Hydraulic Properties in Di ﬀ erent Soil Architectures of a Small Agricultural Watershed: Implications for Runo ﬀ Generation

: Soil architecture exerts an important control on soil hydraulic properties and hydrological responses. However, the knowledge of hydraulic properties related to soil architecture is limited. The objective of this study was to investigate the inﬂuences of soil architecture on soil physical and hydraulic properties and explore their implications for runo ﬀ generation in a small agricultural watershed in the Three Gorges Reservoir Area (TGRA) of southern China. Six types of soil architecture were selected, including shallow loam sandy soil in grassland (SLSG) and shallow loam sandy soil in cropland (SLSC) on the shoulder; shallow sandy loam in grassland (SSLG) and shallow sandy loam in cropland (SSLC) on the backslope; and deep sandy loam in grassland (DSLG) and deep sandy loam in cropland (DSLC) on the footslope. The results showed that saturated hydraulic conductivity ( K sat ) was signiﬁcantly higher in shallow loamy sand soil under grasslands (8.57 cm h − 1 ) than under croplands (7.39 cm h − 1 ) at the topsoil layer. Total porosity was highest for DSLC and lowest for SSLG, averaged across all depths. The proportion of macropores under SLSG was increased by 60% compared with under DSLC, which potentially enhanced water inﬁltration and decreased surface runo ﬀ . The landscape location e ﬀ ect showed that at the shoulder, K sat values were 20% and 47% higher than values at the backslope and footslope, respectively. It was inferred by comparing K sat values with 30 min maximum rainfall intensity at the watershed, that surface runo ﬀ would be generated in SSLC, DSLG, and DSLC sites by storms, but that no overland ﬂow is generated in both sites at the shoulder and SSLG. The signiﬁcantly higher K sat under grasslands in comparison to croplands at the backslope indicated that planting grasses would increase inﬁltration capacity and mitigate runo ﬀ generation during storm events. The ﬁndings demonstrated that croplands in footslope positions might be hydrologically sensitive areas in this small agricultural watershed. , The results of our study may provide a novel insight into the prediction of hydrological processes as a result of landscape evolution. also revealed that on a granitic soil hillslope, decreased water inﬁltration and enhanced runo ﬀ was predicted at the footslope position, while on the shoulder, the ﬂow pathway was dominated by subsurface ﬂow and deep percolation into the soil proﬁle. Di ﬀ erent hydrological responses to landscape position and management are important for further hydrological modeling in the mountainous watershed.


Introduction
The knowledge of soil hydraulic properties, including hydraulic conductivity and water retention, is of great importance for understanding hydrological responses such as water infiltration, surface runoff, and water storage. Soil architecture refers to the organization of soil from the microscopic to the megascopic scales and encompasses three interlinked components (solid components, pore space, and their interfaces) at each scale [1,2]. It is commonly recognized by two general categories: (1) soil architecture within a soil profile, such as mineral structure, aggregates, soil horizons, and pedons; and (2) soil architecture in the landscape including soil catena, soilscape, soil sequence, and pedosphere. Soil architecture is often a reflection of natural soil-forming processes affected by parent material, The reference soil profiles were excavated to the C horizon. The definition of soil thickness in this study refers to the measured thickness of the soil above the parent material (C) horizon. Undisturbed soil cores (5 cm diameter and 5 cm height) were independently sampled in 10 cm increments up to a 40 cm soil depth for each site with three replicates (180 total soil sample cores). Soil core samples were labeled, carefully trimmed, sealed in self-sealing plastic bags, and refrigerated at 4 °C until they were analyzed. Soil bulk density and saturated hydraulic conductivity (Ksat) were The reference soil profiles were excavated to the C horizon. The definition of soil thickness in this study refers to the measured thickness of the soil above the parent material (C) horizon. Undisturbed soil cores (5 cm diameter and 5 cm height) were independently sampled in 10 cm increments up to a 40 cm soil depth for each site with three replicates (180 total soil sample cores). Soil core samples were labeled, carefully trimmed, sealed in self-sealing plastic bags, and refrigerated at 4 • C until they were analyzed. Soil bulk density and saturated hydraulic conductivity (K sat ) were determined for these Water 2019, 11, 2537 4 of 18 soil cores in the laboratory. Undisturbed soil core subsamples were also collected to measure water retention properties. After the collection of soil cores, disturbed soil samples were also collected in plastic bags at the same location to determine the organic carbon content and soil texture. All samples were collected from July to August, 2017. Basic soil properties of the selected sites are shown in Table 2.  SSLG1  backslope  grassland  3  30  292  2  SSLG2  backslope  grassland  3  35  307  1  SSLC  SSLC1  backslope  cropland  3  25  277  7  SSLC2  backslope  cropland  3  30  294  1  SSLC3  backslope  cropland  3  28  280  5  DSLG  DSLG1  footslope  grassland  3  >100  252  3  DSLG2  footslope  grassland  3  90  246  2  DSLG3  footslope  grassland  3  85  244  2  DSLC  DSLC1  footslope  cropland  3  86  253  1  DSLC2  footslope  cropland  3  >100  221  1  DSLC3  footslope  cropland  3  82  241  3 Note: SLSG: shallow loam sandy-shoulder-grassland; SLSC: shallow loam sandy-shoulder-cropland; SSLG: shallow sandy loam-grassland-backslope; SSLC: shallow sandy loam-cropland-backslope; DSLG: deep sandy loam-grassland-footslope; DSLC: deep sandy loam-cropland-footslope.

Laboratory Analysis
The constant head method, as described by [31], determined K sat . Soil cores were covered with cheesecloth at the bottom and then saturated gradually in tubs from the lower end for approximately 48 h. Then, another same-sized empty core cylinder was tightly fixed on the top of the soil core as a reservoir, and a constant head to cylinder reservoir was maintained using a Mariotte bottle. A graduated cylinder was used to measure the volume of water eluted from the sample core at 2 min intervals until a steady flow rate was reached. The water temperature (T, • C) during the experiment was also recorded. Equation (1) was used to calculate the saturated hydraulic conductivity at the experimental temperature (K T ). To make comparisons of hydraulic conductivity at different temperatures, K T was converted to saturated hydraulic conductivity at 10 • C (K sat ) using Equation (2) [31].
where K T is the saturated hydraulic conductivity measured at temperature T (cm h −1 ), 60 is a conversion factor that changes the unit of K sat from centimeters per minute to centimeters per hour, Q is the volume of percolating water (cm 3 ), L is the height of the soil sample (5 cm), S is the soil column's cross-sectional area (19.6 cm 2 ), h is the thickness of water layer (5 cm), and t is the time interval of data recording (min). Finally, the undisturbed soil cores were weighed and oven-dried at 105 • C for 24 h and then weighed again to measure the final water content and dry soil bulk density. The wet sieving-pipette method was used to determine soil texture (Gee and Or 2002). The soil organic carbon content was measured by the potassium-dichromate (K 2 Cr 2 O 7 ) digestion method [32].
The water retention properties of different soil architectures were determined using the centrifuge methods [33], performed on undisturbed soil core subsamples. For every soil architecture, samples were initially saturated for 24 h and then weighed to measure the soil saturated water content before applying centrifuge methods. The samples were slowly drained using nine water potential steps. After each step, the soil samples were weighed and returned to the apparatus to apply to a higher speed of rotation. This procedure was repeated until the last established water potential (−1500 kPa) was finished. Subsequently, the samples were dried in the oven at 105 • C for 24 h until a constant weight was reached. The soil water contents were fitted to the van Genuchten model [34] for water retention curve fitting: where θ is the volumetric water content (cm 3 cm −3 ); θ r and θ s are the fitted residual volumetric water content (cm 3 cm −3 ) and measured saturated volumetric water content (cm 3 cm −3 ), respectively; h is the water pressure head (cm); and α and n are both shape parameters (cm −1 and dimensionless, respectively) obtained from the program RETC [35]. Effective pore size distributions were estimated from soil water retention data using the capillary rise equation [36]: where r represents the effective pore radius (mm), σ the water surface tension (M L −2 ), β the contact angle between the solid phase and the water-air interface, ρ the density of water (M L −3 ), and g the gravitational constant (L T −2 ). Soil pores were classified into four size fractions: macropores (effective diameter >1000 µm), coarse mesopores (effective diameter 60 to 1000 µm), fine mesopores (effective diameter 10 to 60 µm), and micropores (effective diameter <10 µm), as mentioned in [37]. Total porosity was equal to the saturated water content at 0 kPa water potential.

Statistical Analysis
All statistical analyses were performed with IBM SPSS program (Version 21.0, SPSS Inc., Chicago, IL, USA), and the figures were depicted by OriginPro 2016 and Microsoft Excel 2013. Kolmogorov-Smirnov (K-S) test was used to test the normality of the variables. Result of the K-S test showed that K sat was logarithmic normally distributed. Therefore, log-transformed K sat was used for further statistical analysis. Analysis of variance (ANOVA) was used to test for significant differences among landscape position, land management, and soil depth. We tested the hypothesis that landscape position and management had significant effects on soil hydraulic properties. The significance of treatment effects was tested using a 95% confidence level.

Bulk Density
The bulk densities for the six types of soil architecture are shown in Table 3. The results showed that soil architecture and landscape position had significant effects on bulk density at the topsoil (0-20 cm) (p < 0.05). The descending order of mean bulk densities at the 0-20 cm soil depth was . At the footslope, the bulk density was significantly lower than that at the shoulder and backslope positions at the 0-20 cm soil depth (p < 0.05). However, bulk densities showed no significant differences at the 0-20 cm soil depth between soils at the shoulder and backslope (p > 0.05). At the sites of grass cover, the bulk densities varied in the range of 1.09-1.42 g cm −3 at the first (0-10 cm) depth. The mean bulk densities increased with increasing soil depth at all sites, as shown in Table 3, which was similar to the research findings reported by [38]. Table 3. Bulk density for the sampling sites at each depths (mean ± standard deviation). Different letters represent a significant difference at the 0.05 probability level. Soil architecture SLSG 1.28 ± 0.08ab

Saturated Hydraulic Conductivity
The results of the saturated hydraulic conductivity (K sat ) for all types of soil architecture at four sampling depths are illustrated in Figure 2. The results showed that K sat values changed (0.09-11.32 cm h −1 ) significantly among soil types and depths for the study sites. The largest K sat was measured at SLSG in the first soil depth, which was 1.36 and 1.90 times higher than that at the SSLG and DSLG sites, respectively. The . The land management effects on K sat were significant (p < 0.05), with the grassland sites having the largest values. At the 0-10 cm soil depth, the K sat value of grassland averaged over landscape position was 21% higher (p < 0.05) than that under the cropland. This was probably attributed to more pronounced grass roots that generated abundant connected macropores in the soil profile. cm soil depth, Ksat varied in the range of 0.25-4.28 cm h −1 (SLSG), 0.07-2.24 cm h −1 (SLSC), 2.08-7.25 cm h −1 (SSLG), 0.11-7.23 cm h −1 (SSLC), 0.25-4.35 cm h −1 (DSLG), and 0.09-8.53 cm h −1 (DSLC). The land management effects on Ksat were significant (p < 0.05), with the grassland sites having the largest values. At the 0-10 cm soil depth, the Ksat value of grassland averaged over landscape position was 21% higher (p < 0.05) than that under the cropland. This was probably attributed to more pronounced grass roots that generated abundant connected macropores in the soil profile.

Soil Water Retention
The results of soil water retention at all measured soil water pressures are shown in Table 4 and Figure 3. Water content was significantly higher (p < 0.05) for the DSLG and DSLC sites than values for other sites (SLSG, SLSC, SSLG, and SSLC) at all soil depths and suction pressures used. However, the average soil water contents showed no significant differences (p > 0.05) among SLSG, SLSC, SSLG, and SSLC sites at all measured soil water pressures ( Figure 3). Water contents varied significantly with soil depth (p < 0.05). A descending trend was found in saturated water content for all soil architectures. Averaged over land use and soil depths, the footslope location had significantly (p < 0.05) higher saturated water content (0.31 cm 3 cm −3 ) than that in the backslope positions (0.27 cm 3 cm −3 ). However, no significant differences in saturated water content showed between the footslope and shoulder position (p > 0.05). Soil water retention in the cropland at the four sampling depths for all suction pressures were significantly (p < 0.05) higher than for the grassland except at the −1500 kPa pressures (Table 4).    Figure 4 and Table 5 shows the pore size distribution results for all types of soil architectures, landscape positions, and land uses at four depths. Results showed that the macroporosity in SLSG and SLSC was significantly (p < 0.05) higher than that in SSLG, SSLC, DSLG, and DSLC at the first soil depth (Figure 5a). However, no significant differences in macroporosity were found between SLSG and SLSC at four soil depths (p > 0.05). Macroporosity at the SLSG sites at the first soil depth was 22% higher than that at the SSLG sites. No significant differences in macroporosity were showed among SSLC, DSLG, and DSLC at all depths (p > 0.05). The larger macroporosity leads to the higher Ksat values observed for the SLSG site, as reported above. Coarse mesoporosity was significantly higher in SLSG and SLSC in comparison to other types of soil architecture at the first depth (p < 0.05). However, coarse mesoporosity showed no significant differences among soil architectures for the deeper (20-40 cm) soil depths (p > 0.05). Fine mesoporosity for DSLG and DSLC was significantly (p < 0.05) lower at 0-30 cm depths than other study sites. Significantly higher values of microporosity were found in the DSLC sites at all depths compared with other sites (P < 0.05). Microporosities in SSLG were significantly (p < 0.05) lower at all soil depths than that in SLSC, SSLC, DSLG, and DSLC. However, no significant differences in microporosity were found between SSLG and SLSG (P > 0.05). At the sites under the grass cover, the mean microporosity (0.216 cm 3 cm −3 ) at the DSLG sites was significantly (p < 0.05) higher than that at the SLSG (0.146 cm 3 cm −3 ) and SSLG (0.143 cm 3 cm −3 ) sites. Averaged over land management and soil depth, the shoulder position had the highest macroporosity, which was 31% and 52% larger than the backslope and footslope position,  Table 5 shows the pore size distribution results for all types of soil architectures, landscape positions, and land uses at four depths. Results showed that the macroporosity in SLSG and SLSC was significantly (p < 0.05) higher than that in SSLG, SSLC, DSLG, and DSLC at the first soil depth (Figure 5a). However, no significant differences in macroporosity were found between SLSG and SLSC at four soil depths (p > 0.05). Macroporosity at the SLSG sites at the first soil depth was 22% higher than that at the SSLG sites. No significant differences in macroporosity were showed among SSLC, DSLG, and DSLC at all depths (p > 0.05). The larger macroporosity leads to the higher K sat values observed for the SLSG site, as reported above. Coarse mesoporosity was significantly higher in SLSG and SLSC in comparison to other types of soil architecture at the first depth (p < 0.05). However, coarse mesoporosity showed no significant differences among soil architectures for the deeper (20-40 cm) soil depths (p > 0.05). Fine mesoporosity for DSLG and DSLC was significantly (p < 0.05) lower at 0-30 cm depths than other study sites. Significantly higher values of microporosity were found in the DSLC sites at all depths compared with other sites (p < 0.05). Microporosities in SSLG were significantly (p < 0.05) lower at all soil depths than that in SLSC, SSLC, DSLG, and DSLC. However, no significant differences in microporosity were found between SSLG and SLSG (p > 0.05). At the sites under the grass cover, the mean microporosity (0.216 cm 3 cm −3 ) at the DSLG sites was significantly (p < 0.05) higher than that at the SLSG (0.146 cm 3 cm −3 ) and SSLG (0.143 cm 3 cm −3 ) sites. Averaged over land management and soil depth, the shoulder position had the highest macroporosity, which was 31% and 52% larger than the backslope and footslope position, respectively. A similar trend was found for coarse and fine mesoporosity. However, the footslope position had higher microporosity compared to the shoulder and backslope position.

Discussion
The variations in soil hydraulic properties among landscape positions were due to the differences in soil thickness and architectures. Many researchers have investigated the effects of soil thickness on soil hydraulic properties and water infiltration [30][31][32][33][34][35][36][37][38][39][40][41]. The authors of [42] evaluated the physiochemical and hydraulic properties of surface soils at varying microtopographic positions in the Ozark Highlands. They found that the infiltration rates were approximately twofold higher at the microtopographic low position (thicker soil) than that at the high position (shallower soil). In this study, we observed that the saturated hydraulic properties at the shoulder position with a shallower soil thickness was higher than that at the backslope and footslope positions (thicker soil thickness). However, in a study conducted by [43], they characterized the hydraulic properties in different landscape and conservation practices along a claypan soil catena and found that the backslope (shallow soil) position had lower saturated hydraulic conductivity than the summit and footslope (thick soil) positions. They also concluded that water retention was higher at the backslope position among the landscape positions. These differences from our study could partly be explained by the claypan horizon existing in their study soil profiles. Besides, it was observed that macro-and mesoporosity at SLSG and SLSC sites was significantly higher than that at DSLG and DSLC sites (Table 5). We think that shallow soil thickness resulted in more distinct soil horizonation, which contributed to more subsurface flow movement of the infiltrated water at the SLSG and SLSC sites (Figure 5a,b).
The effect of landscape position on soil architecture was also reflected in soil texture. Soil profiles at DSLG and DSLC sites showed homogeneity, and the clay content was relatively higher than other sites. However, at SLSG and SLSC sites, high rock fragments content and low clay content resulted in more heterogeneous soil profiles ( Table 2). The authors of [29] reported that the clay content was the major controlling factor in topsoil thickness effects on water retention. The authors of [44] found

Discussion
The variations in soil hydraulic properties among landscape positions were due to the differences in soil thickness and architectures. Many researchers have investigated the effects of soil thickness on soil hydraulic properties and water infiltration [30][31][32][33][34][35][36][37][38][39][40][41]. The authors of [42] evaluated the physiochemical and hydraulic properties of surface soils at varying microtopographic positions in the Ozark Highlands. They found that the infiltration rates were approximately twofold higher at the microtopographic low position (thicker soil) than that at the high position (shallower soil). In this study, we observed that the saturated hydraulic properties at the shoulder position with a shallower soil thickness was higher than that at the backslope and footslope positions (thicker soil thickness). However, in a study conducted by [43], they characterized the hydraulic properties in different landscape and conservation practices along a claypan soil catena and found that the backslope (shallow soil) position had lower saturated hydraulic conductivity than the summit and footslope (thick soil) positions. They also concluded that water retention was higher at the backslope position among the landscape positions. These differences from our study could partly be explained by the claypan horizon existing in their study soil profiles. Besides, it was observed that macro-and meso-porosity at SLSG and SLSC sites was significantly higher than that at DSLG and DSLC sites (Table 5). We think that shallow soil thickness resulted in more distinct soil horizonation, which contributed to more subsurface flow movement of the infiltrated water at the SLSG and SLSC sites (Figure 5a,b).
The effect of landscape position on soil architecture was also reflected in soil texture. Soil profiles at DSLG and DSLC sites showed homogeneity, and the clay content was relatively higher than other sites. However, at SLSG and SLSC sites, high rock fragments content and low clay content resulted in more heterogeneous soil profiles ( Table 2). The authors of [29] reported that the clay content was the major controlling factor in topsoil thickness effects on water retention. The authors of [44] found that soil with high clay content increases preferential flow and decreases the exchange of lateral flow. In this study, the largest water content was found at the DSLC sites at all measured soil water pressures (Table 4). Moreover, higher spatial heterogeneity was caused by high rock fragments at the SLSG, SLSC, and SSLG sites. Rock fragments increase the interaggregate voids and accelerate the water flow. Higher saturated hydraulic conductivities were observed in SLSG, SLSC, and SSLG with the higher rock fragments content. The same results were obtained by [45]. They observed high extra-interphase or macroporosity between soil matrix and rock fragments, thus increasing water infiltration through the soil.
Many researchers have demonstrated that land management significantly affects soil hydraulic properties [46][47][48][49]. Considering that the selected sites are located on the same hillslope developed upon granites, the significantly higher K sat under grassland in comparison with cropland (Figure 2), provides a pronounced difference in K sat between these grasslands and croplands. It often results from the tillage or compaction of the soil in the topsoil layer [50]. In the tillage layer at the SSLC and DSLC sites, the structural macropores were destroyed by cultivation [51], giving rise to the significant difference in K sat between the topsoil layer and the deeper soil layer (Figure 2). Land management practices are of great importance to soil physical and hydraulic properties, which can be attributed to the influences of compaction, tillage, structural evolution, and consolidation [46,48,52]. On average, grassland soils demonstrated K sat values 28% higher than cropland soils ( Figure 2). Several other types of research [26,30] have reported similar results of differences between grassland and cropland. A higher percentage of difference was reported in Mexico, where the average K sat of soil under switchgrass was 73% greater than row crop management [30]. The authors of [26] found that K sat values for natural prairie were almost one order of magnitude higher than those for rotation farmland. The authors of [53] reported that the alfalfa grasslands significantly improved the soil infiltration capacity. Such variations in hydraulic properties of topsoil are of great importance for hydrological responses such as infiltration, overland runoff, and water quality.
This study also showed a higher total porosity in cropland (0.30 cm 3 cm −3 ) than grassland soils (0.28 cm 3 cm −3 ), by a percentage of 7%. The value of water content in this study was much lower than the results by other authors [38,43,54]. The authors of [54] reported that the average saturated water contents of prairie and row crop soils were 0.57 and 0.51 cm 3 cm −3 . The pronounced differences in water contents between the two studies may be a result of contrasting soil textures in the study soils. The soils in this study consisted of sand and loamy sand, which have poor water retention capacities compared to claypan soils in the study of [54]. This also indicates that the effects of land management are similar regardless of soil textures. Our study was carried out in summer, but the results should be complemented, especially for croplands, by monitoring the temporal variations in soil hydraulic properties. The soil properties for croplands are modified regularly by agricultural activities such as tilling, plowing, leveling, sowing, and irrigating [55][56][57][58]. Vertical variation of Ksat combined with rainfall intensity has been shown to affect the dominant runoff pathways [59][60][61][62], including overland flow (ORF, appears when rainfall intensity exceeds the soil infiltration capacity), subsurface flow (SSF, occurs at lateral flow structures), and deep-water infiltration (DF, water penetrates into the soil profile). In this study, the mean I30 (maximum rainfall intensity in 30 min, 6.3 mm h −1 ) of all erosive rainfalls during 2016 (data were from the collection of the Zigui Soil and Water Conservation Experiment Station, not shown here) was chosen to predict the dominant runoff pathways under different soil architectures. Results indicated that in the shoulder position, SLSG, SLSC, and SSLG sites could not generate overland flow for the condition of mean I30. The percentages of deep infiltration at the SLSG sites (16%) and the SSLG sites (37%) were comparable ( Figure 5). In the backslope, the movement of water in SSLG and SSLC profiles was dominated by accumulation or in the form of subsurface flow, with the proportion of 63% and 75%, respectively. A relatively higher incidence of overland flow in the footslope position was found in this study, which indicated that the increased soil thickness resulted in enhanced surface soil water repellency. This is inconsistent with the findings by other researchers [6,63]. With grass cover, the overland runoff was decreased by 23.7% compared with the cropland in the footslope. It has been suggested that appropriate land management is highly recommended for thicker soils to reducing runoff generation as well as soil erosion. The study highlights the differences in soil hydrological processes under six typical soil architectures in a granitic landscape of a mountainous watershed, providing insights into the modeling of hillslope hydrology.
Soil hydraulic properties data are widely applied in environmental and agricultural activities, such as irrigation and drainage planning, water balance calculation, and nutrients leaching prediction [64]. The knowledge of soil and water interaction is very important for evaluating the soil's role in water quantity and quality and the water's role in soil quantity and quality [65]. Soil architecture across microscopic to megascopic scales largely controls soil hydraulic properties and water I is the infiltration that appeared in the soil profile, and the number in the parenthesis is the infiltration rate. K sat is saturated hydraulic conductivity, cm h −1 . SLSG: shallow loam sandy-shoulder-grassland; SLSC: shallow loam sandy-shoulder-cropland; SSLG: shallow sandy loam-grassland-backslope; SSLC: shallow sandy loam-cropland-backslope; DSLG: deep sandy loam-grassland-footslope; DSLC: deep sandy loam-cropland-footslope.
Vertical variation of K sat combined with rainfall intensity has been shown to affect the dominant runoff pathways [59][60][61][62], including overland flow (ORF, appears when rainfall intensity exceeds the soil infiltration capacity), subsurface flow (SSF, occurs at lateral flow structures), and deep-water infiltration (DF, water penetrates into the soil profile). In this study, the mean I 30 (maximum rainfall intensity in 30 min, 6.3 mm h −1 ) of all erosive rainfalls during 2016 (data were from the collection of the Zigui Soil and Water Conservation Experiment Station, not shown here) was chosen to predict the dominant runoff pathways under different soil architectures. Results indicated that in the shoulder position, SLSG, SLSC, and SSLG sites could not generate overland flow for the condition of mean I 30 . The percentages of deep infiltration at the SLSG sites (16%) and the SSLG sites (37%) were comparable ( Figure 5). In the backslope, the movement of water in SSLG and SSLC profiles was dominated by accumulation or in the form of subsurface flow, with the proportion of 63% and 75%, respectively. A relatively higher incidence of overland flow in the footslope position was found in this study, which indicated that the increased soil thickness resulted in enhanced surface soil water repellency. This is inconsistent with the findings by other researchers [6,63]. With grass cover, the overland runoff was decreased by 23.7% compared with the cropland in the footslope. It has been suggested that appropriate land management is highly recommended for thicker soils to reducing runoff generation as well as soil erosion. The study highlights the differences in soil hydrological processes under six typical soil architectures in a granitic landscape of a mountainous watershed, providing insights into the modeling of hillslope hydrology.
Soil hydraulic properties data are widely applied in environmental and agricultural activities, such as irrigation and drainage planning, water balance calculation, and nutrients leaching prediction [64]. The knowledge of soil and water interaction is very important for evaluating the soil's role in water quantity and quality and the water's role in soil quantity and quality [65]. Soil architecture across microscopic to megascopic scales largely controls soil hydraulic properties and water infiltration processes [1,2]. In our study, an in-depth understanding of soil hydraulic properties in real soil-landscape systems explicitly reflects the universal heterogeneity of soil architecture in nature from an aspect of the soil-water feedback response. Runoff generation pathways inferred by precipitation and the vertical variability of saturated hydraulic conductivities provide possible ways to predict surface and subsurface flow [61,66]. Moreover, the results provide important information for improving the accuracy of hydrological modeling in an agricultural watershed and, in turn, promote sustainable agricultural management and water resources management.

Conclusions
The study assessed the effects of six different soil architectures (SLSG, SLSC, SSLG, SSLC, DSLG, and DSLC) on bulk density, saturated hydraulic conductivity (K sat ), soil water retention characteristics, and pore size distributions at different landscape positions (shoulder, backslope, and footslope) in a small agricultural watershed. The results showed that the shallow loam sandy soil in grassland (SLSG) on the shoulder had significantly higher K sat , greater macroporosity than the other selected sites. This may largely increase infiltration and reduce overland runoff in SLSG. The difference was controlled by the high rock fragments content and structural variations in different soil depths. The results of water retention indicated that deep sandy loam in cropland (DSLC) had markedly increased water content at all measured soil water pressures compared with shallow sandy loam in cropland (SSLC) and deep sandy loam in grassland (DSLG). The DSLC sites also had lower K sat and higher microporosity, which increased the potential to generate surface runoff. The study also demonstrated that land management systems had significant effects on hydraulic properties. The grassland system showed a remarkable improvement in soil hydraulic properties compared to the cropland system. The grassland had 54% higher K sat and 60% greater macroporosity than DSCL. Land management was the dominant factor influencing hydraulic properties in the surface horizon. Reduced K sat in deeper horizons was controlled by the compaction and illuvial clay content. Our study also revealed that on a granitic soil hillslope, decreased water infiltration and enhanced runoff was predicted at the footslope position, while on the shoulder, the flow pathway was dominated by subsurface flow and deep percolation into the soil profile. Different hydrological responses to landscape position and management are important for further hydrological modeling in the mountainous watershed.