Effects of preferential flow on soil nutrient transport in karst slopes after recultivation

In response to the global food shortage, a large amount of abandoned land in karst areas has been reclaimed as cultivated land, causing severe nonpoint source pollution. Preferential flow-driven soil nutrient transport on karst slopes remains poorly studied, though it is a major factor in nonpoint source pollution, as it responds to changes caused by reclamation. We explored the characteristics of soil preferential flow differences in recultivated land, grassland, and shrubland from returning farmland by dye tracer experiments and quantitatively examined the effect of preferential flow on nutrient transport. Under the condition of 40 mm precipitation, the preferential flow paths (PFPs) of the three types of plots were mainly distributed from 0 to 40 cm. The total porosity in the 20–40 cm soil layer was significantly reduced by reclamation, and the number of preferred flow paths in the 20–40 cm soil layer was significantly reduced from 60 to less than 10, which was significantly less than that in grassland and shrubland. But, reclamation results in the transport of more soil nutrients by preferential flow. The contribution rate of preferential flow to other nutrient indexes in the reclaimed land, in contrast to grassland and shrubland, was lower than zero, except for organic matter and total potassium. Moreover, when the PFP was connected to the rock–soil interface, the soil water can leak underground through the rock–soil interface quickly. Therefore, our findings indicated that reclamation reduces the distribution depth of the PFP. Still, the connection of soil preferential flow to the rock–soil interface increases the transport of soil nutrients to deep fissures and even underground rivers, thereby causing recultivated land to become one of the main sources of groundwater pollution in karst areas.


Introduction
Agricultural nonpoint source pollution is one of the leading causes of water pollution and ecological degradation worldwide (Qian et al 2021), and it is considered a severe threat to the safety of the aquatic ecosystem and drinking water (Zhu et al 2019). Therefore, nonpoint source pollution has long been the focus of research in hydrology and water resource science (Xu et al 2019). Karst areas account for 12% of the Earth's land area, and their aquifers provide water for 20%-25% of the world's population (Ford and Williams 2013). The karst areas in Southwest China, which are the most widely distributed karst areas in the world, affect the drinking water safety of hundreds of millions of people in the Yangtze River basin (Chang and Zhu 2021). However, due to serious soil erosion and barren soil, unreasonably large amounts of fertilizers are applied to improve crop yield on cultivated land, leading to nonpoint source pollution, which is particularly serious in karst areas dominated by agriculture (Longyang 2019). Moreover, in response to global food shortages, mountainous karst areas, which are the primary agricultural areas, promote the large-scale reclamation of abandoned land for transformation into cultivated land. According to incomplete statistics, Guizhou Province and Chongqing Municipality, centrally located in the karst area of Southwest China, reclaimed more than 600 km 2 and 500 km 2 of abandoned land, respectively, in 2022 (Chongqing Daily 2022, Department of Agriculture and Rural Affairs of Guizhou Province 2022). Large reclamation areas lead to more serious nonpoint source pollution in karst areas.
Understanding the mechanism of soil nutrient transport and accurately estimating nutrient diversion loss provide a basis for addressing nonpoint source pollution (Mencaroni et al 2021). Preferential flow is a typical hydrological phenomenon of water transport from homogeneous to heterogeneous areas in unsaturated soils (Sanders et al 2012), accounting for 11%-54% of the total water flux in regional hydrology; it can be used to reflect the process of soil water transport  and has an essential impact on water use efficiency and water conservation (Franklin et al 2021). In addition, preferential flow can control the migration of water and solutes, bypassing the soil matrix to reach deeper soils and thus significantly influencing regional nonpoint source pollution (Flury et al 1994). In some soils, preferential flow can transport more than 70% of water and solute (Beven and Germann 2013). An enhanced understanding of preferred flow processes is needed to better understand the underground transport of various pollutants.
Extensive studies have shown that the influence of land use on soil pore properties is considered one of the most critical factors strongly affecting the preferential flow transport process (Baez 2019). Land use type affects plant residue and root characteristics, resulting in differences in soil macropore properties (Budhathoki et al 2022). Zhang et al (2021) reported that the restoration of vegetation from herbaceous to woodland would increase the soil porosity of the 0-50 cm soil layer. Udawatta et al (2008) showed that perennial vegetation, compared with row crops, can increase soil porosity. In addition, the characteristics and formation of soil macropores can change with tillage practices (Pires et al 2019). Fuentes et al (2004) reported that conventional agricultural tillage, compared to no-tillage, results in changes in pore structure that destroy larger pores and restrict water flow. However, some scholars have argued that soil macropores under no-tillage are generally smaller than those under traditional tillage (Zaraee and Afzalinia 2016). Thus, the understanding of the impact of land use types and tillage methods on soil pore characteristics is inconsistent. This paradox may be due to differences in specific climates, land cover, and management practices. What are the effects of vegetation restoration on soil porosity and the preferential flow of abandoned land in karst areas? It is worth further research to help diagnose changes due to agricultural management practices and to design appropriate management guidelines.
The main goals of this research by dye tracer experiments were to (a) determine the characteristics and influencing factors of soil preferential flow path (PFP) after recultivation, (b) determine the differences in soil nutrients between the preferential flow area and the matrix flow area, and (c) determine the effect of preferential flow on soil nutrient transport. These results are significant for further understanding the process and mechanism of nutrient transport and nonpoint source pollution in karst areas.

Study site
Dye infiltration experiments were performed at Huaxi, Guiyang, Guizhou, Southwest China, between June 2018 and July 2018 (figure 1). The study catchment is a typical agricultural region in the karst area, with a shallow soil thickness (average depth 30 cm) and bedrock outcrops (outcrop rate is 30%). Beneath the shallow soil is limestone with developed shallow karst fissures. Shallow karst fissures are usually filled by soil, which is an important hydrological path of karst slopes (Carrière et al 2020). The study area has a subtropical monsoon humid climate with a mean annual temperature of 14.9 • C and a mean annual rainfall of 1178.3 mm. Three types of land use were selected to explore the preferential flow on soil nutrient transport in karst slopes after land use was returned to farmland (table 1). These included grassland (abandoned for five years from cultivated land, GL), shrubland (five years of continuous abandonment from abandoned grassland, SL), and cultivation land (four years of recultivation from abandoned grassland, CL).

Dye-tracer infiltration
Dye tracer experiments were carried out at three subsites with land use types of cultivation land, grassland, and shrubland. Three replicates of the typical dual hydrological structure profile were determined by the chain pain method in each subplot for the dye infiltration experiments (Yan et al 2020) (figure 2(a)). Before the dyeing experiment, the aboveground parts of the plants in the experimental site were removed with branch shears, leveling the surface soil. Next, a steel frame (length × width × height: 50 cm × 50 cm × 10 cm) was embedded into the soil in the center of the plot. Brilliant blue (FCF, C 37 H 34 N 2 Na 2 O 9 S 3 ) was chosen as a dye tracer because it dissolves easily in water and is environmentally friendly. Additionally, the aqueous solution FCF (4.0 g l −1 ) was irrigated with an intensity of 40 mm h −1 for 60 min at the plot. No ponding was found for any plots throughout the experiment. Then, the experimental plots were covered with plastic film to prevent evaporation or rain over 24 h and to ensure that the infiltration process was complete. Three 1.0 m deep soil vertical sections were dug in each plot after the end of infiltration with   a horizontal interval of 10 cm (figure 2(b)). The dye-stained patterns of the soil sections were photographed using a digital camera (Canon EOS 80D) in the vertical orientation.

Distinguishing between preferential and matrix flow zones
In this study, soil samples were collected from the preferential flow zone (stained area, PA) and the matrix flow zone (unstained area, MA) after the dyestained patterns of each vertical profile were recorded (figure 3). MA and PA were distinguished based on the staining of the vertical and horizontal soil profiles. When no bright blue color appears in the vertical section corresponding to the unstained part of the horizontal section of the soil, there is essentially no PFP in this unstained region, implying that this region is a soil MA. The zone where both horizontal and vertical segments are stained is the PA.

Soil sampling and analysis
In each plot, three replicates of undisturbed soil samples from the PA and the MA were collected at depths of 10, 20, 30, and 40 cm by cutting cylinders (inner diameter × height: 50.46 mm × 50.00 mm; volume, 100 cm 3 ). A small shovel was used to collect all the disturbed soil samples. The undisturbed soil sample was used for the measurements of field capacity (FCW), capillary water holding capacity (CWC), total porosity (TCP), and noncapillary porosity (NCP). After the physical properties of the soil were determined, the undisturbed and disturbed soil were mixed into a mixed sample into a mixed sample and be used to determine the soil particle size distribution (PSD) and soil nutrient properties. The weighing method measured the FCW, CWC, TCP, and NCP (Arnold 1986). The PSD was measured by the pipette method (Liu 1996), the soil organic matter (OM) was measured by the potassium dichromate titration method, the total nitrogen (TN) and alkali hydrolysable nitrogen (AN) were determined by the Kjeldahl method and Conway method, respectively, the total phosphorus (TP) and available phosphorus (AP) were analyzed by the Mo-Sb-Vc colorimetric method, and the total potassium (TK) and available potassium (AK) were determined by flame photometry dissolved with acid and flame photometry extracted with NH 4 OAc, respectively (Institute of Soil Science of Chinese Academy of Sciences 1978).

Image processing
Image processing was performed according to the following: cut out, shot corrected, image binarization, and extracting the dye coverage (DC) with MAT-LAB (Janssen and Lennartz 2008). To eliminate the influence of the boundary effect, Photoshop CS5 (Adobe Systems Incorporated, California, USA) was used to cut the section 10 cm from both sides of the boundary. The height of the image was based on the dimension of the largest infiltration depth among all dyed sections. Shot correction: Photoshop CS5 was used to correct the stained image, which could not be photographed orthogonally, and an image of 1 mm × 1 mm was defined as 1 pixel. Image binarization and the application of Image-Pro Plus 6.0 turned the stained areas black, while the rest stayed white. Moreover, black was assigned '0' , and white was assigned '255' .
2.6. Analytical methods of dyed images with the PFP 2.6.1. DC DC is usually used to evaluate the development degree of preferential flow. DC is obtained from the following function according to (Flury et al 1994): where DC is the dye coverage (%), D is the total dyed area of the soil profile, and ND is the nondyed area of the soil profile.

Index of the extent of preferential flow (IP)
The index of the extent of preferential flow (IP) converted from the variation coefficient of the DC in the preferential flow area is used to quantitatively describe the change in dyeing morphology after the removal of the matrix flow areas and is calculated as follows: where H t is the total vertical length of the dyed section, mm; H un is the vertical length of the matrix flow area of the staining profile, mm; a is the unit soil layer height, 1 mm; and IP is the variation coefficient of the DC in the preferential flow area.

Contribution rate of the preferential flow (CP)
We used the contribution rate (CP) to quantitatively determine the effect of soil preferential flow on soil nutrient transport: where C iPA is the soil nutrient content of the PA, and C iMA is the soil nutrient content of the MA.

Statistical analysis
Statistical analysis was performed using STATIST-ICA 12.5 (StatSoft Inc., Tulsa, USA), and significant differences in the soil nutrient properties and the preferential flow contribution rate were determined by repeated measures general linear models (GLMs). The soil nutrient properties differed between the PA and MA according to discriminant function analysis (DFA). Origin 2021 (Origin Lab, Northampton, Massachusetts, USA) was used to analyze Spearman's correlation coefficients to explore the correlation between soil physical properties and the brilliant blue FCF DC of soil profiles in different plot types.

Dye pattern of the vertical soil section
In three plots, the area stained with FCF in the soil profiles was mainly concentrated in the 0-40 cm soil layer and decreased with soil depth (figure 4). In addition, as the FCF stained the rock−soil interface of the poorly permeable soil layer, which caused the area stained by the FCF to sharply decrease at 20-40 cm of the soil layer. Figure 5 showed that TCP and NCP decreased with soil depth, but the trend was different by plot type. With the deepening of the soil layer, the TCP and NCP of CL decreased rapidly, while the TCP and NCP of SL increased first and then decreased. Especially for the soil layer below 20 cm, the TCP of the CL plots decreased sharply to 10% and was 10%-20% less than that of the GL and SL plots. Figure 6 showed that the number of PFP was lowest in the CL plots and highest in the SL plots. In the CL plot, the mean DC increased with soil depth in the 0-20 cm soil layer (IP was 12.94%), and then the mean DC decreased sharply in the 20-40 cm soil layer (IP was 29.51%). In the GL and SL plots, the mean DC was also mainly distributed in the 0-40 cm soil layer.  The DC of the two types of plots decreased slowly in three steps with soil depth. Figure 7 showed that the number of PFP had a positive effect on the DC in the three types of plots.

Differences in soil nutrient properties between the preferential flow zone and matrix flow zone
As showed in figure 8 and table 2, the soil nutrient properties significantly differed between the PA and MA (p < 0.001; DFA). The largest squared Mahalanobis distance between the PA and MA was in the CL plot, and the shortest distance was in the GL plot. The difference in the soil nutrient properties between the PA and MA was due to the interaction effect of plot type and soil depth (p < 0.001, GLM, figure 9). OM content in the three plots' PA was higher than in the MA. Compared with the total nutrients, AN, AP, and AK in the PA of the three plot types were significantly lower than those in the MA, except for the AP in the PA of GL and SL.

Soil preferential flow pathways contribute to soil nutrient properties
Preferential flow significantly affected soil nutrient transport and differed significantly between each plot type (p < 0.001, figure 10(a)). The Cp of OM was higher than 0 in all three plot types. In general, the Cp of soil nutrients was significantly lower than 0 in the CL and GL plots, except for the Cp of TK and TP, which was higher than 0 in the CL and GL plots, respectively. Additionally, the transport effect  Figure 9. Difference in the soil nutrient properties between the preferential flow and matrix zones. Note: the mean value is the average value of 0-40 cm in the same area. OM is soil organic matter, TN is total nitrogen, TP is total phosphorus, TK is total potassium, AN is alkali hydrolysable nitrogen, AP is available phosphorus, and AK is available potassium. PA is the preferential flow zone; MA is the matrix flow zone.
of preferential flow on nutrients was affected by the plot type and the soil depth (figures 10(b)-(d)). In the CL plot, except the Cp of TK generally being higher than 0, most of the soil total nutrients (TN and TP) and available nutrients (AN, AP, and AK) were less than 0 and significantly (p < 0.05) decreased with soil depth. The Cp of TN, TK, and AK in the SL plot was higher than 0, significantly increasing with soil depth first and then decreasing. Contrary to the Cp of TN and TK, the Cp of TP was lower than 0 and increased with soil depth.

Effect of recultivation on preferential flow
PFP can reflect the process of water migration, which plays a vital role in the hydrological process of slopes (Beven 2018). Soil texture and porosity are critical factors for controlling the preferential flow (Bachmair et al 2009). In this study, the preferred flow path were also significantly affected by TCP and NCP and soil texture ( figure 11). An increase in soil porosity and sand content significantly increases the DC and IP.
Previous studies have suggested that the biochemical effects of plant roots during the transition from herbaceous to shrubbery may improve soil texture and increase soil porosity, thereby promoting water infiltration throughout the soil profile .
Conversely, when forestland is converted into cultivated land, the bulk density increases, the porosity decreases, and plowed layers are formed, resulting in reduced soil permeability, which promotes preferential transverse flow in the surface layer (Schwen et al 2014).

Nutrient transfer in relation to preferential flow pathways
In this study, the OM content in PA was generally significantly higher than that in MA, but the difference decreased with the depth of the soil layer (figure 10).
The reason for this difference may be that soil macropores are the main channel for preferential flow and are mainly root canals left after root death and decomposition (Guo et al 2019). Meanwhile, root decomposition also releases OM and NPK, which makes the soil nutrient content in the PA relatively higher than that in the MA (McCourty et al 2018).
The root counts and biomass generally decrease with soil depth, which leads to a decrease in the preferred flow path in deep soil and, thus, a decrease in the difference in nutrient content between the PA and the MA. It is also possible that the water-soluble carbon in the soil collects in the PFP with the movement of soil moisture, making the OM content in the PA higher than that in the MA . For this reason, the OM content in the PA was higher than that in the MA. And the Cp of TN, TK, and AK in the SL plot was higher than 0 and showed a significantly increasing in soil depth first and then decreasing. Moreover, preferential flow promoted the solute transport of nutrients under the combined effect of soil depth (figure 12). DC had positive correlation with CP of soil nutrients except TK, which was significantly correlated (p < 0.01) with CP of TP and AN. Compared with GL and SL plots, the difference of nutrient content between PA and MA in CL plot was larger due to the transport of preferential flow. Long-term farming activities and soil erosion have led to a lack of nutrients in cultivated land (Krauss et al 2020). The artificial application of water-soluble fertilizer effectively supplemented soil nutrients, causing the soil nutrients in the CL plot to be more easily transported by preferential flow than those in the GL and SL plots (He Figure 11. Heatmap of Pearson's correlation coefficients (r) between soil physical properties and the dye coverage of the brilliant blue FCF dye of soil profiles in different plot types. Note: DC is dye coverage, IP is the index of the extent of preferential flow, NCP is non-capillary porosity, TCP is total porosity, FCW is field capacity, CWC is capillary water holding capacity, SAND is sand content, SLIT is slit content, CLAY is clay content. et al 2020). Preferential flow-driven nutrient migration led to a lack of soil nutrients, and when water containing a large amount of dissolved nutrients was supplied to groundwater, groundwater eutrophication occurred (Sun et al 2022). Therefore, reclaimed land involved in farming has become the main source of non-point source pollution in karst areas (Zhang et al 2020).

Hydrological connectivity facilitates soil nutrients loss
Hydrologic connectivity determines the whereabouts of soil nutrients driven by preferential flow. It is worth noting that the soil−rock interface is a significant factor in the PFP, enhancing the hydrological connectivity on the karst slope. Sohrt et al (2014) reported that the rock−soil interface flow is an essential preferential flow type in which runoff percolates faster and to greater depths than water infiltrating directly from the soil surface. We found that when the soil flows preferentially connected to the rock−soil interface, water first preferentially infiltrated from the rock−soil interface (figure 13). Katuwal et al (2021) also found that after the finger flow, macropore flow, and fissure flow in the soil were connected with the geotechnical interface, the soil runoff preferentially leaked from the rock−soil interface to the deeper soil. These actions are attributable to long-term soil creeping and sliding on the bedrock surface, which causes the bedrock surface to be very smooth (Zhang et al 2011). This situation results in a weak connection between the bedrock and the soil and forms a channel conducive to water infiltration (Zhao et al 2020). When PFP is connected to the rock−soil interface, Figure 13. Characteristics of the rock-soil interface inside the fissure. the nutrient quickly migration from the soil through the rock−soil interface driven by preferential flow, leaks into the deep fissures, and even reaches the saturated zone to replenish groundwater ( figure 14). When water containing a large number of dissolved nutrients was supplied to groundwater, groundwater eutrophication occurred (Sun et al 2022). Reclamation may reduce the distribution depth of PFPs in the soil compared to grassland and shrubland. However, the rock-soil interface connected to the soil's PFP causes more nutrients to be lost to the deep subsurface in reclaimed land. As a result, reclaimed land will be the main source of groundwater pollution in karst areas.

Conclusion
This study quantitatively disclosed the characteristics of soil preferential flow and its effect on soil nutrient transport in shallow fissure soil after recultivation. The results demonstrated that the preferential soil flow paths of the karst slope with surface and underground dual hydrological structures were generally distributed in the 0-40 cm soil layer. Rainfall could quickly infiltrate into the soil inside the shallow karst fissure (20-40 cm) through the PFP in the surface soil layer (0-20 cm). The effect of recultivation on the depth of preferential flow distribution was not apparent, but it significantly affected the distribution of the number of PFP. In contrast to the continuously abandoned grassland and shrub, recultivation decreased the TCP and NCP in soil layers >20 cm, drastically reducing the PFP for soil layers deeper than 20 cm. In addition, the transport effect of preferential flow on soil nutrients was influenced by plot type, soil depth, and nutrient type. The effect of preferential flow on the available soil nutrients was more substantial than that of total nutrients, making more soil available for nutrient transport by preferential flow in reclaimed land with frequent artificial fertilization than grassland and shrubland. In brief, under the transmission of preferential flow, soil nutrients in recultivated land may cause leakage into deep fissures and even underground aquifers. Therefore, our findings indicated that reclaimed land would become a primary source of groundwater pollution in karst areas.

Data availability statement
The data that support the findings of this study are available upon request from the authors.