Influence of Vegetation Restoration on Soil Hydraulic Properties in South China

Over the past several decades, vegetation restoration has been carried out extensively in South China. Theoretically, the process of vegetation restoration is usually accompanied by changes in soil properties. However, the effects of vegetation restoration on soil hydraulic properties are poorly documented in humid subtropical China. In this study, we compared soil hydraulic properties across three undisturbed subtropical forests, i.e., Pinus massoniana forest (PF), mixed Pinus massoniana/broad-leaved forest (MF), and monsoon evergreen broad-leaved forest (BF), which represented a vegetation restoration sequence in South China. Our results showed that vegetation restoration decreased the bulk density while increasing the total porosity and the soil organic matter (SOM). The clay content and capillary porosity of soil in the middleand late-recovery-stage forests were significantly higher than those in the early stage, which was consistent with the soil water-holding capacity. The saturated hydraulic conductivity (KS) values of BF were always significantly higher than those of the other forests. In the whole soil profile, the water-holding capacity and KS in the topsoil (above 30 cm depth) were significantly higher than those in the deep soil for all forests. Further analyses indicated that the SOM was the main factor that affected KS, and the relationship of them could be fitted by a linear equation. Overall, our study revealed vegetation restoration ameliorates soil hydraulic properties in humid subtropical China. And the role of SOM in improving soil hydraulic properties should be emphasized in future forest ecosystem management.


Introduction
Water is the key component of terrestrial ecosystems, and forests serve as a natural reservoir [1,2]. Studies have confirmed that the soil layer, as the main aquifer of forest ecosystems, plays an important role in the regulation of water movement in mountainous areas [3,4]. Soil hydraulic properties such as saturated hydraulic conductivity and soil water retention characteristics affect runoff generation, the patterns of infiltration, and water retention, etc. [5,6]. Thus, understanding the variability of soil hydraulic characteristics are of great significance to catchment water management, such as water conservation and soil erosion control [1,7,8].
Soil physicochemical properties (e.g., soil texture, bulk density, organic matter content) are the basis for the formation of soil hydraulic characteristics [5,6]. The dominant factors affecting soil hydraulic characteristics could differ vastly among different regions. Neris et al. found that the soil

Soil Sampling
Soil sampling was conducted between July and December 2019. Three random replicate plots with an area of 10 × 10 m were established in each forest. Three pits of 1.3 × 1 m were dug for soil collection in each plot. The soil layer was measured from 0 to 90 cm at intervals of 15 cm. Three undisturbed soil cores from each layer were collected with 100 cm 3 cylindrical metal cores for the measurement of the saturated hydraulic conductivity (KS, mm/min), bulk density (BD, g/cm 3 ) and soil porosity (%). Five points were established in the four corners and center of each plot to collect disturbed soil samples with a 3.5 cm diameter soil auger. The samples from the five points of each layer within a plot were fully mixed as a composite sample for the measurement of soil organic matter (SOM, g/kg), soil specific gravity (SSG) and soil particle size distribution (clay, silt and sand, %). Forest conservation of Dinghushan began in 1950 [21]. The forest has an area of 1156 ha and grows on a soil derived from sandstone and shale weathering. Soils are classified in the ultisol group and udult subgroup according to the soil classification system of United States Department of Agriculture (USDA) [32], highly acidic (pH 4-4.9) and rich in humus in the surface layer. The process of vegetation restoration in Dinghushan could be explained by Markov process [33]. Based on the vegetation composition characteristics of each recovery stage, Pinus massoniana forest (PF), mixed Pinus massoniana/broad-leaved forest (MF), and monsoon evergreen broadleaf forest (BF) could represent early, middle and advanced recovery stages, respectively. PF predominately occupies the periphery of the reserve, approximately 22 ha, with a single tree species, i.e., Pinus massoniana. MF, approximately 557 ha, is located between the central area and periphery of the reserve. BF is predominately located in the central area of the reserve, covering approximately 218 ha. The characteristics of the experimental sites are shown in Table 1.

Soil Sampling
Soil sampling was conducted between July and December 2019. Three random replicate plots with an area of 10 × 10 m were established in each forest. Three pits of 1.3 × 1 m were dug for soil collection in each plot. The soil layer was measured from 0 to 90 cm at intervals of 15 cm. Three undisturbed soil cores from each layer were collected with 100 cm 3 cylindrical metal cores for the measurement of the saturated hydraulic conductivity (K S , mm/min), bulk density (BD, g/cm 3 ) and soil porosity (%). Five points were established in the four corners and center of each plot to collect disturbed soil samples with a 3.5 cm diameter soil auger. The samples from the five points of each layer within a plot were fully mixed as a composite sample for the measurement of soil organic matter (SOM, g/kg), soil specific gravity (SSG) and soil particle size distribution (clay, silt and sand, %). Another three undisturbed soil cores were collected from 0-10, 10-20, 20-40, 40-60, and 60-100 cm soil depths in each pit for the measurement of a soil water retention curve (SWRC).

Measurements
The weights of cylindrical metal cores (m 0 ) were first recorded before soil sampling. Soil cores with fresh soil samples were placed in a plastic container, and water was added to the plastic container until the water level reached approximately 0.5 cm. The soil cores were kept steeped for 12 h. After that, the soil cores were placed on dry sand for 2 h at room temperature, and the resulting weight was recorded (m 1 ). Then, the soil cores were moved back to the plastic container, and water was added until the water level just reached the top of the soil cores. After the soil cores were kept saturated for 24 h, the K S was measured by the constant-head method based on Darcy's law. Initially, an empty cylinder of the same size was tightly secured to act as a reservoir, and a Mariotte bottle was used to keep a constant head ( 1, x FOR PEER REVIEW 4 of 16 g), soil specific gravity (SSG) and soil particle size distribution (clay, silt and sand, %). ree undisturbed soil cores were collected from 0-10, 10-20, 20-40, 40-60, and 60-100 cm in each pit for the measurement of a soil water retention curve (SWRC).
ements eights of cylindrical metal cores (m0) were first recorded before soil sampling. Soil cores soil samples were placed in a plastic container, and water was added to the plastic ntil the water level reached approximately 0.5 cm. The soil cores were kept steeped for 12 at, the soil cores were placed on dry sand for 2 h at room temperature, and the resulting s recorded (m1). Then, the soil cores were moved back to the plastic container, and water until the water level just reached the top of the soil cores. After the soil cores were kept or 24 h, the KS was measured by the constant-head method based on Darcy's law. Initially, ylinder of the same size was tightly secured to act as a reservoir, and a Mariotte bottle was p a constant head ( ⊿ h) in the core reservoirs [34]. Next, the water was allowed to from the upper surface of the soil sample, and the outflow water was caught using a le and weighed in 30-min intervals. Last, the steady-state flow, Q, was defined as a change ller than 0.05 g over five consecutive readings [5]. The KS at the experimental temperature ted using Equation (4). Finally, the soil cores were placed in an oven to dry at 105 °C until eight (m2). osite samples of approximately 1 kg were air-dried and sieved through a 2 mm sieve for ng measurements. The SSG and SOM contents were determined by the specific gravity od [35] and the K2Cr2O7·H 2SO4 wet oxidation method [36], respectively. The particle size n was measured by the laser diffraction technique using a Mastersizer 2000 (Malvern ts, Malvern, UK) [12]. According to the classification system of USDA, the soil particle size ied as sand (2-0.05 mm), silt (0.05-0.002 mm), and clay (<0.002 mm). The relevant formulas ws: CP ( ) BD 100 ( ) is the soil bulk density (g/cm 3 ); V is the volume of the cylindrical metal core (cm 3 ); TP is il porosity (%); CP is the capillary porosity (%); KS is the saturated hydraulic conductivity 10 is a unit conversion factor that converts the KS from centimeters per minute to s per minute; Q is the stable percolation volume of water (cm 3 ); L is the length of the sample he cross-sectional area of the sample (cm 2 ); ⊿h is the difference in the water head (cm); t is terval (min); and T is the experimental temperature (°C). K10 is the KS measured at 10 °C. KS denotes the saturated hydraulic conductivity at 10 °C. WRC was measured in the laboratory using a centrifuge (CR21G, Hitachi, Tokyo, Japan) the measurement, samples were first saturated in water for 24 h. Then, the soil samples ifuged at 20 °C from low speed to high speed in order and weighted at water balance. In we mainly measured the weights of the samples at pressure heads of 102, 204, 408, 612, 040, 4080, 6120, 8160, and 10,200 cm H2O. After that, the samples were oven-dried at 105 °C obtain the soil dry mass and to calculate the soil volumetric water content (cm 3 /cm 3 ). To he van Genuchten (VG) model was used to fit the data and to derive the VG equation for each sample [38]. According to the SWRC, the volumetric water content at pressure 0 and 15,000 cm H2O was calculated, representing the field water capacity (FWC, cm 3 /cm 3 ) h) in the core reservoirs [34]. Next, the water was allowed to flow down from the upper surface of the soil sample, and the outflow water was caught using a plastic bottle and weighed in 30-min intervals. Last, the steady-state flow, Q, was defined as a change in flux smaller than 0.05 g over five consecutive readings [5]. The K S at the experimental temperature was calculated using Equation (4). Finally, the soil cores were placed in an oven to dry at 105 • C until constant weight (m 2 ).
Composite samples of approximately 1 kg were air-dried and sieved through a 2 mm sieve for the following measurements. The SSG and SOM contents were determined by the specific gravity bottle method [35] and the K 2 Cr 2 O 7 ·H 2 SO 4 wet oxidation method [36], respectively. The particle size distribution was measured by the laser diffraction technique using a Mastersizer 2000 (Malvern Instruments, Malvern, UK) [12]. According to the classification system of USDA, the soil particle size was classified as sand (2-0.05 mm), silt (0.05-0.002 mm), and clay (<0.002 mm). The relevant formulas are as follows: Forests 2020, 11, x FOR PEER REVIEW 4 of (SOM, g/kg), soil specific gravity (SSG) and soil particle size distribution (clay, silt and sand, % Another three undisturbed soil cores were collected from 0-10, 10-20, 20-40, 40-60, and 60-100 c soil depths in each pit for the measurement of a soil water retention curve (SWRC).

Measurements
The weights of cylindrical metal cores (m0) were first recorded before soil sampling. Soil cor with fresh soil samples were placed in a plastic container, and water was added to the plas container until the water level reached approximately 0.5 cm. The soil cores were kept steeped for h. After that, the soil cores were placed on dry sand for 2 h at room temperature, and the resultin weight was recorded (m1). Then, the soil cores were moved back to the plastic container, and wat was added until the water level just reached the top of the soil cores. After the soil cores were ke saturated for 24 h, the KS was measured by the constant-head method based on Darcy's law. Initial an empty cylinder of the same size was tightly secured to act as a reservoir, and a Mariotte bottle w used to keep a constant head ( ⊿ h) in the core reservoirs [34]. Next, the water was allowed flow down from the upper surface of the soil sample, and the outflow water was caught using plastic bottle and weighed in 30-min intervals. Last, the steady-state flow, Q, was defined as a chan in flux smaller than 0.05 g over five consecutive readings [5]. The KS at the experimental temperatu was calculated using Equation (4). Finally, the soil cores were placed in an oven to dry at 105 °C un constant weight (m2).
Composite samples of approximately 1 kg were air-dried and sieved through a 2 mm sieve f the following measurements. The SSG and SOM contents were determined by the specific gravi bottle method [35] and the K2Cr2O7·H 2SO4 wet oxidation method [36], respectively. The particle si distribution was measured by the laser diffraction technique using a Mastersizer 2000 (Malve Instruments, Malvern, UK) [12]. According to the classification system of USDA, the soil particle si was classified as sand (2-0.05 mm), silt (0.05-0.002 mm), and clay (<0.002 mm). The relevant formul are as follows: where BD is the soil bulk density (g/cm 3 ); V is the volume of the cylindrical metal core (cm 3 ); TP the total soil porosity (%); CP is the capillary porosity (%); KS is the saturated hydraulic conductivi hAt (4) where BD is the soil bulk density (g/cm 3 ); V is the volume of the cylindrical metal core (cm 3 ); TP is the total soil porosity (%); CP is the capillary porosity (%); K S is the saturated hydraulic conductivity (mm/min); 10 is a unit conversion factor that converts the K S from centimeters per minute to millimeters per minute; Q is the stable percolation volume of water (cm 3 ); L is the length of the sample (cm); A is the cross-sectional area of the sample (cm 2 ); Forests 2020, 11, x FOR PEER REVIEW 4 of 16 (SOM, g/kg), soil specific gravity (SSG) and soil particle size distribution (clay, silt and sand, %). Another three undisturbed soil cores were collected from 0-10, 10-20, 20-40, 40-60, and 60-100 cm soil depths in each pit for the measurement of a soil water retention curve (SWRC).

Measurements
The weights of cylindrical metal cores (m0) were first recorded before soil sampling. Soil cores with fresh soil samples were placed in a plastic container, and water was added to the plastic container until the water level reached approximately 0.5 cm. The soil cores were kept steeped for 12 h. After that, the soil cores were placed on dry sand for 2 h at room temperature, and the resulting weight was recorded (m1). Then, the soil cores were moved back to the plastic container, and water was added until the water level just reached the top of the soil cores. After the soil cores were kept saturated for 24 h, the KS was measured by the constant-head method based on Darcy's law. Initially, an empty cylinder of the same size was tightly secured to act as a reservoir, and a Mariotte bottle was used to keep a constant head ( ⊿ h) in the core reservoirs [34]. Next, the water was allowed to flow down from the upper surface of the soil sample, and the outflow water was caught using a plastic bottle and weighed in 30-min intervals. Last, the steady-state flow, Q, was defined as a change in flux smaller than 0.05 g over five consecutive readings [5]. The KS at the experimental temperature was calculated using Equation (4). Finally, the soil cores were placed in an oven to dry at 105 °C until constant weight (m2).
Composite samples of approximately 1 kg were air-dried and sieved through a 2 mm sieve for the following measurements. The SSG and SOM contents were determined by the specific gravity bottle method [35] and the K2Cr2O7·H 2SO4 wet oxidation method [36], respectively. The particle size distribution was measured by the laser diffraction technique using a Mastersizer 2000 (Malvern Instruments, Malvern, UK) [12]. According to the classification system of USDA, the soil particle size was classified as sand (2-0.05 mm), silt (0.05-0.002 mm), and clay (<0.002 mm). The relevant formulas are as follows: h is the difference in the water head (cm); t is the time interval (min); and T is the experimental temperature ( • C). K 10 is the K S measured at 10 • C. Hereafter, K S denotes the saturated hydraulic conductivity at 10 • C.
The SWRC was measured in the laboratory using a centrifuge (CR21G, Hitachi, Tokyo, Japan) [37]. Before the measurement, samples were first saturated in water for 24 h. Then, the soil samples were centrifuged at 20 • C from low speed to high speed in order and weighted at water balance. In this study, we mainly measured the weights of the samples at pressure heads of 102, 204, 408, 612, 816, 1020, 2040, 4080, 6120, 8160, and 10,200 cm H 2 O. After that, the samples were oven-dried at 105 • C for 24 h to obtain the soil dry mass and to calculate the soil volumetric water content (cm 3 /cm 3 ). To this end, the van Genuchten (VG) model was used to fit the data and to derive the VG equation parameters for each sample [38]. According to the SWRC, the volumetric water content at pressure heads of 300 and 15,000 cm H 2 O was calculated, representing the field water capacity (FWC, cm 3 /cm 3 ) and the wilting water content (WWC, cm 3 /cm 3 ), respectively. The available water content (AWC, cm 3 /cm 3 ) was the difference between the FWC and the WWC. The relevant formulas are as follows: where θ is the volumetric water content (cm 3 /cm 3 ), θ S is the saturated water content (cm 3 /cm 3 ), θ r is the residual water content (cm 3 /cm 3 ), α is the scaling parameter related to the inverse of the air entry pressure (cm −1 ), n is the curve-shape parameter related to the pore size distribution, and h is the metric potential (cm H 2 O).

Statistical Analysis
The SWRC was fitted by RETention Curve software (Version6.0, University of California, Riverside, CA, USA). We calculated the basic statistical parameters, such as the mean and standard error of the soil properties. The coefficient of variation was calculated for the SWRC parameters. The primary statistical characteristic analysis was carried out using SPSS software (20.0). The normality of the soil properties was tested using the Kolmogorov-Smirnov test at the p = 0.05 significance level before statistical analysis. One-way analysis of variance (ANOVA) was applied to compare the differences in soil properties among various vegetation types and soil depths. When the ANOVA results were significant according to the F values, Duncan's test at p < 0.05 was performed to compare the means of the soil variables. Pearson's correlation analysis, multiple regression analysis and path analysis were conducted to investigate the relationships among soil properties.

Soil Organic Matter and Bulk Density
As shown in Figure 2a, the SOM content of each forest showed significant surface enrichment in the soil profile. The SOM content of the surface layer (15 cm) significantly increased along vegetation restoration with values ranging from 29.9 g/kg to 54.8 g/kg. The ratio of the subsurface (30 cm) value to the surface value was 49.1%, 40.9%, and 42.8% in BF, MF and PF, respectively. When averaged across the 0-90 cm depth, the SOM content was ranked as BF (25.3 g/kg) > MF (16.0 g/kg) > PF (12.1 g/kg). The soils collected from BF had significantly higher SOM contents than the others. and the wilting water content (WWC, cm 3 /cm 3 ), respectively. The available water content (AWC, cm 3 /cm 3 ) was the difference between the FWC and the WWC. The relevant formulas are as follows: where θ is the volumetric water content (cm 3 /cm 3 ), θS is the saturated water content (cm 3 /cm 3 ), θr is the residual water content (cm 3 /cm 3 ), α is the scaling parameter related to the inverse of the air entry pressure (cm −1 ), n is the curve-shape parameter related to the pore size distribution, and h is the metric potential (cm H2O).

Statistical Analysis
The SWRC was fitted by RETention Curve software (Version6.0, University of California, Riverside, CA, USA). We calculated the basic statistical parameters, such as the mean and standard error of the soil properties. The coefficient of variation was calculated for the SWRC parameters. The primary statistical characteristic analysis was carried out using SPSS software (20.0). The normality of the soil properties was tested using the Kolmogorov-Smirnov test at the p = 0.05 significance level before statistical analysis. One-way analysis of variance (ANOVA) was applied to compare the differences in soil properties among various vegetation types and soil depths. When the ANOVA results were significant according to the F values, Duncan's test at p < 0.05 was performed to compare the means of the soil variables. Pearson's correlation analysis, multiple regression analysis and path analysis were conducted to investigate the relationships among soil properties.

Soil Organic Matter and Bulk Density
As shown in Figure 2a, the SOM content of each forest showed significant surface enrichment in the soil profile. The SOM content of the surface layer (15 cm) significantly increased along vegetation restoration with values ranging from 29.9 g/kg to 54.8 g/kg. The ratio of the subsurface (30 cm) value to the surface value was 49.1%, 40.9%, and 42.8% in BF, MF and PF, respectively. When averaged across the 0-90 cm depth, the SOM content was ranked as BF (25.3 g/kg) > MF (16.0 g/kg) > PF (12.1 g/kg). The soils collected from BF had significantly higher SOM contents than the others.
In general, the BD increased with increasing soil depth (Figure 2b). The BD of BF, MF and PF increased from 1.2 g/cm 3 , 1.3 g/cm 3 , and 1.4 g/cm 3 at the 0-15 cm depth to 1.4 g/cm 3 , 1.5 g/cm 3 , and 1.6 g/cm 3 , respectively, at the 75-90 cm depth. The value of the surface soil was significantly lower than that of other soil layers. When averaged across the 0-90 cm depth, there was a significant difference in the BD between forests, i.e., BF (1.3 g/cm 3 ) < MF (1.4 g/cm 3 ) < PF (1.5 g/cm 3 ). Values are the means ± SE. Lowercase letters above the columns represent statistically significant differences among stand types for the same soil layer (Duncan's test, p < 0.05). Abbreviations: SOM, soil organic matter; BD, bulk density; PF, Pinus massoniana forest; MF, mixed Pinus massoniana/broad-leaved forest; BF, monsoon evergreen broad-leaved forest. In general, the BD increased with increasing soil depth (Figure 2b). The BD of BF, MF and PF increased from 1.2 g/cm 3 , 1.3 g/cm 3 , and 1.4 g/cm 3 at the 0-15 cm depth to 1.4 g/cm 3 , 1.5 g/cm 3 , and 1.6 g/cm 3 , respectively, at the 75-90 cm depth. The value of the surface soil was significantly lower than that of other soil layers. When averaged across the 0-90 cm depth, there was a significant difference in the BD between forests, i.e., BF (1.3 g/cm 3 ) < MF (1.4 g/cm 3 ) < PF (1.5 g/cm 3 ).

Soil Particle Composition
As shown in Figure 3, there was no significant difference in all levels of particles among the soil layers in PF. When averaged across the 0-90 cm depth, the proportion of sand was significantly higher (50.3%), and the silt and clay contents were significantly lower (26.4% and 23.3%, respectively) in PF than in the other forests. The sand content of MF decreased from 48.4% to 27.7% as soil depth increased, which was significantly lower than that of PF. The average silt content of MF (28.0%) was similar to that of PF for the 0-90 cm depth. The change trend of the clay content of MF in the vertical section was opposite to that of the sand content, with a mean value of 35.6%, similar to BF (34.4%). When averaged across the 0-90 cm depth, the sand content of BF (30.6%) was significantly lower than that of other forests, and the total proportion of the silt and clay was 69.4%. Values are the means ± SE. Lowercase letters above the columns represent statistically significant differences among stand types for the same soil layer (Duncan's test, p < 0.05). Abbreviations: SOM, soil organic matter; BD, bulk density; PF, Pinus massoniana forest; MF, mixed Pinus massoniana/broadleaved forest; BF, monsoon evergreen broad-leaved forest.

Soil Particle Composition
As shown in Figure 3, there was no significant difference in all levels of particles among the soil layers in PF. When averaged across the 0-90 cm depth, the proportion of sand was significantly higher (50.3%), and the silt and clay contents were significantly lower (26.4% and 23.3%, respectively) in PF than in the other forests. The sand content of MF decreased from 48.4% to 27.7% as soil depth increased, which was significantly lower than that of PF. The average silt content of MF (28.0%) was similar to that of PF for the 0-90 cm depth. The change trend of the clay content of MF in the vertical section was opposite to that of the sand content, with a mean value of 35.6%, similar to BF (34.4%). When averaged across the 0-90 cm depth, the sand content of BF (30.6%) was significantly lower than that of other forests, and the total proportion of the silt and clay was 69.4%.

Soil Pore Distribution
As shown in Figure 4, the TP decreased with increasing soil depth, and the corresponding value of surface soil was significantly higher than that of the other soil layers. The TP of BF was always significantly higher than that of the other forests within the 0-90 cm layer, and there was no significant difference in the TP of MF and PF in the soil layers below 60 cm. When averaged across the 0-90 cm depth, the TP was 49.5%, 44.7%, and 41.7% in BF, MF, and PF, respectively.
The CP was the main component of soil porosity in the three forests. The average CP/TP values were 0.82, 0.88, and 0.80 in BF, MF, and PF, respectively. The CP of all soil layers in PF was significantly lower than that in the other forests, and the corresponding values of BF and MF were similar. In general, the CP of BF and MF tended to decrease as the soil depth increased, and the CP of the 15-30 cm soil layer in BF was the highest (42.6%). When averaged across the 0-90 cm depth, the CP values were ranked as BF (40.3%) > MF (39.3%) > PF (33.0%). The mean NCP of MF (5.4%) was the lowest among the forests at the 0-90 cm depth, while the mean NCP of PF (8.6%) was close to that of BF (9.2%).

Soil Pore Distribution
As shown in Figure 4, the TP decreased with increasing soil depth, and the corresponding value of surface soil was significantly higher than that of the other soil layers. The TP of BF was always significantly higher than that of the other forests within the 0-90 cm layer, and there was no significant difference in the TP of MF and PF in the soil layers below 60 cm. When averaged across the 0-90 cm depth, the TP was 49.5%, 44.7%, and 41.7% in BF, MF, and PF, respectively.    The CP was the main component of soil porosity in the three forests. The average CP/TP values were 0.82, 0.88, and 0.80 in BF, MF, and PF, respectively. The CP of all soil layers in PF was significantly lower than that in the other forests, and the corresponding values of BF and MF were similar. In general, the CP of BF and MF tended to decrease as the soil depth increased, and the CP of the 15-30 cm soil layer in BF was the highest (42.6%). When averaged across the 0-90 cm depth, the CP values were ranked as BF (40.3%) > MF (39.3%) > PF (33.0%). The mean NCP of MF (5.4%) was the lowest among the forests at the 0-90 cm depth, while the mean NCP of PF (8.6%) was close to that of BF (9.2%).

Soil Water-Holding Characteristics
SWRC (θ-h relationships) was determined at pressure heads from 102 to 10,200 cm H 2 O. The measured data were well fitted by the VG model, which had high determination efficiencies of more than 99%. The data in Figure 5 show that BF and MF had significantly larger moisture retention capacities of soil at any given pressure head compared to PF, especially at the 0-40 cm depth. There was no significant difference in the water-holding capacity between forests at the 40-100 cm depth.

Soil Water-Holding Characteristics
SWRC (θ-h relationships) was determined at pressure heads from 102 to 10,200 cm H2O. The measured data were well fitted by the VG model, which had high determination efficiencies of more than 99%. The data in Figure 5 show that BF and MF had significantly larger moisture retention capacities of soil at any given pressure head compared to PF, especially at the 0-40 cm depth. There was no significant difference in the water-holding capacity between forests at the 40-100 cm depth. The parameters (α, n) of SWRC in the different forests varied with the soil layer ( Table 2). The mean α value of PF (0.005) was higher than that of BF (0.003) and MF (0.002). The coefficient of variation for α between forests increased in the order MF (0.  The parameters (α, n) of SWRC in the different forests varied with the soil layer ( Table 2). The mean α value of PF (0.005) was higher than that of BF (0.003) and MF (0.002). The coefficient of variation for α between forests increased in the order MF (0.203) < BF (0.211) < PF (0.248). There was little change in the range of n values between the different soil layers of each forest, and the coefficient of variation ranged between 0.013 and 0.014. The mean n value among forests showed a decreasing sequence of PF (1.193) > MF (1.174) > BF (1.171).
As shown in Table 3, when averaged across the 0-100 cm depth, the θ S between forests ranked as BF (0. 43

Saturated Hydraulic Conductivity
As shown in Figure 6, the K S values in different forests decreased as the soil depth increased, with ranges of 0.37 to 1.57 mm/min, 0.04 to 0.86 mm/min, and 0.03 to 0.24 mm/min in BF, MF, and PF, respectively. Across all forests, the K S values in the topsoil were significantly higher than those in the subsoil, and the K S values tended to be stable in the soil layers below 30 cm or 45 cm. For the same soil layer among different forests, the K S values of BF were always significantly higher than those of the other forests. There were significant differences in the K S values in the topsoil between forests, i.e., BF (1.57 mm/min) > MF (0.86 mm/min) > PF (0.24 mm/min). However, there was no significant difference in the K S values between MF and PF below the surface layer. When averaged across the 0-90 cm depth, the K S value of BF (0.75 mm/min) was significantly higher than that of the other forests and was 6.94 times that of PF. The K S values of MF and PF were 0.24 and 0.11 mm/min, respectively.  Table 4 shows the relationship between different soil properties. The soil water-holding characteristics of the above three forests were mainly affected by pore distribution and texture. The KS was closely related to the soil porosity, silt content, BD and SOM. Furthermore, the absolute value   Table 4 shows the relationship between different soil properties. The soil water-holding characteristics of the above three forests were mainly affected by pore distribution and texture. The K S was closely related to the soil porosity, silt content, BD and SOM. Furthermore, the absolute value of the correlation coefficients showed a decreasing sequence of BD (0.905) > SOM (0.904) >TP (0.878) > CP (0.638) > silt (0.538) > NCP (0.504). Due to the strong interaction between soil properties, multiple stepwise regression analysis was used to select the optimal factors that influenced the K S . Eight factors, sand (X 1 ), silt (X 2 ), clay (X 3 ), BD (X 4 ), TP (X 5 ), CP (X 6 ), NCP (X 7 ), and SOM (X 8 ), were taken as the independent variable factors, and K S was the dependent variable Y. When the independent variable factors silt (X 2 ) and SOM (X 8 ) were included, the model had the highest coefficient of determination (R 2 ) of 0.92, which was statistically significant. The regression equation was expressed as:

Relationship between Soil Properties and K S
where Y denotes K S (mm/min). This indicated that SOM and silt contents were the main drivers of K S in this study area. To determine the direct and indirect effects of the above two soil properties on K S , the path analysis method was further used for analysis. The residual path coefficient was 0.28, and the Durbin Watson statistic was 1.16, indicating that the result of path analysis was reliable. As shown in Figure 7, the direct path coefficient of SOM (0.82) for K S was greater than that of the silt content (0.32), while the indirect path coefficient of SOM for K S based on the silt content was 0.08. Figure 6. Distribution of KS along the soil profile (a) and the differences in average KS in the different forests (b) in Dinghushan. Values are the means ± SE. Lowercase letters above the columns represent statistically significant differences among stand types (Duncan's test, p < 0.05). Abbreviations: KS, saturated hydraulic conductivity; PF, Pinus massoniana forest; MF, mixed Pinus massoniana/broadleaved forest; BF, monsoon evergreen broad-leaved forest. Table 4 shows the relationship between different soil properties. The soil water-holding characteristics of the above three forests were mainly affected by pore distribution and texture. The KS was closely related to the soil porosity, silt content, BD and SOM. Furthermore, the absolute value of the correlation coefficients showed a decreasing sequence of BD (0.905) > SOM (0.904) >TP (0.878) > CP (0.638) > silt (0.538) > NCP (0.504). Due to the strong interaction between soil properties, multiple stepwise regression analysis was used to select the optimal factors that influenced the KS. Eight factors, sand (X1), silt (X2), clay (X3), BD (X4), TP (X5), CP (X6), NCP (X7), and SOM (X8), were taken as the independent variable factors, and KS was the dependent variable Y. When the independent variable factors silt (X2) and SOM (X8) were included, the model had the highest coefficient of determination (R 2 ) of 0.92, which was statistically significant. The regression equation was expressed as: X X Y 8 2 0.026 0.031 1.029

Relationship between Soil Properties and KS
where Y denotes KS (mm/min).
This indicated that SOM and silt contents were the main drivers of KS in this study area. To determine the direct and indirect effects of the above two soil properties on KS, the path analysis method was further used for analysis. The residual path coefficient was 0.28, and the Durbin Watson statistic was 1.16, indicating that the result of path analysis was reliable. As shown in Figure 7, the direct path coefficient of SOM (0.82) for KS was greater than that of the silt content (0.32), while the indirect path coefficient of SOM for KS based on the silt content was 0.08.   α indicates scaling parameters related to the inverse of the air entry pressure, n indicates curve-shape parameters related to pore size distribution and θ S indicates the saturated water content. Abbreviations: K S , saturated hydraulic conductivity; FWC, field water content; WWC, wilting water content; AWC, available water content; TP, total porosity; CP, capillary porosity; NCP, noncapillary porosity; BD, bulk density; SOM, soil organic matter; ** p < 0.01; * p < 0.05.

Discussion
Soil development is closely related with topography, parent rock, vegetation, climate, and time [39]. The three undisturbed forests mentioned above are similar in topography and elevation, and the soil all develop on the same parent rock [28,40], so the difference in soil properties have mainly been affected by the biome [31]. Along with the process of natural vegetation restoration, the SOM content significantly increased, which has been confirmed by a large number of studies [6,15,41]. The SOM of forests mainly derives from litterfall input, and the accumulation of SOM is beneficial to the improvement of the soil physical structure [11,42,43]. The increase in SOM promotes microbial activity and the growth of the root system. The physical interpenetration of root growth is conducive to the development of soil pores, and the chemical conditions created by the exudation of organic acids by roots and microorganisms promote the decomposition of soil particles, which ultimately lead to a decrease in BD [44,45]. Along the soil profile, the BD of surface soil in different forests was significantly lower than that of other soil layers. A possible explanation for this was that the large amount of organic matter from decomposing litter first returned to the surface layer, and the improvement in the surface soil structure was the strongest. In the deep layer, the SOM decreased greatly, and the soil compaction caused by deep root growth increased the BD [10].
In this study, soil particles developed towards fine grains during the forest recovery process, consistent with the results of Błońska et al. [7] and Oktavia et al. [46]. It was worth noting that the clay content of soil in BF was similar to that in MF. The alternation of dry and wet seasons here was beneficial to the leaching of soil colloids, which drove the migration of clay particles from topsoils to deep soils. Soil particle composition was one of the main factors that determined the soil pore structure. Analysis data showed that the CP was significantly correlated with all soil particles in this area. Moreover, the CP decreased with deeper soil layers in different forests, which might also be related to the variation trend of SOM along the soil profile. The increase in SOM was conducive to soil agglomeration, which promoted soil pore formation [11]. Generally, the transformation of the soil structure would affect the soil water-holding capacity, and then the soil moisture condition changes. The soil water-holding capacity was highly correlated with the clay particles, and could be indicated by the parameter α in the VG model [47]. In this study, the correlation coefficient between CP, clay and the α value was 0.75 (p < 0.01) and 0.53 (p < 0.05), respectively. The improvement of the local soil water-holding capacity was mainly realized by the increasing quantity of clay particles and CP. It could be seen from the water-holding characteristics between forests that the water storage potential of the middle-and late-recovery-stage forests increased significantly, providing a stable water environment for vegetation growth. A good soil moisture environment might be one of the important factors promoting vegetation development in this area. Overall, vegetation restoration significantly improved the soil water storage capacity and stable soil water supply capacity, which was also reported by Owuor et al. [3] and Li et al. [42].
Vegetation restoration also gradually increased the K S , which was consistent with the research results of Hassler et al. [48], Leite et al. [1] and Li et al. [42]. The SOM and silt particles were the main factors that affected K S in this region, and the direct effect of SOM on K S was the strongest. Zema et al. also reported that the SOM was one of the key parameters in driving the soil hydraulic characteristics of pure even-aged Spanish black pine stands [6]. In general, SOM affected the hydraulic properties by improving the quality of the soil colloids, and the condition and volume of soil pores played a controlling role in water permeability [47,49,50]. A study by Hao et al. showed that K S was mainly affected by soil porosity and water-stable aggregates in subtropical forests [51]. In this study, the correlation coefficient between TP and K S was 0.878 at a significant level. We assumed that the SOM content were conducive to the reduction of BD and thereby had an effect on K S by increasing the soil porosity. Except for the surface layer, the K S of MF in the other soil layers was higher than that of PF, but the difference was not significant, which might be related to the gradual decrease of SOM content below the surface layer of MF. A low organic matter content beneath the surface soil led to poor stability of the soil structure and high dispersion of soil particles, easily forming a dense and thick shell during water flow [41,52]. In addition, the proportion of CP in MF was the highest among forests, and the clay particles gradually moved down and blocked the pores in the process of water flow, which greatly weakened the permeability of the soil [10,53].
The improvement of the K S reduced surface runoff and erosion, thus effectively replenishing soil moisture and groundwater resources [3,6]. It could be seen from Figure 8 that the accumulation rate of SOM was fast before the middle recovery stage and then slowed down. The K S increased with the increase in SOM, and the relationship of them could be fitted by a linear function (y = −0.149 + 0.029x, R 2 = 0.815). We could predict that the regional eco-hydrological benefits will gradually improve with the maturation of the stands planted in Guangdong in 1980s, which is of positive significance for the response to the extreme precipitation pattern in the future [21]. MF was the highest among forests, and the clay particles gradually moved down and blocked the pores in the process of water flow, which greatly weakened the permeability of the soil [10,53]. The improvement of the KS reduced surface runoff and erosion, thus effectively replenishing soil moisture and groundwater resources [3,6]. It could be seen from Figure 8 that the accumulation rate of SOM was fast before the middle recovery stage and then slowed down. The KS increased with the increase in SOM, and the relationship of them could be fitted by a linear function ( x y 0.029 0.149 + − = , R 2 = 0.815). We could predict that the regional eco-hydrological benefits will gradually improve with the maturation of the stands planted in Guangdong in 1980s, which is of positive significance for the response to the extreme precipitation pattern in the future [21]. Overall, vegetation restoration had significant effects on the amelioration of soil quality. The increase amount of litter return played an important role in leading to such a change. Thus, more efforts should be taken to protect the forest litter. Furthermore, the transformation of soil hydraulic performance by vegetation depended on the change in soil depth. For surface soil layer, the soil waterholding capacity and KS improved obviously when the vegetation was restored to the middle stage. Therefore, the form of mixed forests should be further emphasized in future afforestation practice in order to improve the eco-hydrological benefits in South China.

Conclusions
Vegetation restoration has continuously improved the soil physicochemical properties and hydraulic characteristics in South China. The continuous accumulation of SOM has decreased the bulk density and improved the total porosity and soil hydraulic properties. The clay content and capillary porosity of soil in the middle-and late-recovery-stage forests were significantly higher than those in the early stage, which was consistent with the soil water-holding capacity. The KS values of BF were always significantly higher than those of the other forests. In the whole soil profile, the waterholding capacity and KS in the topsoil (above 30 cm depth) were significantly higher than those in the deep soil. Further analyses indicated that the SOM was the main driver of KS, and the relationship of them could be fitted by a linear equation. Overall, vegetation restoration gradually ameliorated the soil hydraulic properties alongside soil structural improvement, especially the surface layer. The SOM played a key role in improving soil quality in South China. The results of the study could provide sufficient strategies for eco-environment rehabilitation and forest management.  Overall, vegetation restoration had significant effects on the amelioration of soil quality. The increase amount of litter return played an important role in leading to such a change. Thus, more efforts should be taken to protect the forest litter. Furthermore, the transformation of soil hydraulic performance by vegetation depended on the change in soil depth. For surface soil layer, the soil water-holding capacity and K S improved obviously when the vegetation was restored to the middle stage. Therefore, the form of mixed forests should be further emphasized in future afforestation practice in order to improve the eco-hydrological benefits in South China.

Conclusions
Vegetation restoration has continuously improved the soil physicochemical properties and hydraulic characteristics in South China. The continuous accumulation of SOM has decreased the bulk density and improved the total porosity and soil hydraulic properties. The clay content and capillary porosity of soil in the middle-and late-recovery-stage forests were significantly higher than those in the early stage, which was consistent with the soil water-holding capacity. The K S values of BF were always significantly higher than those of the other forests. In the whole soil profile, the water-holding capacity and K S in the topsoil (above 30 cm depth) were significantly higher than those in the deep soil. Further analyses indicated that the SOM was the main driver of K S , and the relationship of them could be fitted by a linear equation. Overall, vegetation restoration gradually ameliorated the soil hydraulic properties alongside soil structural improvement, especially the surface layer. The SOM played a key role in improving soil quality in South China. The results of the study could provide sufficient strategies for eco-environment rehabilitation and forest management.  Acknowledgments: Special thanks to the Innovative and Entrepreneurial Training Program for College Students of South China Agricultural University and Guangdong Key Laboratory for Innovative Development and Utilization of Forest Plant Germplasm for their support. We also thank Ge Sun for his comments on the early version of the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.  10 the saturated hydraulic conductivity measured at 10 • C VG van Genuchten ANOVA One-way analysis of variance