VZJ sign up for etocs
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 1 February 2008
Published in Vadose Zone J 7:1-9 (2008)
DOI: 10.2136/vzj2006.0177
© 2008 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Kechavarzi, C.
Right arrow Articles by Nikolopoulos, P.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Kechavarzi, C.
Right arrow Articles by Nikolopoulos, P.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Kechavarzi, C.
Right arrow Articles by Nikolopoulos, P.
Related Collections
Right arrow Multiphase Fluid Flow
Right arrow Variably Saturated Fluid Flow

ORIGINAL RESEARCH

Laboratory Study of Immiscible Contaminant Flow in Unsaturated Layered Sands

C. Kechavarzia,*, K. Sogab, T. Illangasekarec and P. Nikolopoulosb

a Natural Resources Dep., Cranfield Univ., MK43OAL, Cranfield, UK
b Dep. of Engineering, Univ. of Cambridge, CB21PZ, Cambridge, UK
c Colorado School of Mines, Golden, CO 80401. USA
* Corresponding author (c.kechavarzi{at}cranfield.ac.uk).

All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.


Received 16 December 2006.



   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Methods and Materials
 Results and Discussion
 Conclusions
 REFERENCES
 
Little quantitative experimental data are available describing the behavior of immiscible contaminants in unsaturated heterogeneous porous media. Such data are, however, essential to the fundamental understanding of the processes governing nonaqueous phase liquid behavior and for the validation of modeling tools. The effect of macro-heterogeneity on light nonaqueous phase liquid (LNAPL) flow and distribution in the unsaturated zone was investigated experimentally by simulating LNAPL spills in layered soil systems consisting of sands with various textures. Two multiphase flow experiments were conducted in a two-dimensional flume (180 x 120 x 8 cm). The vertical distribution of water and LNAPL pressure were measured using hydrophilic and hydrophobic tensiometers. An image analysis technique was used to estimate the saturation distribution of the fluids in a two-dimensional vertical plane. The experiments show that LNAPL entrapment, which contributes to long-term soil and water contamination, depends strongly on the initial water saturation and water pressure at the layer interfaces and on the texture contrasts between the soil layers, which lead to permeability and capillary barrier effects. Thus, the knowledge of the initial water pressure and saturation distribution in unsaturated layered soil formations is critical to the correct prediction of LNAPL infiltration and drainage.

Abbreviations: LNAPL, light nonaqueous phase liquid • NAPL, nonaqueous phase liquid • Pc-S, capillary pressure-saturation


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Methods and Materials
 Results and Discussion
 Conclusions
 REFERENCES
 
Subsurface contamination by organic wastes and hydrocarbons has become a prevalent environmental problem. These contaminants in the form of nonaqueous phase liquids (NAPLs) are only sparingly soluble in water and thus act as long-term sources of groundwater contamination. Releases of NAPLs from accidental spills and leakage are common at industrial sites. Specifically, the spilling of petroleum hydrocarbons that are lighter than water (LNAPLs) is frequent at refineries and storage facilities.

When LNAPLs are released into the soil, some knowledge of the extent of the polluted area is useful in designing strategies for recovery or remediation. In natural soil formation, however, the movement and eventual distribution of such immiscible contaminants are complicated by soil heterogeneity. This can lead to poor remediation effectiveness unless extensive site sampling is performed. The distribution of LNAPL in heterogeneous unsaturated soils depends on the texture variability of the soil formation (Davis et al., 2004). For example, horizontal bedding in the formation results in lateral spreading and entrapment at the textural interfaces. The soil water content in the unsaturated zone also controls the migration and spatial distribution of LNAPL (Schroth et al., 1998a). The soil water content varies in space and time due to spatial variability of soil properties and also with precipitation and fluctuation of the water table. Hence, following a spill, LNAPLs can be expected to be found at residual saturations, in the form of isolated ganglia in coarse-textured units or at texture interfaces in the form of pools. Pooling can also occur on the top of the capillary fringe or over impermeable soil units. These heterogeneously distributed LNAPL sources result in continuously evolving contaminated groundwater plumes with soluble LNAPL components partitioning into infiltrating water in the unsaturated zone and flowing groundwater in the saturated zone. Using laboratory simulations of NAPL flow in heterogeneous synthetic aquifers created in intermediate scale test tanks, Illangasekare et al. (1995a) reported the existence of zones of immobile entrapped NAPL with saturation as high as 75% due to the interface effects between different soil formations.

Hence, from the perspective of site investigation, monitoring, and remediation, knowledge of NAPL saturation distribution in the unsaturated zone is essential. This knowledge relies on the understanding of the contaminant behavior under complex hydrogeological field conditions based on physical and mathematical modeling. Physical modeling at the laboratory scale under controlled conditions is useful for calibrating and validating models, which are used for predicting NAPL distribution at the field scale. The drawback is that laboratory experiments are limited to given sets of boundary conditions. To test the predictive capability of multiphase flow models for general field application, a large number of experiments is needed. In addition, these experiments are complex and time consuming. Thus, early experimental studies of NAPL behavior in the unsaturated zone have been restricted mainly to homogeneous aquifers, often under one-dimensional conditions (Eckberg and Sunada, 1984; Schiegg, 1990; Reible et al., 1990; Thomson et al., 1992, Host-Madsen and Jensen, 1992; Lenhard et al., 1993, Van Geel and Sykes, 1994). Subsequently, a number of two-dimensional laboratory studies of NAPL flow in heterogeneous unsaturated systems were reported (Schwille, 1988; Pantazidou and Sitar, 1993; Hofstee et al., 1998; Schroth et al., 1998a; Wipfler et al., 2004). In addition, Illangasekare et al. (1995a,b) performed experimental simulations of NAPL flow in heterogeneous aquifers for a wide range of boundary conditions. Their experimental studies investigated the effect of layer thickness and contrast in soil texture on NAPL flow pattern in multilayered aquifers. They also studied the effect of coarse- and fine-sand lenses, acting as sinks and barriers, on NAPL entrapment and movement including horizontal migration in systems with a sloping water table. In their experiments, the final NAPL saturation distribution was determined with a gamma attenuation system. Also of significant interest is the experimental investigation by Schroth et al. (1998a,b), who studied the effect of water saturation at the interface of sloping layers with contrasting textures on NAPL flow pattern. Although their observations were mostly qualitative, their study highlighted the importance of initial water saturation conditions on NAPL spatial distribution in layered systems. Nonetheless, there is a lack of two-dimensional heterogeneous studies where both the saturation and the pressure of all the fluids are measured under transient conditions. Transient data sets are essential to assess the adequacy of the empirical constitutive hydraulic relations on which current multiphase flow models rely (Oostrom and Lenhard, 1998; Kechavarzi et al., 2005).

The objective of this experimental study is to investigate quantitatively the effect of textural interfaces on NAPL and water distribution in the vadose zone during three-phase transient flow. A two-dimensional experimental setup was designed to simulate the spill of an LNAPL in unsaturated layered soils. The two-dimensional saturation distribution of both LNAPL and water during transient flow was estimated using an image analysis method (Kechavarzi et al., 2000). Vertical water saturation variations were also continuously monitored using resistivity probes to enable comparison with the image analysis data. In addition, LNAPL and water pressure were measured using a set of tensiometers. Two experiments were conducted. For both experiments, the soil models consisted of a fine-sand layer embedded in a coarse-sand formation; however, the contrast in texture between the layers and the soil formation was varied. In the subsequent sections of this paper, these experiments are referred to as Exp. 1 and Exp. 2, respectively.


   Methods and Materials
TOP
 ABSTRACT
 INTRODUCTION
 Methods and Materials
Results and Discussion
 Conclusions
 REFERENCES
 
Two-Dimensional Soil Tank Models
The two-dimensional flow flume used for the experiments was 180 cm high, 120 cm long and 8 cm wide. The back wall of the flume was made of 19-mm thick Perspex sheet drilled with sensor ports, and the front wall was made of 20-mm thick tempered glass. The soil model in Exp. 1 consisted of a 22-cm thick horizontal layer of uniform fine sand embedded 33 cm below the soil surface in a uniform coarse-sand deposit. The soil model in Exp. 2 consisted of a horizontal layer of the same fine sand, approximately the same thickness (24 cm), embedded 31 cm below the soil surface in a coarse well-graded sand deposit. The geometry and the dimensions of the soil models are shown in Fig. 1 . The saturated soil models were prepared by pouring the sand laterally above de-aired water. The water level in the tank was raised progressively and maintained close to the soil surface (10 cm above the surface) during pouring to avoid excessive particle segregation in water. Before sand pouring, a gravel layer was laid at the base of the tank and two drainage wells were placed on each side of the model. The saturated soil models were drained by lowering the water table to a prescribed height (Fig. 1), which was kept constant throughout the experiments. The models were left to drain for 4.0 d. Four liters of a test LNAPL were then injected under a constant pressure of 1.8 kPa via a line source placed on the center line of the tank approximately 5 cm below the sand surface. The boundary conditions in the two-dimensional flume were a no-flow condition for the water and LNAPL phases at the top of the models and along the two vertical sides from the water table to the top of the model. The pressure at the top of the model and on the two vertical sides (where the drainage wells are located) is atmospheric. Therefore, no liquid flow will occur across those boundaries as long as the capillary pressures are positive, which is the case in practice under unsaturated conditions. Along the bottom of the model and along the two vertical sides, below the water table, a constant water pressure condition was present because of the fixed position of the water table. The water table was used as the datum.


Figure 1
View larger version (32K):
[in this window]
[in a new window]

  		FIG. 1. Schematic of the flow cell experiments (to scale). (a) Exp. 1, (b) Exp. 2.

 
Porous Media and Fluid Properties
The fine and coarse sands used were uniform silica sands corresponding to British Standards (BS-4500) fraction E (particle size ranging from 90 to 150 µm) and fraction D (particle size ranging from 150 to 300 µm), respectively. The well-graded sand was a natural quarry sand called RH65 (particle size ranging from 100 to 600 µm). The particle-size distribution of the sands was obtained by dry sieving. The permeability was measured with a constant head permeameter. The characteristics of the sands are given in Table 1 .


View this table:
[in this window]
[in a new window]

  		TABLE 1. Characteristics of porous media.

 
The test LNAPL used was Soltrol 220 (Phillips Petroleum Inc., Oklahoma City, OK). Soltrol 220, which has negligible solubility in water, a low vapor pressure, and low toxicity, has been used extensively in previous studies and has well-defined properties. It has a density of 0.79 g cm–3 and a viscosity of 0.0048 Pa s. The interfacial tensions between Soltrol 220 and water and Soltrol 220 and air are 0.036 and 0.026 N m–1, respectively (Oostrom and Lenhard, 1998). The colorless LNAPL was dyed red with a hydrophobic dye (Oil RedO, Aldrich Chemical Company Inc., Milwaukee, WI).

The static two-phase capillary pressure–saturation (Pc–S) relationships of the tested sands (air–water, NAPL–water, and air–NAPL) were determined with a porous plate method similar to that described by Host-Madsen and Jensen (1992), Thomson et al. (1992), and Van Geel and Sykes (1994). The measured Pc–S curves, which represent the main drainage branch of the Pc–S relationships, are shown in Fig. 2 . The van Genuchten model (van Genuchten, 1980) was fitted to the data, and the model parameters for the test sands are summarized in Table 2 .


Figure 2
View larger version (15K):
[in this window]
[in a new window]

  		FIG. 2. Two-phase capillary pressure–saturation relations fitted with the van Genuchten model for each sand: (a) fraction E fine sand, (b) fraction D coarse sand, and (c) RH65 well-graded sand.

 

View this table:
[in this window]
[in a new window]

  		TABLE 2. Van Genuchten model parameters based on two-phase capillary pressure–saturation (Pc–S) data.

 
Water–LNAPL Pressure and Saturation Measurements
Water and LNAPL pressures were measured with hydrophilic and hydrophobic tensiometers, respectively. The tensiometers consisted of high nonwetting fluid entry pressure porous stones, 10 mm long and 5 mm in diameter, attached to cylindrical brass tubes (30 mm long and 6 mm in diameter) connected to 1 bar pressure transducers. The ceramic stones used for the LNAPL tensiometers were chemically treated to render them hydrophobic by following the procedure outlined by Lenhard and Parker (1988). Six water tensiometers (Pw1–Pw6) were used for each experiment. Six LNAPL tensiometers (Po1–Po6) were used for Exp. 1, and seven (Po1–Po7) for Exp. 2. Their positions are indicated in Fig. 1.

The method used to estimate water and LNAPL saturation simultaneously using a multispectral image analysis technique, including the possible sources of errors, is described in detail in Kechavarzi et al. (2000). Images of the tank containing the LNAPL plume were taken within two different narrow spectral bands (10 nm large and centered at 500 and 970 nm) using a near-infrared digital camera (Kodak DCS420, Rochester, NY) under constant lighting conditions. Using independent calibration measurements, it was shown that the optical density defined for the reflected luminous intensity was a linear function of the LNAPL and the water saturations for each spectral band and for any two- and three-fluid phase system. This meant that, providing that the relations between optical density and saturation were calibrated for given experimental conditions, two images taken within two different spectral bands provided two equations that could be solved for the water and the LNAPL saturation. The images of the flow domain were discretized into rectangular matrices of 44 rows and 21 columns (924 elements) for both experiments. The average LNAPL, water, and air (by default) saturations were calculated for each element of the matrices using the calibration equations. Contour plots of the saturation of the fluids were made on each matrix. Also, vertical fluid saturation profiles could be plotted for any column of the matrices. Four images taken during water drainage and nine images taken during LNAPL flow were analyzed for each experiment. To assess the overall error on the LNAPL saturation for a given image (i.e., at a given time), the sum of the LNAPL volumes measured for all the elements of the matrix was compared to the known injected LNAPL volume at that time. The average relative error over the nine images between the volume of injected LNAPL and the volume estimated using the image analysis method was 8.7% for Exp. 1 and 5.1% for Exp. 2.

Miniature resistivity probes were used to acquire continuous vertical water saturation data that are complementary to the two-dimensional data obtained with the image analysis method at discrete times. The probes consisted of four in-line platinum wires, 1 mm in diameter, inserted into a plastic tube and soldered to a four-strand cable. The overall length of the probes was 11 mm. The resistivity was calculated from the measured potential drop across the inner pins. The resistivity of the multiphase system was a function of water saturation since water was the major conductive material of the system. Hence, water saturation could be calculated from the measured resistivity of the three-phase system using a calibration equation developed separately for a simple two-phase system. Details of the measurement technique as well as the calibration procedure can be found in Kechavarzi and Soga (2002). Thirteen probes were used for Exp. 1 and 12 for Exp. 2. They are labeled R1 to R13 in Fig. 1. For each experiment and for each image, the water saturation data measured with the image analysis method at the locations where the probes were positioned were compared to the resistivity data. The two methods were found to agree within 8.5% of water saturation.


   Results and Discussion
TOP
 ABSTRACT
 INTRODUCTION
 Methods and Materials
 Results and Discussion
Conclusions
 REFERENCES
 
Initial Conditions before the LNAPL Spills
Water drainage of the saturated models was initiated at time t = 0. Figure 3 shows the vertical water saturation profiles measured on the center line of the flume with the image analysis method and the corresponding water pressure profiles for Exp. 1 and Exp. 2 at t = 4.0 d. The profiles represent the initial water saturation and pressure conditions before the LNAPL spill.


Figure 3
View larger version (24K):
[in this window]
[in a new window]

  		FIG. 3. (a) Water saturation profile after water drainage for Exp. 1 and (b) corresponding water pressure profile; (c) water saturation profile after water drainage for Exp. 2 and (d) corresponding water pressure profile.

 
As shown in Fig. 3b and Fig. 3d, the water pressure distribution for both experiments was not hydrostatic before the LNAPL spill at t = 4.0 d. Kechavarzi et al. (2005) showed that for a homogeneous fine-sand formation (fraction E) and the same boundary conditions, hydrostatic distribution was achieved at t = 4.0 d and that water saturation was residual in the upper part of the profile after only 1.0 d of drainage. This discrepancy arises because the capillary barrier effect occurring at the lower textural interface between the fine-sand layer and the bottom coarse-sand deposit. The capillary barrier effect is controlled by the contrast in the hydraulic properties of the layer and the soil formation. Because of the lower air-entry pressure of the coarse sands compared with the fine sand (Fig. 2), the bottom coarse-sand deposits start to drain before the fine-sand layer. As the water saturation below the lower textural interface decreases, the relative permeability of water becomes low. This results in a capillary barrier effect with slow water drainage across the interface, preventing static conditions from being achieved even after long periods of time. As shown in Fig. 3a and Fig. 3c, the capillary barrier effect is characterized by a discontinuity in water saturation across the lower textural interface. This is because, as long as the water phase remains continuous, there must be continuity in pressure across the interface. Equality in pressure across the interface implies that a discontinuity in saturation exists because the two sands are characterized by contrasting Pc–S relations (Fig. 2). The magnitude of the discontinuity depends on the contrast between the Pc–S relations and on the value of the capillary pressure at the interface. It is worth noting, however, that for Exp. 1 the gradual decrease in water saturation across the lower interface from 0.49 in the fraction E sand to 0.33 and 0.16 in the fraction D sand (Fig. 3a) was unexpected. By continuity in pressure and according to the water pressure measured at Pw3 and Pw4, the value of Pc at the interface is between 70 and 78 cm. Hence, the water saturation calculated using the Pc–S relations (Fig. 2) should be between 0.29 and 0.49 immediately above the interface and between 0.15 and 0.16 immediately below it. Whereas the water saturation above the interface was predicted correctly using the Pc–S relations, the measured value of 0.33 immediately below the interface indicates a transition zone with intermediate retention properties. This was likely the result of interlocking of fine- and coarse-sand particles that occurred when pouring the fine uniform sand over the coarse uniform sand. This was not observed for Exp. 2.

Experimental Results of the LNAPL Spills
The data generated during the experiments are the LNAPL and water pressure measured continuously at discrete locations on the center line of the flume with the tensiometers. Water saturation was also measured continuously at discrete locations on the center line of the flume with the resistivity probes. In addition, the two-dimensional water, LNAPL, and air saturation (by default) distribution at discrete times was measured with the image analysis method. For conciseness, we limited the graphical description of the data to LNAPL saturation contours and vertical water and LNAPL saturation and pressure profiles at selected times representing key stages in the migration of the LNAPL. Also to illustrate the effect of the layers on the permeability of LNAPL, the LNAPL permeability in each sand type was approximated by the expression developed by Parker et al. (1987) and plotted as a function of depth for both experiments:

Formula 1[1]
where kro is the relative permeability of the LNAPL; Formula 1t=Formula 1, and Formula 1w=Formula 1, with Sr, St, Sw, and So being the irreducible water saturation, the total liquid saturation, the water saturation, and the LNAPL saturation, respectively; and m = 1 – 1/n, where n is the parameter obtained for each sand type by fitting the van Genuchten model (van Genuchten, 1980) to the air–water Pc–S data (Table 2).

Experiment 1
Figure 4a shows four of the contour plots of the LNAPL saturation measured with the image analysis method for Exp. 1. The corresponding water and LNAPL saturation profiles measured on the center line of the flume are plotted on Fig. 4b. Figure 5 represents the LNAPL and water pressure profiles at the same times. Figure 6a shows the LNAPL permeability profiles at the same times.


Figure 4
View larger version (48K):
[in this window]
[in a new window]

  		FIG. 4. (a) Light nonaqueous phase liquid (LNAPL) saturation contours for Exp. 1; (b) corresponding LNAPL and water saturation profiles measured on the center line of the flume during LNAPL flow.

 

Figure 5
View larger version (18K):
[in this window]
[in a new window]

  		FIG. 5. (a) Light nonaqueous phase liquid (LNAPL) and (b) water pressure profiles for Exp. 1 (note that x axes have different scales).

 

Figure 6
View larger version (25K):
[in this window]
[in a new window]

  		FIG. 6. Light nonaqueous phase liquid (LNAPL) permeability profiles for (a) Exp. 1 and (b) Exp. 2.

 
For clarity we define the time at which the LNAPL spill was initiated as t = 0. The experiment lasted 2.15 d. It took 0.24 d to inject the 4 L of LNAPL. At the beginning of injection, the LNAPL plume in the upper coarse-sand deposit was narrow and oblong (Fig. 4a), indicating that the LNAPL pressure gradients due to the injection head and the capillary forces were small compared with gravitational potential gradients and that the flow was mainly gravitational. The maximum measured total liquid saturation in the coarse sand during injection (St = 0.86 at t = 0.03 d) occurred below the injection source, and because of the positive LNAPL pressure head at the source, the sand there was likely to be apparently saturated with liquid. Therefore, Sa = 1 – St = 0.14 below the source likely corresponds to entrapped air.

The arrival of the LNAPL front at the upper textural interface was characterized by a reduction in the LNAPL permeability (Fig. 6a), large LNAPL pressure gradients (Fig. 5a) across the interface (gradient of 1.5 at t = 0.07 d between Po1 and Po2), an increase in LNAPL saturation at the front (Fig. 4b), and an increase in the width of the plume (Fig. 4a). The large LNAPL pressure gradients were the result of the larger capillary pressure between air and LNAPL in the fine sand compared with that in the coarse sand, which resulted in the lateral spreading of the plume in the fine-sand layer. The LNAPL saturation across the upper interface was high in both sands (0.76 and 0.81 on the center line at t = 0.07 d in the coarse and the fine sand, respectively). The LNAPL pressure measured at Po1 and Po2 at t = 0.07 d was –11 cm and –16 cm, respectively. This indicates that the capillary pressure between air and LNAPL across the interface was below the air-entry value of the air–LNAPL fluid system for both sands (Table 2 and Fig. 2). If it is assumed, following the reasoning by Parker et al. (1987), that the total liquid saturation can be approximated from the two-phase air–LNAPL Pc–S curves (Fig. 2), then both sands should be apparently saturated with liquid. The measured total liquid saturation at Po1 and Po2 at t = 0.07 was 0.85 and 0.91, respectively. Therefore, the entrapped air saturation at these two locations was 0.15 and 0.09, respectively. This indicates that hysteresis and nonwetting fluid entrapment was significant during the LNAPL infiltration stage. The nearly saturated liquid flow across the interface explains the lack of marked discontinuity in LNAPL saturation that would be expected during unsaturated flow in layered soils with contrasting Pc–S curves. In addition, the contrast in saturated permeability between the sands was relatively small, and the permeability barrier effect was limited to a small increase in the plume width in the fine sand. Thus, there was no accumulation and preferential spreading of LNAPL above the interface. The LNAPL gauge pressure remained negative and the LNAPL flow remained mainly vertical.

Because of the change in interfacial tension when switching from a two-phase to a three-phase system, the water pressure increased at all tensiometer locations at the arrival of the LNAPL front (Fig. 5b). The water pressure was maintained at approximately a constant value until t = 0.24 d, at which time the water pressure started to decrease again. This time corresponds to the beginning of LNAPL redistribution. The decrease in interfacial tension between water and LNAPL compared with water and air resulted momentarily in an increase in vertical water pressure gradients and in vertical water displacement (Fig. 4b). However, above the upper textural interface, water displacement was small because the water saturation was residual. At t = 2.15 d, the water pressure had decreased and the water pressure profile (Fig. 5b) was identical to the initial water pressure profile before the LNAPL spill (Fig. 3b).

As the LNAPL penetrated the layer and moved toward the lower interface, the relative permeability of the LNAPL was reduced further because of the increasing water saturation (Fig. 6a). In addition, water was being displaced, resulting in an increase in water saturation below the LNAPL front (Fig. 4b). At t = 0.13 d, the LNAPL front was just above the lower textural interface, where the water saturation on the center profile increased from 0.49 to 0.58 (Fig. 4b). For the LNAPL to cross the interface and penetrate the underlying air-filled pores in the coarse sand, the LNAPL pressure has to increase and exceed the entry pressure of the coarse sand. The pressure at Po3 was –14 and –16 cm at t = 0.13 d and t = 0.24 d, respectively, which is just below the air-entry pressure of the air–LNAPL Pc–S curve of the coarse sand (Table 2 and Fig. 2b). This increase in LNAPL pressure resulted in low LNAPL pressure gradients in the fine-sand layer between Po2 and Po3 (0.8 and 0.7 at t = 0.13 d and t = 0.24 d, respectively). Despite the increase in LNAPL pressure above the lower interface (Fig. 5a), the capillary pressure between water and LNAPL remained low because the water pressure increased at the same time. Hence, the water saturation at the interface remained relatively high. There was visibly no water being displaced vertically across the interface (Fig. 4b). However, two-dimensional water saturation contours (data not shown) suggest that some water was being displaced horizontally along the interface from the center of the flume. At t = 0.24 d, the water saturation on the profile at the interface had reduced to 0.43, whereas the LNAPL saturation was 0.48. The LNAPL pressure across the interface (–16 cm at Po3 and –13 cm at Po4 at t = 0.24 d) was lower than the air-entry pressure of the air–LNAPL Pc–S curves of both sands, which means that both sands were apparently saturated with liquid. The total liquid saturation across the interface at t = 0.24 d was 0.91 in the fine-sand layer and 0.83 the coarse-sand layer. However, the capillary pressure between water and LNAPL (51 cm at Po3–Pw3 and 45 cm at Po4–Pw4) resulted in a higher LNAPL saturation in the coarse sand below the interface (0.69 at t = 0.24 d) than in the fine sand above the interface (0.48 at t = 0.24 d). This corresponds to an increase in LNAPL relative permeability across the interface (Fig. 6a) and low pressure gradients (Fig. 5a), as shown by the pressure at Po3 and Po4 (gradient of 0.7 at t = 0.24 d).

At t = 0.24 d, the redistribution of the LNAPL began. The LNAPL saturation behind the drainage front continued to decrease during redistribution with the LNAPL saturation in the upper coarse-sand deposit, decreasing more rapidly than in the fine-sand layer. The LNAPL plume continued its migration in the bottom coarse sand deposit under gradients close to gravitational gradients (Fig. 5a), and its width reduced (Fig. 4a). At t = 0.44 d (data not shown), the LNAPL reached the top of the capillary fringe and began to spread horizontally above the saturated zone in the form of a lens, eventually depressing the fringe (t = 2.15 d in Fig. 4).

The final LNAPL saturation profile (t = 2.15 d in Fig. 4b) is similar to that observed by Hofstee et al. (1998) and Oostrom et al. (2003), who performed comparable three-phase experiments in columns. Notably, the LNAPL saturation remaining at the lower textural interface was small and close to residual. This is in contrast with Walser et al. (1999), who performed a two-phase air–LNAPL column experiment where the final wetting fluid (i.e., LNAPL) distribution showed a large amount of LNAPL remaining at the interface of a fine-sand layer overlaying a coarse-sand layer. This is because, in their experiment, no water was present initially at the interface, and the LNAPL acted as the wetting fluid. In the three-phase system, the total liquid saturation at the lower interface is controlled by the capillary pressure between LNAPL and air. For any significant LNAPL flow to take place across the interface in the larger air-filled underlying pores, the capillary pressure between LNAPL and air has to decrease to values lower than the entry pressure of the coarse sand. This corresponds to an increase in LNAPL saturation as LNAPL accumulates at the interface and to a large total liquid saturation. However, because of the large initial water saturation at the interface, the increase in LNAPL saturation and the subsequent LNAPL saturation retained at the interface will be small compared with a system with residual water saturation or no water at all as in Walser et al. (1999). In this experiment, if the water pressure distribution had been hydrostatic, the water saturation at the lower interface would have been residual according to the two-phase air–water Pc–S curve of the fine sand (Fig. 2), and the volume of LNAPL retained at the interface would have been larger. Hence, predicting the LNAPL saturation distribution by assuming water to be hydrostatic even after a long period of drainage would lead to overestimating the amount of LNAPL retained at the interface and underestimating the volume of LNAPL reaching the capillary fringe.

Experiment 2
Experiment 2 differed from Exp. 1 by the texture of the sand deposit in which the fine-sand layer (fraction E) was embedded. In Exp. 1, the coarse-sand deposit (fraction D) containing the layer was uniform and had a saturated permeability 3.5 times larger than the fine sand (fraction E). In Exp. 2, the well-graded sand deposit (RH65) had a similar D50 to that of the coarse sand (fraction D) used in Exp. 1, but its saturated permeability is close to that of the fine-sand layer (Table 1). Thus, the difference between the air-entry pressure of the layer and the deposit was similar in both experiments, but the contrast in particle size distribution and in permeability was different.

Figure 7 represents the contour plots of the LNAPL saturation measured with the image analysis method at four different times for Exp. 2 with the corresponding water and LNAPL saturation profiles measured on the center line of the flume. Figure 8 represents the LNAPL and water pressure profiles at the same times. Figure 6b is the variation in LNAPL permeability as a function of depth at the same times calculated using Eq. [1].


Figure 7
View larger version (48K):
[in this window]
[in a new window]

  		FIG. 7. (a) Light nonaqueous phase liquid (LNAPL) saturation contours for Exp. 2; (b) corresponding LNAPL and water saturation profiles measured on the center line of the flume during LNAPL flow.

 

Figure 8
View larger version (18K):
[in this window]
[in a new window]

  		FIG. 8. (a) Light nonaqueous phase liquid (LNAPL) and (b) water pressure profiles for Exp. 1 (note that x axes have different scales).

 
Experiment 2 lasted 2.00 d. It took 0.28 d to inject the 4 L of LNAPL during Exp. 2 as opposed to 0.24 during Exp. 1; and it took 0.03 d for the plume to reach the fine-sand layer in Exp. 1, and 0.07 d in Exp. 2. This difference is due to the lower permeability of the RH65 sand compared with the fraction D sand. In the top coarse-sand deposit, the lower permeability of the RH65 sand resulted in higher pressure gradients and a larger horizontal spreading of the LNAPL plume (Fig. 7a) than in Exp. 1 (Fig. 4a). Also, the LNAPL front during injection was less sharp than in Exp. 1 (Fig. 4b and 7b) because the RH65 sand was less uniform than the fraction D sand.

The coarse RH65 sand above the upper interface was less saturated with fluid than in Exp. 1 (So = 0.54 at t = 0.12 d). However, similar to Exp. 1, the capillary pressure between air and LNAPL at the interface, as indicated by the pressure at Po2 (–11 cm at t = 0.12 d), was lower than the air-entry pressure of the fraction E sand (Table 2 and Fig. 2). This explains the sharp increase in LNAPL saturation at the interface in the fine sand (So = 0.78 at t = 0.12 d), which was apparently saturated with liquid (St = 1 at t = 0.12 d). Hence, in contrast to Exp. 1, there was no marked effect of the layer on the geometry of the LNAPL plume in Exp. 2 (Fig. 7a). This is because both sands in Exp. 2 had the same saturated permeability and, as opposed to Exp. 1, the relative permeability of the LNAPL was larger in the fine sand at the interfaces than in the RH65 sand (Fig. 6b). This resulted in smaller pressure gradients (Fig. 8a) across the upper interface (gradient of 1.2 at t = 0.12 d between Po1 and Po2) and less vertical resistance to the flow of LNAPL. Therefore, as opposed to Exp. 1, there was no permeability barrier effect at the upper interface.

At the lower interface of the layer, the initial water saturation was not as high as in Exp. 1 (0.33 as opposed to 0.49) because the capillary barrier effect during water drainage was not as pronounced as in Exp. 1. Therefore, the LNAPL relative permeability (Fig. 6) and the LNAPL saturation at the interface (Fig. 7) were not as low as in Exp. 1. There was a capillary barrier effect on the LNAPL saturation distribution that was similar to the effect observed on water during drainage. Similar to Walser et al. (1999), this resulted in some LNAPL remaining at the interface. After redistribution (t = 2.00 d), the LNAPL saturation at the lower textural interface of the layer was still 0.36, whereas it was close to 0.20 in the rest of the layer.

At t = 2.00 d, LNAPL pressure gradients (Fig. 8a) across the upper textural interface and in the fine-sand layer were still high (between 1.2 and 1.5). In the bottom coarse-sand deposit, they varied between 0.9 and 1.2, indicating that flow was mainly gravitational. However, across the lower interface, the hydraulic gradient between Po3 and Po4 was close to 0, indicating very little LNAPL flow across the lower interface.

Similar to Exp. 1, there was an increase in water pressure during LNAPL imbibition (Fig. 8b). The water pressure began to drop at the beginning of LNAPL redistribution and the start of LNAPL drainage (t = 0.28 d). At t = 2.00 d, the water pressure had decreased (Fig. 8b) but remained slightly higher than the initial water pressure profile before the LNAPL spill (Fig. 3d). Some vertical water displacement occurred (Fig. 7b), but it was more pronounced in the lower part of the flume where initial water saturation was higher. After redistribution, the water saturation below the lower textural interface remained larger than residual, whereas large volumes of water were displaced in the capillary fringe, which had been depressed by the LNAPL plume (Fig. 7b).

Compared with Exp. 1, Exp. 2 shows that the contaminant behavior depends on the contrast in texture and permeability in layered systems. The contrast in the hydraulic functions characterizing the layers affects the initial water pressure and saturation distribution on which the LNAPL behavior largely depends.


   Conclusions
TOP
 ABSTRACT
 INTRODUCTION
 Methods and Materials
 Results and Discussion
 Conclusions
REFERENCES
 
Two two-dimensional multiphase flow experiments were conducted to investigate the effect of textural interfaces on water and LNAPL pressure and saturation distribution during infiltration and drainage. The LNAPL saturation and pressure distribution depends on the contrast in texture and in permeability between the soil layers. The initial water saturation and pressure distribution is also determined by the contrast in the hydraulic properties of the soil layers and strongly affects the LNAPL behavior at the textural interfaces. The results show an important capillary barrier effect on water and LNAPL flow at the interface of fine-sand layers overlaying coarse-sand deposits. The low unsaturated hydraulic conductivity of the lower coarse-sand layers resulted in very small flow across the interfaces and in some cases semipermanent retention of the fluids, preventing static pressure conditions from being achieved even after long periods of time. The initial water pressure distribution was not hydrostatic. This led to high initial water saturation at the lower interface of the fine-sand layer, which had a strong influence on the LNAPL final saturation distribution. Therefore, assuming static initial conditions in layered systems can lead to erroneous predictions of final LNAPL distribution. This has important implication because water saturation distribution at the field scale is usually unknown before a spill event and is affected by rain, recharge water, and water table fluctuations. Furthermore, very small variations in texture can have a nonnegligible effect on saturation distribution. Those variations in texture are likely to be numerous and larger in the field where finer-textured soils are commonly encountered.

Finally, monitoring the pressure and saturation distribution of NAPL, water, and air during transient flow in three-fluid phase flow experiments is essential for testing the predictive capability of numerical models. These experiments can be used to calibrate numerical models, and numerical studies should be performed to investigate the influence of different boundary and initial conditions, as well as varied contrasts in soil texture and geometry of heterogeneities on LNAPL flow in the vadose zone.


   ACKNOWLEDGMENTS
 
Funding of this work was provided by the Marie Curie Research Training Grant ERBFMBICT972658 from The European Commission, DG XII, under the Training and Mobility of Researchers (TMR) program.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Methods and Materials
 Results and Discussion
 Conclusions
 REFERENCES
 




This article has been cited by other articles:


Home page
 Vadose Zone JHome page
I. Hincapie and P. Germann
Gravity-Driven Viscous Flow in Sand Boxes Assessed with Neutron Radiography
Vadose Zone J., November 17, 2009; 8(4): 891 - 901.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Kechavarzi, C.
Right arrow Articles by Nikolopoulos, P.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Kechavarzi, C.
Right arrow Articles by Nikolopoulos, P.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Kechavarzi, C.
Right arrow Articles by Nikolopoulos, P.
Related Collections
Right arrow Multiphase Fluid Flow
Right arrow Variably Saturated Fluid Flow

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome