LNAPL Transmissivity, Mobility and Recoverability—Utility and Complications

Abstract

Light non-aqueous phase liquid (LNAPL) mobility, transmissivity (T_n), and recoverability are controlled by a suite of factors, namely release characteristics, the environment in which it resides, its age, physical/chemical properties, and others. LNAPL plumes are dynamic, but typically trend toward field stability after a release stops. Unlike groundwater mobility, finite LNAPL releases represent an ever-diminishing mobile mass due to degradation, residualization, and other factors that will be subsequently examined. For these and other reasons, mobility aspects vary both spatially and temporally. T_n is a useful, but complex, related parameter. It is commonly used to determine the applicability of LNAPL hydraulic recovery. But since it is not constant, a cutoff must be carefully considered, along with its manner of determination. Further, even if a plume is theoretically “recoverable” on a T_n rate basis that has very little relationship to whether that action would be expected to have any net environmental benefit. LNAPL hydraulic recovery is inherently self-limiting and has little net benefit unless a plume is mobile (and contained) or a significant mass recovery percentile is possible, which is rare.

12.1 Introduction

For plumes generated by fuel and related light non-aqueous phase liquid (LNAPL) releases, arguably the most important element is the genesis of that footprint over time. In the early stages of a release when LNAPL is migrating, it represents a direct potential risk to receptors controlled by the rate and ultimate locations of that transport. In the early stages of a release, LNAPL transmissivity (T_n) and mobility will generally be larger than later in its lifespan, often by orders of magnitude. The LNAPL release and its mobility is also affected by a variety of other natural processes including biodegradation, physical weathering, partitioning (collectively termed natural source zone depletion), residualization, and others aspects discussed elsewhere in this text.

Transmissivity, as a parameter in groundwater flow mechanics, has been around for much of the history of our field. In the mid-1800s, Henry Darcy framed aspects of water flow through sand as related to the pressure gradients and the properties of the sand (Simmons 2008). In the current era, framing T_n is an important element in determining a similar thing, the potential rates of LNAPL flow or recovery through geologic media. But unlike the perfect sand/water conditions of Monsieur Darcy, aspects of T_n are vastly more complex, and it is, at best, a distant relative to the transmissivity of groundwater mechanics. Its roots are in the underpinnings of Darcy’s Law, petroleum and agricultural applications in multiphase mechanics and other related technical fields of interest.

Although controlled by the same physical processes, the LNAPL transport behavior in the vadose zone differs substantially from that around the water table or saturated zone where water contents are significantly greater and gradient interactions differ. Once a fuel release has stabilized, it represents a typically depleting source mass (often called the “source term”) for dissolved and/or vapor-phase impacts, as well as being a source for potential methane generation (note, vapor-phase transport is also a multiphase process). LNAPL releases follow the gradients created by those release and related characteristics, like volume, rate of input, subsurface, and LNAPL properties, and so on, as discussed elsewhere in this text.

If an LNAPL release reaches the water table, its behavior differs substantially from the ambient groundwater conditions since the plume genesis is unrelated to the boundary conditions controlling groundwater and is vastly more sensitive to pore-scale lithologic variability. Many intuitive field observations may be at odds with the controlling realities. For instance, large thicknesses of floating LNAPL in the water table region are sometimes thought to represent the worst-case areas of a plume. But what if that large thickness is due to capillary and other resistive forces that do not allow the product to dissipate (i.e., the product is locally immobile)? What if the mobility in some sandy zones is vastly greater even though the observed thickness is quite small? What if the direction of LNAPL migration has little or sometimes nothing in common with the groundwater gradients? Yes, LNAPL generally follows the hydraulic gradients, but specifically the LNAPL gradients, which may differ substantially from groundwater gradients. Many apparently logical assumptions can lead to erroneous conclusions when it comes to LNAPL and multiphase mechanics, particularly in the real-world (applied, not theoretical).

Regardless of T_n or recoverability values, hydraulic LNAPL recovery can never “clean up” an aquifer, it is simply not possible. This is because of aspects related to LNAPL residualization (defined as non-mobile retention in the pore space in this chapter) and other factors. Residual LNAPL generally represents a persistent mass in the subsurface, the degree of persistence depending on setting, mass in-place, fuel characteristics, and others. Further, there are multiphase, formation, and operations-dependent limitations to hydraulic recovery such that even if there were no residualization, the method would still fail to recover all the LNAPL in-place. At best, it can only reduce or eliminate the spread of a mobile LNAPL plume and perhaps reduce its longevity in the environment if a sufficient percentage of the total mass is recovered (which almost never occurs). We will see that while sometimes useful, hydraulic recovery is generally ineffective at many sites to reduce risk, substantially reduce care and monitoring, or otherwise provide a net environmental benefit to those actions. That is certainly not to say that cleanups are not important in many instances, but rather that pump and treat is quite limited in its capacity to effect such cleanup goals, as will be clearly demonstrated here. Hydraulic recovery of LNAPL, in most cases, is like hitting a nail with a feather. One can do so, but to what beneficial effect?

The discussion will first provide a theoretical framework and overview identifying key parameters and other aspects of T_n that define the parameter. The second part of this chapter discusses the field and laboratory-based derivations of T_n and some of the complications in this task; other texts provide a description of how to perform these actions, which will not be repeated here. The last major section of this chapter deals with observed real-world complications and how nuances in porous media structure and makeup sometimes conflict with commonly applied multiphase theory and those effects with regard to T_n. Finally, the implications of T_n and its application to mobility and recoverability will be covered.

We hope to show the practical challenges to LNAPL characterization, the implementation, and effectiveness of LNAPL recovery in remediation applications so that practitioners can answer the question: “Is hydraulic recovery LNAPL useful in this situation, or should other methods of remediation be considered?” The author’s suggestion? Use T_n and these related processes to explain important elements to the environmental protection equation, but test the key underlying assumptions against multiple lines of evidence, particularly field-based observations and data. It is usually an enlightening process that will point out deficiencies in the LNAPL conceptual site model (LCSM).

12.2 Quantitative Definition of T_n

As with groundwater flow in fully saturated aquifers, the LNAPL transmissivity is a bulk term that provides an indication as to the relative ease with which the liquid (LNAPL or oil in this case) may pass through a unit thickness of material. The USGS defines aquifer transmissivity as: “The rate at which water of the prevailing kinematic viscosity is transmitted through a unit width of an aquifer under a unit hydraulic gradient. It equals the hydraulic conductivity multiplied by the aquifer thickness” (USGS 2004).

The concepts and application of aquifer transmissivity was originally developed primarily for the evaluations of well hydraulics in confined aquifers (Freeze and Cherry 1979). As should be immediately clear to the reader, the scale and application of transmissivity to LNAPL migration and recovery is vastly smaller and coupled with nonlinear behaviors (as follows below). This chapter will delve into just a few of those complications following the development of idealized applications of T_n. This chapter will not redevelop the underlying fundamentals of transmissivity, storativity, specific yield, and other analogous aquifer characteristics. However, it is important to understand that each of these aspects play a role in LNAPL mobility and recovery and should be considered when appropriate to the evaluation at hand. Additional discussions of these basics are provided in several of the cited references.

To understand LNAPL transmissivity and its relationship to mobility, it is useful to define the parameter by its associated quantitative factors. Analogous to groundwater transmissivity, the LNAPL transmissivity can be defined similarly as shown by Eq. 12.1 below (Huntley 2000). Note “oil” is classically used as a general term for immiscible LNAPL, as in the “oil” phase of a 3-phase system consisting of the air/water, oil/water, oil/air phase couplets (e.g., see Parker 1989, an excellent overview of multiphase transport).

$$T_{n} = \int\limits_{{z_{w} }}^{{z_{n} }} {k_{rn} k \frac{{\rho_{n} g}}{\mu }{\text{d}}z}$$

(12.1)

where k is the porous media intrinsic permeability, k_rn is the relative permeability of the LNAPL, ρ_n and μ are the LNAPL density and kinematic viscosity, respectively, and g is gravitational acceleration. Integration is between the base of the flowing LNAPL interval and the uppermost distribution. In an ideal equilibrium setting, that would be equal to the oil/water and oil/air interfaces represented by LNAPL thickness in a monitoring well.

Of the parameters shown in Eq. 12.1, only the fluid factors may be relatively constant in the local formation (i.e., could be moved out of the integral), assuming there is a single LNAPL source in the mobile interval. Even there, properties like viscosity and interfacial tension may change with weathering. Intrinsic permeability varies as a function of lithology, which tends to be naturally heterogeneous. The relative permeability, as discussed elsewhere in this text, varies as a function of saturation, which in turn varies as a function of capillary pressure, wettability, and other factors. The integration of these factors is across the vertical interval in which LNAPL flow occurs and not where immobile residual is present. These facets generally apply to the other fluid phases (water and gas) as developed elsewhere in this text, but this chapter is specifically focused on the mechanics of the LNAPL (aka, oil phase).

The value of T_n is usually most sensitive to the intrinsic and relative permeability ranges of a particular setting. Intrinsic permeability may range from 10^‒8 to 10⁵ Darcy in natural earth materials (Freeze and Cherry 1979), a span of 13 orders of magnitude, although end-member values are relatively rare. The more significant complication with respect to T_n is the relative permeability scalar that varies nonlinearly from 0 to 1 as a function of the effective LNAPL saturation. LNAPL saturation, in turn, varies nonlinearly as a function of soil and fluid capillary characteristics, as discussed elsewhere in this text. A simplified form of the LNAPL relative permeability scalar is the square of the LNAPL effective saturation, shown in Eq. 12.2 below (Charbeneau et al. 2000). Later in this discussion, we will take a look at both theoretical and applied complexities of these and other controlling factors:

$$k_{{{\text{rn}}}} \approx \overline{S}_{{{\text{n}}}}^{2}$$

where k_rn is the LNAPL relative permeability scalar, and $\bar{S}_{n}$ is the effective LNAPL phase saturation, defined in Chaps. 2 and 3.

Finally, while less sensitive (linear effects), the density and viscosity of the LNAPL also contributes to the observed T_n, sometimes in unexpected ways. The kinematic viscosity of petroleum and its products varies over many orders of magnitude if one include crude oils, from less than 1.0 to more than 1 million centipoise (API 2006). The relative density of petroleum products is less variable, generally ranging from about 0.7–0.99 for common fuels; of course, any non-aqueous phase liquid (NAPL) with a density greater than water is defined as a dense NAPL (DNAPL). While the concepts discussed here often apply to DNAPL, it is less often observed as a free-phase lens that meets the testing and analysis requirements discussed for T_n below. Further, the concepts apply in related ways to air and water phase movement in a three-phase environment, again as discussed elsewhere in this text. Three-phase flow cannot be described without the coupling of capillarity, relative permeability, saturation, and underlying soil and fluid properties. T_n is merely one rather narrow facet of the broader family of related multiphase considerations.

Now that T_n has been defined by each of its integral parameters, we can see that if natural earth materials vary in permeability over roughly 10 orders of magnitude, then T_n should vary by several more orders (in theory). Later in this chapter we will discuss field observations and observe how the T_n field results align with these theoretical considerations and where they may not. First, however, let us review the implications of the mathematics of T_n from a theoretical and heuristic standpoint.

12.3 Theoretical Implications of T_n Factors

A suite of theoretical interrelationships is developed in this section that inspect the most sensitive and important factors affecting the values of T_n. This will help the reader understand the wide variability that might be expected and some of the practical implications of T_n and its application to site-specific LNAPL issues. As noted, the author views this discussion as heuristic, as these theoretical conditions may exist under ideal laboratory conditions, but as will be seen subsequently, rarely under field conditions. Nonetheless, the underlying theory is well-grounded in multiphase studies and applications.

The method of many of these T_n considerations is to invoke the assumption of vertical hydrostatic equilibrium (VEQ; i.e., no vertical head gradients in oil or water and that fluid levels in a well represent the formation heads and pressures, as described in Chap. 2). The VEQ assumptions provide an analytic basis for the comparisons that will be shown here (see a full VEQ discussion in Lenhard and Parker 1990; Farr et al. 1990). As the reader may suspect, whether VEQ is present or not is yet another facet to consider. Where T_n and relative permeabilities are small, it is mathematically and volumetrically unlikely that VEQ is ever attained. Subsequent discussion of field complications will touch on field challenges to some of these simplifying boundary assumptions.

12.4 Effect of Soil Type

The type of porous media and its capillary and permeability characteristics are synergistically related to both the effective LNAPL hydraulic conductivity (K_n) and T_n. Because capillary characteristics control the LNAPL saturation (S_n) as a function of the pore throat diameter distribution and other factors, and because capillarity is nonlinear, these facets exhibit strong interplay in the resulting K_n and T_n values. It is instructive to look at both parameters in parallel.

For five soils, ranging from predominantly fine-grained to clean coarse-grained materials,¹ the calculated theoretical VEQ LNAPL saturation distributions for 1 m of gasoline-free product in an observation well are shown in Fig. 12.1, along with the volume per unit area of free-phase LNAPL annotated. The differences in these profiles as a function of fuel type will be shown below. Table 12.1 provides the van Genuchten (1980) parameters used to generate each profile, based on API’s LNAPL Parameters Database ranges (Beckett and Joy 2003), except for the irreducible residual water (S_wr) content was taken from Schaap et al. (2001)¹. Note, the capillary fringe extends above 1 m but is not engaged in idealized flow (not part of T_n).

Fig. 12.1

VEQ gasoline saturation profiles for various soil types for 1 m free product

Table 12.1 Input parameters to VEQ estimates

As noted above, the relative permeability with respect to LNAPL may be approximated as the square of its effective saturation (Charbeneau 2000). This particular function is selected for ease of explanation, and the reader should be aware that there are many relative permeability functions that might apply from agricultural and petroleum applications (e.g., Burdine 1953; Mualem 1976; Delshad and Pope 1989). As will be discussed in the real-world observations section, each of these functions is an empirical approximation and depends on the lithologic characteristics of the site-specific porous media. Due to the expense of relative permeability testing, these are not commonly performed for environmental applications. Also, to be discussed, there are likely laboratory and scaling issues that may need to be considered. Figure 12.2a and b show the effective permeability (k_n = k_i * k_rn) toward LNAPL on arithmetic and semi-log scales, respectively. The reason for plotting on semi-log scale is visually clear, one cannot “see” the small values of effective permeability on an arithmetic scale when contrasting soil types. Obviously, the contrasts in soil capillary properties and the resultant LNAPL saturations are amplified by the relative permeability phenomenon.

Fig. 12.2

a and b: Calculated effective permeability for a range of soils on arithmetic and semi-log scales. Intrinsic permeability ranges taken from Freeze and Cherry (1979)

The effective permeability tensor is highlighted because it provides a direct relationship to effective conductivity and its vertical integral, T_n. Several attributes of the effective permeability plots become immediately clear. First, at small S_n values (or correspondingly small capillary pressures), the k_n is very small for all soil/porous media types. It is of course smallest in fine-grained materials, but for any soil/porous media type becomes negligible at some point on the curve. The effect is compounding with regard to T_n, because T_n is integrated over a mobile-zone thickness. As that LNAPL thickness and associated saturation decreases, so does T_n as a linear function of that thickness and a nonlinear function of saturation and relative permeability, as shown in Fig. 12.3, plotted again on semi-log scale.

Fig. 12.3

Calculated gasoline transmissivity as a function of elevation and saturation above the oil/water interface

As expected, the T_n values span many orders of magnitude. At low capillary pressures (or well thickness) and saturations, values become quite small regardless of soil type. Using the coarse soil Type 5 as an example, there would be about 525,000 times more flow potential at 1 m versus 0.1 m for the same porous media properties; “potential” because neither case may actually be mobile due to other impeding factors (see related discussions in this text). At the other end of the spectrum, because of differing capillary properties and saturation distributions, Soil Type 1 has a contrast of about 6000 times more flow for 1 m thickness versus 0.1 m. This exponential variability and other observations from Fig. 12.3 indicate that in real-world settings, T_n is a highly variable and complex parameter. It also implies that VEQ conditions may not always occur, as the time to fill a wellbore at low transmissivities, regardless of soil type, will be quite long and likely be overprinted by hydrologic variability. Unless the hydrogeologic system is relatively static (and most are not), VEQ assumptions and estimates themselves should be used with caution. More discussion on this and other complications will be discussed in the real-world observations section below.

Fig. 12.4

VEQ LNAPL saturation estimates for two layered soil systems

Lastly, it is useful to consider how heterogeneous systems might behave (at least theoretically). If a system is under VEQ conditions, then the capillary pressure profile will be linearly uniform across any sequence of soil/porous media types. For explanation purposes, we will look at two heterogeneous profiles. The first will be a three-layer system of coarse, mid-range, and fine-grained soils from the bottom-up (i.e., Soil Types 5, 3, and 2 above). The second will be a two-layer system with the finest-grade material comprising the bottom 0.5 m, and the coarse soil the upper 0.5 m. Each of these examples is again for 1 m of free-phase LNAPL under VEQ conditions.

The vertical position of soils upon one another has a significant impact on the estimated S_n distribution and the associated specific volume and T_n values. Because capillary pressures are smaller at the base of the LNAPL column, the coarse-grained soil contains much less LNAPL in Case 1 than in Case 2 even though the LNAPL thickness and soil characteristics are identical (Fig. 12.4). The resultant effect on the T_n values is shown in Fig. 12.5, which amplifies the noted saturation contrasts. In practical terms, the LNAPL in Case 2 is effectively perched on the lower permeability materials below that, even under ideal VEQ conditions, do not meaningfully contribute to the integrated T_n value. It is also interesting that, because of low saturations and intrinsic permeability, the upper two soil types for Case 1 add almost nothing to the T_n profile (too small to visualize). Similarly, one can see almost and 8-order increase in T_n above the 0.5 m mark for Case 2, supporting the observation above that this LNAPL condition is essentially perched on the finer-grained material.

Fig. 12.5

Gasoline transmissivity for case 1 and 2 layered soil conditions under VEQ conditions

12.5 Effect of LNAPL Properties

As a final set of theoretical considerations, we will look at the role that LNAPL fluid characteristics play in T_n and its contrasts for different petroleum products and properties. From a T_n and mobility/recoverability standpoint, product viscosity is the most variable factor of petroleum products (linear effect). However, because the product density drives the capillary pressure between oil/water, that factor has a nonlinear influence in the expected T_n values. Another parameter, the interfacial fluid tension (IFT) between water and oil, often varies significantly in the field from ideal laboratory measurements; biodegradation and other aspects are likely causes, but more study is needed. Since the ratios of IFT are used to scale capillary properties for each phase couplet (oil/water, water/air, and oil/air), this too has a nonlinear impact to estimation of LNAPL saturation and T_n, as is explored here. Additional discussion of capillary processes and their basis and applications are provided elsewhere in this book.

Figure 12.6 provides contrasts in T_n values for the same 1 m of assumed LNAPL thickness, for five petroleum types, and using Soil #4. The fluid parameters for these products are provided in Table 12.2 (from API 2006). For the interested reader, Environment Canada has an oil properties database that has an extensive set of LNAPL properties (https://www.etc-cte.ec.gc.ca/databases/oilproperties/Default.aspx). As seen in this figure, the combination of density and viscosity differences combine to create order-of-magnitude differences in the expected T_n values as a function of product type. Per Eq. 12.1 above, the larger the viscosity, the lower the K_n and T_n values (inversely proportional on a linear basis). The density contrasts, although relatively small, roughly 0.7–0.9, affects the calculated capillary pressure between water and oil. That lower capillary pressure then nonlinearly decreases the calculated saturation with increases in density.

Fig. 12.6

LNAPL transmissivity under VEQ conditions for various petroleum products

Table 12.2 LNAPL physical properties

12.6 Transience of LNAPL Transmissivity

From a theoretical standpoint, T_n is transient, even absent complicating field aspects discussed subsequently and elsewhere in this book. As LNAPL releases develop in the subsurface, there is a sequential set of temporal and spatial changes that generally occur. In early time, the release will be located near its point of origin, as shown below in a multiphase model-derived schematic of LNAPL plume development for a finite release (Fig. 12.7, upper left panel). The associated LNAPL saturations and observed thicknesses are large early in time, as would be the resulting local area T_n. At this early time in the release progression, there is no LNAPL arrival yet at distal locations, where the T_n is temporarily zero. When there is a potential for significant recovery and migration control benefits, it is almost always in these early stages of a release.

Fig. 12.7

Modeled LNAPL saturation distribution in a homogeneous scenario at example time stamps. Note that the plume is initially concentrated near its release point, migrating over time with saturations changing accordingly. The example does not consider processes, such as water table fluctuations

As the LNAPL migrates, mass redistributes more widely, occupying a larger volume of porous materials. Along with that there are new LNAPL arrivals at further distances away from the release along with a corresponding decrease in mass and saturations in the release area. That causes a transiently increasing T_n with distance for some period, and decreasing T_n in the release zone. The transient sequencing of this scenario is shown in Fig. 12.8 for some example post-release times. As should be noted, different release and subsurface characteristics will generate analogous results, but with different migration characteristics and transient T_n conditions. Beyond the direct mass transfer effects discussed here, there are other field factors that transiently affect the T_n (natural mass losses, residualization, smear zone progression, and others). Some of those are discussed further below in the field observations discussion and in other sections of this text. Remember, any process that influences LNAPL saturation, hydraulic conditions or other controlling properties discussed herein will also have a direct effect on the T_n and mobility conditions. For periodic releases at some active facilities, these facets become quite complex, but the principles discussed still apply.

Fig. 12.8

Transient changes in T_n as a function of plume migration and genesis. Note that these types of changes will occur virtually everywhere, but vary significantly as a function of release setting and the characteristics of the fuels and porous materials involved

12.7 Summary of Theoretical Observations

In summary, it is clear that T_n has linear and exponential sensitivities, both with respect to porous media and LNAPL properties. The K_n is similarly variable and that is important because it is a key component in discrete mobility potential within the more permeable and saturated zones in the subsurface, which often coincide due to related porous properties of permeability and capillarity. Darcy’s Law can be used as a screening estimate by combining K_n (which is not a simple scaling of T_n), LNAPL gradients, and the effective porosity to result in an estimated average linear pore velocity. While heuristically useful, one must use such methods while remaining cognizant of other multiphase processes that often limit LNAPL plume migration, particularly for older plumes. In other words, plumes can reach field stability even when there is a conductivity and apparent gradient.

The above theoretical background leads to the following conclusions and implications, absent other complicating factors like water table fluctuation, residualization, natural source zone depletion (NSZD), and others. Additional (and sometimes non-intuitive) field- and data-based observations follow in the subsection below.

• Porous media that has low permeability and associated low LNAPL saturations (due to capillary properties) will not contribute meaningfully to T_n or K_n values and associated mobility or recoverability, particularly where higher permeability materials also reside in the LNAPL continuous-phase interval.

• Soil and oil types combine to produce order-of-magnitude variabilities in T_n/K_n values, even where initial observations of LNAPL thickness in monitoring wells are similar.

• Because of the nonlinear sensitivity to S_n, the T_n values will necessarily decrease over time with recovery actions, residualization, migration, and other phenomena (except under ongoing releases and/or transient hydraulic effects).

• While there is no thickness exaggeration implied by LNAPL observed in wells, only a portion of that total LNAPL saturation column will contribute meaningfully to T_n and mobility because lower saturation zones will have very low effective permeabilities. Therefore, the implications regarding mobility and recoverability are dependent primarily on the zones of high effective K_n.

• Because T_n is proportional to the time it will take to fill a given wellbore volume, caution needs to be applied when invoking VEQ assumptions. At small saturations and/or in fine-grained materials, VEQ likely does not occur under generalized field conditions. Even in high permeability materials, if there is significant water table variability, equilibrium may never occur.

• At some combination of saturation, permeability, and fluid properties, the effective mobility of LNAPL becomes de minimis. For instance, a value of practical impermeability in some clay liners is 1 × 10^‒6 cm/s, or a transmissivity across 1 m of 1 × 10^‒4 cm²/s.

• Similar to the point above, at some lower limit of T_n, there is no practical way to field test that condition through general pumping or well-scale hydraulics. If a T_n estimate is needed in such conditions, it would be derived from its other fundamental parameters (which can be challenging, as discussed below). But, at such low T_n values both recoverability and likely mobility and will be negligible.

• As observed in the discussion of transience, the time where hydraulic recovery has the best opportunity to contain plume migration and recover the maximum mass possible is immediately following the release. The passing of time significantly diminishes the practicality of LNAPL recovery under most conditions (except active/ongoing releases). Many historic LNAPL plumes have been shown through field study and plume persistence to be practicably unrecoverable, regardless of the T_n values measured in the field. Other remediation methods may work, but hydraulic recovery will not, at least to any net environmental benefit.

12.8 Estimation of LNAPL Transmissivity

T_n may be estimated using analogous principles to aquifer testing, either as single-well tests or under pumping conditions. T_n may also be estimated through bench or laboratory petrophysical testing by determining the parameters of its integration defined in Eq. 12.1. While this is straightforward in concept, the complications discussed above, coupled with field observations in the next section, suggest that this determination is more complex than sometimes recognized. This portion of the chapter will focus on these complications rather than the nuts and bolts of performing T_n quantification testing. Other resources such as the API LNAPL Transmissivity Workbook (Charbeneau et al 2012), the ASTM Guide for Estimation of LNAPL Transmissivity  (ASTM 2021), and Huntley (2000) all provide excellent descriptions of LNAPL T_n testing and parametric derivation.

One of the more common testing techniques for determining T_n is the LNAPL baildown test (aka, LNAPL slug test; ASTM 2021). It is relatively simple to execute. One quickly removes free product from a correctly screened observation well and then monitors its rate of return. The faster the LNAPL recovers, the greater the T_n (in general). The resources above detail the procedures and field considerations. Our discussion here is about dissecting the physical underpinning so that the complexities and limitations can be considered by the practitioner. As background, the author and the late Dr. David Huntley (Professor Emeritus, SDSU) have conducted many of these field tests since about the mid-1980s and have the benefit of observing both expected and unexpected results. The intent here is to provide some key observations and potential limitations to these methods of T_n estimation.

Foremost in our consideration is the fact that T_n is not a constant of the aquifer (Beckett and Huntley 2015), as it is for single-phase flow of groundwater across a large aquifer saturated thickness. For a finite LNAPL release, the transmissivity is expected to generally decrease over time due to a number of factors. One is progressive residualization and entrapment of the mobile LNAPL due to groundwater level variability, hysteresis, and occlusion. LNAPL plumes are also continuously losing mass and changing composition due to both partitioning and degradation mechanisms (e.g., Huntley and Beckett 2002; Garg et al. 2017). As the reader now recognizes, mass losses generally imply a reduction in LNAPL saturation and that causes associated declines in relative permeability, K_n, and the associated T_n value (see Eq. 12.1 above).

Further to T_n not being a constant are the conditions of testing. Baildown or pumping tests done under one set of free product and water level conditions will produce a different result than under others. One of those key factors is occlusion of LNAPL by rising water levels in unconfined settings. Figure 12.9 provides a general schematic of how that works. In Frame 1, T_n is at its greatest value before any residualization or occlusion. In Frame 2, T_n will be smaller based on the new residual mass held in the formation above. In Frame 3, T_n will be quite small because of occluded 2-phase mass below no longer in hydraulic communication with the well. In Frame 4, the T_n is zero because there is no flowing free-phase product, and, in Frame 5, there is an approximate return to conditions in Frame 2.

Fig. 12.9

Schematic of LNAPL residualization and occlusion with changing water table conditions. Both residualized and occluded LNAPL are described as “residual LNAPL.” Figure courtesy of Richard Jackson

To emphasize this important consideration, a baildown test for T_n in Frame 1 may imply an unacceptable mobility risk, whereas testing in Frame 4 is not necessary because no free-phase LNAPL means no mobility or recoverability (T_n = 0). But that would be inaccurate and highly misleading if the water table falls again as in Frame 5. Hydrologic conditions and the vertical distribution of LNAPL must be considered as part of the T_n characterization, along with plume mobility and stability aspects. The history of both release and hydrologic conditions are also important factors. So, unless one is in a demonstrably stable hydrologic environment, which is rare, one value of T_n will not confidently define conditions.

Figure 12.9 nicely captures the overall mechanics of unconfined water table variability on free product observations within a well. Under actual field conditions, these observations become even more complex, as expected. Figure 12.10 shows a long-period of fluid level gauging in an unconfined monitoring well where the estimated T_n is shown for a few different corrected water level elevations. When the LNAPL is fully submerged, the T_n ~ 0, and otherwise varies over several orders of magnitude depending on the observed free-phase thickness. While the LNAPL mobility/recoverability is negligible for about six years, from 1992 to 1998, it then returns to a temporary high mobility state when the water table falls to a low-level in early 2000.

Fig. 12.10

Hydrograph of fluid levels in an unconfined monitoring well showing both the inverse relationship between LNAPL thickness and the corrected groundwater elevation. The estimated LNAPL transmissivity associated with different thicknesses is also shown, based on laboratory and field data for this location

The lesson is clear and often repeated. T_n is dependent on the overall timing and conditions of the aquifer acting on the LNAPL release, as shown in this example. The LNAPL plume mobility state may be negligible for many years, only to increase dramatically on a change in conditions, particularly falling water tables. Clearly, a single set of T_n measurements at one point in time will not likely reflect the full range of potential mobility and recoverability conditions. If T_n testing is performed at periods of high entrapment, those will indicate a misleadingly small mobility and recoverability value. This is also why the quality of the LNAPL conceptual site model (LCSM, discussed elsewhere) is so important. If in-situ LNAPL conditions are not adequately understood, there is no context to evaluate whether and when T_n values will be at their greatest, commonly corresponding to the greatest mobility risk. Simply, T_n is transient and so is any risk associated with potential LNAPL plume mobility, even at late stages. One needs to look for transient events, not field steady-state conditions, if we wish to properly understand the mobility and risk context.

Other conditions of T_n testing include a number of variables and some fairly restrictive assumptions in the analytic methods of determining those values. Those include the linear and nonlinear parametric aspects discussed above, as well as practical aspects that may include those listed below (and others). In the final subsection of the chapter, we will review how some of those conditions affect interpretations regarding T_n estimation, LNAPL mobility, recoverability, and other aspects.

• Assumed well and LNAPL continuity with the formation

  • Potential for PVC swelling

  • The effects of well filter packs and other construction materials

  • Drilling rinds that may limit hydraulic connectivity

  • Screen interval relative to the location of the mobile S_n profile in the formation

• Potentially confined LNAPL conditions

• Local area versus plume-wide characteristics

• Variations in permeability and saturation conditions that are not identified in well logs

• Variations in petrophysical and fluid properties at scales too small to visually identify

• Potential for multiple releases creating differing LNAPL conditions and physical properties

• Differential weathering of the LNAPL body and associated variations in liquid properties

• Hydrologic overprinting and development of a “smear zone” that contains mobile and immobile LNAPL and a resulting variable mobility state

• Release of occluded “immobile” LNAPL on changes in phase conditions, such as a falling water table.

Lastly, all aquifer testing analytic approaches have a number of constraining assumptions. For instance, the Bouwer–Rice slug test method is commonly used for T_n derivation from baildown tests (see references above). This slug test analytic method is based on the steady-state confined radial flow model, and the boundary requirements are: (a) aquifer/LNAPL zone has infinite areal extent; (b) aquifer and LNAPL zone is homogeneous and of uniform thickness; (c) aquifer/LNAPL potentiometric surface is initially horizontal; (d) the volume of LNAPL recovered to initiate the test is instantaneous; and (e) the return flow is steady state. Clearly, (a), (b), and (e) are never fully met, and (c) and (d) are often not met during LNAPL slug testing, introducing non-ideal conditions and error into most T_n derivations.

Blended numerical analytic approaches to T_n derivation have been developed (e.g., Zhu et al. 1993). These couple the capillary saturation analytic models (discussed elsewhere in this text) with a nonlinear parameter estimator to result in estimates of T_n and associated multiphase soil properties. One can also use multiphase models (e.g., MAGNAS3 1994; T2VOC 1995) to analyze test results by setting background saturation and physical property conditions that are then transiently overprinted by the initiation of the bail or pumping tests. Like an aquifer pump test, one then varies the controlling parameters to result in a model match to the observed LNAPL recovery. The author has found good utility in such numerical investigations, but those typically require a substantial work and computing effort to derive informative results. But that effort often results in better understanding via the struggle to adequately explain the data response. In any case, uncertainties and variability of in-situ field LNAPL conditions limit the resolution of even these more comprehensive solutions. The reader recognizes clearly by now that the number of controlling parameters is such that evaluations results are often non-unique (i.e., more than one parameter suite may explain the observed test results).

An alternate approach to estimating K_n and T_n is to measure the individual parameters comprising these porous media properties. Saturation testing is a direct physical measurement, as are fluid properties. Capillary testing can provide the relationship between pressure and saturation, usually on a 2-phase system that can then be scaled to the other fluid couplets through surface and interfacial tension ratios (see the capillary theory discussion elsewhere in this text). Relative permeability functions can be measured or assumed. Collectively then, one can estimate K_n directly for a given soil type, LNAPL type and saturation, with T_n being the integral of that over the usually variable LNAPL saturation profile in question. While that sounds simple enough, laboratory and field experience has shown there to be many issues with the scaling and application of petrophysical parameters to field conditions. Some of these issues will be discussed below in the field observations subsection of this chapter.

Taken collectively, it should be clear that T_n or K_n values and their derivation are probably best viewed as approximations of a highly complex set of variables and often nonlinear conditions. As will be shown in the next section, it is recommended that one inspect multiple lines of data when constructing an understanding of LNAPL plume conditions and the potential risks posed. Like most hydrogeologic endeavors, one will rarely find that inserting parameters into equations will result in representative estimates of T_n or other important attributes of LNAPL plumes. That will only result in decimal points on a derivation that is itself often invalid.

12.9 Field and Laboratory Testing Observations

This subchapter will provide some of the field and laboratory testing observations that the author and others have developed over the last few decades. As a researcher and primary author of the API LNAPL Parameters Database (Beckett and Joy 2003), the author has benefited by observing thousands of LNAPL plumes and overseen the LNAPL characterization and testing at many of those. The mix of the provided observations is meant to develop a deeper understanding of the real-world complexities that can impact multiphase mechanics and its application to important aquifer and environmental protection problems. Each aspect of the observations provided is important, and it is likely a site-specific matter as to when one area of observations become more important than others. It is also the case that while the observations themselves are often clear in their implication, the underlying causes may not.

12.10 Intrinsic Permeability and Fluid Type

It is a common applied assumption, as implied in the equations and discussions above, that intrinsic permeability is an invariant property of the porous media (as indicated by the term “intrinsic”). However, that assumption often conflicts with measurements. It has been recognized in the petroleum production industry that the oil permeability of different reservoirs can vary as a function of their mineralogical makeup. One supposes that should be similar in environmental and aquifer restoration applications since we encounter analogous geologic materials, albeit under differing conditions.

It is rare to find agreement between air-based and water-based intrinsic permeability measurements. Figure 12.11 shows some typical permeability laboratory data selected at random from the author’s archives for a wide variety of soil types and permeability ranges. In all but one sample (of 29), the water-based permeability is less than the air-derived value, often by an order of magnitude or more. It is also clear that, in general, lower permeability/finer-grained materials exhibit larger contrasts in the measured permeability values between water and air. Indeed, it is in finer-grained materials, particularly those containing clay, where the polarity of water can interact with the porous media and produce the observed result. This working hypothesis explaining the observations would certainly benefit from additional scientific inquiry (as is true of much else in this text). But we see these effects in the field based on multiphase flow characteristics that are explainable only under non-ideal conditions.

Fig. 12.11

Laboratory permeability measurements showing the common contrast between water- and air-derived intrinsic permeability values

While one can easily define an effective air transmissivity (e.g., Beckett and Huntley 1994), the purpose here are the implications to T_n and LNAPL mobility. From the observations above, one might suspect a similar outcome when one measures permeability to water relative to a petroleum product. Petrophysical laboratories often use kerosene for such tests. Figure 12.12 provides an example of permeability tests conducted on exactly the same core samples and under precisely the same test conditions. This site consists of mixed alluvial sediments with a significant clay/silt content in some horizons. As observed, the permeability contrast is quite large, up to several orders of magnitude for this example. The author has seen this outcome almost universally in samples tested in this way and suggests the reader try the same experiment. The primary implication is obvious; LNAPL is commonly much more mobile, particularly in fine-grained materials, than would otherwise be expected by applying standard ranges of intrinsic permeability. From a practical standpoint, it also means that field tests will tend to produce a smaller range of T_n variability than theory alone might suggest (more to follow on this issue). Coupled with this observation are difficulties in scaling laboratory-based measurements to the field (again, to be discussed).

Fig. 12.12

Example contrast of intrinsic permeability measured on the same samples separately with water and kerosene

12.11 Interfacial Tensions

Elsewhere in this text the reader will have been introduced to interfacial and surface tensions and how those are used to scale the capillary couplets between the phases of interest (air/water, water/oil, and oil/air). The IFT factors are often buried in other calculations or models and, in the author’s experience, are viewed as relatively unimportant or otherwise being known through literature ranges. However, field conditions can substantially alter the IFT values, particularly for the oil/water couplet. The API LNAPL Parameters Database (API 2006) provides oil/water IFT values for 28 fluid pairs. Excepting one likely outlier at 65 dynes/cm, all IFT values are significantly smaller than a common literature value of about 50 dynes/cm; the measured range is from 8.0 to 39.4 dynes/cm. Because the IFT scaling parameter is used in the capillary functions, the ranges of IFT create another nonlinear effect. Figures 12.13 and 12.14 show this for four IFT values (low, average, and high values from the API LNAPL Database [API 2006]) relative to the commonly assumed value of 50 dynes/cm (S_n and T_n shown).

Fig. 12.13

Gasoline saturation under different IFT conditions for soil type 4. Note the zone above the free-oil surface (1 m) is shown because of the substantial mass present for low IFT values. The integrated LNAPL mass under the curves is given in kilograms

Fig. 12.14

Gasoline transmissivity as a function of elevation above the oil/water interface for varying IFT values, gasoline saturation under different IFT conditions for soil type 4

First, as the IFT value decreases, the LNAPL saturation and mass increase. As we know from above, that means that the relative permeability then increases exponentially, resulting in much higher LNAPL mobility than would be estimated using an assumed literature value of 50 dynes/cm. It is well known that petroleum hydrocarbons degrade in the environment, often producing polar compounds. While more research is needed, the author has observed generally smaller IFT values in plumes that have been weathered and exhibit a substantial footprint of polar compounds. One would generally expect that as the LNAPL interfaces become more polar through biodegradation, the IFT would decrease relative to non-polar LNAPL (i.e., fresh). Again, the effects of IFT are obvious and substantial, but more study is needed on how it changes, rates, where (at phase interfaces or throughout the LNAPL body), its transient behavior and other aspects. As noted previously, the nonlinear aspects of multiphase theory make it important to understand the implications of parameter variability in the field and how that impacts mobility, recovery, and risk conditions.

12.12 Real-World Heterogeneity

Most people working in the field of groundwater contamination recognize the importance and ubiquitous presence of heterogeneity. This subsection will share data and observations from field sites that demonstrate, as suggested by theoretical considerations above, a complexity and scale that has direct implications to T_n, mobility, recoverability, and risk. In brief, the scale of complexity is such that most qualitative and quantitative evaluation approaches will not be reflective of this scale, but rather represent forms of averaging. As with other aspects discussed throughout this text, significant additional research is necessary if we are to better understand these processes to ultimately result in better groundwater protection and restoration for our public stakeholders.

Although this subsection is titled “real-world heterogeneity”, it is useful to recall some interesting bench-scale observations. John L. Wilson and his colleagues at New Mexico Tech did some pioneering bench-scale modeling work in the late 1980s and early 1990s, including scanning electron microscopic (SEM) logging of some of those experiments (Wilson et al. 1990). One of the key observations of this wide-ranging investigation is that even in relatively homogeneous pore fields, heterogeneous LNAPL behavior is the norm. Only the pore pathways with continuous phase saturation and connectivity will contribute to mobility and the associated T_n values. For anyone wishing to “see” NAPL behavior, this study suite is exceptional and highly recommended.

12.13 T_n Field Observations

In this section, several sets of T_n tests will help to demonstrate the following observations:

• LNAPL thickness and T_n do not often correlate in the field, often due to heterogeneity;

• T_n from pumping test can be much larger than from baildown, and if the point of T_n is recoverability, then it likely needs to be measured under field pumping conditions;

• The quandary of stability; we often have clearly stable LNAPL plumes, yet T_n and gradients are both present that indicate mobility if one applies a simple Darcy approach. Balancing factors should be considered, such as pore-entry limitations, NSZD, and others when assessing stability.

In a uniform and relatively homogeneous system, T_n would be strongly correlated to the initial LNAPL thickness observed in a monitoring well. Figure 12.15 provides the results of T_n testing at a large release site as a function of the observed initial LNAPL thickness in each test well. This example site is in a dune sand system and is as homogeneous as one will ever find in the real world. For example, extensive permeability testing of these sands determined a range from about 2–13 Darcy, but within a factor of two for the 25th and 75th percentile values. Visually, one cannot distinguish any difference in permeability, and all boring logs simply identify a clean fine- to medium-grained sand. Further, the LNAPL released is also relatively uniform. In other words, there was just a single LNAPL type released. Yet review of the test results shows, contrary to theory, that there is no strong relationship between initial LNAPL thickness and the resulting T_n value. In fact, some of the smallest T_n values are associated with the largest thicknesses (no real surprise to our inner geologist).

Fig. 12.15

LNAPL transmissivity results at a site with uniform dune sand aquifer and generally uniform LNAPL characteristics. Note, the wells names are pseudonyms to protect the privacy of the site owner

What may explain the non-intuitive observations above? The multiphase mechanics discussed previously demonstrated that a wide variety of factors have nonlinear effects on the T_n results. For instance, although all the LNAPL released is of a very similar character, it was released over time allowing an opportunity for environmental changes in certain areas. For instance, the IFT values are variable and range from about 10–20 dynes/cm. The discussion above demonstrated the nonlinear effects of IFT on expected LNAPL saturation and T_n values. There is also the potential effects of fuel weathering and compositional changes that, in turn, affect the viscosity of the LNAPL in the subsurface. This site has a strong polar hydrocarbon footprint. Fluid measurements at this site suggest viscosity ranges from about 5–30 cP, or equivalently, influencing the T_n values by a factor up to six-fold (recall viscosity is a linear factor). The LNAPL density range, however, is quite uniform.

While there are many possible reasons for the observed T_n contrasts, one can easily observe that ideal theoretical outcomes are not always present in the field, even at the most uniform of sites. If direct theory application does not hold at this example “homogeneous” site, it is hard to imagine it being representative for almost any site.

Contrasting the above with a more typical site is also informative. This example site is in a large alluvial basin, and the soil types at the water table/LNAPL interval range from silty and clayey sands to coarser clean sand intervals. As shown in Fig. 12.16, more typical heterogeneous sites have wide range of T_n values as a function of the initial thickness (note the log scale because of that range). One can observe that the greatest T_n values correspond to two of the smallest initial thicknesses. Conversely, the smallest T_n values correspond to an observed well thicknesses of more than 3 m (or ~ 10-ft) of LNAPL. This is geology working in a multiphase way. From a mobility, risk, and recovery perspective, one may not have suspected that some of the smallest LNAPL thicknesses actually represent the largest mobility and risk potentials (i.e., thought is required).

Fig. 12.16

LNAPL baildown test results plotted relative to increasing initial well LNAPL thickness (Log F.P., increasing to the right). Note, well names are pseudonyms for the privacy of the site responsible party

A final observation regarding the example in Fig. 12.16 (and others like it) is as follows. Although the range of T_n values is relatively wide, covering a little more than two orders of magnitude, that range does not remotely begin to approach the end-members that the theoretical discussion suggested. While this is a topic for more elaborate discussion elsewhere (and research), it has been the author’s experience that T_n values from a wide range of sites and LNAPL types fall within a relatively narrow range that is inconsistent with theory, at least as commonly applied. Recall that earth materials are expected to have permeability ranges over many orders of magnitude, and those will be amplified by the capillary, relative permeability and fluid property aspects. We do not see that in field results. Why?

There are several possible explanations that combine to comprise the observed results. It was shown earlier how fine-grained materials have a much larger permeability to LNAPL than to water (Fig. 12.12). That would narrow the range of field T_n results in the direction observed above. Secondary permeability and other features in some sediments are present, but their effect on LNAPL mobility and T_n are commonly not recognized. The author worked on a site in a marine clay environment. Fuel impacts were not observed, directly beneath the tanks and pumps, from ground surface through about 30 m of vadose zone. Several feet of free-phase LNAPL was found within the water table zone. In this case, layering in the marine clay (fissility) allowed essentially a fracture-like transport regime to occur from the release source to groundwater. The only hint of impacts was very faint petroleum odors, but field laboratory sampling exhibited no detections in the vadose zone. If one had simply logged the subsurface without extending the investigation to the water table, it might have been reasonably (but wrongly) assumed that the thick zone of clay would impede any possible LNAPL transport downward. These and other field nuances control actual risks, and theory alone can be misleading.

In summary, the real-world exhibits heterogeneities in porous materials, in their fine-scale distributions, in secondary features, in LNAPL characteristics spatially and temporally, and others. Many of these aspects are not well studied, and some tend to be unsupportive of certain aspects of multiphase theory as commonly applied (per text above). The author’s own take-away from all this is that LNAPL plumes will generally move farther and faster than the underlying multiphase mechanics suggest. Finite releases also typically stop migrating in relatively short periods, and old plumes are generally stable plumes (but, clearly with exceptions). In the pursuit of useful aquifer and environmental protections, we must all keep our eyes open to these possibilities. Those include the unexpected transport behaviors mentioned, as well as degree of heterogeneity that is probably best described as fractal in many settings (recall Wilson et al. (1990) groundbreaking experimentation and SEM work).

12.14 The (F)Utility of LNAPL Recovery

In this section, we will look at the utility and (often) futility of LNAPL hydraulic recovery as part of the T_n paradigm. This will be a combination of real-world observations coupled with the heuristic tendencies one might expect based on some of the theoretical underpinnings. A phrase that the author and late colleague Dr. David Huntley have long used is: “It is the LNAPL you leave behind and its characteristics that define the net benefit of any hydraulic recovery action.” Dr. Huntley also used to say: “It’s the LNAPL, stupid.”, at times addressing this author. Meaning, do not overlook the importance of the LNAPL characteristics of multiphase, multicomponent conditions that are the key to understanding potential risk and potential benefit of any remedy action. This section considers only hydraulic recovery because of its prevalence and relationship to T_n.

T_n is often used as a metric for deciding on whether or not LNAPL hydraulic recovery might be practicable. For instance, the ITRC LNAPL guidance (ITRC 2009) on this matter suggests a cutoff range of 0.1–0.8 ft²/day, below which recovery is deemed likely impracticable. The API Interactive LNAPL Guide (API 2006) provides a number of LNAPL recovery nomographs as a function of equilibrated LNAPL well thicknesses (API, 2006) showing when recovery is theoretically possible (and where not). Figure 12.17 shows one example chart for an initial equilibrated (VEQ) LNAPL thickness of 2.5-ft for a variety of different fuel products. Given the discussion above on observed T_n variability, charts like this based on ideal theoretical conditions should be used with caution and assessed for whether site-specific T_n values align with the projections of recovery/mobility thresholds.

Fig. 12.17

LNAPL recovery probability chart showing the cutoff between potentially recoverable and non-recoverable product as a function of the LNAPL viscosity and the hydraulic conductivity of the soil materials. Source; API Interactive LNAPL Guide (API 2006)

As the reader now knows, recovery conditions are proportional to the associated T_n value, which quickly diminishes as the saturations and observed thickness decreases. Said another way, LNAPL recovery always reduces T_n and is a self-diminishing process (absent new or ongoing releases). LNAPL recovery is also one of the most commonly applied remediation techniques. For instance, a review of the Los Angeles Basin refinery system found that hydraulic recovery was active (or had been historically) at almost all of the sites (Beckett et al. 2005). The same is true of many other refineries and large-scale terminals across the globe.

12.15 Recoverability Assessment of a Recent Release

This subsection will review a site-specific assessment of LNAPL mass and recoverability for a then recent release (relative to the time of the study; Beckett and Lyverse 2004). A synopsis of the study is provided followed by a few key findings as they relate to mobility and recoverability.

12.15.1 General Site Background and Findings

A field-based assessment of LNAPL and dissolved-phase mobility was undertaken after a pipeline valve rupture in early 2000 (light sweet crude) released approximately 13,000–16,000 barrels. The study is unique in that the time of release is precisely known, and it is a single release event with a uniform product and environmental sampling has provided an exceptional field-based tracking of the LNAPL plume migration. A comprehensive investigation using borehole geophysics (cone penetration testing [CPT], laser-induced fluorescence [LIF]) followed by continuous coring and petrophysical work produced a detailed understanding of site conditions. Consistent with the discussion above, based on observational data the LNAPL plume moved much farther and faster than would have been predicted based on soil and oil characteristics. The plume ceased migration after less than four years and was determined to be hydraulically non-recoverable at that stage of development.

Excavation and soil sampling work began within a short-time after the release was identified. That work was followed by boring and well installations, two time-separated CPT/LIF investigations, continuous coring with high resolution photography under white and UV light, and extensive petrophysical testing to define the soil and oil multiphase characteristics.

The site subsurface materials consist of marly sands and interbedded finer-grained materials. The laboratory-derived percentage of fines (silt and clay) ranged from about 40–85%. Oil saturations in the LNAPL interval ranged from non-detect to a maximum of about 17%. The LNAPL zones were targeted for sampling using the results of the LIF logs, coupled with high resolution core photography. Even within the LNAPL interval, it was observed that there were non-detects in some zones, which is at odds with the concept of a continuous-phase interval. Heterogeneity in material properties causes heterogeneity in saturations, and the real-world is often non-uniform in its behaviors.

The field-measured hydraulic conductivity of the water table zone ranged from about 0.15 to 40 ft/day, with a geometric mean of 4.5 ft/day, somewhat larger than might be expected based on the fine fractions present. The oil T_n ranged from about 0.02 to 0.4 ft²/day at the time of measurement about two years after the release and under declining mobility conditions, as developed further below.

The sweet crude has a density of about 0.84 g/cc, and a viscosity of about 4.5 cP, and an oil/water IFT of about 25 dynes/cm. It also has a large aromatic component that was used, in part, to track and understand plume development since the dissolved-phase plume is sourced by the LNAPL. Measured capillary properties indicated a relatively large capillary rise, or alternatively, a relatively high-water retention characteristic; an example characteristic curve is shown in Fig. 12.18. In general, given the observed and measured characteristics described, one would not expect high rates of LNAPL migration, but those were observed quite clearly in the data, as further discussed.

Fig. 12.18

Capillary characteristic curve from a boring sample collected in the zone of LNAPL migration. Note the relatively large capillary fringe equivalent of approximately 100 cm

Because site investigation activities occurred coincident with the LNAPL plume migration, we were able to track that movement via a combination of data observations. One was new arrivals of free product at monitoring locations. Another was observed changes between two time-separated CPT/LIF investigation events (see below). Finally, increases in groundwater petroleum concentrations were also used to interpret the migration characteristics. Fig. 12.19 shows the general growth of the LNAPL plume over time, from February 2000 through the end of 2002, by which time, the majority of plume movement had ceased. As seen and expected, the growth in early time was the most significant. We estimate that the plume began its migration at rates of about 30–50 ft/day expansion. This is based primarily on field observations and limited sampling as the early movement did not yet have dedicated wells and other investigative installations. By August 2001, when many more investigatory locations were available, the approximate rate of movement was about 1 ft/day, falling to about 0.1–0.2 ft/day by August with negligible migration after December 2002, or about three years following the release. The estimated incremental migration rates for the various available time stamps are shown in Fig. 12.20. Note the time gap from early 2000 to August 2001 corresponds to the time where investigation locations were being installed and data collected and assimilated.

Fig. 12.19

LNAPL plume area outlines based on available data suites at sequential points in time

Fig. 12.20

Estimated incremental LNAPL migration rates based on the spatial distribution of LNAPL impacts. The progression is represented by a fourth-order polynomial fit

The behaviors shown in Figs. 12.19 and 12.20 are heuristically consistent with the multiphase mechanics discussed previously. LNAPL mobility is greatest in early time and diminishes as the plume spreads into a larger volume of porous materials with an associated fall in average saturations and free-phase thickness. However, while the overall trends are consistent with theoretical expectations, the quantitative details are not (once again). For instance, the initial spreading rates on the order of tens of feet per day dwarf the range of groundwater flow. By the end of three years following the release, the long-axis of the plume had spread to about 1000-ft in length, or about 333-ft per year on average to give a general sense. The linear groundwater velocity is on the order of 20–80 ft/yr. This is a fine-grained setting and combined with the capillary properties and measured saturations that would otherwise suggest very limited LNAPL mobility, quite contrary to actual observations. For instance, if one used the analytic tools available in the API Interactive LNAPL Guide (2006), the estimated LNAPL velocity for this site’s parameter ranges would be less than a tenth of a meter per day (accounting for the site parameter ranges). The key observation, once again, is that LNAPL generally migrates faster and farther than one might anticipate using solely theoretical relationships. This plume also ceased migration much faster than would be anticipated by theory alone and parameter application within the related equations or transient models.

Finally, there are several key site investigation observations that provide insights that could not be realized from standard site investigation techniques alone (such as borings, visual logging and related). Foremost among the critical data were the combination of CPT/LIF investigations, coupled with the laboratory core photography, to isolate core sampling target intervals that would be most relevant to the various technical questions regarding mobility, mass recovery, and potential environmental risks. First, the relative changes in the LNAPL plume morphology were clearly demonstrated by in-situ changes documented by the two separate CPT/LIF investigations. Figure 12.21 shows a plan view of the LNAPL plume distribution as indicated by LIF responses.

Fig. 12.21

Indicated smear zone thickness (ft) distribution in plan view. The smaller grey-shaded plume was defined in an August 2001 LIF investigation, the colored isopach by LIF results from December 2002

These same CPT/LIF data also provide a strong basis, when coupled with petrophysical photologs and test results, to evaluate the fine-scale distributions of lithologies and LNAPL within those. Figure 12.22 shows a geologic cross-section from prior site work by the well-qualified geologist who logged the borings. Figure 12.23 is approximately the same cross-section showing the fine-scale detail available from the combined CPT, LIF, and correlated petrophysical results. The contrast between the two could not be more pronounced. Figure 12.22, while accurate to its level of resolution, is not particularly useful in assessing the LNAPL properties or nuances that control this system. The tools and data suite in Fig. 12.23 is directly informative to the LNAPL mobility and risk questions under consideration.

Fig. 12.22

Geologic cross-section from the site by a qualified geologist. This is the best that the “eye” can do

Fig. 12.23

Combined geologic, CPT, LIF, and petrophysical-related cross-section. This bears only slight resemblance to that above. LIF intensity is in percent of calibration, and tip/sleeve ratios (%) were correlated to continuous cores for the lithologic interpretation

In summary, it required the full suite of data collected, from geophysical, to continuous cores, to petrophysical and chemical analytics to adequately define the plume conditions and make a risk determination. Clearly there is an interpretive component as well, which is part of the fun and challenge to multiphase mechanics. Where such data are unavailable (e.g., like in Fig. 12.22), there are many degrees of freedom around those interpretations, and the author highly recommends a holistic and thoughtful approach. One will not generally produce useful answers to mobility and risk questions by entering a static suite of parameters into a calculator or model. Clearly, LNAPL mobility/recoverability is vastly more complicated than that.

12.15.2 LNAPL Mobility and Recoverability

With the site background in mind, we can now consider the question of recoverability. As noted above, the T_n values in March 2002 were all below the upper limit of the ITRC recovery T_n cutoff value of 0.8 ft²/day, and three of the five results were below the lower limit (see Fig. 12.24). As will be seen, the plume is indeed not recoverable by hydraulic methods to any practicable degree. However, it should also be recognized that this LNAPL plume was still mobile and laterally migrating at the time of these measured transmissivity values. Per the discussion above, this is why the conductivity, gradients and mobility potential are related, but represent a much different question than recoverability using T_n as its proxy. In other words, mobility can be present even at low T_n values if the discrete effective LNAPL conductivity and gradients are sufficient to allow migration to occur (other factors apply as well). The key question for this example site was whether there was any potential risk that might result from further migration. There was not because the plume was demonstrated to be stabilizing and could not reach any distal receptors. The shallow groundwater in this area is not potable. However, risk considerations are a discussion for a different chapter in this text.

Fig. 12.24

LNAPL transmissivity values based on baildown tests at five monitoring locations. Once again, there is no relationship between T_n and initial LNAPL thickness, and all values are less than the upper ITRC cutoff value. Note the log scale

From the site transmissivity values shown in Fig. 12.24, hydraulic recovery is unlikely to have any benefit. Recall, however, that the transmissivity is only relevant to the continuous phase LNAPL intervals, not the total of the plume body. From the data above, it can be determined that the phase-continuous LNAPL interval is about 10% of the total smear zone volume. Within that interval, about 17% of the LNAPL may be recoverable under ideal conditions. That assessment is based on two-phase and three-phase residual testing of capillary pressure-fluid saturation relationships and model evaluations of the recovery decay with distance from a pumping well (e.g., Fig. 12.25). The contrast in geometry between the phase-continuous interval and the total smear zone is shown in cross-section in Fig. 12.26.

Fig. 12.25

Model simulation of LNAPL recovery at its asymptotic endpoint after 10 years of simulated LNAPL recovery. As seen, the radial effectiveness decays rapidly with distance and a substantial mass remains in the free-phase LNAPL zone. This particular simulation is analogous to, but is not site specific (i.e., heuristic)

Fig. 12.26

Cross-section showing the outline of the LNAPL smear zone and within that, the zone of LNAPL phase continuity, which represents about 10% of the whole. At one time, there must have been phase continuity within the total smear zone, though not necessarily at the same point in time

Combined, the data and evaluations indicate that something on the order of only 2–4% of the LNAPL plume is recoverable as of the time plume migration ceased. This is consistent with site hydraulic recovery testing (trench and well pumping) that combined recovered less than 300 gallons of oil; recall, as much as 700,000 gallons were released. It does not require additional calculations to demonstrate that only a diminish fraction of the oil is recoverable and of too small a volume to be of any benefit. However, it is completely a different question as to whether hydraulic containment might have value. As noted, the plume was migrating over about a three-year period. If there had been risks associated with that movement and the final plume distribution, then hydraulic containment with gradients sufficient to overcome the LNAPL migration gradients would be one protection measure to be considered. Cutoff trenches and others were also considered. In the end, the oversight agency concurred that the LNAPL plume had stabilized and there were no risks presented by this particular plume and its context. Obviously, this may not be the outcome at other release sites. As noted previously, even immobile LNAPL plumes can potentially present risks through fluxes in groundwater, vapor, or as a source for methane generation (to name just a few).

The conclusions and observations from this example site are compelling. A plume less than three years of age had already become impracticable to recover by hydraulic methods. The T_n implied under early migration conditions is many orders of magnitude greater than that observed at near-stable conditions. The residual smear zone is vastly larger than the free-phase interval and that will be the controlling mass relative to plume longevity. What might that suggest about other, older LNAPL plumes that have had vastly longer periods of residualization, entrapment, and weathering? In the next subsection, we will consider some hypothetical situations.

12.16 Laboratory-Derived versus Field LNAPL Transmissivity

A final case study is briefly presented here. At a site with an older suite of fuel releases, a detailed field investigation was conducted. Many of the approaches were similar to the case study above, and the approaches will not be developed in detail. Continuous coring, petrophysical, CPT/LIF, and other investigation methods were applied to result a comprehensive LNAPL CSM. That CSM determined that the LNAPL was predominantly residualized, stable, and contained about one million gallons of free, occluded, and residual NAPL. Based on the data and multiphase evaluations, it was determined that less than 3% of the LNAPL in-place was recoverable, as shown in Fig. 12.27. The zone of potential recovery is also geographically limited, as heuristically anticipated by multiphase theory.

Fig. 12.27

Distribution of the recoverable LNAPL fraction based on detailed multiphase parameter collection, mass estimates, modeling, and related evaluations

Lastly, a comparison was possible at this study site between the T_n values derived by baildown testing, versus the calculated values derived from the laboratory parameters described previously that define T_n (see Eq. 12.1). The laboratory-derived values were determined to be orders of magnitude smaller than the field measurements. The differentials ranged from a low of about 5 (field values 5 times greater than the laboratory value at that location) up to about 250 times greater. It is commonly recognized that laboratory scaling and other issues often produce results that are not field-comparable. The complications of multiphase mechanics make this laboratory testing issue even more prevalent in the author’s study and testing experience. While petrophysical testing is valuable and provides insights, the parameter derivations generally cannot be put into a simple equation and result in realistic estimates of T_n (or mobility).

12.17 Flux and Longevity Considerations on LNAPL Recovery

This last section will briefly explore physical and chemical considerations with respect to LNAPL mobility and recoverability. In other sections of this book, the physical and chemical properties and variability of LNAPLs were discussed. As is clear from this chapter, each of those has a direct or indirect effect on the mobility and recoverability of plumes in various settings. As shown by the example above and many others not discussed, the age of the LNAPL plume is often inversely related to its mobility and recoverability. Plumes that are stable, meaning not expanding geographically, are also not generally recoverable by hydraulic means. This is because the processes that cause stabilization are essentially the same that limit recoverability (residualization, weathering, pore entrapment, submergence, etc.). Before exploring these limitations, an editorial note is warranted. This discussion is about applying good science to set expectations on aquifer restoration measures. We have already observed that there are significant limitations to hydraulic recovery approaches and, where applicable, would often be part of a broader remedy solution. Recovery limitations do not imply that remediation is not viable, but rather that hydraulic recovery specifically may be incapable of achieving cleanup objectives and other options, or combined approaches, may be more appropriate. Where mitigation measures are necessary, recovery should not be the only method considered, particularly if there is no practical benefit. In short, remedy evaluations are about determining the most appropriate and effective measures and hydraulic recovery may often not be in that category.

For ease and consistency of comparison, we will explore the theoretical recovery aspects for a few different sets of conditions. The methods described in API #4715 (Hunley and Beckett 2002) will be used for the comparisons, using the toolkit in the API Interactive LNAPL Guide (API 2006). As should be clear by now, these comparisons are heuristic. Real-world plumes will be more complex than indicated by the simplified models applied. In general, these VEQ-based models will underpredict recovery in fine-grained materials and overpredict the same in coarser soils, per the earlier observations that field T_n/mobility conditions tend to coalesce around a much narrower range of values than indicated by multiphase theory. Further, these methods only consider the free-phase, vertical equilibrium LNAPL interval, not the full plume mass, which is critical to site-specific considerations as shown above. In other words, take the lessons learned and assume actual field conditions will be much less amenable to recovery attempts than what these models might suggest.

The evaluations here will use the three coarsest soil types defined in the theoretical discussion above; (i) Sandy loam; (ii) Sand mixtures; and (iii) clean coarse sands. Finer-grained materials will not be evaluated because they will not be predicted to have any recoverable LNAPL volumes. For each of these three soil types, LNAPL skimming and dual-phase (water and LNAPL) pumping recovery will be estimated on a mass basis, as well as inspecting for component chemical changes that might occur. Each recovery action is assumed to run for three years. For estimates, we will look at benzene and naphthalene as indicator compounds to give a sense for how recovery might influence the composition and longevity of these compounds.

Numeric values of mass are in kilograms for an LNAPL plume with dimensions of 30 × 30 m and an equilibrated initial thickness of 1 m. The percentages in the bottom section are the estimate percent of the initial mass recovered.

Table 12.3 presents the key mass and mass reduction results. On a bulk basis, one can quickly observe two key outcomes. Diesel is not as recoverable as gasoline, both because of its property differences, as well as differences in its distribution relative to the water table. Second, there is a vast difference in recovery potential, for this same initial LNAPL thickness (1 m), between these soils. The loamy sand presents a negligible recovery mass under these conditions. This observation is why there is no value or need to inspect finer-grained materials. Recovery in fine-grained materials will rarely have a net benefit for all the collective reasons discussed previously (keeping in mind that field conditions can and do differ from theory).

Table 12.3 Mass and mass change percentages for each condition

The estimated LNAPL recovery curves for each scenario are shown in Fig. 12.28 for the three years of recovery. One can observe the significant differences in both the cumulative totals between the different soil types and fuels, but also that the asymptotic break in each differs substantially. The smaller the initial and incremental T_n (and saturations), the less LNAPL that will be recovered and the longer it will take to reach an asymptote. This closely parallels general field observations of asymptotic recovery for hydraulic systems. The asymptote means that additional recovery is small and perhaps impracticable. It does not imply, however, that the recovery action has had any net benefit, it just means it is finished.

Fig. 12.28

Cumulative LNAPL volume recovery curves for each scenario (three soils and two fuels). Note that the recovery in the loamy sand is near zero for the diesel condition

As noted previously, it is the LNAPL mass left behind that often defines the net benefit of hydraulic recovery as compared to other mitigation measures. This is particularly true if the LNAPL plume in question is stable and no longer mobile in the environment. As discussed in the earlier case example, although the LNAPL was not practicably recoverable, there could have been a net benefit to hydraulic containment or other measures if the temporarily mobile plume had posed a risk.

For each of these example estimates, the benzene and naphthalene concentrations over time were tracked (methodologies per Huntley and Beckett 2002; API 2006). At the end of the three-year period, the estimated time reduction to reach 5 ug/l benzene and 20 ug/l naphthalene was calculated, as shown in Fig. 12.29. This chart shows that there is no net benefit to recovery in the loamy sand for either gasoline or diesel fuel relative to the longevity of chemical impacts. There is some potential benefit in the intermediate soil, more for gasoline than for diesel. The clean sands had the greatest time reduction by a factor of about 4–5 depending on the compound and fuel type.

Fig. 12.29

Estimated chemical longevity as a percentage of the original for each fuel, soil type, and hydraulic recovery method. A value of 100% means the chemical longevity is unchanged relative to initial ambient conditions

The methodology of API (2002) accounts for multiphase, multicomponent partitioning, fluxes, and transport processes. It does not account for potential mass losses other than those partitioned fluxes or as mass removal from a recovery action (e.g., NSZD is not considered, nor are other complex processes). For those reasons, the absolute times of depletion are not as relevant as the comparative values noted. A blending of these types of evaluations with actual plume-specific chemical depletion observations is one way to assess some of the chemical longevity aspects. Qualitatively, however, one can still make some simple observations. Whatever the time to reach background concentrations may be, it could be scaled by the reduction factors above as a preliminary approximation.

While the heuristic evaluations above are quite useful, probably the most important consideration is not part of a simple VEQ-based estimate, but rather one that also incorporates estimates of the smear zone volume and residual mass. Taking the case example above, the free-phase zone was approximately 10% of the total volume of impacts at the time approaching plume stabilization (about three years). In such cases, even recovery of an idealized 80% of the mass in coarse-grained sands would be insignificant relative to the total mass in-place. Under a condition like that, perhaps 5–10% of the total mass might be recovered. Further, as shown in Fig. 12.25, more robust numerical modeling and field observations combine to indicate that simple uniform recovery around a well does not occur. The recovery effectiveness will decay with distance away from the recovery well, often quite strongly. That is why the methods described in API (2002) and API (2007) are idealized and not likely reflective of the actual field limitations to recovery (at least in most cases). The author performed numerical verification work for both these API toolkits and notes that in general, they are heuristic for the reasons described in prior sections (they are theory-based only with stringent assumptions). As noted above, one would expect such approaches to overestimate recoverability in coarse-grained materials and light-end fuels and the opposite for fine-grained materials and heavy-end products. The evaluation methods, as shown by demonstration, are useful in considering potential comparative outcomes and sensitivities to then lead to field testing and demonstration of actual conditions. Those actual conditions will, in general, be substantially more complex, nuanced and will look nothing like the theoretical estimates.

12.18 Conclusions

There are many conclusions and observations provided throughout this chapter. A few that come together as a result of the whole, coupled with decades of detailed field and laboratory investigations, are as follows:

• Hydraulic recovery has the highest potential benefits in plumes that were recently released and are still migrating in the environment. Little net benefit will be realized for plumes that are mostly residualized and stable in the environment because the processes that lead to that condition also severely limit LNAPL recovery.

• It is the mass left behind after recovery efforts (or others) that generally define the net benefit of the remedy action. Little benefit is expected when insignificant mass can be recovered by hydraulic methods, excepting certain containment objectives may have benefits if properly applied.

• T_n and mobility are highly transient and dependent on various in situ conditions. Fulsome consideration of those conditions and their implications are often warranted. Simplistic application of multiphase equations or models can provide heuristic value, but are unlikely to be representative of actual conditions. The value is dependent on the quality of the interpretations and LCSM.

• Old plumes are commonly stable, residualized, and non-recoverable plumes. If there are risks presented by such conditions, it is recommended that mitigation measures other than hydraulic recovery be considered, as it will do nothing to meaningfully alter those underlying conditions.

• Be aware of transient site conditions, like LNAPL entrapment, that may only temporarily immobilize plumes. Make assessments of risk and plume status holistic and thoughtful to ensure potential changes in conditions have been appropriately considered.

• Build LNAPL CSMs with a focus on processes, material properties, saturation distributions, and related observations. When there is apparent conflict between estimates and observations, clearly observations are favored and require plausible explanations.

• In places and jurisdictions where LNAPL recovery is expected because of perceived mass reduction benefits, challenge those assumptions if there is no net benefit to the action (in favor of actions with a net benefit). The purpose of mitigation measures should always lean toward resource restoration and risk protection (among others). LNAPL recovery usually achieves none of those goals.

• Tangential to the above, it is now widely recognized that LNAPL plumes degrade, sometimes at fairly high rates though natural source zone depletion mechanisms. Where LNAPL plumes are stable and NSZD rates are greater than initial LNAPL recovery rates, there is likely no benefit to hydraulic recovery. A plume in this life-stage is typically mostly non-recoverable and that recovery, if applied, would be expected to diminish rapidly to less than the NSZD rates. If risks need to be managed, apply a different method.