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


     

Published online 1 August 2007
Published in Vadose Zone J 6:483-495 (2007)
DOI: 10.2136/vzj2006.0172
© 2007 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 Similar articles in ISI Web of Science
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 ISI Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Fagerlund, F.
Right arrow Articles by Niemi, A.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Fagerlund, F.
Right arrow Articles by Niemi, A.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Fagerlund, F.
Right arrow Articles by Niemi, A.
Related Collections
Right arrow Spatial Variability
Right arrow Multiphase Fluid Flow
Right arrow Experiment Design

ORIGINAL RESEARCH

Nonaqueous-Phase Liquid Infiltration and Immobilization in Heterogeneous Media: 2. Application to Stochastically Heterogeneous Formations

F. Fagerlunda,*, T.H. Illangasekareb and A. Niemia

a Dep. of Earth Sciences, Uppsala Univ., Villavägen 16, 75236 Uppsala, Sweden
b Center for Experimental Study of Subsurface Environmental Processes (CESEP), Environmental Science and Engineering, Colorado School of Mines, Golden, CO 80401-1887
* Corresponding author (ffagerlund{at}mines.edu).

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 24 November 2006.



   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Theory and Methods
 Results and Discussion
 Conclusions
 REFERENCES
 
We performed a series of well-controlled laboratory experiments investigating the infiltration and subsequent immobilization of nonaqueous-phase liquid (NAPL) in saturated heterogeneous media. A system of two distinct aquifer zones separated by a dipping interface was considered. Heterogeneity was represented by a spatially correlated random field with known geostatistical parameters in one zone in combination with a homogenous packing of the other zone. The effects of heterogeneity on NAPL flow and entrapment in each of the two zones were investigated. The time-varying NAPL saturations were continuously monitored in space and the final static entrapment–saturation distribution was accurately measured. The immobilized-NAPL distribution contributes to plume generation from source zones. The results show that capillary barriers produced by the small-scale heterogeneity strongly influenced the migration paths and the final distribution of NAPL both in space and across different saturation ranges. The NAPL was immobilized both by snap-off to discontinuous blobs and ganglia and by capillary barriers at textural interfaces. Heterogeneity generally increased entrapment, because spatial variations in capillary properties caused NAPL to be entrapped at higher saturations. Heterogeneity in the finer formation provided points of entry into this formation where the NAPL subsequently could spread as the pressure built up. The NAPL was immobilized at high saturations because high displacement pressures in the fine materials inhibited flow at low saturations. The accessibility for water flow through NAPL occurrences and thereby also the dissolution of NAPL is limited by (i) high entrapped NAPL saturations that decrease the aqueous-phase relative permeability and (ii) the location of NAPL inside a formation with low average permeability.

Abbreviations: DNAPL, dense nonaqueous-phase liquid • GPR, ganglia/pool ratio • NAPL, nonaqueous-phase liquid


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Theory and Methods
 Results and Discussion
 Conclusions
 REFERENCES
 
We performed a two-part experimental study of dense nonaqueous-phase liquid (DNAPL) infiltration and immobilization in heterogeneous saturated media. In the first part of the study (Fagerlund et al., 2007), the methodology to continuously monitor NAPL saturations in space and time was developed. The methodology was demonstrated for a reference case with a packing configuration consisting of two homogeneous layers with a dipping interface. The second part of the study focused on the generation of a comprehensive data set for various combinations of heterogeneity of the two layers. This data was used to investigate the effects of stochastic geological heterogeneity on the flow and entrapment of DNAPLs in saturated media.

While the motivation for this research is described in more detail in Fagerlund et al. (2007), the background is briefly summarized here. Understanding and possible characterization of the subsurface distribution of DNAPLs are important for the assessment and prediction of dissolved mass flux generation from source zones (Parker and Park, 2004; Phelan et al., 2004; Soga et al., 2004; Bradford et al., 2003; Saenton et al., 2002; Dekker and Abriola, 2000; Oostrom et al., 1999; Powers et al., 1994, 1998; Mayer and Miller, 1996). The location of the NAPL with respect to the groundwater flow field and the interfacial area between the NAPL and the aqueous phase influence the rate of mass transfer. During remediation, mass is removed from the DNAPL source zone, changing the entrapment architecture. Therefore, the architecture of the NAPL entrapment is important to the dissolution process, the severity of the contamination problem as determined by the plume concentrations at downgradient receptors, and successful remediation.

In natural systems, geological heterogeneity at various scales controls the migration, immobilization, and final distribution of DNAPLs as well as the characteristics of the source zones of groundwater contamination. Larger scale features such as lenses and interfaces between different subsurface deposits may direct the general movement of mobile DNAPL and give rise to pools perched on top of low-permeability units. On smaller scales, heterogeneity may produce preferential flow channels and control entrapment of DNAPL as blobs and ganglia. Unstable migration by fingering may also occur. Capillary forces vary with the heterogeneity of the medium and capillarity can therefore have a large influence on migration and entrapment.

In many field situations, the geological heterogeneity encountered is stochastic in character, described through its geostatistical characteristics. Yet our present-day understanding of and capability to model NAPL spreading and immobilization in such systems is limited, mostly due to the lack of well-controlled experimental data. Existing field studies (Kueper et al., 1993; Poulsen and Kueper, 1992) do not provide sufficient characterization of the porous medium to allow, e.g., direct comparison with numerical modeling. Controlled laboratory experiments, on the other hand, have so far concentrated on fundamental but simplified cases of geological heterogeneity. The behavior of DNAPL at horizontal interfaces between homogeneous sands or at horizontal lenses has been studied by Kueper et al. (1989), Illangasekare et al. (1995), Hofstee et al. (1998), and Oostrom et al. (1999). A more complex system was studied by Compos (1998), who investigated NAPL flow and entrapment in a spatially correlated heterogeneous permeability field, constructed by rectangular blocks of five different sands. In all five studies, however, the dynamic behavior of the NAPL was only monitored by visual inspection of the NAPL front, and NAPL saturations were measured only at the end of the experiments, using dual-energy gamma-attenuation techniques.

In this study, the effects of known heterogeneity on NAPL migration and immobilization in a formation consisting of two sand zones with a dipping interface were studied. Building on the reference case presented by Fagerlund et al. (2007), two experiments with complex heterogeneity were studied. In the first experiment, the coarser zone where the NAPL was injected was packed following a pregenerated stochastically heterogeneous permeability field, while the finer zone was homogeneous. In the second experiment, the finer zone was packed as a heterogeneous field and the coarser zone was homogeneous. Using the methodology developed in Fagerlund et al. (2007), NAPL saturations were monitored continuously in space and time and the process of NAPL immobilization in these heterogeneous systems was analyzed. The data set presented is intended both for improving our fundamental understanding of the dominant processes and as the basis for validating the relevant conceptual and numerical models under conditions of complex geological heterogeneity.


   Theory and Methods
TOP
 ABSTRACT
 INTRODUCTION
 Theory and Methods
Results and Discussion
 Conclusions
 REFERENCES
 
Entrapment at Capillary Barriers
In the saturated zone, NAPL (the nonwetting phase) is entrapped through two main processes. When water imbibes and NAPL saturations decrease, some of the NAPL is snapped off and becomes immobile as discontinuous blobs and ganglia, occluded by water. The NAPL can also become immobile in a continuous state, however, entrapped at capillary barriers.

For a given medium, capillary pressure, Pc (M L–1 T–2), can be related to fluid saturations, S, and the saturation history. The Brooks and Corey (1964) and van Genuchten (1980) models are two common functional forms to describe Pc–S relations. Regarding the immobile wetting phase in the smallest pore spaces (irreducible saturation) as essentially part of the solid medium, the effective water saturation, Sw, and effective NAPL saturation, Sn, are defined as

Formula 1[1]
where Sw is the water saturation, Sn is the NAPL saturation, and Swr is the irreducible or residual water saturation. For a NAPL–water system, where water is the wetting phase, the capillary pressure between the NAPL and water, Pcnw (M L–1 T–2), is, according to the Brooks and Corey (1964) model,

Formula 2[2]
or

Formula 3[3]
where Pd (M L–1 T–2) is the displacement pressure and {lambda} is a curve-fitting parameter related to the pore size distribution. In the alternative form (Eq. [3]), the displacement pressure for a reference system (e.g., air and water) Pdref (M L–1 T–2) is scaled using the scaling factor ß, which, following Parker et al. (1987), usually is taken as the ratio of interfacial tensions between the reference system and system of interest (in this case NAPL–water). Both Pd and {lambda} are parameters that are related to the solid medium. A region of higher Pd constitutes a capillary barrier to NAPL (nonwetting phase) flow. The value of Pcnw increases with higher Pd and lower Formula 3w. Thus, for NAPL flow from a region of lower Pd (coarser sand) to a region of higher Pd (finer sand), the water saturation in the low-Pd region needs to be low enough (i.e., Sn needs to be high enough) to produce equal or higher Pcnw. The values of Pd and {lambda} are, however, subject to hysteresis and dependent on the saturation history. When NAPL displaces water (drainage), Pd and {lambda} are generally higher than when water displaces NAPL (imbibition).

Entrapment by snap-off can be related to saturation history by the Land (1968) model, which has been used in hysteretic models of Pc–S relations by, for example, Parker and Lenhard (1987), Lenhard (1992), and Gerhard and Kueper (2003). The entrapment model by Land (1968) is described in Fagerlund et al. (2007).

Experimental Setup
The experimental setup was generally the same as for the reference case and is described in more detail in Fagerlund et al. (2007), where also the concept of the upside-down equivalent system is described and a conceptual figure of the experimental flume is presented. To investigate the effects of heterogeneity in each of the two sand zones, two experiments, referred to as Exp. 2 and 3 (Exp. 1 is the reference case in Fagerlund et al. [2007]), were conducted. In Exp. 2, the lower coarser zone was packed as a heterogeneous field represented by a spatially correlated random field with known geostatistical parameters, while the upper coarser zone was kept homogeneous. In Exp. 3, the upper finer zone was heterogeneous and the lower zone was homogeneous. For these heterogeneous experiments, the flume had inner dimensions of 71 by 53 by 4.7 cm. As for the reference case, the dipping interface between the two sand zones was set at an angle of 3.25° and there was a slow water flow from right to left. The NAPL injection parameters and hydraulic gradients for the different combinations of homogeneous and heterogeneous sand zones are summarized in Table 1. The reference case is included for comparison. The light NAPL used was a mixture of Soltrol 220 spiked with iodoheptane (10% w/w) and colored red with Sudan IV (0.1% w/w; see Fagerlund et al., 2007).


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

  		TABLE 1. Experiment configurations and fluid flux parameters.

 
The origin of the coordinate system was defined as the center of the lead collimator in the left constant-head well, in the lower left corner of the flume. With x as the horizontal and z as the vertical axis, the injection coordinates are (x,z) = (50.5,5.9) for Exp. 2 and (x,z) = (51.2,5.7) for Exp. 3. The NAPL was injected along a line source throughout the depth (y dimension) of the flume, creating a point source in two dimensions. The injection locations are shown in Fig. 1 and 2.


Figure 1
View larger version (31K):
[in this window]
[in a new window]

  		FIG. 1. Spatial distribution of sand types in Exp. 2. The color scale shows the different sand types defined by effective sieve size: dark red (2.36 mm) is the coarsest sand; blue (0.212-mm sand) is the finest. The injection point is shown as a white x on black background.

 

Figure 2
View larger version (30K):
[in this window]
[in a new window]

  		FIG. 2. Spatial distribution of sand types in Exp. 3. The color scale shows the different sand types defined by effective sieve size: red-orange (1.18 mm) is the coarsest sand; dark blue (0.136 mm) is the finest. The injection point is shown as a white x on black background.

 
Stochastic heterogeneity was implemented by discretizing the heterogeneous zone into blocks of 3 cm in the horizontal and 1 cm in the vertical direction. The blocks were packed according to a pregenerated spatially correlated random permeability field, using discrete sands with known properties. The correlation lengths of the fields were 6 and 2 cm in the horizontal and vertical directions, respectively, following an exponential variogram. Packing was done using six different sands, the properties of which are given in Table 2. Permeability and air–water retention properties for all sands used have been measured by Sakaki and Illangasekare (personal communication, 2006). The scaling factor ß needed for an air–water reference system (see Eq. [3]) was taken as {sigma}aw/{sigma}nw = 0.072/0.036 N m–1 = 2.0, where {sigma}aw and {sigma}nw [M T–2] are the air–water and NAPL–water interfacial tensions, respectively. Using this scaling method, the displacement pressures (Pd) in the NAPL–water system can be obtained by dividing by 2 the values given in Table 2. The tank was packed wet, i.e., deaired water was added first, followed by the sand. In the heterogeneous zones, one horizontal layer was packed at a time, maintaining only a small water depth of 1 to 2 cm above the sand to avoid layering. During the packing, the tank was tilted to the same angle as the interface, which facilitated the packing of these tilted heterogeneous zones.


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

  		TABLE 2. Properties of the sands used in this study; Brooks–Corey pore-size distribution index ({lambda}) and air–water displacement pressure (Pd) refer to primary drainage in an air–water system.{dagger}

 
Figure 1 shows the spatial distribution of sand types used in Exp. 2. The lower coarser heterogeneous zone consisted of 24 x 31 = 744 blocks. In the flume, the whole field was dipping at an angle of 3.25° and the edge blocks were cut to fit the tilted permeability field between the vertical boundaries (screens) at the sides. The heterogeneous permeability field was built using sands with effective sieve sizes of 2.36, 1.18, and 0.6 mm, a mixture of 0.6 and 0.3 mm, and 0.212 mm; the relative amounts of these sands were 3, 22, 44, 25, and 6%, respectively. The calculated average permeability, k (L2), was 1.75 x 10–10 m2 [mean ln(k) = –22.5], which is the same as for pure 0.6-mm sand, and the variance of ln(k) was 1. The upper zone was a homogeneous mix of 0.6- and 0.3-mm sands (proportions 2:1). To obtain the desired angle of dip, there was a wedge of 0.212-mm sand underneath the heterogeneous field.

In Exp. 3, the upper finer heterogeneous zone consisted of 24 x 26 = 624 blocks, while the lower coarser zone was homogeneous 0.6-mm sand. The distribution of the different sands used is shown in Fig. 2 . The heterogeneous zone was built using sands with effective sieve sizes of 1.18 and 0.6 mm, a mixture of 0.6 and 0.3 mm, 0.212, and 0.136 mm; the relative amounts of these sands were 5, 26, 41, 24, and 4%, respectively. Its average permeability was 6.56 x 10–11 m2 [mean ln(k) = –23.45], which is the same as for the mix of 0.6- and 0.3-mm sand, and the variance of ln(k) was 1. Above the heterogeneous zone there was a wedge of 0.212-mm sand.


   Results and Discussion
TOP
 ABSTRACT
 INTRODUCTION
 Theory and Methods
 Results and Discussion
Conclusions
 REFERENCES
 
Porosity
Before the NAPL infiltration experiments, the spatial distribution of porosity was measured as described in Fagerlund et al. (2007). A measurement spacing of 3 by 1 cm was used to obtain one measurement point at the center of each heterogeneous sand cell. The spatial distributions of porosity measured using the x-ray system are shown in Fig. 3 .


Figure 3
View larger version (30K):
[in this window]
[in a new window]

  		FIG. 3. Spatial distributions of porosity in the sand flume for Exp. 2 and 3; interface between sand zones shown with a white line.

 
Experiment 2—Heterogeneous Coarse Zone
Spatial Distribution of Nonaqueous-Phase Liquid
In Exp. 2, 600 cm3 of NAPL was injected at (x,z) = (50.5,5.9) in the lower heterogeneous zone for 2.63 h. At the beginning of the experiment, digital images captured the spatial distribution of NAPL saturation better than x-ray measurements, as during this period the NAPL migration was too fast to allow an adequate number of x-ray measurements be taken within a sufficiently short time to obtain a good spatial picture. Therefore, the evolution of spatial NAPL distributions given in Fig. 4 is presented as a combination of digital images for the early part and contour plots based on x-ray attenuation measurements for the latter part.


Figure 4
View larger version (82K):
[in this window]
[in a new window]

  		FIG. 4. Evolution of spatial nonaqueous-phase liquid distribution with time in Exp. 2; x-ray measurement points shown as black x's.

 
As can be seen in Fig. 4a, 4b, and 4c, heterogeneities controlled the path of the NAPL as it migrated upward toward the interface. At 12.38 h (Fig. 4d), the NAPL had started to pool and spread along the interface. At this point, the movement was slow enough to produce a detailed spatial plot based on x-ray measurements. As the NAPL pooled at the interface, the buoyancy force acting on the NAPL was not large enough to drive it through the finer materials upward along the dip to the left; instead the pool extended downward along the dip in the coarser sands toward the right. After approximately 33 h, the NAPL started to move out of the experiment domain into the constant-head well on the right-hand side.

The general movement of mobile NAPL was upward toward the interface and then out through the right constant-head well. At 474 h (~20 d), shown in Fig. 4f, the NAPL was immobile everywhere in the tank. High NAPL saturations were still found in the coarsest sands (2.36 and 1.18 mm). To exit these sands, the NAPL generally had to move through the finer 0.6-mm sand. At such material interfaces, a capillary barrier preventing the flow of NAPL at low saturations exists. In the 0.6-mm sand, NAPL saturations stabilized at values around 0.15 everywhere, except in a thin pool at the interface where higher saturations were found.

It can be concluded that the NAPL was immobilized through two forms of entrapment. In the third coarsest 0.6-mm sand, the NAPL could generally migrate upward without being stopped by capillary barriers to finer sands. Without the presence of capillary barriers, NAPL was immobilized when small blobs and ganglia were snapped off and left behind as discontinuous, immobile NAPL as water imbibed and the NAPL saturation decreased. This type of immobilized NAPL can be referred to as discontinuous entrapped NAPL or ganglia. The second form of entrapment is seen in the coarsest 2.36- and 1.18-mm sands as well as in the 0.6-mm sand at the interface to the finer zone. Here NAPL was entrapped because its movement was restricted by the presence of capillary barriers. Because entrapment occurred without snap-off into discontinuous pieces, this type of immobile NAPL can be referred to as continuous entrapped NAPL or pools.

In Fig. 4f, it can be seen that several larger blobs, disconnected from the main NAPL body, had formed, e.g., at locations (x,z) = (57,17) and (60,20). This shows that, at least in heterogeneous media, the NAPL may be entrapped as relatively large isolated islands as water imbibes, expelling the NAPL. The probable reason is that the imbibing water may split the continuous NAPL body; the NAPL retracts toward each side of the split, leading to the formation of such NAPL islands. The formation of aqueous-phase flow channels around such NAPL islands may promote groundwater-flow bypassing of entrapped NAPL occurrences.

Nonaqueous-Phase Liquid Volume
The total NAPL volume, Vn(t) [L3], as estimated from spatial measurements of Sn, is shown as triangles in Fig. 5 . Based on the period between 12 and 30 h, before the NAPL started to exit the tank and when a large number of spatial measurements were available, the total volume was slightly underestimated compared with the known injected volume of 600 cm3. As for the reference case (Exp. 1, Fagerlund et al., 2007), a slight difference may have resulted in the fine-tuning of the measurement apparatus. Application of a correction factor of 1.02 to all measurements of Sn results in a better estimate of Vn(t) (plotted as circles in Fig. 5). After approximately 33 h, the NAPL started to exit through the right constant-head well, which resulted in a decrease in the recorded total volume. After approximately 100 h, the NAPL volume in the tank stabilized at around 330 cm3, indicating that the NAPL had become immobilized in the experiment domain and no longer exited through the well.


Figure 5
View larger version (14K):
[in this window]
[in a new window]

  		FIG. 5. Total nonaqueous-phase liquid volume Vn(t) inside the experiment domain. Estimates based on original measurements shown as triangles, estimates based on calibrated values shown as circles.

 
Figure 6 shows how the NAPL volume was distributed throughout the different types of sand. The figure shows total volumes recorded in the different sands, which depend, e.g., on the NAPL saturations, the relative abundance of the sands (see Fig. 1), and the migration path of the NAPL. It can be seen that throughout the experiment, the proportions of NAPL in the different sands remained roughly the same. The 0.6- and 1.18-mm sands contained about 40% each and the 2.36-mm and the mixture of 0.3- and 0.6-mm sands roughly 10% each, while essentially no NAPL entered the 0.212-mm sand. How the NAPL distributed itself in the different sands can also be seen by comparing Fig. 4 and 1. Comparing the photograph in Fig. 4c to Fig. 1, it can be seen how the NAPL moved around the areas of the finest sands (the 0.212-mm sand and the mixture of 0.3- and 0.6-mm sands), indicating that only very small NAPL volumes should be found in these sands. The NAPL volume recorded in the 0.3- and 0.6-mm sand mixture (Fig. 6) seems to be somewhat overestimated and, therefore, part of the NAPL shown as residing in this mixture in Fig. 6 should instead be included in the NAPL in the coarser 0.6- and 1.18-mm sands. The probable main cause is smoothing of the sharp front at the edge of the NAPL body in the interpolation used to map the spatial distribution (Fig. 4d–f). The edge of the NAPL body was mainly located along material interfaces between the 0.6- and 0.3-mm sand mixture and either the 0.6-mm sand (most common) or the 1.18-mm sand.


Figure 6
View larger version (17K):
[in this window]
[in a new window]

  		FIG. 6. Total nonaqueous-phase liquid volumes (Vn) in the different types of sand: 2.36-mm sand: red empty circles; 1.18-mm sand: purple squares; 0.6-mm sand: blue triangles; 0.6- and 0.3-mm sand mixture: black x's; and 0.212-mm sand: green filled circles.

 
Nonaqueous-Phase Liquid Saturation as a Function of Time
Figure 7 shows the development of NAPL saturations with time for five selected locations in the flume. The red symbols (circles) correspond to observations in the coarsest 2.36-mm sand. It can be seen that the NAPL behavior and final saturation in this sand can be very different at different locations in the flume and depends on the presence of capillary barriers. The red empty circles correspond to a location close to the interface, where high NAPL saturations still existed when the NAPL had become immobile at the end of the experiment (see Fig. 4f). The red filled circles correspond to a point on the right edge of the plume, in the lower part of an "arm" going to the right, where NAPL saturations were initially high but from where the NAPL pulled back during the later stages of the experiment (see Fig. 4). Comparing measurements throughout the flume, it can also be seen that the peak saturations reached in the beginning of the experiment were generally higher for the coarser sands. The reason is that under the same pressure, the NAPL can access a larger proportion of the pore space when the pore sizes are larger and capillary pressures lower. In Fig. 7, the purple squares correspond to a point in the second coarsest 1.18-mm sand, the blue triangles to the third coarsest 0.6-mm sand, and the black crosses to the medium coarse 0.6- and 0.3-mm sand mixture.


Figure 7
View larger version (23K):
[in this window]
[in a new window]

  		FIG. 7. Nonaqueous-phase liquid (NAPL) saturation with time at five points: in 2.36-mm sand at (x,z) = (41.9,27.2), red empty circles, and at (50.2,14.7), red filled circles; in 1.18-mm sand at (41.7,24.2), purple squares; in 0.6-mm sand at (52.8, 8.5), blue triangles; in a mixture of 0.6- and 0.3-mm sand at (43.8,8.0), black x's.

 
At most points, the NAPL saturations exhibited piecewise linear trends with time, which can also be observed for the points shown in Fig. 7. When the NAPL was not entrapped at very high saturations (e.g., empty circles), the common trend is that the NAPL saturation first rose sharply when the front arrived; after that, Sn decreased linearly at a relatively slow rate until 30 to 40 h after the start of injection. At that point, the rate of decrease became significantly faster, again following an approximately linear trend. Finally, the decrease in Sn abruptly stopped, and Sn stabilized at a constant value.

The change in the rate of decline in Sn is roughly correlated to the time when NAPL started to exit the flume (~33 h). Therefore, a possible explanation for this behavior is that up to the time when the NAPL had found its way out to the constant-head well, the speed of migration was controlled by the front propagation into uncontaminated, fully water-saturated areas. Until this point, water was displaced by NAPL at the NAPL–water front (drainage) and the displacement (entry) pressure here (see also Eq. [2]) affected the pressure gradient across the entire NAPL migration path. The drainage process at the front thus controlled the speed of the NAPL migration. Once the gravel-filled constant-head well was reached, however, there was essentially no capillary resistance to NAPL flow at the front. The migration could therefore speed up and was then controlled by the resistance (effective NAPL permeability) and capillary barriers during the displacement of NAPL by water (imbibition) along the continuous path of NAPL that went all the way to the boundary.

The abrupt stop in saturation decline occurred when the NAPL was snapped off from the continuous NAPL body, and thus became an immobile, discontinuous ganglion. At the point (41.9,27.2)—red empty circles, there is a strong capillary barrier from the coarser 2.36-mm sand where the measurement was taken, to the finer 0.6-mm sand above. Even though continuous NAPL existed over the barrier, the barrier seems to have remained relatively strong during the imbibition process. At the end of the experiment when NAPL saturations had decreased in the lower parts of the flume, there was not enough pressure in the NAPL (buoyancy force) to drive it over the barrier and it therefore remained entrapped at the barrier in a high-saturation, continuous state.

Saturation History and Entrapment
The recording of NAPL saturation as a function of time allows study of the effects of saturation history on the final entrapped saturations. Figure 8 shows the relationship between entrapped NAPL saturations, Snt, and the maximum saturation previously reached, Snmax, for all measurement points. The entrapped saturation is the average saturation during a period from 440 to 500 h when the NAPL had become immobile throughout the experimental flume.


Figure 8
View larger version (28K):
[in this window]
[in a new window]

  		FIG. 8. Entrapped (late-time) nonaqueous-phase liquid (NAPL) saturation as a function of the maximum NAPL saturation previously reached. Measurements taken in 2.36-mm sand are red; in 1.18-mm sand, purple; in 0.6-mm sand, blue; in 0.6- and 0.3-mm sand mixture, black; and in 0.212-mm sand, green.

 
For the third coarsest 0.6-mm sand (blue), a large number of points are located in a band where Snt is between ~0.1 and ~0.2. These points are probably not affected by capillary barriers and Sn can decrease until it becomes discontinuous. Points above this trend are likely to be affected by capillary barriers and here the NAPL thus can be immobilized before it becomes discontinuous. For the second coarsest 1.18-mm sand (purple), a trend of ganglion entrapment may exist roughly between Snt = 0.05 and ~0.18, but it is difficult to say at what value of Snt capillary barriers start to have an effect. The wide ranges of values of Snt that can be observed regardless of Snmax values indicate that the NAPL was entrapped in pools in which saturation may vary. For the 2.36-mm sand (red), relatively few measurement points exist and, being the coarsest sand, most are affected by strong capillary barriers and high entrapped saturation values are observed. In general, at points near the 45° line of equal Snt and Snmax, strong capillary barriers exist, preventing the expulsion of NAPL by imbibing water after Snmax has been reached. Both the highest maximum and highest entrapped saturations occurred in the coarsest sands.

As can be seen when comparing Fig. 1 and 4f, high entrapped saturations in the 0.6-mm sand existed primarily in a pool zone along the interface to the upper finer zone. When all points within 4 cm of the interface are excluded from the analysis, a clear trend in the Snt–Snmax relationship appears, shown in Fig. 9 . Below the pool zone, NAPL could generally move on from areas of 0.6-mm sand without the influence of capillary barriers. The NAPL was entrapped primarily as a discontinuous phase and the models of ganglion (snap-off type) entrapment can be fitted to these data. Shown as a solid black line in Fig. 9, the Land (1968) model yields a RMSE of 0.035, while the constant-entrapment model (dotted red line) has RMSE = 0.040. The maximum entrapped saturation, Snmax, which is a fitting parameter of the Land model, is 0.17, which is the same value that was obtained in the reference case (Fagerlund et al., 2007). For the constant model, the fitted constant value of Snt is 0.15.


Figure 9
View larger version (13K):
[in this window]
[in a new window]

  		FIG. 9. Entrapped saturation (Snt) as a function of maximum saturation (Snmax) in 0.6-mm sand below the pool zone at the interface. Fitted ganglion entrapment models: solid black line, Land (1968) model; dotted red line, constant entrapment model.

 
Nonaqueous-Phase Liquid Distribution across the Saturation Range
Figure 10 shows how the NAPL was distributed across the saturation range for two different times. The color divisions on the bars show volume fractions contained within different sand types. At 12.38 h, the bulk of the NAPL was found at high Sn. The coarser the sand, the stronger was the tendency for the NAPL to exist primarily at high saturations. In the relatively finer 0.6-mm sand and the 0.6- and 0.3-mm sand mixture, the NAPL is more evenly distributed across the saturation range than in the coarser 1.18- and 2.36-mm sands. At 474 h, the NAPL had become immobile and the distribution had shifted toward lower saturations. In the 0.6-mm sand, the distribution has a peak between Sn = 0.10 and 0.17, which corresponds to NAPL entrapped as discontinuous ganglia. In the coarser 1.18- and 2.36-mm sands, the NAPL still predominantly existed at high saturations, although even for these sands there was a slight shift toward lower saturations. For the 1.18-mm sand, there is a band of slightly higher saturations at Sn = 0.10 to 0.16, which possibly corresponds to discontinuous ganglion entrapment.


Figure 10
View larger version (36K):
[in this window]
[in a new window]

  		FIG. 10. Nonaqueous-phase liquid (NAPL) distribution as a function of saturation in different sand types at (a) 12.38 and (b) 474 h: 2.36-mm sand, dark red; 1.18-mm sand, red-orange; 0.6-mm sand, yellow; 0.6- and 0.3-mm sand mixture, cyan; and 0.212-mm sand, blue.

 
The saturation distribution can also be characterized by its ganglia/pool ratio (GPR) (Lemke et al., 2004; Fagerlund et al., 2007), shown in Fig. 11 (circles). Judging from Fig. 10 (474 h), the cutoff value between ganglia and pools is most sensitive to NAPL in the 0.6-mm sand, and is therefore based on Snmax for that sand and taken as 0.17. Even in the 0.6-mm sand both below and above the cutoff value of Sn = 0.17, some mischaracterization of pools and ganglia, respectively, will occur due to the spread in Sn for ganglion entrapment seen for 0.6-mm sand in Fig. 9. The two effects counteract each other, however, and the net effect on GPR is deemed to be relatively small. To demonstrate the sensitivity of the GPR, it has been calculated also for cutoff values of 0.15 (squares) and 0.19 (triangles).


Figure 11
View larger version (14K):
[in this window]
[in a new window]

  		FIG. 11. Ganglia-to-pool ratio (GPR) for a cutoff value of 0.17 (circles). The GPRs for cutoff values 0.15 (squares) and 0.19 (triangles) indicate sensitivity.

 
Experiment 3—Heterogeneous Finer Zone
Spatial Distribution of Nonaqueous-Phase Liquid
Figure 12 shows the evolution of spatial NAPL distribution with time for Exp. 3. Here again, digital images are used to show the front propagation at very early times, whereas at later times x-ray attenuation measurements are given. Care was taken to create an extremely homogeneous packing in the lower zone, which is the reason why the spreading toward the interface was very uniform. The 500 cm3 of NAPL was injected during 4.35 h at the location (x,z) = (51.2,5.7). At the end of the injection (Fig. 12b), the NAPL had reached the interface, where it found a point of entry and started to enter into the upper heterogeneous finer zone.


Figure 12
View larger version (70K):
[in this window]
[in a new window]

  		FIG. 12. Evolution of spatial nonaqueous-phase liquid distribution (Sn) with time in Exp. 3. X-ray measurement points shown as black x's.

 
After 9 h (Fig. 12c), high NAPL saturations existed in the coarser parts (1.18-mm sand) of the heterogeneous zone. A relatively large amount of NAPL remained in the lower homogeneous zone and some spreading along the interface to the right also occurred. After 25 h (Fig. 12d), more NAPL had moved up into the heterogeneous zone. To reach the higher parts of the upper zone, the NAPL had to go through the 0.6- and 0.3-mm sand mixture (see Fig. 2). To do so, however, the NAPL pressure needed to exceed the entry pressure of this sand. Therefore the NAPL slowly distributed itself inside the upper layer and, as more NAPL was supplied from below, saturations and pressure increased. At about 48 h, the NAPL had found its way to the left constant-head well through a small flow channel and started to exit the experiment domain. The first part of this channel can be seen in Fig. 12e, which shows the NAPL distribution after 66 h at coordinates (x,z) = (10,36) to (13,36). The channel, however, extends all the way out to the metal-mesh screen at (5.1,36).

At 385 h (Fig. 12f), the thin flow channel to the left well was no longer active and the NAPL had become immobile everywhere in the tank. The NAPL saturation in the homogeneous 0.6-mm sand below the interface takes a value on the order of 0.15 everywhere, which is consistent with the two previous experiments. In the upper heterogeneous zone, higher NAPL saturations existed, especially in areas of 1.18-mm sand, but relatively high saturations also occurred in 0.6-mm sand, e.g., in the region at the left side where the flow channel begins. Snap-off at larger scale, as was seen in Exp. 2, can be observed on the right side of the NAPL body where the pool just below the interface had retracted, leaving two small blobs behind. Also, a relatively large NAPL occurrence that contained high saturations, and which had its center located approximately at (x,z) = (52,30), had been disconnected from the main NAPL body.

Nonaqueous-Phase Liquid Volume
The total NAPL volume in the flume, Vn(t), estimated from the point measurements of the length of the NAPL, ln, is shown as red triangles in Fig. 13 . When the total injected volume could be recorded properly by the measurement system, the estimate of Vn(t) stabilized at a constant value until at approximately 48 h, the NAPL started to exit the experiment domain through the thin flow channel to the left well, seen in Fig. 12 above. As in the two previous experiments, application of a correction factor to all measurements of ln, in this case 0.96, results in a correct volume estimate at early times, marked with blue circles in Fig. 13. Here the amount of NAPL that exited into the left well at the end of the experiment was also measured as 153 ± 5 cm3. This value corresponds well to the calibrated Vn estimates at the end, when Vn had stabilized at 345 cm3, and supports the assumption that a linear correction of ln (and Sn) is appropriate.


Figure 13
View larger version (13K):
[in this window]
[in a new window]

  		FIG. 13. Total nonaqueous-phase liquid volume Vn(t) inside the experiment domain. Estimates based on original measurements shown as triangles, estimates based on calibrated values shown as circles.

 
Figure 14 shows how the NAPL was distributed within the different types of sands in the flume. Once the NAPL got into the upper heterogeneous zone, it could move from the 0.6-mm sand to other sands. Most of the NAPL lost from the experiment domain was lost from the 0.6-mm sand in the lower homogeneous zone. From here, the NAPL moved up to the heterogeneous zone until only entrapped NAPL, at low saturation and presumably in a discontinuous state, remained in the lower zone. The small flow channel through the 0.6- and 0.3-mm sand mixture to the left constant-head well allowed NAPL to exit the heterogeneous zone as long as the NAPL pressure was high enough to overcome the displacement pressure in this flow channel. Throughout the heterogeneous zone, however, capillary barriers resisting NAPL flow through areas of finer material remained relatively strong, resulting in entrapment at high saturations. No NAPL got into regions of 0.136-mm sand and only insignificant volumes entered the 0.212-mm sand. Both the 1.18-mm sand and the 0.6- and 0.3-mm sand mixture had peaks in Vn relatively close the time of maximum NAPL pressure, which should have occurred at the time that the flow channel to the well developed (48 h) and for some time afterward.


Figure 14
View larger version (19K):
[in this window]
[in a new window]

  		FIG. 14. Nonaqueous-phase liquid volumes (Vn) in the different sand types: 1.18-mm sand, purple squares; 0.6-mm sand, blue triangles; 0.6- and 0.3-mm sand mixture, green empty circles; 0.212-mm sand, cyan filled circles; and 0.136-mm sand, black x's.

 
Nonaqueous-Phase Liquid Saturation as a Function of Time
Figure 15 shows the NAPL saturation as a function of time at four representative points. Blue triangles correspond to a point in the upper part of the homogeneous zone. The behavior is characterized by a sharp rise to Snmax, followed by a roughly linear decline as the NAPL moved into the heterogeneous zone. Eventually Sn abruptly stabilized at a constant value (Snt) when the NAPL at that point was disconnected from the main (continuous) NAPL body. A second point located in the 0.6-mm sand, in the far left of the heterogeneous zone, is marked with blue circles. Here maximum saturation was reached late because the point is far from the injection point. The decline in Sn was slow because, to be expelled from here, the NAPL needed to move out through the narrow flow channel to the constant-head well (see above, e.g., Fig. 12e). When there was no longer enough NAPL pressure to overcome the capillary barrier present in the channel, Sn stabilized at a constant value in a continuous state. Purple squares correspond to a point in the 1.18-mm sand, in a branch of the plume that was not developed until the NAPL pressure had risen in the heterogeneous zone. The branch, which can be seen, e.g., in Fig. 12e and 12f, but not in Fig. 12d (25 h), was a dead end for further NAPL migration. Nonetheless, Sn decreased slightly even here when the pressure decreased elsewhere in the NAPL plume and the "overpressure" in the branch allowed the NAPL to redistribute back to the main NAPL body. Eventually and interestingly, as can be seen in Fig. 12f, the whole branch was snapped off from the rest of the plume, thus losing the possibility for any further decrease in Sn. For measurement points located in the 0.6- and 0.3-mm sand mixture, the maximum saturation typically was low, as most pores in this sand were difficult for the NAPL to enter, and the entrapped saturation was similar to Snmax.


Figure 15
View larger version (21K):
[in this window]
[in a new window]

  		FIG. 15. Nonaqueous-phase liquid (NAPL) saturation with time at four points: in 1.18-mm sand at (x,z) = (50.7,29.6), purple squares; in 0.6-mm sand at (53.1,18.5), blue triangles, and at (18.0,35.5), blue circles; and in a mixture of 0.6- and 0.3-mm sand at (32.9,33.6), black x's.

 
Saturation History and Entrapment
Figure 16 shows the relationship between entrapped (late-time) saturation, Snt, and the maximum saturation previously reached, Snmax, at all points of measurement in the flume. There are two clear trends. In the heterogeneous zone, the entrapped saturation is relatively close to the maximum, whereas in the homogeneous lower zone entrapped saturations are mostly between 0.1 and 0.2 regardless of Snmax. Maximum saturations in the different sands are comparable or somewhat higher than in Exp. 2. Higher Snmax is likely to be a result of the pressure that built up in the heterogeneous zone until the NAPL found its way out to the left well through the previously described small channel. Evidence of this pressure buildup can also be seen by comparing maximum saturations in the 0.6-mm sand (blue dots in Fig. 16) for the two trends mentioned above, i.e., for the 0.6-mm sand in the homogeneous and heterogeneous zones. In the lower homogeneous zone, the maximum recorded Snmax is roughly 0.75, while in the heterogeneous zone, Snmax values of up to 0.85 can be observed, indicating that the NAPL pressure had reached higher values. Also in the 0.6- and 0.3-mm sand mixture and the 0.212-mm sand, higher Snmax and Snt values are recorded than in Exp. 2.


Figure 16
View larger version (28K):
[in this window]
[in a new window]

  		FIG. 16. Entrapped (late-time) nonaqueous-phase liquid (NAPL) saturation as a function of the maximum NAPL saturation previously reached. Measurements taken in 1.18-mm sand, purple; in 0.6-mm sand, blue (homogeneous zone–empty circles, heterogeneous zone–filled); in a mixture of 0.6- and 0.3-mm sand, black; in 0.212-mm sand, green; and in 0.136-mm sand, yellow.

 
Figure 17 shows the Snt–Snmax relationship for points of measurement in the lower homogeneous zone, thus isolating the trend associated with these points also seen in Fig. 16. As for Exp. 1 (Fagerlund et al., 2007) and 2, saturation-history-dependent entrapment for the 0.6-mm sand can be studied. The Land (1968) model is fitted with Snmax = 0.19 and a RMSE of 0.027. The constant model is fitted with Snt = 0.16 and RMSE = 0.031. The slightly higher entrapment observed here compared with the two previous experiments may be a result of a slightly higher degree of homogeneity in the 0.6-mm sand. Having been more evenly distributed, more of the NAPL was retained. The RMSE of the Land model is higher than for Exp. 1 (Fagerlund et al., 2007), which seems to result from the fact that for Snmax values between 0.15 and 0.3, Snt is more variable for Exp. 3 (Fig. 17). The RMSE is, however, lower than for Exp. 2, where the differences between different locations of measurement were generally larger. The difference in RMSE between the constant model and the Land model is relatively small (RMSE difference = 0.004) compared with Exp. 1 (0.009). The number of spatial points used for this analysis in Exp. 3 is more than double that of Exp. 1, but in Exp. 1 the coverage with time at each point was, on the other hand, significantly better (closer spacing between measurements in time).


Figure 17
View larger version (13K):
[in this window]
[in a new window]

  		FIG. 17. Entrapped saturation (Sn) as a function of maximum saturation (Snmax) in 0.6-mm sand below the pool zone at the interface. Fitted ganglion entrapment models: solid line, Land (1968) model; dotted line, constant entrapment model.

 
Nonaqueous-Phase Liquid Distribution across the Saturation Range
The NAPL distributions across different saturations at 9 and 385 h are shown in Fig. 18 . In the beginning, all NAPL resided in the 0.6-mm sand of the lower homogeneous zone, with a saturation distribution similar to that seen in Exp. 1 (Fagerlund et al., 2007). Once the NAPL reached the interface and started to enter into the upper heterogeneous zone, saturations in the other sand types could be observed. After 9 h, the distribution was similar to that of Exp. 1 (for the same time) but with the addition of small amounts of NAPL occurring at higher saturations inside the upper zone. As the NAPL pooled on its way in through the coarser parts of the upper heterogeneous zone, higher saturations in the 0.6-mm sand compared with the lower homogeneous zone can be observed, indicating that the NAPL pressure was also higher. The NAPL in the 1.18-mm sand was mostly found at high saturations.


Figure 18
View larger version (30K):
[in this window]
[in a new window]

  		FIG. 18. Nonaqueous-phase liquid (NAPL) distribution in the different sands as a function of saturation at 9 and 385 h: sand size 1.18 mm, red-orange; 0.6 mm, yellow; 0.6- and 0.3-mm mixture, cyan; and 0.212 mm, blue.

 
At 385 h, when the NAPL had become immobile, the saturation distribution was similar to the distribution at the end of Exp. 1 in the sense that a large peak in the distribution can be seen at the Sn interval where the NAPL became discontinuous in the 0.6-mm sand. In Exp. 3, however, a lot more NAPL was entrapped at high saturations (in a continuous state) inside the upper zone.

In Exp. 3, Snmax for the Land model was 0.19. This value has been used to define a constant cutoff value for the GPR, shown as circles in Fig. 19 . This value is consistent with the peak in the Sn distribution (Fig. 18, 385 h) associated with discontinuous ganglion entrapment, which ended at roughly Sn = 0.20. Sensitivity of the GPR is demonstrated by showing the GPR for cutoff values of 0.17 and 0.21 in Fig. 19. The sensitivity of the GPR is larger toward lower cutoff values, as large amounts of NAPL occurred within a narrow interval from the peak. Toward higher cutoff values, the sensitivity is less because only small amounts of NAPL occurred at saturations just higher than 0.20 and there is a natural end of the peak zone. Compared with Exp. 2, the GPR was higher at the end of the experiment when the NAPL had become immobile. The reason is that here a large part of the entrapment occurred in the homogeneous lower zone, where the NAPL had to become discontinuous before it was entrapped. In general, heterogeneity will greatly favor the occurrence of immobile pools.


Figure 19
View larger version (16K):
[in this window]
[in a new window]

  		FIG. 19. Ganglia/pool ratio (GPR) for a cutoff value of 0.19 (circles); GPRs for cutoff values 0.17 (squares) and 0.21 (triangles) indicate sensitivity.

 
Effects of Heterogeneity on Flow and Entrapment
In Exp. 1 with two homogeneous zones (Fagerlund et al., 2007), the general movement of the NAPL was up toward the interface and then along it in the upwardly dipping direction. The pool that developed at the interface was thin and elongated. The pressure in the pool was not large enough to overcome the entry pressure and produce penetration into the finer zone. There were no capillary barriers resisting NAPL movement along the interface, meaning that the NAPL could spread along the interface until all NAPL was immobilized as saturations decreased to the point where the NAPL became discontinuous. Throughout the plume, the NAPL was immobilized in a discontinuous state within a similar saturation range; however, a relationship between immobile and maximum saturations existed.

In Exp. 2, the heterogeneities in the lower coarser zone produced a different path and entrapment behavior on the way to the interface. The NAPL preferably moved through the coarser sands. Two forms of entrapment were observed. In the medium-coarse 0.6-mm sand, apart from at the interface, the NAPL was entrapped as discontinuous ganglia as in Exp. 1. In the coarsest sands (2.36 and 1.18 mm) as well as in the 0.6-mm sand in the pool at the interface, the NAPL was entrapped at capillary barriers to finer sands. Here the NAPL was entrapped at higher saturations in a continuous state. The existence of heterogeneity in the coarser zone means that there were capillary barriers along the dipping interface. Therefore the interface was not smooth like in Exp. 1 and a pool of higher saturation and pressure could develop. The NAPL pool along the dipping interface resided in coarse and medium sands and was essentially bounded by an L-shaped boundary of interfaces to finer sands, with the long leg of the L along the interface and the short leg pointing down on the left side of the flume. The short leg of the L was long enough to prevent NAPL flow toward the left up to the point when, instead, the NAPL started to exit the experiment domain through the right constant-head well. The pressure in the pool never became large enough to produce significant penetration into the finer zone, even though the finer zone in Exp. 2 was not as fine as in Exp. 1.

In Exp. 3, the movement toward the interface was very similar to that of Exp. 1, although the extra homogeneous packing of Exp. 3 seemed to generate slightly higher saturations of NAPL entrapped as discontinuous ganglia left behind along the path of NAPL movement. The difference may be due to preferential flow of NAPL, leaving less entrapped NAPL behind, which to a small extent occurred in Exp. 1 but was essentially eliminated in the homogeneous zone of Exp. 3. At the interface, the heterogeneities present in the finer zone provided coarser areas where the NAPL could immediately start to enter. A little pooling and spreading along the interface occurred, but nearly all free NAPL followed the path into the upper layer. Here the NAPL spread upward along channels of coarser material. In such vertical channels, higher pressures can build up compared with horizontal pools along smooth interfaces. The NAPL could therefore enter into finer pores and eventually penetrate all the way to the constant-head boundary in a channel through fine material. Below the interface, the NAPL was entrapped as discontinuous ganglia, whereas in the heterogeneous finer zone, it was entrapped at higher saturations in a continuous state.

In both Exp. 2 and 3, the heterogeneous zones were designed so that the dimensions of these zones were at least 12 times the correlation lengths in both the horizontal and vertical directions. This was done to assure that a sufficiently large region representing the underlying statistics would be sampled by the plume. In both cases, however, the NAPL migration in the heterogeneous zone is still deemed to be strongly dependent on the specific realization of the stochastic field. For example in Exp. 2, for another realization with the same statistical characteristics, the NAPL may have moved up the dipping slope and out through the left constant-head well, resulting in a quite different spatial distribution of the NAPL and different spatial moments. The distribution of entrapped NAPL across different saturations and the GPR on the other hand are likely to be less dependent on the specific realization in question and more dependent on the general geostatistical characteristics of the field. The behavior of NAPL for multiple realizations could be further investigated by means of numerical modeling studies, after first using the presented data as a basis for model validation. Further studies could, e.g., investigate the possibility of relating NAPL entrapment characteristics and their statistics to geostatistical characteristics of the permeability field. Thereby, knowing the geostatistical characteristics of a given site, information about the immobile NAPL distribution could be inferred and be used, e.g., as a basis for upscaling the entrapped-NAPL characteristics into larger scale NAPL-dissolution models.


   Conclusions
TOP
 ABSTRACT
 INTRODUCTION
 Theory and Methods
 Results and Discussion
 Conclusions
REFERENCES
 
Flow and immobilization of NAPL in saturated heterogeneous formations was studied in a two-dimensional intermediate-scale sand tank. In a system consisting of two distinct zones separated by a dipping interface, effects of stochastically generated heterogeneity in each of the two zones were studied.

A detailed data set on the infiltration, redistribution, and immobilization process was developed. The NAPL saturations were carefully monitored in space and time following a methodology developed in Fagerlund et al. (2007). The precise packing of the flume in combination with precise measurements at a large number of exactly known locations allow a unique and exact monitoring of dynamic NAPL behavior in formations where complex, small-scale heterogeneities are present.

The observations made about the NAPL migration and entrapment can shortly be summarized as follows:

  1. Heterogeneity produces variations in capillary resistance to NAPL flow that largely governs the NAPL flow path as the NAPL preferentially moves into the coarser pores where capillary pressures are lower. When NAPL is immobilized in heterogeneous media, two types of entrapment occur. In areas from which the NAPL can move away without being affected by capillary barriers, the NAPL is entrapped as snapped off, discontinuous blobs and ganglia. Here the trend in the relationship between Snt and Snmax is well predicted by the Land (1968) model, although some random spreading around the trend occurs. In addition to being entrapped as discontinuous ganglia, NAPL is also entrapped in a continuous state in the presence of capillary barriers. In areas influenced by strong capillary barriers, Snt is close to Snmax.
  2. Compared with the homogeneous case, in the heterogeneous zones more NAPL is entrapped and it is entrapped across a wider range of saturations. In homogeneous media, snap-off is the dominating entrapment mechanism and as entrapment by snap-off occurs within a narrow range of saturations, a large part of the entrapped NAPL is found within this narrow range. Also, in cases where both types of entrapment occur, a peak in the entrapped saturation distribution can be seen within this range. The upper limit of the peak range seems to constitute a reasonable cutoff value for the GPR.
  3. Local heterogeneity may allow NAPL to enter a formation that, on average, is so fine that the NAPL would not enter if it was homogeneous. Once the NAPL has entered, pressures may build up in the flow channels, enabling further advancement and penetration into fine-grain areas. Because of the domination of finer materials, the NAPL entrapped inside such a formation may be accessible only to very slow aqueous-phase flow, which means that dissolution will be slow.

It can be concluded that heterogeneity strongly affects NAPL migration and immobilization in saturated media, both with respect to the migration path and spatial distribution and with respect to distribution of NAPL across different saturations. The NAPL saturation affects aqueous-phase effective permeability and, especially in regions where immobilized NAPL saturations are high, bypassing of the water flow through fully saturated finer material can be expected. High NAPL saturations are found in heterogeneous media in the presence of capillary barriers. Generally, the location of the NAPL occurrences in the aqueous-phase flow field is important for the accessibility to water flow and thereby also for the rate of dissolution. Therefore, heterogeneities that allow NAPL to enter into averagely fine formations may be particularly important to subsequent dissolution and the longevity of contamination from entrapped sources. The data set presented can be used to validate conceptual and numerical models for further study of flow and entrapment of NAPLs in complex heterogeneous systems.


   ACKNOWLEDGMENTS
 
We thank the Swedish Rescue Services Agency (Räddningsverket), the Swedish Road Administration (Vägverket), and the U.S. National Science Foundation (award no. DMS-0222286) for providing funding for this research. Furthermore, F. Fagerlund thanks Dr. T. Sakaki and Dr. M. Komatsu for measurements of sand permeabilities and capillary pressure–fluid saturation curves, Dr. K. Glover, Dr. D. Rodriguez, and J. Gago for training and assistance with the x-ray-attenuation measurement system, and Dr. M. Mathew for assistance with flume packing and modeling.


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




This article has been cited by other articles:


Home page
 Vadose Zone JHome page
F. Fagerlund, T.H. Illangasekare, and A. Niemi
Nonaqueous-Phase Liquid Infiltration and Immobilization in Heterogeneous Media: 1. Experimental Methods and Two-Layered Reference Case
Vadose Zone J., August 1, 2007; 6(3): 471 - 482.
[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 Similar articles in ISI Web of Science
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 ISI Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Fagerlund, F.
Right arrow Articles by Niemi, A.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Fagerlund, F.
Right arrow Articles by Niemi, A.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Fagerlund, F.
Right arrow Articles by Niemi, A.
Related Collections
Right arrow Spatial Variability
Right arrow Multiphase Fluid Flow
Right arrow Experiment Design

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