Impacts of changes in vegetation on saturated hydraulic conductivity of soil in subtropical forests

Saturated hydraulic conductivity (Ks) is one of the most important soil properties that determines water flow behavior in terrestrial ecosystems. However, the Ks of forest soils is difficult to predict due to multiple interactions, such as anthropological and geomorphic processes. In this study, we examined the impacts of vegetation type on Ks and associated mechanisms. We found that Ks differed with vegetation type and soil depth, and the impact of vegetation type on Ks was dependent on soil depth. Ks did not differ among vegetation types at soil depths of 0–10 and 20–30 cm, but was significantly lower in managed forest types (mixed evergreen broad-leaved and coniferous forests, bamboo forests, and tea gardens) than native evergreen broadleaf forests at a depth of 10–20 cm. Boosted regression tree analysis indicated that total porosity, non-capillary porosity, and macro water-stable aggregates were the primary factors that influenced Ks. Our results suggested that vegetation type was a key factor that influences hydraulic properties in subtropical forest soils through the alteration of soil properties, such as porosity and macro water-stable aggregates.

specific soil parameters on the resulting K s is not always certain. Here, our objectives are to test: (1) whether differences in forest vegetation, resulting from changes in management objectives, affect soil K s across multiple soil depths and; (2) how changes in soil K s might be associated with the physicochemical attributes of soil. To address our first objective, we used analysis of variance to test the effect of forest vegetation types on soil K s . For the second objective, we used boosted regression tree (BRT) analysis, which resembles an additive regression model and can achieve higher accuracy and less bias in predictions than traditional multiple regression models 30 . In particular, BRT analysis is good for handling multi-collinearity concerns and violations of parametric assumptions 31,32 .

Results
Both vegetation type and soil depth affected soil K s and other characteristics ( Table 1). The effect of vegetation type on K s was significantly dependent on soil depth as indicated by the significant interaction effect between vegetation type and soil depth (P = 0.03). In the top soil layer (0-10 cm), soil K s was higher in bamboo forests and tea gardens than in native and mixed forests; in the deep soil layers (10-20 cm and 20-30 cm), native forests had significantly higher soil K s than other vegetation types, with the lowest values in tea gardens (Fig. 1).
The soil bulk density was impacted by changes in vegetation type and soil depth, but not by the interaction between vegetation type and soil depth (P < 0.001, P < 0.001 and P = 0.97, respectively). A similar trend was seen for total porosity and capillary porosity, which were impacted by vegetation type and soil depth, but not by their interaction. Total soil nitrogen was impacted considerably by changes in vegetation type with P = 0.01, while the impacts of soil depth and the interaction between vegetation type and soil depth were not significant. The non-capillary porosity of soil was impacted by the interaction between vegetation type and soil depth with P = 0.09, but not by these individual factors (P = 0.21 and P = 0.36, respectively). Meso and micro water-stable aggregates were impacted only by soil depth with P = 0.04 and P = 0.02, respectively (Table 1, Fig. 1).
Both root length density and root surface area density varied across vegetation types and soil depths. Root length density was impacted by changes in vegetation type, soil depth, and the interaction between them (P < 0.001, P < 0.001 and P = 0.04, respectively). Root surface area density was impacted by changes in vegetation type and soil depth (P = 0.04 and P < 0.001, respectively), but not by their interaction (P = 0.59). Root length density of bamboo forests was higher than other vegetation types in all soil depths, while root surface area density was lower in 0-10 cm soil depth. The root length density of bamboo forests in 0-10 cm and 10-30 cm was lower ( Table 1, Fig. 1).
Correlation analysis showed the correlation between all soil properties (Fig. 2). All soil properties contributed to differences in K s , with non-capillary porosity, total porosity, and macro water-stable aggregates exhibiting the greatest contributions (Fig. 3). BRT analysis indicated that non-capillary porosity, total porosity, and macro water-stable aggregates contributed 25.1%, 24.5%, and 16.8% of the BRT model explained variations in K s , respectively (Fig. 4). The other factors of capillary porosity, bulk density, total organic carbon, meso water-stable aggregates, and micro water-stable aggregates, had relatively minor contributions in this model (Fig. 4).

Discussions
Soil K s is a critical factor for plant growth that involves air-filled porosity, plant-available water, and so forth 33 . Hence, the improvement of K s is essential in order to avoid runoff and soil erosion 34 . As we anticipated, K s differed among the four vegetation types, with the disparities resulting primarily at the 10-30 cm soil depth. This result was similar to previous research, in which changes in vegetation type were shown to alter K s significantly 35  www.nature.com/scientificreports www.nature.com/scientificreports/ effects of vegetation type on soil K s were probably by means of root distribution and morphological characteristics such as root biomass and distribution in soil [36][37][38][39] . The root system affects soil texture via mechanical forces such as insertion or extrusion in soil 40 . Root length density and root surface area density both showed a decreasing trend with respect to depth in different vegetation types. In native forests and mixed forests, the main roots are obvious and the root system might extend to the lower depths of the layer, while in bamboo forests, the main roots are not obvious and the underground rhizome expands near the soil surface. In tea gardens, roots do not extend as deeply in soil compared to the other two types, nor do they expand like bamboo rhizome.
The roots distribution characteristics also affect the soil texture by adjusting litter input from the soil or its surface over time 41 , thus affecting soil organic carbon (SOC) and other soil physicochemical characteristics 3,4 . This stimulates belowground microbial biomass and rhizospheres 42,43 , and the effects of the physicochemical attributes of soil on K s via microbial community activities [44][45][46] . So to alter the physiochemical characteristics of soil, by means of producing solid, gas, and gel phases in order to adjust the fraction of the total spatial volume that is available for water flow, and hence the K s .
The effects of soil depth on soil K s differed with vegetation types though K s in native forests did not vary with soil depth. This is in alignment with prior researches [13][14][15] , which may have been the result of distinct vertical distributions of the physicochemical characteristics of soil and root distribution 14 . This might partly explain the increasing trend in soil depth from 10-20 cm to 20-30 cm. The decreasing root length and surface area densities were weakened by the effects of roots via mechanical forces or litter input characteristics. The higher K s in 0-10 cm might be attributed to the great probability in contacting a fresh litter of leaves or branches.
In this study, the variation of the K s value was higher in the tea garden. This may be attributed to the higher distribution density of the tea stems and the complexity of the root distribution underground, which could affect the K s value 28,37 . We also observed that the K s value at the 10-20 cm depth for the vegetation types other than native forests was lower than at the other soil depths. We attributed this to soil disturbances during vegetation conversion at that time, which might have had the effect of compacting the 0-20 cm depth layer. However, since the relation between K s and root distribution was not clarified in this study, we propose to explore this area further in future research.
In this study, total soil porosity, non-capillary porosity, and macro water-stable aggregates were the principal factors that influenced K s . A key parameter in this study was bulk density, from which the calculations on total soil porosity were derived. This was similar to a study in which differences in K s between samples were found to be correlated with bulk density and macro porosity 47 . The characteristics of pores in soils, such as their dimensions, distribution, and interconnections have been known to impact K s . It was found in many studies that lower www.nature.com/scientificreports www.nature.com/scientificreports/ bulk density was aligned with higher K s , and vice versa 48,49 , while water stable macro aggregates were positively correlated with K s 50 . It was found that the K s values were reduced in soils with smaller aggregates in contrast to those with large aggregates 12 . This was likely attributed to the impacts of different fresh organic matter, which were produced by different vegetation types 51 . Vegetation generated litter may simulate soil aggregation 52 , which subsequently influences bulk density and porosity 53,54 , while bulk density and porosity are closely correlated to adjustments in K s 55,56 . It remains a scientific challenge to describe in detail the complex continuous soil space 28 . However, we may conclude that soil pore characteristics are important factors.
In conclusion, our results show that change in vegetation type is a driving factor that strongly influences the hydraulic properties of soils in subtropical forests. Vegetation type, soil depth, and their interaction were observed to influence K s significantly, and the effects of soil depth on K s varied for different types of vegetation. The K s of native forests did not significantly differ at soil depths from 0-30 cm. For the other vegetation types, the K s at the 10-20 cm depth was significantly lower than that at 0-10 cm and 20-30 cm depths. There are multiple factors that impact K s ; however, total soil porosity, non-capillary porosity, and macro water-stable aggregates comprised the primary factors in this study. Soil K s is strongly influenced by changes in vegetation type, indicating that shifts in aboveground vegetation may strongly impact the water dynamics of soil. Based on our data, we suggest that the restoration of the native evergreen broad-leafed forests will assist in the retention and maintenance of soil hydrologic properties. Additional research will be required to confirm other factors and mechanisms that influence K s , such as the role of root systems and microbial communities in the processes that follow changes in vegetation species.  sampling. In June 2013, we randomly sampled five native evergreen broad-leaved forest stands, five mixed forest stands, five bamboo forest stands, and four tea garden stands at elevations ranging from 1250 to 1450 m, which resulted in a total of 19 sampled stands. All of the sample stands resided on well-drained mesic sites with slopes inclines of less than 5% to minimize the effects of inherent site conditions on soil characteristics 57,58 . In each stand, we established a sample plot of 20 × 20 m. Using a knife and a trowel, we extracted soil samples at depths of 0-10 cm, 10-20 cm, and 20-30 cm by digging a 15 × 15 cm section at each sampling point to enable the analysis of soil physicochemical characteristics 59,60 , which resulted in a total of 57 samples. For the determination of soil K s , bulk density, and capillary porosity analysis, we extracted soil samples with a metal corer (5.5 cm in diameter x 5 cm in height) at each sampling point 61 , which resulted in a total of 171 samples (57 samples for K s analysis, 57 samples for soil bulk density analysis and capillary porosity analysis, and 57 samples for other physicochemical properties).

Materials and Methods
To further understand how roots affect the impact of vegetation type on K s , we did supplementary sampling of soil with a metal corer (5.5 cm in diameter × 5 cm in height) in December 2018. Similar to the early sampling, we randomly sampled five native evergreen broad-leaved forest stands, four mixed forest stands, three bamboo forest stands, and three tea garden stands at elevations ranging from 1250 to 1450 m, which resulted in a total of 15 supplementary sampled stands.  www.nature.com/scientificreports www.nature.com/scientificreports/ saturated hydraulic conductivity measurements. K s was determined based on the constant hydraulic head method by imposing a stable hydraulic head to the top of the cores that were sampled at each of the sampling points, which were saturated with water prior to experiments in the laboratory 1 .

Analysis of soil physicochemical and roots properties.
Soil bulk density was determined by drying the samples in an oven at 105 °C until a constant weight was attained, and then adjusting for root and stone volume 58 . Soil samples for other physicochemical analyses were air-dried, sieved (2 mm mesh) in the laboratory, and then stored in air-tight plastic bags. Total organic carbon (TOC) content was measured using the sulfuric acid-potassium external heating method 62 . Total nitrogen and total phosphorus were simultaneously determined using a Bran + Luebbe Autoanalyser 3 Continuous Flow Analyzer (Bran + Luebbe GmbH, Norderstedt, Germany). Root length density and root surface area density were analyzed with the Win RHIZO root system (Regent Instruments, Québec, Canada). Before the analysis, all roots were washed out from the metal corer and then scanned with EPSON LA (Seiko Epson Corporation, Nagano-ken, Japan). soil porosity. The total porosity was calculated using the following equation 63 : where P t is the total soil porosity (%); 100 is the unit conversion factor; D b is the soil bulk density (g cm −1 ); and D p is the soil particle density (g cm −1 ), which was assumed to be 2.65 g cm −1 according to China's standard 64 . The soil capillary porosity was determined based on the water suction method, with the surface of the water located just below the tops of the soil cores 63 . Each soil core was initially weighed and placed onto a salver via filter paper until it attained a constant weight. Following weighing, the soil samples were allowed to drain completely under gravity. The soil samples were subsequently weighed again; their capillary water contents were determined by the differences in weight between the saturated and drained states.
where P c is the capillary porosity (%); P n is the non-capillary porosity (%); 100 is the unit conversion factor; W c is the soil capillary water content (%); V is the volume of the soil core (cm 3 ).
Water stable aggregate measurements. Water stable aggregate was measured using a routine wet-sieve method via a mechanical sieving procedure 65 . Briefly, for each soil sample, 200 g of air-dried soil was placed on a series of sieves to determine the dry aggregate size distribution (combined in three nest sizes in the order of >2 mm, 0.5-2 mm, and <0.5 mm) prior to wet-sieving. Subsequently, 50 g samples were prepared according to their dry-sieving percentages by the weight of aggregates at each size distribution for wet-sieving. The samples were immersed in water for 10 minutes and then placed under oscillation at 30 rpm for 30 min. The aggregate fractions that remained on each sieve were removed with aqua distillate into aluminum bins, to be oven-dried at 105 °C for 24 h. The aggregate fractions were then weighed to calculate the aggregate weights from each size class 58,66 .

Data analysis.
To examine the impact of land use type and soil depth on the K s and other soil characteristics, an analysis of variance (ANOVA) was performed following a split plot design, with soil layers nested within the sample plot. We modelled the fixed effects of vegetation type, soil layer, and their interaction on K s with plot as the random factor using maximum likelihood with the lme4 package 67 . ANOVA assumption tests were done with the lmerTest package 68 . Shapiro -Wilk's test was conducted. In this study, the Shapiro -Wilk's test involving capillary porosity and non-capillary porosity failed, so a Box-Cox transformation was performed by the following equation 69 : where V trans is the transformed value of capillary porosity or non-capillary porosity; V origin is original value of capillary porosity or non-capillary porosity; and λ is the parameter of box-cox. We used boosted regression tree analysis (BRT) to elucidate how K s was potentially affected by soil physicochemical characteristics. Furthermore, we examined Pearson's correlation between potential factors and K s to reduce the fitting predictors. Then we fitted all BRT models using the adjusted settings for ecological modeling: tree complexity = 5, learning rate = 0.0001, bag fraction = 0.7. All analyses were performed using BRT with the R package gbm 70 . www.nature.com/scientificreports www.nature.com/scientificreports/