the direction and strength of trade-offs varying with climate regime, species functional traits such as rooting depth and water-use efficiency, restoration age, successional stage, and soil depth. This complexity makes clear that our understanding of soil carbon–water responses to vegetation restoration and land-use change remains incomplete.

As the key interface between vegetation and soil, the litter layer is both a primary source of SOC and a major pathway regulating rainfall infiltration and evaporative water loss [22]. It therefore offers a critical entry point for explaining how soil carbon–water coupling responds to different vegetation restoration or land-use changes. Litter influences soil moisture by intercepting rainfall, modulating infiltration, and reducing evaporation [23]. Concurrently, decomposition products serve not only as substrates for microorganisms but also as binding agents that promote aggregate formation and structural stabilization [24]. These processes improve soil structure by increasing porosity and water-holding capacity and by forming organo–mineral complexes, thereby affecting the inputs, transformation, and physical protection of organic carbon and fundamentally shaping the synchronized dynamics of soil carbon and water [25,26]. As the basic units that maintain soil structure and function, aggregates further mediate the linkage between litter inputs and soil carbon–water relations; they physically protect SOC while supporting water infiltration and retention [27,28]. In forest soils, aggregate formation and stability are regulated by biotic factors such as litter quality, root activity, and the soil microbial community [29,30]. In mountainous plateaus with strong rainfall seasonality and rugged topography, aggregate stability is particularly critical because it determines resistance to raindrop splash and overland flow, thereby influencing the retention of carbon and water [31]. Integrating measurements of litter mass, decomposition rates, aggregate stability, and related soil properties can thus provide a mechanistic framework for interpreting carbon–water trade-offs or synergies. However, few studies to date have explicitly incorporated litter–soil interactions into analyses of soil carbon–water coupling.

From a geographic perspective, studies on how soil carbon–water coupling responds to vegetation restoration or land-use change have largely focused on the arid and semi-arid regions of northern China, where trade-offs between SOC gains and water consumption tend to dominate [8,32]. By contrast, research remains scarce in humid subtropical montane systems such as southwestern China, despite their comparable importance for regional carbon and water cycles. The central Yunnan Plateau is characterized by rugged topography, highly weathered soils, and pronounced wet–dry seasonality, which together create strong spatiotemporal heterogeneity in soil carbon–water processes. Historically, intense human–land conflicts and severe forest degradation have left zonal vegetation dominated by secondary forests [33]. Over recent decades, large-scale afforestation has been implemented to restore degraded lands and raise forest cover, primarily using drought-tolerant, fast-growing species such as Alnus cremastogyne, Pinus yunnanensis, and Acacia dealbata. Although these species are well adapted to infertile and relatively dry soils [34,35,36], it remains unclear whether they can simultaneously meet the dual objectives of water conservation and carbon sequestration [37]. To date, most studies in this region have examined either SOC or SWS in isolation [38,39,40], with few directly assessing their coupling or trade-offs. Consequently, substantial knowledge gaps persist regarding the dynamics and drivers of soil carbon–water synergies or trade-offs during vegetation restoration in humid subtropical mountains, which limit sustainable forest management and the effectiveness of ecosystem restoration in the region.

A simple way to quantify the magnitude of the soil carbon–water trade-off is to compute the root-mean-square deviation (RMSD) between the two standardized benefits [20]. After min–max normalization of SOCS and SWS, each sampling point is placed in a two-dimensional space and its Euclidean distance to the 1:1 ideal diagonal is calculated; RMSD is then obtained by aggregating distances at the point level, with smaller values indicating more balanced SOC and water contributions and thus better coupling, and larger values indicating a more pronounced trade-off [41]. As a simple, unit-free metric that is robust to nonlinear responses, RMSD has proven effective for diagnosing balance among ecosystem-service benefits and has been widely applied to inform ecological management, landscape planning, and decision-making [18,42,43]. This study aims to characterize the distributions of SOCS and SWS within the 0–60 cm profile across representative plant communities and stand-age gradients on the central Yunnan Plateau, and—using RMSD—to evaluate soil carbon–water trade-offs and the drivers of their variation, including litter traits, aggregate stability, and related soil physicochemical properties. Specifically, we seek to: (i) describe SOCS and SWS distributions across plant communities; (ii) assess the direction and intensity of their trade-offs; and (iii) quantify the relative contributions of environmental factors to SOCS and SWS variation. We hypothesize that vegetation restoration generally increases SOCS and SWS, but that their trade-off is context dependent, with direction and intensity differing among communities and being regulated by litter characteristics and aggregate stability.

2. Materials and Methods

2.1. Study Site

The study was conducted in the Shuanghe-Monande Municipal Nature Reserve, Anning City, central Yunnan Province, China (24°28′25″–24°40′36″ N, 102°15′40″–102°30′22″ E) (Figure 1). The reserve lies located in a low-relief mountainous area, with elevations ranging from 1930 to 2513.4 m (mean: 2070 m). The region experiences a low-latitude plateau subtropical monsoon climate, characterized by a mean annual temperature of 13.1 °C and annual precipitation of 1008.4 mm, with 70%–80% occurring during the rainy season (May–October) and the dry season spanning November–April.

Soils are derived from weathered colluvial materials of basalt, classified as Acrisols or Ferralsols according to the Food and Agriculture Organization of the United Nations (FAO) soil classification system, corresponding to Ultisols in the United States Department of Agriculture (USDA) Soil Taxonomy [44]. The natural vegetation is a semi-humid evergreen broadleaf forest, which has been heavily degraded during the 1950s–1960s, resulting in widespread secondary forests. Current vegetation types include SEBF dominated by Cyclobalanopsis glaucoides and Lithocarpus dealbatus, mixed conifer-broadleaf forests combining Quercus variabilis, Quercus aliena, and Pinus yunnanensis, as well as extensive plantations of Pinus yunnanensis, Pinus armandii, Alnus cremastogyne, and Acacia dealbata. Understory vegetation mainly consists of Rhododendron decorum, Rhododendron spiciferum, Vaccinium mandarinorum, Vaccinium bracteatum, Ternstroemia gymnanthera, Viburnum cylindricum, and Myrica nana.

2.2. Plot Establishment and Sampling

In May 2019, based on prior reconnaissance and considering representative forest types and stand ages in the study region, we selected the following stands in the central Yunnan Plateau: a secondary evergreen broadleaf forest closed to use for approximately 20 years (SF); an Alnus cremastogyne plantation representing the native zonal broadleaf type (ACF) with a comparable restoration age; an Acacia dealbata plantation representing a non-zonal broadleaf type (ADF); Pinus yunnanensis plantations at 9, 22, and 60 years (PY9, PY22, PY60) representing conifer plantations; and a 35-year-old mixed conifer–broadleaf stand (MF). Sloping rainfed farmland (SRF; maize–potato–wheat rotation) served as the reference. Except for SF and MF, which arose through natural regeneration and supplementary planting, all forest types were established by afforestation on former SRF; for every plantation, the pre-afforestation land use was SRF. Stand ages were determined from local planting/replanting records and verified in the field. Field inspection confirmed that all selected plots had been under long-term protection, with no recent management or disturbance (e.g., thinning, fertilization, grazing, understory clearing, or fire; Supplementary Table S1). All field sampling was conducted in May, i.e., at the end of the dry season and immediately prior to the onset of the monsoon. At this time, soils typically reach their annual minimum water content after prolonged drought, which maximizes contrasts in soil water storage among forest types and reduces artifacts associated with transient moisture conditions, thereby better reflecting the intrinsic water-holding capacity of each forest type. May is also a peak period of litter return in the region.

For each forest type, three 20 m × 20 m plots were established. Within each plot, four 0.5 m × 0.5 m subplots were randomly positioned to collect litter from the undecomposed (OL) and partially decomposed (OF) horizons, as well as soil from 0 to 20, 20–40, and 40–60 cm. Within each subplot, materials from the same horizon or depth interval were combined and homogenized to form a composite sample. In total, 42 litter samples were collected (7 forest types × 2 litter horizons × 3 replicates; sloping rainfed farmland, SRF, had no accumulated litter), and 72 samples each of composite bulk soil, bulk-density cores (ring cutter), undisturbed soil blocks, and aluminum-tin samples were obtained (8 land use/forest types × 3 depths × 3 replicates).

2.3. Litter Accumulation and Water Holding Capacity Measurements

On-site measurements were taken for each litter sample, including fresh weight ($W_f$) and litter layer thickness. Litter decomposition intensity (LDI) was calculated as the proportion of the OF layer accumulation relative to the total accumulation. All litter samples were dried to constant weight ($W_0$) in an oven at 85 °C, and the litter accumulation per unit area (W, t hm⁻²) was subsequently determined. The water-holding capacity of the litter was assessed using a static immersion method under laboratory conditions, quantifying water absorption rate and maximum water-holding capacity.

To capture the rapid changes during the initial adsorption phase (≤2 h) and the slower approach to equilibrium at later stages (≥4–24 h)—and given that water uptake remains essentially stable beyond 24 h—oven-dried litter was placed in nylon mesh bags (45 cm × 30 cm) and submerged in water for preset durations of 0.25, 0.5, 1, 2, 4, 6, 8, 12, and 24 h. After each interval, samples were drained and weighed ($W_t$). Three replicates were conducted for each sample. The water retention rate ($R_t$, %) was calculated as:

$$R_t=(W_t-W_0)/W_0\times 100\%$$

(1)

After 24 h of immersion, the litter water content reached saturation, which was defined as the maximum water retention rate ($MWR$, %) and corresponding maximum water holding capacity ($W_m$).

Initial water absorption rate (IWAR) was defined as the water retention rate within the first 0.5 h of immersion, calculated using Equation (2):

$$IWAR=R_t/0.5$$

(2)

Based on Balocchi et al. [45], the effective water interception capacity (P, %) was estimated as 85% of $MWR$ after subtracting the initial moisture content ($R_0$, %):

$$P=0.85\times MWR-R_0$$

(3)

The effective stock capacity (ESC) was calculated using Equation (3):

$$ESC=P\times W$$

(4)

where $P$ is the effective interception rate (%), W is litter biomass (t ha⁻¹), and $ESC$ is the effective stock capacity (t ha⁻¹).

2.4. Soil Parameter Measurement

Soil samples were dried to constant weight in an aluminum box at 105 °C to determine soil water content (SWC, %). The mixed samples were air-dried, ground, and sieved through 2 mm, 1 mm, and 0.25 mm sieves for the analysis of soil physicochemical properties. Soil indicators, including soil organic carbon (SOC), pH, bulk density (BD), capillary porosity (CP), non-capillary porosity (NCP), total and available nitrogen and phosphorus content, were determined following the standard methods described by Carter and Gregorich [46]. Repeated measurements and rigorous quality control procedures were implemented throughout the experiment to minimize errors and ensure data reliability.

Undisturbed soil samples were air-dried, gently broken along natural aggregates to clods ≤10 mm, and sieved through a 10 mm mesh. Aggregate size distribution was determined using combined dry and wet sieving methods [29]. Dry sieving separated aggregates into four size classes: >5 mm, 2–5 mm, 0.25–2 mm, and <0.25 mm, and the mass percentage of each fraction was measured. Based on these proportions, a 50 g composite sample was prepared and subjected to wet sieving to obtain the proportion of water-stable aggregates. Aggregate stability was evaluated using mean weight diameter (MWD, mm) and percentage of aggregate destruction (PAD, %) as follows [28]:

$$\mathrm{M}\mathrm{W}\mathrm{D}={\sum }_{i=1}^n\overline{R_i}\times W_i$$

(5)

$$\mathrm{P}\mathrm{A}\mathrm{D}={\displaystyle \frac{{DR}_{>0.25}-{WR}_{>0.25}}{{DR}_{>0.25}}}\times 100\%$$

(6)

where $W_i$ is the mass percentage of aggregates in the ith size class (%); $\overline{R_i}$ is the mean diameter of aggregates in the ith size class (mm), assigned as 7.5, 3.5, 1.125, and 0.125 mm for the respective fractions; ${DR}_{>0.25}$ is the proportion of dry-sieved aggregates > 0.25 mm; $W{R}_{>0.25}$ is the proportion of water-stable aggregates > 0.25 mm.

2.5. Calculation of Coupled Coordination Degrees and Trade-Offs Between SOCS and SWS

In this study, we use SWS to denote the depth-equivalent quantity of water contained in the soil profile at sampling (mm). SWS is distinct from plant-available water capacity and field capacity; instead, it reflects the instantaneous amount of water present, computed from volumetric water content integrated over depth [13]:

$$\mathrm{S}\mathrm{W}\mathrm{S}={\sum }_{i=1}^n{SWC}_i\times {{BD}_i\times H}_i\times 10$$

(7)

Similarly, the SOC stock (SOCS, kg C·m⁻²) were calculated as follows:

$$SOCS={\sum }_{i=1}^{n}0.01\times {SOC}_i\times {BD}_i\times \left(1-{ST}_i\right)\times △d_i$$

(8)

where ${SWC}_i$ and $H_i$ are the SWC and thickness of the $i$ soil layer (cm), respectively; ${SOC}_i$ is the soil organic carbon concentration of the ith layer (g kg⁻¹); ${BD}_i$ is the soil BD of the ith layer (g cm⁻³); $△d_i$ is the soil layer thickness (cm). The proportion of gravel (>2 mm) was negligible and not considered in this study.

The trade-off between SOCS and SWS was assessed using the root mean square deviation (RMSD) method [43]. SOCS and SWS were first standardized to a 0–1 scale using min-max normalization:

$${\mathrm{X}}_{\mathrm{s}\mathrm{t}\mathrm{d}}=\left({\mathrm{X}}_{\mathrm{o}\mathrm{b}\mathrm{s}}-{\mathrm{X}}_{\mathrm{m}\mathrm{i}\mathrm{n}}\right)/\left({\mathrm{X}}_{\mathrm{m}\mathrm{a}\mathrm{x}}-{\mathrm{X}}_{\mathrm{m}\mathrm{i}\mathrm{n}}\right)$$

(9)

where ${\mathrm{X}}_{\mathrm{s}\mathrm{t}\mathrm{d}}$ is the standardized value; ${\mathrm{X}}_{\mathrm{o}\mathrm{b}\mathrm{s}}$ is the observed value; ${\mathrm{X}}_{\mathrm{m}\mathrm{i}\mathrm{n}}$ and ${\mathrm{X}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ are the global minimum and maximum computed across all observations (i.e., soil depths, and land use/forest types).

Standardized SOCS and SWS were plotted in a two-dimensional coordinate system, with SOCS on the x-axis and SWS on the y-axis. Root-mean-square deviation (RMSD) was calculated as [43]:

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{D}=\sqrt{{\displaystyle \frac{\sum _{\mathrm{i}=1}^{\mathrm{n}}\hspace{1em}{\left(\mathrm{X}{1}_{\left(\mathrm{i}\right)}-\mathrm{X}{2}_{\left(\mathrm{i}\right)}\right)}^2}{\mathrm{n}-1}}}$$

(10)

where $\mathrm{X}{1}_{\left(\mathrm{i}\right)}$ and $\mathrm{X}{2}_{\left(\mathrm{i}\right)}$ are the standardized values of SOCS and SWS for sample i; n is the number samples. Overall ecosystem service gains were calculated as the average of standardized SOCS and SWS, assuming equal importance of both services. The RMSD for an individual sample is calculated as ${\left(\mathrm{X}{1}_{\left(\mathrm{i}\right)}-\mathrm{X}{2}_{\left(\mathrm{i}\right)}\right)}^2$.

2.6. Statistical Analysis

In R 4.4.3 [47], data normality and homogeneity of variances were assessed using the Kolmogorov–Smirnov (K–S) test and Levene’s test implemented in the nortest package. Subsequently, one-way analysis of variance (ANOVA) from the stats package combined with the least significant difference (LSD) method in the agricolae package was used to examine the significance of differences in SOCS and SWS among forest types and soil depths. Given that soil texture is primarily governed by parent material and long-term pedogenic processes, its spatial heterogeneity effectively reflects the inherent environmental background differences among sampling sites (e.g., parent material, climate, and historical development). Therefore, soil texture was incorporated as a key covariate to control major environmental heterogeneity beyond vegetation restoration and topographic variables (elevation and slope). Furthermore, linear mixed-effects models (LMEs) were constructed using the lme4 packages to evaluate the effects of forest type, soil depth, and their interaction on SOCS and SWS, with plot identity specified as a random intercept. In addition, partial correlation analysis was conducted with the ppcor package to disentangle the independent relationships between soil carbon and water storage, their trade-off intensity, and environmental factors, while controlling for forest type, soil texture, and topographic covariates.

3. Results

3.1. Distribution Patterns and Relative Gains of SOCS and SWS

SOCS and SWS differed by forest type, soil depth, and their interaction (p < 0.05; Figure 2a,b). The SOCS of SRF across the entire 0–60 cm profile was 7.65 Mg·hm⁻². Compared with SRF (Figure 2a), coniferous stands were 44.13%–166.57% higher in SOCS and broadleaf stands were 136.81%–274.76% higher. Among forest types, SF and ACF showed the highest SOCS—27.26 and 28.63 Mg·hm⁻², respectively—significantly greater than the other stands (p < 0.05). Along the soil profile, ADF showed increases only in the surface layer and remained significantly lower than SF and ACF at 40–60 cm (p < 0.05). In SRF and coniferous stands, 48.01%–60.07% of SOCS occurred at 0–20 cm, with only 7.11%–22.02% at 40–60 cm; by contrast, broadleaf forests held 25.48%–29.48% of SOCS in the deepest layer.

For SWS (Figure 2b) across the entire 0–60 cm profile, differences between SRF and other stands were generally small, but ACF (148.97 mm) was significantly higher than SRF (109.09 mm), PY9 (110.82 mm), ADF (100.60 mm), and MF (108.73 mm) (p < 0.05; Figure 2b), with the remaining stands at intermediate levels. Across the profile, SWS increased from the surface to mid-depth and then decreased in SRF (p < 0.05), whereas in the other forest types, SWS increased with increasing depth (p < 0.05).

3.2. Trade-Off Between SOCS and SWS

Significant differences were observed in the coupling relationship between soil carbon and water storage across different forest types and soil depths (Figure S2). Among the forest types, significant negative correlations between SOCS and SWS were found only in coniferous forests and mixed coniferous-broadleaf forests (MF) (p < 0.05), while no significant correlations were detected in SRF or broadleaf forests (p > 0.05), indicating a negative trade-off in soil carbon–water relationships in coniferous forests. Furthermore, a significant positive correlation between SOCS and SWS was observed in the 0–20 cm soil layer (p < 0.05), whereas no significant correlations were found in the 20–40 cm and 40–60 cm layers (p > 0.05).

In terms of the gains and trade-off intensity between soil carbon and water storage, SWS gains were greater than SOCS gains in SRF and coniferous forests, while the opposite pattern was observed in broadleaf and mixed forests. The root mean square deviation (RMSD) increased gradually from SRF to PY9, PY22, and PY60, but decreased again in MF. The RMSD of ADF was similar to that of SRF and PY9, while the lowest RMSD values were recorded in ACF and SF (Figure 3a). Across all forest types (Figure 3b), the dominant benefit transitioned from SOCS gains in the surface layer (0–20 cm) to SWS gains in the deeper soil layers (40–60 cm).

3.3. Controlling Factors of SOCS and SWS

Correlation analysis revealed distinct stratification in environmental factors influencing SOCS, SWS, and their trade-off intensity (C–W) across forest types (Figure 4, Figures S3 and S4). Throughout the 0–60 cm profile, SOCS exhibited highly significant negative correlations with BD and PAD (r = −0.62 and −0.68, p < 0.01), and significant positive correlations with silt content, MWD, LDI, IWAR, TN, and C/N (r = 0.53–0.91, p < 0.05). Stratified by depth, SOCS showed weakened correlations with silt content and ESC in the 0–20 cm layer, but a stronger association with slope. In the 20–40 cm layer, correlations with BD, aggregate stability indices, and C/N diminished, while relationships with textural parameters such as silt and clay content became more pronounced. The dominant factors influencing SOCS in the 40–60 cm layer remained consistent with those observed across the entire profile. SWS throughout the 0–60 cm depth was primarily positively influenced by topographic factors (elevation and slope) and silt content. In the 0–20 cm layer, IWAR and C/N exerted additional positive effects alongside elevation. Within the 20–40 cm layer, SWS correlated positively with silt content but negatively with CP and PAD. In the 40–60 cm layer, SWS maintained positive correlations with elevation and slope, but negative correlations with LDI and C/N. C–W was significantly negatively correlated with LDI throughout the soil profile. It correlated positively only with ESC in the 0–20 cm layer, whereas in the 20–40 cm layer, it related negatively to TN and positively to BD. No significant environmental correlates of C–W were detected in the 40–60 cm layer.

Partial correlation analysis further revealed that after controlling for forest type, topographic and soil textural factors showed no significant effects on SOCS, SWS, or C–W (Figure S4). In contrast, when topographic and textural covariates were controlled, litter properties, aggregate stability, and soil physicochemical properties remained significantly correlated with SOCS. C–W was negatively correlated with SOCS across the entire soil profile, positively correlated with ESC and C/N in the 0–20 cm layer, negatively correlated with SOCS and TN but positively correlated with BD in the 20–40 cm layer, and only positively correlated with PAD in the 40–60 cm layer. SWS was significantly influenced only by SOCS and C/N in the 40–60 cm soil layer, while no significant environmental correlates were detected in the remaining layers.

4. Discussion

4.1. Forest Type Controls on Soil Organic Carbon and Water Storage

Within the 0–60 cm soil profile, our results showed that SOCS was significantly higher in all forest types than in SRF, consistent with previous findings [33]. Notably, broadleaf forests exhibited significant increases in SOCS at 20–40 cm and 40–60 cm (Figure 2a), a depth distribution that differs from many earlier reports in which organic carbon accumulation was concentrated in the 0–20 cm surface layer [48,49]. Although Wu et al. reported that SOCS differences among forest types may be insignificant after vegetation restoration [33], we found that SOCS in broadleaf and mixed conifer–broadleaf forests generally exceeded that in conifer plantations of different ages. This pattern is plausibly related to the characteristics of conifer litter, which, despite its greater accumulation and interception capacity, has high lignin content and low decomposition rate (Figure S1), thereby limiting microbial breakdown and the transfer of carbon into mineral soils [50]. In contrast to studies reporting strong increases in SWS following restoration [8,32], total SWS across 0–60 cm in our study did not generally exceed that of SRF; only ACF showed a significant increase, and at 20–40 cm only SF was higher (Figure 2b). This accords with Yang et al. [51] and Wu et al. [33] and may reflect a new post-restoration balance between plant water use and infiltration regulation [52]. In general, native species consume less water during growth than exotic species, and broadleaf forests tend to provide greater water conservation than conifer forests [11,52].

Previous research indicates that with increasing stand age, SOCS may remain unchanged [53], decline [21], or decrease and then increase [54]. Across all forest types in our dataset, stand age did not have a significant effect on SOCS or SWS. However, in Pinus yunnanensis stands, SOCS and SWS increased progressively from young to mature forest (PY9 to PY60; Figure 2), suggesting a positive age effect on carbon and water storage in conifer plantations, with gains concentrated in surface soils and weaker responses at depth. Changes in SOC depend on the balance between carbon inputs and outputs during stand development [2]. In the early years after afforestation, strong soil disturbance and erosion, together with limited litter and root inputs, typically result in lower SOCS in young stands [55]. With stand development, the relative litter decomposition rate remains similar, but IWAR and ESC increase (Figure S1), indicating greater litter accumulation and potentially lower soil respiration [56], which together promote SOCS accrual. Soil moisture is critical for tree growth [4] and is likely a principal limitation on productivity in the water-limited central Yunnan region. In our P. yunnanensis stands, SWS increased with stand age, in contrast to the age-related declines in soil moisture reported for apple orchards on the Loess Plateau [21] and for Korean pine stands [43]. This discrepancy may relate to the progressive increase in canopy cover in our study (Table S1), which reduces direct solar radiation and soil evaporation while enhancing canopy interception and infiltration. By contrast, studies reporting age-related declines in SWS often involve lower canopy cover, excessive deep-root water use, and persistent drought conditions [21].

4.2. Effects of Litter and Soil Physicochemical Properties on Soil Carbon and Water Storage

Correlation analyses between soil carbon–water storage and environmental factors indicate that litter inputs and shifts in soil physicochemical properties following land-use conversion play a central role in determining SOCS and SWS, with pronounced depth-dependent effects (Figure 4, Figures S3 and S4). Across soil layers, SOCS was significantly higher where LDI, IWAR, and ESC were greater, together with higher TN and C/N ratios (all p < 0.05). These patterns suggest that litter traits mediate the influence of vegetation recovery on SOCS. Litter can enhance infiltration and surface soil water retention [57] while accelerating the transfer of organic matter into mineral soils and its occlusion within aggregates, thereby increasing SOCS and SWS and partly alleviating their trade-off. Numerous studies report that greater litter cover markedly reduces runoff and erosion, and that nutrient-rich broadleaf litter with superior water-holding and slow-release properties more effectively promotes aggregate formation and pore-network optimization than conifer needles, enhancing both soil water storage and the stabilization of soil organic carbon [58,59]. Complementarily, rapid litter decomposition stimulates microbial activity and the release of extracellular enzymes and organic acids, while soluble C and N sustain soil carbon turnover and microbial metabolism [60], together altering soil cementation, improving aggregate architecture, and ultimately fostering SOC stabilization [61]. When decomposition is vigorous and aeration increases, however, the risk of SOC mineralization rises [62], although the concomitant formation of larger pores and higher infiltration capacity can improve physical water retention [60].

In the mid and deep layers, SOCS responded more sensitively to soil structural indicators, and the carbon–water trade-off intensity in the 20–40 cm layer weakened with increasing TN but strengthened with higher BD (Figure 4), consistent with findings from arid and semi-arid regions [11,13,43]. Mechanistically, aggregates protect SOC from microbial decomposition and leaching through both physical occlusion and chemical stabilization [63]: microaggregates restrict enzyme–substrate contact, enhancing long-term carbon retention [64]. Aggregate reconfiguration reshapes the pore spectrum; reduced micropore fractions and increased surface hydrophobicity can weaken capillary water retention and promote preferential flow that accelerates percolation [65]. Declines in BD and expansion of macropores facilitate infiltration and water transport [66], yet may also elevate microbial metabolic entropy, lower carbon use efficiency, and increase the leaching risk of dissolved organic carbon [63], thereby leading to SOC losses. Thus, although structural improvement often coincides with co-benefits in SOCS and SWS [66], we observed that the two do not invariably change in tandem and that responses differ among forest types (Figure 3 and Figure S2). A plausible explanation is that litter inputs increase the proportion and stability of macroaggregates (>2 mm) [63] but concurrently reduce microaggregate abundance and capillary porosity in some contexts, thereby limiting plant-available water; in addition, high-C/N litter decomposes slowly, can induce nitrogen limitation, and may trigger microbial mining of native SOC, intensifying water use during carbon sequestration [60]. By contrast, native broadleaf litter is generally of higher quality and decomposes more rapidly, enabling coordinated gains in SOCS and SWS throughout the profile; deep root systems can also redistribute water across layers and enhance deep carbon inputs [43], together mitigating the carbon–water trade-off.

4.3. Soil Carbon–Water Trade-Offs and Influencing Factors

Because litter traits differ among forest types (Table S1) and water-use intensity varies with stand development [67], the direction of the soil carbon–water trade-off differs markedly across communities (Figure 3 and Figure S2), and the trade-off intensity reflects whether water conservation and soil carbon sequestration can be achieved synergistically during restoration [18]. Our results show that both the coupling direction and the strength of the SOCS–SWS relationship vary significantly with forest type and soil depth. In coniferous and mixed conifer–broadleaf stands, SOCS and SWS were negatively correlated (R² > 0.49, p < 0.05; Figure S2), and the trade-off intensified with stand age, whereas no significant correlation was detected in broadleaf forests. This pattern arises because increases in SOCS were concentrated in surface soils, while gains in SWS were more pronounced at depth. Previous studies indicate that litter and fine-root inputs and decomposition are concentrated near the surface [68], producing a surface-focused SOC profile [21,43]. In contrast, evaporation driven by solar radiation and air movement is stronger at the surface than at depth, which generates a stand-level negative association between SOCS and SWS (Figure S2). This finding echoes Li et al. [69], who observed a shift from positive to negative carbon–water coupling during apple-orchard development, suggesting that persistent deep-soil water depletion can ultimately suppress SOC accumulation. Consistently, Nan et al. reported a depth dependence of vegetation-recovery effects on SOCS and SWS [43], and Yang et al. showed that soil moisture deficits and SOCS exhibit distinct depth-specific dynamics within the same plant community [21].

During vegetation recovery, the relative benefits of SOCS and SWS also differed by forest type and soil depth. SRF and coniferous stands tended to show water-oriented gains, whereas broadleaf and mixed stands emphasized carbon gains; across forest types, carbon gains dominated in the 0–20 cm layer while water conservation became more prominent at depth (Figure 3), consistent with recent findings [20]. Such vertical dislocation of carbon–water benefits may promote SOCS accumulation in deeper horizons but risks undermining the sustainability of SWS. As communities succeed, increased returns of litter and fine roots can facilitate downward translocation and stabilization of SOC [70], even as deep-rooted species continue to pose potential threats to deep soil water. Multiple studies similarly stress that, although deep-rooted species can drive downward carbon inputs via litter and root exudates, they may reduce deep-soil water availability, microbial activity, and carbon turnover rates [71,72]. We further found lower trade-off intensity in broadleaf and mixed stands (Figure 3), indicating greater potential to co-achieve water conservation and carbon accumulation and providing a theoretical basis for selecting forest types with higher water-use efficiency.

At the surface, C–W showed only a weak positive association with ESC, whereas in the mid-layer (20–40 cm) it correlated negatively with SOCS and TN but positively with bulk density, and no significant drivers were detected at 40–60 cm (Figure 4). These patterns imply that where litter is more readily decomposed, increases in SOCS are accompanied by comparatively smaller SWS losses, and that organic matter inputs, nitrogen availability, and soil physical structure jointly shape carbon–water balance. This is in line with Wiesmeier et al. [73], Ding et al. [74], and Jin et al. [18]. The strengthening role of soil structure in mid-depth carbon–water processes also agrees with prior studies [73,75]. Moreover, under dry conditions, soil water becomes the primary constraint on carbon sequestration [13,76]; deep-soil water deficits can weaken plant–microbe interactions, decouple water–carbon linkages, and thereby affect ecosystem sustainability. Future work should focus on the long-term and seasonal dynamics of deep-soil carbon–water relationships across plant communities, explicitly integrating root distribution, water-use strategies, and microbial functional processes to elucidate the drivers of carbon–water coupling and to inform restoration and sustainable management in degraded ecosystems.

4.4. Uncertainties and Limitations

In this study, the pronounced effects of forest type and soil depth on SOCS and SWS may have masked potential influences of topography and soil texture (Figure 4, Figures S3 and S4;

Table 1. Results of LME for SOCS and SWS with forest type, soil depth, and their interaction, controlling for standardized topographic and soil texture covariates. Plot was included as a random effect.

Effect	NumDF	DenDF	SOCS		SWS		C-W
			F-Value	p-Value	F-Value	p-Value	F-Value	p-Value
Forest type	7	14	34.19	<0.001	1.24	0.31	0.91	0.51
Soil depth	2	30	79.31	<0.001	56.82	<0.001	33.69	<0.001
Elevation	1	14	2.29	0.14	0.82	0.37	0.17	0.68
Slope	1	14	1.49	0.23	1.64	0.21	0.69	0.41
Sand content	1	30	0.00	0.997	0.25	0.62	0.37	0.54
Silt content	1	30	1.98	0.17	3.22	0.08	0.03	0.85
Forest type × Depth	14	30	4.36	<0.001	3.57	<0.001	7.38	<0.001

Note: NumDF and DenDF refer to numerator and denominator degrees of freedom. Statistical significance was assessed at p < 0.05. The model accounts for plot-level variation as a random effect.

). Although previous work has reported significant effects of elevation and slope—and the associated climatic differences—on these variables [11], our study area exhibits low relief and limited variation in slope (Figure 1), with little climatic differentiation. Consequently, variation in SOCS and SWS was primarily governed by forest type and soil depth (Table 1). While SWS showed significant correlations with topography and texture in unstratified analyses (Figure S3), partial correlations indicated that, after controlling for forest type or for topographic and textural covariates, significant relationships remained only in the 40–60 cm layer, where SWS correlated with SOCS and the soil C/N ratio (Figure 4). This pattern suggests that SWS is regulated by interacting factors [11]; however, once the interaction between forest type and soil depth is considered, effects of topography and texture on SWS are not significant (Table 1), implying that forest type and depth exert stronger controls than terrain and texture. Even so, the spatial distribution of SWS remains sensitive to topographic variation [19]. Future work should therefore identify optimized community–topography configurations for vegetation restoration on the Yunnan Plateau and in comparable ecoregions.

This study focused exclusively on soil carbon–water services; aboveground carbon stocks, the whole-soil nitrogen pool, and understory biodiversity may follow different response trajectories and warrant parallel assessment [6,9,41]. In addition, sampling was limited to 0–60 cm and thus did not capture potential roles of deeper soil (>100 cm) in carbon–water regulation. Prior studies indicate that horizons below 1 m contain more than 50% of the stable SOC pool [20], and deep soil water can contribute 18%–40% of annual transpiration; in deep-rooted species [77], roots strongly influence deep-soil water use, carbon inputs, and aggregate stability [78]. Future research should incorporate deeper profiles and long-term monitoring to elucidate vertical linkages in carbon–water coupling and their climate sensitivity. Because this work was conducted on the central Yunnan Plateau, the conclusions are most applicable to similar subtropical restoration contexts; extrapolation to drier, wetter, or tropical systems should be made with caution and verified through multi-site, multi-year studies.

5. Conclusions

Using trade-off analysis and partial correlations, this study evaluated SOCS–SWS distributions and coupling within the 0–60 cm profile across SRF and representative forest types in the central Yunnan Plateau. The results suggest that, in water-limited regions comparable to our study area, increasing SOCS without intensifying water stress is best achieved by prioritizing low water use, high-litter-quality mixed conifer–broadleaf stands or native zonal broadleaf forests. By accelerating aggregate stabilization, moderately increasing soil N, and improving soil physical structure, these communities promote carbon–water synergies in surface soils, mitigate structural trade-offs at mid-depth, and avoid excessive water depletion in deeper layers, thereby improving the spatiotemporal coordination of SOCS and SWS. Future work should integrate multiple hydroclimatic and site contexts and combine multi-scale observations with modeling to elucidate feedback among litter quality, soil physicochemical properties, and SOCS–SWS trade-offs across forest types, providing more actionable evidence for restoration pathways that jointly enhance carbon sinks and water conservation.