4
Approaches for Evaluating Natural Attenuation

Documenting that a contaminant has disappeared or that the concentration has become very low in groundwater samples is an important piece of information for proving that natural attenuation is working, but it is not sufficient, even at simple gas station sites. Contaminants can bypass sampling locations due to the dynamic nature of groundwater systems. Also, some mechanisms can cause apparent loss of the contaminant, when in fact the contaminant has moved to a place or changed to a form that is difficult to detect.

Because of the limitations of monitoring only for the loss of a contaminant, the National Research Council’s (NRC’s) Committee on In Situ Bioremediation proposed that two other types of evidence are needed to prove that in situ bioremediation of any type is working (NRC, 1993). The first is sound scientific documentation (laboratory measurements or literature describing such measurements) that the mechanism claimed as responsible for contaminant destruction or control is scientifically feasible in the type of environment at the site. The second is documentation that the proposed mechanism is actually occurring at the site. The key issue is that an observed disappearance of contaminants has to be linked to the mechanism acting at the site. In short, cause and effect must be supported. This same principle applies to natural attenuation.

This chapter describes a weight-of-evidence approach for demonstrating the mechanisms responsible for observed contaminant losses in natural attenuation. Direct field measurements of mechanisms of contaminant transformation or degradation are difficult or impossible. Several types



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 150
Natural Attenuation for Groundwater Remediation 4 Approaches for Evaluating Natural Attenuation Documenting that a contaminant has disappeared or that the concentration has become very low in groundwater samples is an important piece of information for proving that natural attenuation is working, but it is not sufficient, even at simple gas station sites. Contaminants can bypass sampling locations due to the dynamic nature of groundwater systems. Also, some mechanisms can cause apparent loss of the contaminant, when in fact the contaminant has moved to a place or changed to a form that is difficult to detect. Because of the limitations of monitoring only for the loss of a contaminant, the National Research Council’s (NRC’s) Committee on In Situ Bioremediation proposed that two other types of evidence are needed to prove that in situ bioremediation of any type is working (NRC, 1993). The first is sound scientific documentation (laboratory measurements or literature describing such measurements) that the mechanism claimed as responsible for contaminant destruction or control is scientifically feasible in the type of environment at the site. The second is documentation that the proposed mechanism is actually occurring at the site. The key issue is that an observed disappearance of contaminants has to be linked to the mechanism acting at the site. In short, cause and effect must be supported. This same principle applies to natural attenuation. This chapter describes a weight-of-evidence approach for demonstrating the mechanisms responsible for observed contaminant losses in natural attenuation. Direct field measurements of mechanisms of contaminant transformation or degradation are difficult or impossible. Several types

OCR for page 150
Natural Attenuation for Groundwater Remediation of field data, combined with models of the subsurface, generally will be needed to link the observed decreases in contaminant concentration to the underlying mechanisms responsible for contaminant losses. As described at the end of this chapter, the level of detail of data and analysis required will vary substantially depending on the complexity of the site. Leaks from gas stations may require only a small fraction of the analysis necessary at large industrial sites with contaminants that are less well understood than gasoline. Nonetheless, the basic principles of analysis described in this chapter apply to all sites. FOOTPRINTS OF NATURAL ATTENUATION PROCESSES Although the mechanisms that destroy or sequester contaminants in groundwater cannot be observed directly, they leave “footprints.” Footprints occur because the mechanism controlling contaminant fate also consumes or produces other materials, many of which can be measured in groundwater samples. Thus, an observation of the loss of a contaminant, coupled with observation of one or (preferably) several footprints, helps to establish the cause and effect that is so crucial to documenting natural attenuation in field settings. As examples, Box 4-1 describes briefly some types of footprints produced by different contaminants and attenuation mechanisms. Table 4-1 summarizes the footprints important for documenting the varying degrees to which natural attenuation occurred in the case studies in Chapter 3, as well as in two new case studies (Bemidji, Minnesota, and an unnamed field site) described later in this chapter. Table 4-1 illustrates two important features of footprints. First, footprints provide evidence for and against attenuation mechanisms. For example, the conversion of organic material (measured as chemical oxygen demand, COD) to methane provided evidence of the reductive dechlorination of trichloroethene (TCE) at the St. Joseph site, but the lack of COD removal indicated that reductive dechlorination of TCE was unlikely at Edwards Air Force Base. Second, the observation of positive footprints does not necessarily mean that the contaminants are fully controlled. Incomplete removal of the original contaminants (as at the Hudson River site) or formation of hazardous products (as at the St. Joseph site) means that contaminant concentrations are still above regulatory levels, even though a natural attenuation mechanism is at work. Using footprints to link cause and effect is not always straightforward. In some cases, detecting small changes that would prove cause and effect is extremely difficult. As an example, reductive dechlorination of TCE at low concentrations may produce chloride and acid at rates that are overwhelmed by natural background levels. In other cases, footprints can be obscured by reactions that produce or use the footprint materials. One

OCR for page 150
Natural Attenuation for Groundwater Remediation BOX 4-1 Examples of Footprints That Can Indicate Natural Attenuation Mechanisms that cause contaminants to degrade or transform in the subsurface cannot be observed directly, but they leave footprints that can be detected in groundwater samples. The following examples explain how these footprints can be used to document natural attenuation: The aerobic biodegradation of petroleum hydrocarbons consumes oxygen and produces inorganic carbon in well-established ratios. Estimating the oxygen supply rate and correlating it with increases in inorganic carbon can yield a quantitative estimate of the rate of hydrocarbon biodegradation, if the changes in inorganic carbon concentration can be measured properly. The biodegradation of organic contaminants, including hydrocarbons, under denitrifying or sulfate-reducing conditions consumes nitrate or sulfate and produces inorganic carbon and alkalinity. Estimating the supply rates of sulfate or nitrate and correlating them with changes in inorganic carbon concentration and alkalinity can provide evidence for these anaerobic biodegradation reactions. Reductive dechlorination of solvents such as trichloroethene (TCE) and trichloroethane (1,1,1-TCA) releases the chloride ion (Cl−) and strong acid, while it consumes an electron donor. Thus, the release of Cl− can be correlated with the supply rate of an electron donor, such as H2 or an H2 precursor, and a decrease in alkalinity. In many cases, only a small fraction of the electron donor is used to reduce TCE or 1,1,1-TCA. In these cases, consumption of the donor can be a large, easily measured rate, even if Cl− production and an alkalinity decrease are not easy to detect. Precipitation of uranium as UO2(s) due to the reduction of the mobile uranium species UO22+ requires consumption of an electron donor and produces strong acid. Therefore, loss of UO22+ from solution should be accompanied by corresponding losses of an electron donor and a decrease in alkalinity. example is the dissolution of calcareous minerals, which adds alkalinity and inorganic carbon to water and therefore can mask the footprints of biodegradation reactions that change the alkalinity or inorganic carbon concentration. Another confounding factor is transfer of contaminants or footprint chemicals to or from another phase, such as exchange of CO2 or O2 with soil gas. Sampling errors also can confound efforts to document footprints. Because of the possibility of confounding factors, a weight-of-evidence approach, measuring several footprints, generally must be used to document natural attenuation. Even though one type of evidence may be compromised, having several different types can lead to the conclusion that attenuation mechanisms are (or are not) acting based on a weight of

OCR for page 150
Natural Attenuation for Groundwater Remediation TABLE 4-1 Summary of Natural Attenuation Footprints Evaluated in Case Studies Case Study Contaminant(s) Contaminants Controlled? Footprints Traverse City BTEX Yes Depletion of O2; formation of CH4 and Fe2+ Vandenberg Air Force Base MTBE No Insignificant O2 and SO42− concentrations; extension of MTBE plume far beyond BTEX plume Borden Air Force Base Five chlorinated solvents Partially Detection of metabolites of solvent degradation St. Joseph TCE Partially Formation of CH4; detection of degradation by-products (vinyl chloride and ethene) Edwards Air Force Base TCE No Documentation of high NO3− and SO42− concentrations; demonstration that TCE moves with water Dover Air Force Base TCE, 1,1,1-TCA Yes Formation of degradation by-products (cis-1,2-DCE, 1,1-DCA, vinyl chloride, and ethene); CH4 and H2S formation; increase in Cl− concentration Hudson River PCBs Partially Detection of breakdown products; detection of unique transient metabolites; observation of microbial metabolic adaptation and expressed biodegradation genes South Glens Falls PAHs Yes Depletion of O2; detection of unique metabolic by-product; detection of genes for degrading PAHs in site microorganisms; rapid PAH degradation in soils taken from site Pinal Creek Basin Metals, acid Yes now; may not be sustainable Observation of carbonate dissolution leading to pH increase coincident with metal precipitation; observation of manganese oxide precipitates in stream sediments Hanford 216-B-5 Radionuclides Yes Sorbed radionuclides observed in site samples Anonymous Field Site (Borden et al., 1995) BTEX Yes Loss of O2, NO3−, and SO42−; formation of Fe2+ and CH4; increase in inorganic carbon concentration; increase in alkalinity Bemidji Petroleum hydrocarbons Partially Loss of O2; formation of Fe2+, Mn2+, and CH4; formation of intermediate metabolites; observation of selective degradation of petroleum hydrocarbons relative to more stable chemicals NOTE: BTEX = benzene, toluene, ethylbenzene, and xylene; DCA = dichloroethane; DCE = dichloroethene; MTBE = methyl tert-butyl ether; PAHs = polycylic aromatic hydrocarbons; PCBs = polychlorinated biphenyls; TCE = trichloroethene; TCA = trichloroethane.

OCR for page 150
Natural Attenuation for Groundwater Remediation evidence. The greater the degree of uncertainty at a site, the greater will be the need for more and different types of information. Using footprints to document cause-and-effect linkages in natural attenuation requires three steps. The first step is to create a conceptual model of the site. The conceptual model should include a description of the groundwater flow system, estimated locations of the contaminant source and plume, and a list of reactions that might contribute to natural attenuation. The second step is to analyze site measurements to quantify the attenuation processes. This analysis may take a variety of forms, including identification of trends in concentrations of the contaminants and footprint chemicals, a simple mass budget that attempts to correlate changes in the contaminant mass with changes in footprint materials, or comprehensive computer-based models that use mass-balance equations to track contaminants and footprints. The final step is to establish a long-term monitoring program to document that natural attenuation continues to perform as expected. Data collected during long-term monitoring should indicate whether or not the plume is behaving in a manner consistent with the conceptual and quantitative models of the site. The remainder of this chapter describes in more detail how to carry out each of these three steps. CREATING A CONCEPTUAL MODEL The first step in understanding natural attenuation processes at a site involves creating a conceptual model. A conceptual model is an idealized picture of the important features of the flow and transport processes operating at a site. Although the model depicts all of the important features of the system, initially it must be based on simplifying assumptions because data for a more detailed model generally are unavailable in the early stages of site investigation. Because of the necessity to make assumptions, development of the conceptual model must be an iterative process. In the early stages, the conceptual model can be expressed simply in the form of a block diagram or a picture showing a cross section of the site. Initially, information commonly available about a particular site may include existing large-scale maps, reports conducted in early characterization studies, or expert knowledge. Developing a preliminary model based on such existing information can save costs by helping to identify an optimal plan for gathering more data. As understanding of the site increases, preliminary calculations often help to identify the dominant attenuation processes. In some cases, preliminary information can be entered in a computer model that simulates the behavior of the site, even with only “best guesses” for missing param-

OCR for page 150
Natural Attenuation for Groundwater Remediation eter values or with a very simplified model. As data are collected, the conceptual model has to be updated to provide a more complete and accurate picture. As part of this process, the calculations used to create the model should be updated and made more sophisticated. This iterative approach makes full use of all available knowledge at each stage of the process. The purposes are to optimize resources and to systematically document and increase understanding of the system. Benefits include the best possible planning for sampling programs and analyses needed to decide whether natural attenuation is effective at the site. However, numerical answers at each stage of the process should be scrutinized carefully and not be overvalued. Characterizing the Groundwater Flow System The foundation of a site conceptual model always is the site’s hydrogeology. In which direction does the groundwater flow? What is its velocity? Is the flow steady or unsteady over time? Is it homogeneous in space or highly varied by location in the subsurface? Contaminants in the subsurface move with the groundwater. Necessary reactants, such as electron acceptors for bioremediation, are transported with the water. Knowing where and how groundwater flows is therefore essential for tracking contaminants and their footprints. In addition, an observation that contaminants are not moving at the rate expected based on groundwater velocity alone provides a first line of evidence that natural attenuation reactions may be controlling the contamination. Characterizing a site’s hydrogeology involves determining the following: the geometry of the hydrogeologic units and their hydraulic properties; hydraulic heads (essentially, groundwater elevations at different points in the subsurface); and the locations and types of hydrologic boundaries, including the locations and flow rates of the most important sources and sinks for groundwater. The distribution of hydrogeologic units is a key aspect controlling the migration of contaminant plumes. Data from surface topography and vegetation, bore hole cuttings, geophysical surveys, regional geologic studies, and concentrations of different chemicals in the groundwater can be used to create an initial three-dimensional concept of the hydrogeologic units. The properties of these units can be estimated initially from their

OCR for page 150
Natural Attenuation for Groundwater Remediation lithology (the types of geologic materials that make up the particular aquifer) and then refined using results from hydrogeologic tests. Measurements of hydraulic heads in all available wells should then be used to create maps in cross section and plan view showing the groundwater elevations at the site. Hydrologic boundaries to be shown in the conceptual model include surface water bodies, flow divides, recharge wells, pumping wells, and evaporation. Temporal and Spatial Variability in the Flow System Experience shows that the conceptual model for site hydrogeology must account for temporal and spatial variabilities. Frequently, transient flow conditions occur due to natural phenomena, such as seasons and extreme weather events, and to anthropogenic phenomena, such as pumping or irrigation. These transients mean that water levels measured on one day do not necessarily represent other days or the long-term average (King and Barker, 1996). Another common confounding factor is spatial variability in aquifer properties. Homogeneous systems occur only in the laboratory or in models. Heterogeneity in aquifer properties is more pronounced at some sites than others, but at every site it limits the ability to document contaminant fate in the subsurface. Identifying and incorporating temporal and spatial complexities are difficult tasks that require significant amounts of information. When the flow direction shifts, the center of the contaminant plume also shifts. Contaminant concentrations for locations normally near the center of the plume may decrease temporarily, only to rise when the flow direction changes again. In addition, a plume may appear to shrink. For example, during drought years the water table may fall below the level of entrapped residual nonaqueous-phase liquid (NAPL) contaminants in the soil, temporarily removing the source of contamination. In subsequent years with higher rainfall, the concentrations in the plume will rise again as the water table comes in contact with the NAPL. This tendency of plumes to shrink and grow in response to hydrologic variations has implications for natural attenuation investigations. To avoid being misled by transient temporal effects, contaminant losses (and other evidence, as well) must be documented over an area that encompasses the longitudinal axis and fringes of the plume over several years. Special attention should be given to potential contaminant migration pathways presenting the greatest risk. These pathways can be identified by careful site characterization. If significant uncertainty remains regarding the location of such pathways, the inescapable conclusion is that the efficacy of natural attenuation cannot be assessed with confidence.

OCR for page 150
Natural Attenuation for Groundwater Remediation Wide variations in aquifer permeability also complicate the movement of the plume. The most common heterogeneities are discontinuous distributions of sand, gravel, and clay found in aquifers consisting of alluvial or glacial outwash sediments. Alluvial and glacial outwash aquifers are common near the ground surface. In these types of aquifers, the water and contaminants preferentially move through the most permeable zones. For example, a plume may meander as it migrates preferentially through sands and gravels of a buried river channel. Plumes traveling in networks of rock fractures underground are the most difficult to characterize, and methods for characterization are a topic of active research. Heterogeneities also affect the trapping of NAPLs, creating multiple sources of contamination in zones with entrapped NAPLs. For example, NAPLs may migrate into rock fractures, and contaminants from NAPLs may diffuse into low-permeability zones. In effect, each NAPL source dissolves to form its own plume. Therefore, it is possible that groundwater samples taken from different locations at the site in some cases can come from plumes generated by different NAPL sources. The greatest effort should be spent documenting the behavior of the largest and fastest-moving plume, which should be along the connected path with the highest permeability. Thus, samples of the plume in the gravels and higher-permeability sands are key to projecting the maximum extent to which a contaminant plume will spread. Uncertainty in Modeling the Flow System Although sophisticated equipment and analysis techniques are utilized to characterize the subsurface, uncertainty is inevitable in estimates of contaminant behavior because of temporal and spatial variability. The best approach to accounting for this uncertainty is the formulation of multiple conceptual models, each representing a different hypothesis about how the system behaves. Hydrogeologists refer to the different representations of the site in this set as “realizations.” Working with multiple realizations and maintaining an open mind with respect to site interpretation until the data are sufficient to support one realization over the others is essential in accurately characterizing the site. Rather than deciding on one conceptual model of the site and then trying to “prove” it is right, this iterative approach involves assessing data needs and gathering new data to discriminate among realizations. Decisions regarding natural attenuation should differ depending on which realization most accurately represents the actual configuration of units in the subsurface. Sometimes, it is not possible to establish that only one realization represents the site, and the modeler must proceed with the evaluation of multiple realizations.

OCR for page 150
Natural Attenuation for Groundwater Remediation The act of creating and testing realizations provides clues to possible misunderstandings in the conceptual model. For example, the top of Figure 4-1 shows a hypothetical site having three bore holes that intersect zones of differing hydraulic conductivity. If no other information is available, all four of the realizations shown in Figure 4-1 (and many more that are not shown) are reasonable interpretations of the subsurface. The distributions of hydraulic conductivity in each of the realizations affect predictions of groundwater flow (and subsequent predictions of contaminant transport) in different ways. To resolve which realization is the most accurate, additional information is needed. Usually, more is known about a site than just the locations of high-and low-hydraulic-conductivity zones in a few bore holes, and this information can be used to rule out some of the realizations. Simultaneous assessment of all available data reduces uncertainty because some of the realizations, while accurately representing some categories of field data, will not represent other data. To illustrate this point, if each circle in FIGURE 4-1 Several interpretations of the type of connection between zones of different hydraulic conductivity based solely on knowledge of the occurrence of two types of geologic material. Dark zones represent high hydraulic conductivity relative to the white background.

OCR for page 150
Natural Attenuation for Groundwater Remediation Figure 4-2 represents the suite of possible interpretations, then using all of the information together reduces the suite of possible interpretations and the uncertainty in modeling the groundwater flow (and associated contaminant transport). If the data circles do not overlap, none of the realizations can explain all types of information. Then the project team needs to identify shortcomings in the data or create new realizations. The Site-Specific Constructed Model of the Flow System A powerful tool for evaluating realizations in hydrologic systems is the constructed model, which is a set of mathematical equations designed to represent the site’s hydrogeology. In each realization, a different set of numerical parameters is used for different parts of the equations. The foundation of a constructed model is mass balance, which is simply an accounting system to make sure that the mass of a material (such as a contaminant) being modeled in the flow system is neither lost nor created out of nothing. The mass balance is a formal way to set up a budget on a material, and it consists of equations and a means to solve these equations. FIGURE 4-2 Use of multiple types of data to reduce the number of possible interpretations of a contaminated site. Each circle represents possible interpretations of the specific data set. Model realizations that can represent reasonable interpretations of all data sets are retained for further analysis.

OCR for page 150
Natural Attenuation for Groundwater Remediation To set up a mass balance, the modeler defines a domain, which is a volume of the subsurface with specified boundaries. A model domain can be very large (e.g., kilometers on a side) or very small (e.g., meters on a side), depending on the goals of the model. Based on the way the model is to be used, the domain boundaries might be defined by the property lines of the site, the physical extent of the hydrologic system, or an area encompassing a plume. Once the domain is defined, the mass balance states that the mass of the material being modeled changes inside the domain in response to inputs or outputs crossing the boundary and in response to processes that produce, store, or consume material inside the volume. The different realizations of the site are simulated through changes in the model parameters. For example, a very porous zone has a very large value of hydraulic conductivity, while a nonconducting area has a very small value. Likewise, the presence of sources and sinks of water may be represented differently in different realizations. Each model realization is complete when its optimal parameter values (and their associated statistical confidence intervals) are determined. Optimal parameter values generally are estimated by calibrating each of the realizations. The calibration process involves forward modeling: that is, substituting estimated parameter values (for example, hydraulic conductivities, heads at boundaries, and recharge rates) in the constructed model and calculating the simulated values (for example, heads, flow rates, and travel times). The simulated values are then subtracted from values observed in the field. The differences, or errors, are called the residuals, and the weighted sum-of-squared residuals is calculated. These weights reflect the certainty associated with each observation. Often, the weights are the inverse of the variance of the measurement that established the value of the observation. Realizations that give reasonable values for the parameters (e.g., conductivity) and have a low value for the weighted sum-of-squared residuals are retained for further consideration, while the others are eliminated. The modeler also should evaluate the residuals to ensure that they have a mean near zero and are not biased with respect to space, time, or simulated value. The time-consuming nature of the trial-and-error approach limits the number of alternative model realizations that can be considered. Automated techniques now are available for optimizing parameter values. Delineating the Contaminant Source After characterizing the groundwater flow system, identifying the sources of contamination is the next critical step in creating a conceptual model. As described in Chapter 3, the source is the subsurface volume

OCR for page 150
Natural Attenuation for Groundwater Remediation Anderman et al. (1996); Barlebo et al. (1996); Christensen (1997); Christiansen et al. (1995); D’Agnese et al. (1996a,b, 1998); Cooley (1979, 1983a,b); Cooley et al. (1986); Gailey et al. (1991); Giacinto (1994); Kueper (1994); McKenna and Poeter (1995); Olsthoorn (1995); Tiedeman and Gorelick (1993); and Yager (1993). Estimating the Sustainability of Natural Attenuation When analyzing data from a natural attenuation site, a key question often is whether the mechanisms that destroy or immobilize contaminants are sustainable for as long as the source area releases them to the groundwater. More specifically, whether the rates of the protecting mechanisms will continue to equal the rate at which the contaminants enter the groundwater may be a concern. Sustainability is affected by the rate at which the contaminants are transferred from the source area and whether or not the protecting mechanisms are renewable. Unfortunately, most evaluations of natural attenuation to date have not analyzed the sustainability of the reactions. Sustainability is of the greatest concern when the contaminant release rate is high. This may occur when the source area is large, the contaminant concentration in the source is high, and/or the contaminant transfers readily to the groundwater. A large hydrocarbon NAPL and tailings pond are examples. The presence of a major source often can be detected by high contaminant concentrations in the groundwater near the source or in the center of the plume. Even though the source often cannot be located and quantified precisely, mass-budget analyses such as those illustrated in Boxes 4-5 and 4-6 offer a means to estimate the rate at which a contaminant is released to the groundwater. The mass-budget analysis (or solute transport model in more complex settings) also can be used to estimate the long-term rates of the destruction and immobilization reactions based on characteristics of the groundwater, the mineralogy, and the hydrogeology. Some protecting mechanisms are continuous and renewable, but others are not. For example, the long-term supply rates of electron acceptors in the upgradient groundwater or from the soil gas often are predictable and reasonably steady. On the other hand, supply rates of electron donors for reductive reactions normally depend on the long-term existence of a hydrocarbon NAPL or a landfill. Electron-donor supply rates might be predictable and stable if the donor source is identified and long-lasting, but they would decline significantly if the donor source were removed or depleted. Natural attenuation mechanisms that rely on soil minerals to provide sorption sites, electron acceptors, or alkalinity have a finite capacity and

OCR for page 150
Natural Attenuation for Groundwater Remediation BOX 4-11 Use of Inverse Modeling to Select the Most Representative Model At the U.S. Department of Energy’s Kansas City Plant, researchers used inverse modeling to evaluate alternative conceptual models (Anderman et al., 1998). Characterization activities carried out by multiple consulting firms had resulted in a myriad of inconsistent conceptual models of the site. Fifteen equally plausible models were evaluated, representing not only the different views of the firms, but also different levels of model complexity (i.e., number of parameters included in the models to represent the system). A two-layer, steady-state MODFLOW (McDonald and Harbaugh, 1988) model was used to represent the major hydrogeologic units of the alluvial aquifer system in all of the alternative conceptual models. The upper layer consists of approximately 9 m (30 ft) of clayey silt, and the lower layer consists of less than 3 m (10 ft) of basal gravel. Nonlinear regression (using UCODE; Poeter and Hill, 1998) was used to estimate optimal parameter values for each conceptual model by matching field observations of 239 head and 13 flow measurements. Statistics resulting from the regression were used to discriminate between the conceptual models and determine which model best represented the site. Figure 1 illustrates that increasing complexity (i.e., greater number of estimated parameters as displayed on the left panel) of the conceptual models improved the fit (second panel) to observed data for models 2 through 7. However, additional complexity did not improve the fit. Beginning with conceptual model 11, particles were tracked from the source area through the simulated flow field, and the final particle paths were compared to the observed plume movement (third panel). Prior information on the parameter values from aquifer tests and a flow observation representing flow through the entire system were added for conceptual model 15. Although conceptual model 15 did not match the particle paths as well as some of the other models, it was considered the “best” representation of the site because many of the particles followed the observed plume movement; the estimated parameter values were reasonable; and the total flow through the system matched the observed flow better than other models (fourth panel). are not renewable. If the contaminant source is large, as in the Pinal Creek case study of Chapter 3, the contaminant plume will migrate as the continual release of contaminant exhausts the capacity of the minerals. In summary, estimating the sustainability of natural attenuation requires identifying active attenuation mechanisms, distinguishing nonrenewable mechanisms from renewable mechanisms, and comparing release rates of contaminants to the potential rates of transformation and immobilization. Mass budgeting is an important tool for assessing sustainability. However, with mass budgeting, uncertainties will remain in predictions of sustainability, due to uncertainties inherent in all site assessments. There-

OCR for page 150
Natural Attenuation for Groundwater Remediation FIGURE 1 Comparison of conceptual models 2 through 15 (model 1 is omitted because it did not adequately represent the site). The first panel shows the number of parameters estimated. The second panel represents the fit of the model to the field data (low values indicate a closer fit). The third panel displays the percentage of simulated particles that followed the observed path of the plume in the field. The fourth panel indicates the deviation from the observed flow through the entire site. fore, long-term monitoring will be needed to ensure that natural attenuation is continuing to protect public health and the environment. MONITORING THE SITE The final step in documenting natural attenuation is to establish a long-term monitoring plan. If the results from the conceptual model and data analysis lead to the decision that natural attenuation is protective, then long-term monitoring must provide assurance that the site’s protective processes continue to operate over time. Monitoring within the plume

OCR for page 150
Natural Attenuation for Groundwater Remediation is necessary to ensure that reactions that destroy or sequester the contaminants continue to be active. Monitoring downgradient from the plume is necessary to ensure that contaminants are not migrating beyond the zone in which natural attenuation is supposed to take place. Monitoring frequency, intensity, and duration will vary with the complexity of the site, groundwater flow direction and velocity, and plume transport speed. Simple sites contaminated with low concentrations of BTEX will not require the intensity or duration of monitoring necessary at sites contaminated with high concentrations and recalcitrant contaminants. Regardless of the site conditions, the monitoring will have to continue until it demonstrates that natural attenuation has succeeded in achieving the required cleanup goals or that natural attenuation has failed in achieving cleanup standards and a contingency plan has to be implemented. Sites at which natural attenuation is a formal remedy should have an exit strategy specifying when long-term monitoring of natural attenuation can stop. Although long-term monitoring is an essential part of using natural attenuation as a remediation strategy, protocols for long-term monitoring of natural attenuation sites are lacking. Comprehensive data from long-term monitoring of existing natural attenuation sites also are lacking for most sites. Long-term monitoring results from existing natural attenuation sites have to be carefully studied in light of the goals described in the preceding paragraph. Based on these results, guidelines for long-term monitoring of natural attenuation should be developed. CONCLUSIONS Processes that degrade and/or transform contaminants in the subsurface leave footprints that often can be measured. Analysis of these footprints with models of the subsurface should form the basis for determining whether natural attenuation can control contamination at a site. The basic steps to document natural attenuation are to develop a conceptual model of the site’s hydrogeology and biogeochemical reactions; analyze site measurements to quantify the attenuation processes (looking for changes in contaminants and their footprints); establish a long-term monitoring program to document that natural attenuation continues to perform as expected. Although the basic steps are the same for all sites, the level of effort needed to achieve them varies substantially with the complexity of the site and the likelihood that the contaminant is controlled by a natural attenuation process. More uncertainty about site conditions or processes

OCR for page 150
Natural Attenuation for Groundwater Remediation that can control the contaminants increases the level of effort required. Table 4-5 summarizes how characteristics of the site and the contaminants determine the typical level of effort for data gathering and interpretation. The level of effort depends on two factors: (1) the contaminant and (2) the hydrogeology. In general, a higher level of data gathering and analysis is required when the contaminant is less likely to transform (along the rows of the table) and the hydrogeology is more complex (down the columns). The way in which the effort level increases depends on the site and the contaminant. Effort involves a combination of the amount of information that must be gathered and the sophistication of the data analysis (i.e., as summarized in Table 4-3). Table 4-5 offers some general guidelines for levels of data gathering and analysis for different site conditions. The table provides relative indications of effort levels as follows: A level-1 effort is appropriate when all contaminants are in the category of high likelihood of success (as defined in Table 3-6) and the hydrogeology is simple and well understood. In these cases, data gathering and analysis must be sufficient to document that the flow direction is reasonably constant, that contaminant migration is consistent with the flow direction, and that contaminant concentrations decrease with distance from the source. In most cases, at least one footprint should be detected at levels commensurate with the loss of contaminant. For example, a common type of data analysis is to develop contour plots of the hydraulic head in the wet and dry seasons to assess the consistency of flow direction. Contour plots of the contaminant and footprint concentrations often are used to document the principal direction of contaminant migration and that contaminant loss is tied to an attenuation mechanism. A set of two or three vertically nested wells located in the central portion of the plume in plan view can be used to estimate the vertical rate of migration. Even at these relatively simple sites, the sustainability of the attenuation reactions has to be demonstrated through long-term monitoring of contaminants and footprints. A level-2 effort is necessary when the site’s hydrogeology is not simple, the likelihood of success is not high, or the attenuation mechanisms may not be sustainable. If heterogeneities are important, cross-sectional plots of the subsurface geology are needed to show the important lithologic units and their properties. A conceptual model should show how the heterogeneities affect plume migration, and vertical and horizontal plots of concentration data should demonstrate that plume migration is consistent with the conceptual model. If the likelihood of success is not high or the sustainability is uncertain (for example, due to high concentrations from the source), then the postulated reactions that cause contaminant loss have to be documented for the geochemical con-

OCR for page 150
Natural Attenuation for Groundwater Remediation TABLE 4-5 Summary of Typical Effort Required for Site Characterization and Data Interpretation   Likelihood of Success of Natural Attenuation of the Contaminant of Concerna Site Hydrogeology High (e.g., BTEX, alcohols) Moderate (e.g., monochlorbenzene, Pb) Low (e.g., MTBE, TCE, 99Tc) Simple flow, and uniform geochemistry, and low concentrations 1 2 2 Simple flow, and small-scale physical or chemical heterogeneity, and medium-high concentrations 2 2 3 Strongly transient flow, large-scale physical or chemical heterogeneity, or high concentrations 2 3 3 NOTES: Level of effort refers to the number and frequency of samples taken, parameters analyzed in site samples, and type of data analysis (see text); 1 = low effort; 2 = moderate effort; 3 = high effort. BTEX = benzene, toluene, ethylbenzene, and xylene; MTBE = methyl tert-butyl ether; TCE = trichloroethene. aLikelihood of success refers to judgments in Table 3-6. ditions of the aquifer. For example, data should demonstrate that footprints are present in the aquifer or that long-term sorption is occurring. Generally, a mass-budget analysis is needed to show that the postulated natural attenuation reactions are sufficient to destroy or immobilize all of the contaminant and are sustained over time. In some cases, simple mass transport modeling may be needed to interpret whether or not concentrations are decreasing over distance and time. A level-3 effort is needed when the site is highly heterogeneous, flow is strongly transient, the likelihood of success is moderate to low (according to Table 3-5), and/or the potential for sustainability is not high. Extensive effort also may be needed when the site contains complex contaminant mixtures. Extensive effort involves collecting enough data to construct a flow and reactive transport model of the plume. The number and locations of samples and the types of materials assayed (i.e., the contaminants and footprints) must be commensurate with the scope and complexity of the model. The model should simulate the important mass-loss mechanisms, and it should describe the footprints, as well as the contaminants. Outputs from the model should be evaluated with

OCR for page 150
Natural Attenuation for Groundwater Remediation long-term monitoring, and the model should be improved as new data are obtained. Model simulations should show that the mass-loss mechanisms could be sustained for the lifetime of the contaminant source. Model simulations should include cases in which the flow system and the geochemistry are perturbed from the present ones. Although natural attenuation may be a feasible alternative in many cases, Table 4-5 makes clear that documenting natural attenuation may require a great effort if the site characteristics or the controlling mechanisms are uncertain. RECOMMENDATIONS At every regulated natural attenuation site, the responsible company or agency proposing the remedy should document the probable processes responsible for natural attenuation. Observing the disappearance of the contaminant is important to prove that natural attenuation is working, but it is not sufficient by itself. Responsible parties should use “footprints” of natural attenuation processes to document which mechanisms are responsible for observed decreases in contaminant concentration in the groundwater. Footprints generally are changes in concentrations of reactants or products of the biogeochemical processes that transform or immobilize the contaminants. Footprints are well established for some biodegradation reactions—for example, for many petroleum hydrocarbons and chlorinated solvents. Footprints for other contaminants should be based on known biogeochemical reactions. Observing several different footprints and correlating them with decreases in contaminant concentration is necessary evidence for or against natural attenuation and helps overcome confounding factors. Responsible parties should have a conceptual model of the site’s hydrogeology and reactions to show where groundwater and contaminants are moving. The conceptual model includes the groundwater flow, the contaminant source, the plume, and the reactions and chemical species relating to natural attenuation at the site. A good conceptual model guides site investigation and decision making. Responsible parties should gather field data in order to evaluate the validity of the conceptual model and quantify the natural attenuation processes. At the beginning of the site investigation, multiple conceptual models will have to be created. Field data should be used to rule out the models that do not adequately represent the site. Field data also should be used to refine the conceptual model that is ultimately chosen as the best site representation.

OCR for page 150
Natural Attenuation for Groundwater Remediation Responsible parties should analyze the field data at a level commensurate with the complexity of the site and the contaminant type. At the most basic level, graphing and statistical analyses are helpful for formulating hypotheses about trends and possible reactions; they may be adequate for documenting cause and effect for sites with simple hydrogeology and when the reactions affecting the contaminant are well understood. At the next level, mass budgeting is a powerful tool for demonstrating whether or not the footprints of the reactions are commensurate with observed contaminant losses; it is valuable for handling sites with moderate levels of uncertainty. Finally, a solute transport model may be needed when uncertainty is high due to site complexity or poorly understood reactions; these models range in complexity from analytical models to comprehensive numerical models that account for variations in aquifer properties, groundwater flow rates, and contaminant reactions (see NRC, 1990, for details). Responsible parties should repeatedly improve the conceptual model and data analysis for their site. The conceptual model represents an evolving understanding of the site. As new data are collected and analyzed, the conceptual model should be refined. As the conceptual model is refined, new data may be needed. Having a new conceptual model and/or new data often requires that analyses be revisited or modified. Responsible parties should provide a higher level of effort to document natural attenuation for sites at which the uncertainty is greater due to site or contaminant characteristics. Table 4-5 summarizes the conditions that lead to increasing level of effort for site characterization and data analysis. When modeling studies are presented as part of a site assessment, the responsible party should present adequate documentation so that the regulator can assess the quality of the model simulations. This documentation should show whether the model accurately represents the processes and is consistent with the data and conceptual model of the site. The uncertainty of the results should be quantified. Other model quality assurance issues are discussed in Groundwater Models: Scientific and Regulatory Applications (NRC, 1990). A long-term monitoring plan should be specified for every site where natural attenuation is approved as a formal remedy for contamination. Monitoring should take place for as long as natural attenuation is necessary to protect public health and the environment. The required monitoring frequency will have to vary substantially depending on site conditions and the degree of confidence in the sustainability of natural attenuation. Simple sites contaminated with low concentrations of BTEX will not require the same degree of monitoring as complex sites with

OCR for page 150
Natural Attenuation for Groundwater Remediation higher contaminant concentrations and more recalcitrant types of contaminants. Guidelines on long-term monitoring of natural attenuation sites are lacking, and such guidelines have to be developed for different type of sites. REFERENCES Anderman, E. R., M. C. Hill, and E. P. Poeter. 1996. Two-dimensional advective transport in groundwater flow parameter estimation. Ground Water 34(6):1001-1009. Anderman, E. R., A. D. Laase, J. O. Rumbaugh, and J. L. Baker. 1998. The Use of Inverse Modeling to Incorporate Model Uncertainty in Evaluation of Alternative Remedial Actions, Poster Session of the MODFLOW’98 Conference, International Ground Water Modeling Center, Colorado School of Mines, Golden, Colo. Baedecker, M. J., I. M. Cozzarelli, D. I. Siegel, P. C. Bennett, and R. P. Eganhouse. 1993. Crude oil in a shallow sand and gravel aquifer. III. Biogeochemical reactions and mass balance modeling in anoxic ground water. Applied Geochemistry 8:569-586. Baedecker M. J., I. M. Cozzarelli, P. C. Bennett, R. P. Eganhouse, and M. F. Hult. 1996. Evolution of the contaminant plume in an aquifer contaminated with crude oil, Bemidji, Minnesota. Pp. 613-620 in U.S. Geological Survey Toxic Substances Hydrology Program. Colorado Springs, Colo.: U.S. Geological Survey. Barlebo, H. C., M. C. Hill, and D. Rosbjerg. 1996. Identification of groundwater parameters at Columbus, Mississippi, using a threedimensional inverse flow and transport model. Pp. 189-198 in van der Heidje, P. and K. Kovar (eds.) Calibration and Reliability in Groundwater Modeling. Proceedings of the 1996 ModelCARE Conference, Golden, Colo. September. International Association of Hydrologic Sciences, Publ. 237. Bennett, P. C., D. I. Siegel, M. J. Baedecker, and M. F. Hult. 1993. Crude oil in a shallow sand and gravel aquifer. I. Hydrogeology and inorganic geochemistry. Applied Geochemistry 8:529-549. Borden, R. C., C. A. Gomez, and M. T. Becker. 1995. Geochemical indicators of intrinsic bioremediation. Ground Water 33(2):180-189. Cherry, J. A. 1996. Conceptual models for chlorinated solvent plumes and their relevance to intrinsic remediation. In Proc. of the Symposium on Natural Attenuation of Chlorinated Organics in Ground Water. Sept. 11-13. EPA/540/R-96/509. Dallas, Tex.: U.S. Environmental Protection Agency. Christiansen, H., M. C. Hill, D. Rosbjerg, and K. H. Jensen. 1995. Three-dimensional inverse modeling using heads and concentrations at a Danish landfill. In Wagner, B. J., and T. Illangesekare (eds.) Proceedings of Models for Assessing and Monitoring Groundwater Quality, IAHS-INGG XXI General Assembly, Boulder, Colo. Christensen, S. 1997. On the strategy of estimating regional-scale transmissivity fields. Ground Water 35(1):131-139. Clark, C., and B. A. Berven. 1984. Results of the Groundwater Monitoring Program Performed at the Former St. Louis Airport Storage Site for the Period of January 1981 Through January 1983. ORNL/TM-8879. Springfield, Va.: National Technical Information Service. Conant, B. 1998. Chlorinated hydrocarbon plumes discharging to streams: The role of the streambed and near stream flow. Eos Transactions 79:S102. Cooley, R. L. 1979. A method of estimating parameters and assessing reliability for models of steady state groundwater flow. 2. Application of statistical analysis. Water Resources Research 15(3):603-617.

OCR for page 150
Natural Attenuation for Groundwater Remediation Cooley, R. L. 1983a. Incorporation of prior information on parameters into nonlinear regression groundwater flow models. 2. Applications. Water Resources Research 19(3):662-676. Cooley, R.L. 1983b. Some new procedures for numerical solution of variably saturated flow problems. Water Resources Research 19(5):1271-1285. Cooley, R. L., L. F. Konikow, and R. L. Naff. 1986. Nonlinear regression groundwater flow modeling of a deep regional aquifer system. Water Resources Research 22(13): 1759-1778. Cozzarelli, I. M., M. J. Baedecker, R. P. Eganhouse, M. E. Tuccillo, B. A. Bekins, G. R. Aiken, and J. B. Jaeschke. 1999. Long-term geochemical evolution of a crude-oil plume at Bemidji, Minnesota. In Morganwalp, D. W., and H. T. Buxton (eds.) U.S. Geological Survey Toxic Substance Hydrology Program: Proceedings of the Technical Meeting, Charleston, South Carolina, March 8-12. Subsurface Contamination from Point Sources, Vol. 3. U.S. Geological Survey Water-Resources Investigation Report 99-4018C. D’Agnese, F. A., C. C. Faunt, M. C. Hill, and A. K. Turner. 1996a. Calibration of the Death Valley regional groundwater flow model using parameter estimation methods and 3D GIS application. In Kovar, K. et al. (eds.) Proceedings of ModelCARE ‘96 Conference, Golden, Colo. D’Agnese, F. A., C. C. Faunt, M. C. Hill, and A. K. Turner. 1996b. Death Valley regional groundwater flow model calibration using optimal parameter estimation methods and geoscientific information systems. In Kovar, K. and P. van der Heidje (eds.) Calibration and reliability in groundwater modeling, Proceedings of the 1996 Model CARE Conference, Golden, Colo., September. International Association of Hydrologic Sciences Publ. 237. D’Agnese, F. A., C. C. Faunt, A. K. Turner, and M. C. Hill. 1998. Hydrogeologic evaluation and numerical simulation of the Death Valley Regional groundwater flow system, Nevada and California. U.S. Geological Survey Water Resources Investigation Report 964300. Nevada and California: U.S. Geological Survey. Doherty, J. 1994. PEST. Corinda, Australia: Watermark Computing. Eganhouse, R. P., M. J. Baedecker, I. M. Cozzarelli, G. R. Aiken, K. A. Thorn, and T. F. Dorsey. 1993. Crude oil in a shallow sand and gravel aquifer. II. Organic geochemistry. Applied Geochemistry 8:551-567. Ellis, D. 1996. Dupont’s experience with intrinsic remediation of chlorinated solvents, Proceedings of the World Environmental Congress. Essaid, H. I., and B. A. Bekins. 1997. BIOMOC, A multispecies solute-transport model with biodegradation. U.S. Geological Survey Water-Resources Investigations Report 97-4022. Reston, Va.: U.S. Geological Survey. Essaid, H. I., B. A. Bekins, E. M. Godsy, E. Warren, M. J. Baedecker, and I. M. Cozzarelli. 1995. Simulation of aerobic and anaerobic biodegradation processes at a crude-oil spill site. Water Resources Research 31(12):3309-3327. Feenstra, S., and J. A. Cherry. 1996. Diagnosis and assessment of DNAPL sites. Chapter 13 in Cherry, J., and J. Pankow (eds.) Dense Chlorinated Solvents and Other DNAPLs in Groundwater. Guelph, Ontario: Waterloo Press. Feenstra, S., and N. Guiger. 1996. Dissolution of dense non-aqueous phase liquids (DNAPLs) in the subsurface. Chapter 7 in Pankow, J., and J. Cherry (eds.) Dense Chlorinated Solvents and Other DNAPLs in Groundwater. Guelph, Ontario: Waterloo Press. Gailey, R. M., S. M. Gorelick, and A. S. Crowe. 1991. Coupled process parameter estimation and prediction uncertainty using hydraulic head and concentration data. Advances in Water Resources 14(5):301-314.

OCR for page 150
Natural Attenuation for Groundwater Remediation Giacinto, J. F. 1994. An application of MODFLOWP to a superfund case study, in fractured dolomite aquifer near Niagara Falls, New York. USGS WRI Report 92-4189. Reston, Va.: U.S. Geological Survey. Gilbert, R. O. 1987. Statistical Methods for Environmental Pollution Monitoring. New York: Van Nostrand Reinhold Company. Hill, M. C. 1998. Methods and Guidelines for Effective Model. USGS WRI 98-4005. Reston, Va.: U.S. Geological Survey. King, M., and J. Barker. 1996. Methods to quantify source input in field scale studies. In Pp. A-134 In Abstracts with Programs. Vol. 28. GSA Annual Meeting. Kueper, L. K. 1994. Nonlinear-regression flow model of the Gulf Coast aquifer systems in the south-central United States. U.S. Geological Survey Water-Resources Investigations Report 93-4020. Reston, Va.: U.S. Geological Survey. Lorah, M. M., L. D. Olsen, B. L. Smith, M. A. Johnson, and W. B. Fleck. 1997. Natural attenuation of chlorinated volatile organic compounds in a freshwater tidal wetland. USGS Water Resources Investigations Report 97-4171. Reston, Va.: U.S. Geological Survey. McBean, E. A., and F. A. Rovers. 1998. Statistical Procedures for Analysis of Environmental Monitoring Data and Risk Assessment. Upper Saddle River, N.J.: Prentice Hall. McDonald, M. G., and A. W. Harbaugh. 1988. A Modular Three-Dimensional Finite Difference Ground-Water Flow Model. Reston, Va.: U.S. Geological Survey. McKenna, S. A., and E. P. Poeter. 1995. Field example of data fusion for site characterization. Water Resources Research 31:12. NRC (National Research Council). 1990. Ground Water Models: Scientific and Regulatory Applications. Washington, D.C.: National Academy Press. NRC. 1993. In Situ Bioremediation: When Does It Work? Washington, D.C.: National Academy Press. Olsthoorn, T. N. 1995. Effective parameter optimization for groundwater model calibration. Ground Water 33(1):4248. Poeter, E. P., and M. C. Hill. 1998. Documentation of UCODE, a computer code for universal inverse modeling. U.S. Geological Survey Water Resources Investigations Report 98-4080. Reston, Va.: U.S. Geological Survey. Rittmann, B. E., E. Seagren, B. A. Wrenn, A. J. Valocchi, C. Ray, and L. Raskin. 1994. In Situ Bioremediation. 2nd Ed. Park Ridge, N.J.: Noyes Publications. Tiedeman, C., and S. M. Gorelick. 1993. Analysis of uncertainty in optimal groundwater transient, three-dimensional groundwater flow model using nonlinear regression. USGS OFR 91-484. Reston, Va.: U.S. Geological Survey. Wiedemeier, T. H., M. A. Swanson, D. E. Mootoox, E. K. Gordon, J. T. Wilson, B. H. Wilson, D. H. Kampbell, J. E. Hansen, P. Haas, and F. H. Chapelle. 1997. Technical Protocol for Evaluating Natural Attenuation of Chlorinated Solvents in Groundwater. San Antonio, Tex.: Air Force Center for Environmental Excellence, Brooks Air Force Base. Yager, R. M. 1993. Simulated three-dimensional groundwater flow in the Lockport Group, a steady-state conditions. Water Resources Research 22(2):199-242.