1 INTRODUCTION

Agricultural irrigation on the Heifangtai terrace has induced frequent loess landslides, and more than 70 large-scale landslides have occurred since the 1960s (Xu et al., 2014). Numerous studies have been conducted to investigate the loess landslides in Hefangtai. Their results indicate that the infiltration of irrigation water causes a rise in the groundwater table (GWT), which triggers a large number of landslides on the terrace (Cui, Pei, Wu, & Huang, 2018; Qi, Xu, & Liu, 2018; Xu et al., 2011; Xu, Qiao, Wu, Iqbal, & Dai, 2012). Since the groundwater depth in the Loess Plateau is mostly 30-80 m (Zhu, Li, Peng, & Zhang, 1983), the GWT is higher than the gully surface runoff and the confined water head of the bedrock, so there is no lateral groundwater recharge (Li, 2001). Therefore, the rise of the GWT is only induced by vertical infiltration, i.e. atmospheric precipitation and agricultural irrigation. These studies do not agree on how surface water infiltrates into the GWT and recharges the groundwater. Several scholars have suggested that preferential water flow paths such as loess fissures, sink holes, and macropores are the main methods of recharge (Xue, 1995). However, these features are observed only on the edge of the loess tableland in the unloading area (Xu et al., 2011). In addition, continuous water flow cannot occur in these preferential paths, because the water flowing into these paths is quickly absorbed by the surrounding soil due to differences in the hydraulic gradient, which indicates that these preferential paths have difficulty connecting with the groundwater. Li et al. (2013) found that the depth of rainfall infiltration was very limited based on manual drip experiments. In addition, they found that there was an obvious increase in the moisture content of the deep paleosol while the moisture content of the upper soil changed only slightly, which indicates that the unsaturated seepage in the loess recharges the groundwater. However, few scholars have been able to draw on any systematic research in unsaturated seepage.
Zimmermann et al. (1966) used the tracer method for a groundwater recharge study and proposed that soil water is transported by piston flow in the homogeneous vadose zone. Subsequently, the tracer method was widely applied in the study of water flow and solute transport in the vadose zone. Davis et al. (1980) gave a comprehensive review of injected tracers, in which anionic tracers (Cl‾, Br‾, and I‾) were used for a broad range of ground-water tracing applications. Cameron and Wild (1982) used36Cl‾, 15NO3‾, and HTO (titrated water) to label water molecules, studied solute transport under field conditions following the application of irrigation water and under winter rainfall conditions. Bowman (1984a and 1984b) found that low molecular weight anions (Cl‾, Br‾, and NO3‾) do not interact with most natural porous media and seem to be an ideal artificial tracer for soil water studies. Porro and Wierenga (1993) analyzed water flow and solute transport in soil using tritiated water and bromide in a large scale soil column. Cook et al. (1994) estimated the recharge rate of groundwater in southern Australia by tracing the soil water movement with 36Cl, 3H, and chloride. Gates et al. (2011) estimated the groundwater annual recharge amount in Yanan, Zhifanggou agricultural land using a chloride mass balance. Huang et al. (2017) used soil chloride profiles and multiple tracers to investigate groundwater recharge in the arid western Ordos Basin, NW China. Cl‾ and Br‾ are widely used as tracers to analyze water flow and solute transport, and good results have been obtained from soil column tests and field experiments.
Therefore, in order to understand unsaturated seepage in the loess vadose zone, a 6 m high remolded loess column test using chloride (Cl‾) and bromide (Br‾) as the artificial tracers was carried out under the effect of irrigation, and a numerical simulation was crested using COMSOL Multiphysics. The findings are important to further studies in the field of unsaturated seepage and in the investigation of landslide mechanisms in Heifangtai.

2 STUDY AREA

Located in the central part of Gansu Province, China (Fig.1), Heifangtai is situated at the confluence of the Yellow River and the Huangshui River, on the class IV terraces of the Yellow River. It extends east to west from the Huangshui River to Fangtai and north to south from the Moshi Gully to the Yellow River. The lithologic profile of this area is as follows from top to bottom (Fig. 2). (1) The top layer is mantled by Malan loess (26–48 m) with a loose structure and well-developed vertical fissures. (2) An alluvial clay layer (3–19 m) with a dense structure and low permeability is present below the loess layer. (3) Overlying the bottom is a fluvial gravel layer (1–6 m) mixed with sands, which has good permeability. (4) The bottom bedrock layer consists of mudstone and sandstone partings. The primary loess used in the column test was collected from the Heitai area. In order to avoid the influence of crop roots on the loess, the sampling depth was below 4 m. In addition, the saturated gravimetric moisture content of the primary loess was 35%, which was measured in-situ. A series of indoor physical property tests was carried out on the loess to obtain its basic physical properties such as the natural density, moisture content, specific gravity, and dry density (Table 1). Then, the ground up soil was air-dried to obtain the residual moisture content and the particle grading curve (Fig. 3).

3 METHOD

3.1 Soil column test

3.1.1 Test preparations
A large number of scholars have observed loess vertical profiles after rainfall or artificially simulated rainfall infiltration experiments in loess areas (Li, Li, & Vanapalli, 2016; Tu, Kwong, Dai, Tham, & Min, 2009). It has been shown that the depth of direct rainfall infiltration is limited and rarely exceeds 4 m. Therefore, considering the persistence of infiltration under the effect of irrigation and based on a better observation of unsaturated seepage in the loess vadose zone, a 6 m high soil column was applied.
In order to study the infiltration and the ion tracing under the effects of irrigation in the soil column, the device shown in Fig. 4a was designed to carry out the test in the laboratory. The device mainly consists of a soil column, moisture probes, water potential probes, and an automatic data acquisition system. The resolution of the EC-5 soil moisture probe is 0.1%, the resolution of Teros-21 soil water potential probe is 0.1 kPa, and the collection frequency of both probes is 1 min. The Em-50 data acquisition system automatically stored the moisture content data and the matric suction collected by the probes. The soil column was constructed from a transparent plexiglass cylinder, with an outer diameter of 320 mm, an inner diameter of 300 mm, a height of 6 m, and a sealed bottom. Fifteen small holes with a diameter of 1 cm were made in the cylinder’s wall for the probes, and 12 sampling holes with a diameter of 2.5 cm were made for real-time sampling.
3.1.2 Test procedures
After air drying, the loess was sifted using a sieve (d=2 mm). Distilled water was added to the loess, and it was fully stirred until the target gravimetric moisture content of the loess (13.7%) was achieved with a dry density of 1.40 g/cm3, which is approximately the same as that of the in-situ loess. The column was filled with the prepared loess, which was tamped down and shaved every 5 cm. In addition, the probes were placed into the center of the column through the 1 cm holes. The moisture probes were installed along the column at heights of 0.15 m, 0.55 m, 1.15 m, 1.55 m, 2.15 m, 3.15 m, 4.15 m, 5.15 m, and 5.4 m; and the water potential probes were installed along the column at heights of 0.15 m, 1.15 m, 2.15 m, 3.15 m, 4.15 m, and 5.15 m. The sampling holes were installed along the column at heights of 0.35 m, 0.95 m, 1.35 m, 1.95 m, 2.35 m, 2.95 m, 3.35 m, 3.95 m, 4.35 m, 4.95 m, and 5.35 m. The remaining gaps in the 1 cm holes and the sampling holes were sealed with glass glue. After every 20 cm of testing loess was added, the soil column was allowed to stand for 12 hours to reach a uniform moisture content. This was continued until the height of the remolded soil column reached 5.6 m. The column was prepared as shown in Fig. 4b.
After the preparation of the column, two layers of filter paper and a 5 cm gravel layer were placed on the top of the soil. This was done to prevent erosion of the surface loess when water was added. Before the infiltration, an ion chromatograph was used to measure the initial Cl‾ and Br‾ concentrations of the remolded soil sample and the infiltration water sample. The soil samples were air-dried, triturated, and filtered using a sieve (d=1 mm). Thereafter, 50 g of filtered soil sample was added to a 250 ml volumetric flask, and then, the flask was filled with deionized water, stirred, and shaken for 1 hour. The mixture was allowed to settle for 24 hours. After filtering the mixture through a 0.22-μm membrane, the supernatant was stored in a 10 ml reagent tube until analysis using an ion chromatograph. The results indicate that the Cl‾ concentration of the remolded loess was 141 mg/L and its Br‾ concentration was 0 mg/L. No Cl‾ or Br‾ were present in the infiltration water.
To analyze the water flow and solute transport of Cl‾ and Br‾ under the effect of irrigation, a falling head permeability test was designed to control the infiltration in the column, that is, an infiltration head of 16 cm was added at a fixed time every day. The infiltration water was prepared with NaCl (AR) and KBr (AR) ensure a Cl‾ concentration of 50 mg/L and a Br‾ concentration of 100 mg/L. According to the in-situ irrigation in Heifangtai, the infiltration test was carried out for 45 days. The test was divided into three stages according to the advancement of the wetting front and the real-time data measured by the probes.
Stage I (irrigation ⅰ): saturated wetting front advance (1–13 days);
Stage II (non-irrigation): without adding infiltration water (14–24 days); and
Stage III (irrigation ⅱ): the second irrigation of moist soil (25–45 days).
As the infiltration was conducted, samples were collected in real-time using the sampling holes along the column. After each 100 g soil sample was collected, the sampling hole was quickly sealed with glass glue to prevent the influence of external air pressure on the column. Four stages of sampling were conducted. (1) Samples were collected at different depths as the saturated wetting front advanced (1–10 days); (2) samples were collected at different depths after the saturated wetting front reached the bottom of the column on the 11th day; (3) samples were collected at different depths in Stage II on the 20th day; and (4) samples were collected at different depths after the second irrigation stable in Stage III on the 45th day. After processing the samples using the methods described above, the samples were placed in an ion chromatograph to measure their ion concentrations, and the columns’ Cl‾ and Br‾ profiles were obtained over time.

3.2 Numerical simulation of the column test

Due to the limited number of sampling times, the process of solute transport could not be demonstrated in detail, and thus Comsol Multiphysics was used to simulate the process in detail. Based on Comsol Multiphysics, which is a multi-physical field coupling software, a 3D soil column was constructed using a 2D axisymmetric component and a time dependent study was used to investigate the variation in the moisture content and ion concentration with respect of time. The Richards’ Equation interface was used to analyze the water flow in the porous media, and to model the infiltration in the column in combination with the Van Genuchten (1980) Equation. Many efforts to simplify and improve the modeling of water flow through variably saturated media have produced a number of variations on the Richards’ equation since its appearance. The form of the equation that COMSOL Multiphysics solves is very general and allows for time-dependent changes in both the saturated and unsaturated conditions (Bear, 2012; Bear, 2013):
. (1)
Where the pressure, p (Pa), is the dependent variable. In Equation (1), Cm (1/m) is the specific moisture capacity, Se is the effective saturation, S (1/Pa) is the storage coefficient,k s (m/s) is the hydraulic permeability, μ(m2/s) is the fluid dynamic viscosity,k r is the relative permeability, ρ(kg/m3) is the fluid density, g(m/s2) is the acceleration due to gravity, D(m) is the elevation, and Qm is the fluid source (positive) or sink (negative).
The transport of the dilute species interface was used to calculate the ion concentration in the porous media, and the solute transport was modeled by taking into consideration the solute sources. The following equations for the concentrations, ci , describe the transport of solutes in a variably saturated porous medium for the most general case, when the pore space is primarily filled with liquid, but also contains pockets or immobile gas (Bear, 2012; Bear, 2013):
. (2)
On the left-hand side of Equation (2), the first three terms correspond to the accumulation of the species within the liquid, solid, and gas phases, respectively, while the last term describes the convection due to the velocity field u (m/s).ci (mol/m3) is the concentration of species i in the liquid,cP, i is the amount adsorbed onto (or desorbed from) the solid particles, and cG, i is the concentration of species i in the gas phase.ρb (kg/m3) is the bulk density,ρb=(1-εp )ρ . εp is the porosity, and ρ(kg/m3) is the solid phase density. For saturated porous media, the liquid volume fraction θ is equal to the porosity εp , but for partially saturated porous media, they are related by the saturation s as θ =p . The resulting gas volume fraction isav = εpθ = (1-s )εp .
On the right-hand side of Equation (2), the first term introduces the spreading of the species due to mechanical mixing, diffusion, and volatilization into the gas phase. The tensor is denoted as DD (m2/s) and the effective diffusion as De(m2/s). The last two terms on the right-hand side describe the production and consumption of the species, respectively. Where Ri is a reaction rate expression, which can account for the reactions in the liquid, solid, or gas phase, and Si is an arbitrary source term, e.g., due to a fluid flow source or sink.
The research method and physical field were chosen by taking into consideration the solute transport in saturated or partially saturated porous media under the effects of diffusion, convection, dispersion, adsorption, etc. Fig. 4c shows the schematic diagram of the soil column model. The model contains two layers. The upper layer is a 16 cm water layer, which has a set Cl‾ source concentration of 50 mg/L, and the lower layer is a 5.6 m loess layer with a set Cl‾ initial concentration of 141 mg/L. The initial moisture content of the model was defined as the pressure head. In addition, the left side (r = 0) was set as the axisymmetric boundary, the right side as the no-flow boundary, the upper surface as the pressure head boundary, and the interlayer as the pervious water layer boundary. It should be noted that the bottom boundary was set as the small flux boundary becuase the column’s bottom contained sampling holes and sensor holes. The simulation parameters were obtained from the results of the column test, in which the saturation permeability coefficient Ks was determined from the average velocity of the saturated wetting front advance, and the parameters a and n in the VG model were obtained using the pressure plate method. Its soil water characteristic curve (SWCC) fitting is shown in Fig. 5. The remaining parameters were chosen according to the optimal solution of a large number of simulation results. All of the parameters used in the simulation are listed in Table 2.

4 RESULTS

4.1 Moisture content and matric suction

The infiltration began with an obvious saturated wetting front (Fig. 6). In Stage I (Figs. 7a and b), and the maximum moisture content of the soil at different depths would be reached at a certain time and would be greater than the saturated volumetric moisture content (49%). This is attributed to the change of the soil mass from dry to wet in a short time, which caused a large amount of water to accumulate. Therefore, the data measured by the moisture probes were the maximum values within the measuring range, which was greater than the saturated volumetric moisture content of the test loess and caused a significant fluctuation in the moisture content of the soil layer within 2 m depth. Meanwhile, the matric suction decreased rapidly, and then gradually stabilized. Once the water began to flow downward into the dry soil again, the moisture content in the upper part decreased as the accumulated water gradually dissipated. Finally, the moisture content remained at a stable unsaturated value. The infiltration water reached the bottom of the column about 10 days later. Due to the sealed bottom of the container, it continued to accumulate and formed the GWT. A significant saturated wetting front advanced and moisture content variations in Stage I. In order to simulate the in-situ irrigation in Heifangtai, infiltration water was no longer added in Stage II (14–24 days). As shown in Fig. 7c and Fig. 7d, the moisture content of the soil layer above a depth of 3.15 m continued to decrease during Stage II. In addition, the variation in the soil moisture state also caused the moisture content of the upper soil layer to fluctuate significantly. The moisture content in the shallow vadose zone changed significantly with the addition of water. However, the moisture content of the deep soil layer remained the same, and the content at a depth of 4.15 m increased to near saturation at a slow rate, and the matric suction gradually became stable. This indicates that cutting off the water supply was conducive to forming perched water at a depth of 4.15 m. After the 25th day (Stage III), the irrigation head at 16 cm was added again at a fixed time every day. Figure. 7e illustrates that only the probes above 3.15 m had a short response to the second irrigation, and the moisture content no longer fell back after this, but continued to increase. Meanwhile, the moisture content of the perched water layer at 4.15 m, which acted as an impervious layer to divide the soil column into two parts, remained saturated and stable. The moisture content of the upper soil gradually increased but still remained constantly low because of the reapplication of the irrigation head, while there was no change in the high moisture content of the lower soil. Although the apparent moisture content remained unchanged in the deep vadose zone, water still flowed there, i.e., unsaturated seepage occurred, because the irrigation head added to the top of the soil column continued to fall each day.
Through the analysis of the above different infiltration stages, we concluded that unsaturated seepage in the deep vadose zone is characterized by a stable apparent moisture content and a low velocity. In addition, because of these characteristics, the direct infiltration depth of the surface water into the column under the effect of irrigation is less than 4 m, which is the same as the observed results obtained under the condition of rainfall (Li et al., 2016; Tu et al., 2009). The reason for this is that below a depth of 3.15 m, the moisture probes did not respond to the second irrigation, so that the advance rate of the wetting front could not be calculated according to the change in the moisture content in Stage III (Fig. 8). Therefore, traditional methods of monitoring the moisture content and matric suction cannot accurately describe the process of unsaturated seepage.

4.2 Chloride and bromide ion profiles

In addition to the monitoring of the moisture content and the matric suction, the Cl‾ and Br‾ profiles at different infiltration stages were also analyzed using the column test. The Cl‾ initial concentration in the loess was 141 mg/L and the Br‾ initial concentration was 0 mg/L, indicating that as shown in Fig. 9a and Fig. 9b, a large amount of Cl‾ is leached while Br‾ mainly accumulates in the soil layer. On the 11th day of Stage I, the wetting front reached the bottom of the column. At this time, the solute front did not reach the bottom along with the wetting front, and a large amount of Cl‾ was detected near a depth of 4 m. It was observed that the ions in the column were not transported in real time with the wetting front, but had a certain lag time, which is consistent with the results of Porro and Wierenga (1993) and Ghuman (1980), i.e., the solute fronts significantly lag behind the moisture fronts. This phenomenon was also clearly reflected in the advance of the saturated wetting front (1–10 days). Only a small amount of Cl‾ was detected below a depth of 3 m, indicating that the Cl‾ in the soil was almost completely leached out. In the upper layer (above a depth of 3 m), a certain concentration of Cl‾ was detected, which is consistent with the results obtained from the lower layer because the advance of the solute front is highly dependent on the soil moisture content (Ghuman, Verma, & Prihar, 1975; Kirda, Nielsen, & Biggar, 1973). Namely, the moisture content of the upper layer fluctuated significantly as the rapid solute front advanced rapidly, so the Cl‾ was not fully leached from the soil. The moisture content of the lower layer increased gradually to saturation while the solute front advanced slowly, so the Cl‾ could be completely leached out. However, the Br‾ profile did not follow the same pattern, probably because the soil had been accumulating Br‾, and less than half of the prepared Br‾ in the infiltration water was absorbed by the soil in Stage I. In Stage III (Figs. 9c and d), since the Br‾ concentration at each depth was close to 0 mg/L, it was difficult to obtain the trend in Br‾ and use it to describe the unsaturated seepage. In the Cl‾ profile, the Cl‾ in the upper layer was almost completely leached, and its concentration was close to 0 mg/L; while in the lower layer, the accumulation intensified, and then the Cl‾ started to become enriched toward the deep vadose zone. The above ion profile analysis reveals that although the apparent moisture content of the vadose zone was unchanged (Fig.10), the Cl‾ continued to be transported downward and its concentration at different depths varied with the water flow, which demonstrates the existence of unsaturated seepage.

4.3 Simulation results

Excessive sampling would affect the internal pressure of the column, resulting in an imbalance between the water and air in the vadose zone, and consequently, the ion profile obtained from the column test would not fully show the unsaturated seepage. Therefore, the transport of Cl‾ during infiltration was simulated. Fig. 11 shows the temporal variation in the Cl‾ trajectory in the longitudinal section and the concentration variation in the 3D soil column obtained from the simulation. At the critical time, the simulation results were basically consistent with the column test results, e.g., Cl‾ reached a depth of 4 m on the 10thday, and remained at a depth of 5 m on the 45th day. The simulated results are idealized, while the column test includes many human and machine errors, e.g., the influence of external air pressure and temperature on the column during the sampling process, the mechanical error when the ion concentration is detected by an ion chromatograph, etc. However, by comparing the test and simulation results for the Cl‾ at different infiltration stages (Fig. 12), we concluded that the trends of the ion profiles are basically consistent, so the model parameters obtained in the column test are applicable to this simulation.
Fig. 13 summarizes the variation in the Cl‾ profile with respect of time from the numerical simulation. In this figure, it can be seen that Cl‾ is enriched at a certain depth every day. In order to show its enrichment law more simply and accurately, the peak of the Cl‾ concentration was extracted to obtain an enrichment curve (Fig. 14). The daily data is shown in Table 3. In Stage I, the solute front rapidly advanced downward, and the daily peak in the Cl‾ concentration varied significantly due to the large fluctuation in the moisture content in the saturated wetting front mentioned in Section 4.2. Interestingly, the peak concentration began to decrease when the enrichment depth reached about 4 m, which is consistent with the depth at which the perched water formed in the column test. Both the solute front and the wetting front changed significantly at 4 m depth because the water began to flow into the deep vadose zone in the form of unsaturated seepage. The advancing rate of the solute front gradually decreased and the peak concentration decreased at an almost constant rate each day in Stage II. The irrigation head was reapplied at the beginning of Stage III, but the solute front and the peak concentration remained basically the same as that in Stage II, which indicates that the second irrigation did not affect the deep unsaturated seepage. In the simulation, on the 30th day, the enrichment depth of the Cl‾ began to increase at a slow rate, and the decreasing rate of the peak concentration only fluctuated slightly. The reason for this may be that the GWT began to rise during this process because of the accumulation of water at the bottom of the column.

5 Discussion

5.1 Tracing the effects of different ions

Because the initial concentration of Cl‾ and Br‾ in the test loess are different, the infiltration water was prepared with different concentrations of Cl‾ and Br‾. This was done to fully balance the tracing effects. Although a large amount of Cl‾ was leached while the Br‾ mainly accumulated in the soil layer in Stage I, as is shown in the Fig. 15, both of the ion profiles exhibit similar tracing effects on the 11th day, both ions have a strong transport ability with water in the loess, and most of the ions are transported rapidly to the bottom of the column, which indicates that both Cl‾ and Br‾ can be used as soil water tracers in the column test. Davis et al. (1980) and Bowman (1984a) also recommend the use of Cl‾ and Br‾ as common soil water tracers, especially Br‾, because Br‾ is usually present in natural water at levels of less than one percent of the Cl‾ concentration and is not absorbed by most soils. During the second irrigation (Fig. 15b), since the Br‾ concentration at each depth was close to 0 mg/L, it was difficult to obtain the law of Br‾ to describe the unsaturated seepage. The reason for this is that it is difficult for the low concentration of Br‾ to precipitate out and be adsorbed onto the surface of the soil during recrystallization, so a large number of Br‾ was not detected in the soil samples. However, Cl‾ exhibits a regular downward transport and its tracing effects are better (Fig.15a), which may be due to its initial concentration in the loess. Of course, the differences mentioned above in the ion tracing effects discussed in this paper are only for the test loess, but different test results can also be obtained because of the ion adsorption capacity of the soil, the effect of convection, hydrodynamic dispersion, etc. Therefore, no in-depth discussion is presented in this paper. However, the above analysis still suggests the following. (1) The adsorption capacity of Cl‾ in the test loess is stronger than that of Br‾. (2) During the in-situ monitoring of the water flow in the vadose zone in Heifangtai, recommend that Cl‾ be used as the soil tracer and Br‾ be used as the water tracer.

5.2 Characteristics of unsaturated seepage

According to the results of the column test and simulation, under the effect of irrigation, the direct infiltration depth of the surface water into the column is less than 4 m and a large amount of water accumulates at a depth 4 m, forming perched water, which acts as an impervious layer and divides the column into two parts. Then, the infiltration into the deep vadose zone is dominated by unsaturated seepage and its characteristics are as follows: (1) the apparent moisture content remains unchanged; (2) it has a low velocity; (3) it is not affected by the second irrigation; (4) the effect of the rising GWT is small and limited.
According to the moisture content profile (Fig. 10), the apparent moisture content remains unchanged during unsaturated seepage, and the remaining characteristics will be discussed in this section. Solute transport is substantially dominated by hydrodynamic dispersion (Bear, 2013), so for a large velocity, mechanical dispersion plays a significant role and distributes the solute along the direction of the average flow velocity. Its effect is related to the pore velocity. In Stage I (Fig. 14), when the soil moisture content (above a depth of 3 m) fluctuates significantly (Fig. 7a), the average velocity in the soil pores is larger, and the solute front can advance downward quickly; when the soil moisture content (below 3 m depth) slowly increases to saturation (Fig. 7a), the average velocity gradually decreases in the pores, and the advancement rate of the solute front decreases. However, in the Stage III, the infiltration is dominated by unsaturated seepage and the apparent moisture content remains unchanged, so the solute front advances at an almost constant rate, which can be seen from the change in the Cl‾ enrichment depth in Table 3. Therefore, unsaturated seepage occurs at low velocities.
For a low velocity, the proportion of molecular diffusion in the total diffusion increases. The peak concentration decreases at a constant rate in Stage II (Table 3), and when the irrigation head is reapplied at the beginning of Stage III, although the molecular diffusion has been aggravated, the peak concentration remains almost the same as that in Stage II, which indicates that the second irrigation does not affect the deep unsaturated seepage.
As the GWT rises, the Cl‾ enrichment depth starts to increase, while the advancement direction of the solute front remains downward. This occurs because the rate at which the GWT rises is very slow and the dilution only affects the transition zone between the saturated zone and the unsaturated zone. Molecular diffusion always makes the solute concentration uniform in the pore channel, so the decreasing rate of the Cl‾ peak concentration only fluctuates slightly. It can be concluded that the influence of the rising GWT on unsaturated seepage is small and is limited to the transition zone.

6 Conclusions

The main conclusions of this study are as follows.
(1) The column test results show that the unsaturated seepage in the deep vadose zone is characterized by a stable apparent moisture content and a low velocity, so it is difficult to investigate it using real-time monitoring of the moisture content and matric suction. However, based on our analysis of the different infiltration stages, the direct infiltration depth of the surface water into the column is less than 4 m, and as the daily irrigation head falls, the existence of unsaturated seepage can be indirectly proven.
(2) The solute transport in the deep vadose zone was investigated using artificial tracers. When the apparent moisture content remains almost unchanged, the regular downward enrichment of the ions demonstrates the existence of unsaturated seepage. In addition, according to the different tracer effects of the chloride and bromide ions, the adsorption capacity of the chloride ions in the test loess is stronger than that of the bromide ions. We recommend that chloride ions be used as the soil tracer and bromide ions be used as the water tracer during the in-situ monitoring of unsaturated seepage in Heifangtai.
(3) To analyze the chloride and bromide ion profiles using the results of the column test and simulation, since the solute transport is essentially affected by hydrodynamic dispersion, the results suggest that the unsaturated seepage is not affected by the second irrigation, and the effect of the rising GWT is small and is limited to the transition zone. The rise in the groundwater table, which is caused by the stable seepage in the loess vadose zone, induces landslides in Heifangtai terrace.
References
Bowman, R. S. (1984a). Evaluation of some new tracers for soil water studies1Soil Science Society of America Journal,  48 (5), 987-993.
http://doi.org/10.2136/sssaj1984.03615995004800050007x
Bowman, R. S. (1984b). Analysis of soil extracts for inorganic and organic tracer anions via high-performance liquid chromatography. Journal of Chromatography A,  285 (3), 467-477.