In Situ Enhanced Source Removal

5.0 Performance Comparison

5.1 COMPARISON OF TECHNOLOGIES

All of the technologies evaluated in this study attempted to remediate a NAPL-contaminated formation by injecting remedial fluids at one or more locations and extracting these fluids and associated contaminants at another location(s).  The primary difference between the technologies is in the amount and composition of the remedial fluid and the resultant extraction mechanism.  Secondary differences are in the geometry of the remedial design.  The table below provides a summary of the remedial fluid composition, the injected concentrations of the remedial adjuvants,  the design flux in the test system, and the number of pore volumes of remedial fluid injected.  The amount injected does not include flow at the beginning of the study when the flow field was being established, or flows at the end of the study when the remedial fluid was being eluted from the system

Table 5.1.1 Abbreviated Summary of the Treatment Systems
Cell Remedial Fluid Injected flux (l/m) Target volume 
(m3)***		 Injected Volume 
(1000 l)		 Approximate pore volumes *
Cosolvent Solubilization 70% ethanol 
0-10% pentanol 
20% water ‡		 4 23 * 40 * 9
Cosolvent Mobilization 81-96% tert butanol 
0-15% hexanol
4% water		 2.6/4 50.2 27 3
Surfactant Solubilization 5% DowfaxTM
95% water		 3.0 50.5 52 5
Surfactant Mobilization 0.4% AOTTM 100 
3% TweenTM 80 
96.6% water		 1.8 67.3 27 2
Micro-emulsion 0-2.5% pentanol 
2.5% BrijTM
95-97.5% water		 1.8/3.6† 55.9 54 9
Macromolecular 10% hydroxypropyl-cyclodextrin 
90% water		 4.6 42.3 64 8
Steam Steam ** 1.7 69.8 10 1
Sparging/Venting air   66.8    
In-well aeration air 1170† 40.1    
* Pore volume is assumed equal to 20% of total volume
** Referenced to the amount of water at standard temperature and pressure injected as steam
***Total formation volume treated
† Air at standard temperature and pressure

5.2 METHODS OF EVALUATION

5.2.1 Test Cell Description

Each demonstration in this study was performed in a test cell which was constructed by physically and hydraulically isolating a portion of the contaminated aquifer.  Each test cell was equipped with a minimum subset of monitoring points and injection and extraction wells.  Test cell isolation was achieved by driving interlocking, sealable joint sheet piles around the perimeter of the cell, and grouting the joints with a portland cement based grout. The cells were then instrumented with seven 2" wells screened across the contaminated zone  (four injection and three extraction).  Twelve multilevel samplers were installed (three transects of four).  The layout for the sparging/venting cell shown in Figure 1 is typical for all of the cells plan view  sectional view .  Included in Figure 1 is a vertical section showing a typical layout of the multilevel samplers.  These samplers were used to collect water samples throughout the study, yielding a three-dimensional assessment of the treatment process.  The multilevel sampler data is included in the database, but the discussion here will be limited to an analysis of the data collected at the extraction wells.

Three types of performance data were collected to assess the quantity of NAPL or NAPL constituents removed during each technology demonstration.  These were: 1) pre- and post-demonstration soil concentrations of selected target contaminants as measured by core sampling and analysis, 2) pre- and post-demonstration NAPL masses estimated from retardation of partitioning tracers, and 3) masses of selected target contaminants eluted from the test cells as determined by direct monitoring of waste stream concentrations.  Because of the very complex nature of the NAPL at the study site, total NAPL content could not be ascertained by direct analytical methods.  This posed substantial challenges with respect to the design and assessment of the remedial technologies.  The following sections assess and compare the effectiveness of the nine remedial approaches included in this study with respect to each type of performance data.

5.2.2 Core Analysis

Soil cores were collected with a hollow-stem auger both prior to and after remediation to evaluate the performance of the technology.  Cores collected prior to remediation were obtained at the time wells or multilevel samplers were installed.  Generally, cores were collected from the injection and extraction well locations and the centerline of multilevel samplers.  After remediation, cores were collected along two transects between the rows of multilevel samplers. Often, core recovery was poor due to large cobbles in the formation.  Ocassionally, little or no recovery was achieved.  The cores were sub-sampled (fraction less than 10 mm in diameter) and extracted with methylene chloride in the field and shipped to the laboratory for GC/MS analysis.  No attempt was made to correct for the cobbles that could not be included in extracted subsamples.  Field extraction of the core samples was utilized to minimize volatile losses of the chemicals during sample handling and storage. Except for the first study (cosolvent solubilization), all cores were analyzed by GC/MS at Michigan Technological University to standardize the methods and improve comparability between the studies.  Cosolvent solubilization sample analyses were split between the University of Florida and the U.S. EPA's Subsurface Protection and Remediation Devision in Ada, Oklahoma.  The target analytes for the cosolvent solubilization are slightly different than the remainder of the studies.  Samples that were below the detection or quantification limits are included as 0.5 times the quantification limit.

Assessment of the effectiveness of a remedial technology requires quantitation of the mass or volume of contaminant removed as a result of the application of the technology.  Core data obtained as described above was used to estimate the total mass of contaminants in the targeted treatment volume both before and after remediation.  Differences in these estimates were assumed to be the mass of contaminant removed as a result of the remedial activity.  Because no proven or universally accepted methods are available, five different approaches were used for estimating total contaminant masses in the treatment volume using the discrete samples collected during coring.  These approaches are described in Section 4.1.1 of this report.  Although trends obtained from these approaches were generally similar, absolute values were often substantially different.  In addition, only the geostatistical method gave a defensible measure of the uncertainty of the estimated values.

A summary of the performances of the enhanced source technologies included in this study, as determined from core data, is presented in the following table.   The boxcar averaging method was used in compiling these results.  This method was selected because it removes some of the bias associated with variable sampling density within the volume of interest and permits the incorporation of stratigraphic information.

Table 5.2.2.1 Technology Performance (Fraction Removed) Determined from Core Data (Boxcar Averaged Data)
Chemical Co-sol. solub. Co-sol.
mob.		 Surf. solub. Surf. mob. Micro emul. Macro-
mol.		 Sparging
Venting		 In-well aeration Steam
dichlorobenzene 0.96 0.24†   0.93 0.97 0.79 0   0.67
1,1,1-trichlorethane 0.99+ 0.96 0.37 0.69 -0.2 * 0.76 0.13 * 0.46 * 0.98
toluene 0.91 0.94 0.68 0.98 0.4 * 0.71 0.57 * 0.53 0.92
ortho-xylene   0.94 0.59 0.97 0.92 0.69 0.24 0.48 0.79
meta-xylene   0.92     0.89 0.72 0.33   0.84
naphthalene 0.89 0.89 0.60 0.95 0.89 0.79 0.50 0.63 0.71
trimethylbenzene   0.94 0.54 0.96 0.93 0.49 0.38 0.18 0.78
decane 0.97 0.80 0 0.93 0.95 0.26 0.42 0.24 0.72
undecane   0.84 0 0.87 0.98 0.32 0.49 0.16 0.73
ethylbenzene   0.93 0.58 0.97 0.86 * 0.77 0.36    
Mean **  0.94  0.84-0.91  0.42  0.92  0.93  0.63 0.27   0.37  0.79
* Insufficient concentration initially in formation for reliable estimates (mean concentrations < 0.1 mg/kg).
** Ignores data where initial concentration was considered insufficient to make reliable estimates and where data is not available.
† Value significantly biased by two samples that were higher than any samples measured in the pre-remediation cores. Removing these samples changes removal fraction to ~90%.

To evaluate the performance of a technology, one should also determine the effectiveness of the technology for mass removal as a function of remedial fluid or resources used.  For the data in the above table,  this would mean plotting removal fraction times the initial mass of a contaminant as a function of remedial fluid used.  Unfortunately, information obtained from core data provides only a single point on this type of performance curve.  This makes it difficult to rank the performance of the technologies.  Based on the information in the table, five of the technologies merit further consideration as a viable means of aggressively treating the types of contaminated formations represented by this study site.  These technologies are: cosolvent solubilization, cosolvent mobilization, surfactant mobilization, microemulsification, and steam stripping. The macromolecular approach may also have potential since its initial mass of contamination was significantly greater than the mass in the other test cells.  It must be emphasized that technology performance is often dependent upon site and contaminant characteristics; therefore, performances observed in this study should not be interpreted as blanket endorsements or indictments for the technologies.  These results significantly enhance our understanding of the potential effectiveness, applicability, and implementability of these source remediation approaches.

Other criteria, in addition to the ability to extract contaminant, must be evaluated in order to assess the utility of a technology.  Cost, design and operational complexity, impact on and compatability with site properties and conditions, and public and regulatory acceptability must be taken into consideration when determining the most appropriate remedial approach for a specific site.  For example, surfactant mobilization is a promising technology, but requires careful control of aqueous and geochemical conditions.  Finally, with the variability in hydraulic conductivity and the complexity and distribution of NAPL at this site, one should not expect to reach cleanup goals with two or three pore volumes of flushing.  Remedial objectives, in addition to site conditions and technology effectiveness, will determine the extent of flushing required.
 
 

5.2.3 Treatment Performance Evaluated with Elution Data


Each of the studies monitored the flux of the extracted fluid and the concentration of target constituents in the effluent.  The procedure was to establish a stable flow field at the design flow rate.  Once the flow field was established, the fluid was switched from water to the remedial fluid for a predetermined time (based on the number of pore volumes of remedial fluid that were to be passed through the test cell).  For this study, the number of pore volumes of remedial fluid to be used or the length of operation of a technology was driven primarily by budgetary and time constraints.  Ideally, for technology assessment, one would continue a demonstration until predetermined remedial objectives are met.  After the desired volume of remedial fluid had been injected into the cell, the fluid was switched back to water to elute the remedial fluid from the formation.  An attempt was made to maintain a design flow throughout the studies.  Several power outages occurred and equipment problems created stoppages that are not uncommon with field work.  During the initial flushing, the concentration of the target contaminants was monitored.  The concentration of the constituents eluted from the study cells prior to the injection of the remedial fluid is indicative of how a pump-and-treat system would perform with the same flow geometry and flux.  The fluxes in this study were greater than would occur under a natural hydraulic gradient.  Therefore, these systems are more representative of "water flood" systems than pump-and-treat systems.

The evaluation of technology performance from elution data requires:  1) determination of the elution pattern of contaminants from the treated zone, and 2) estimation of the initial mass of contaminant present in the swept volume.  Elution curves were determined by monitoring the concentrations of the selected NAPL constituents in the extraction well effluent streams as a function of time.  Because of the chemically complex NAPL at this site and the presence of large volumes of chemical adjuvants, or the presence of immiscible phases, contaminant concentrations were often difficult to quantify.  Thus, there is substantial uncertainty associated many of the estimates of contaminant mass contained in extracted fluids.

The initial masses of contaminants in the treatment zone were estimated using the core data discussed in the previous section or tracer results which will be discussed in the following section.  The tracer data provides a measure of bulk NAPL present in the swept volume; therefore, information about the chemical composition of the NAPL is required to estimate individual constituent concentrations or masses from these data.  Because of its complexity, it is impossible to completely characterize the NAPL, and very difficult to make reliable projections of the relationship between NAPL content and constituent concentrations.  To avoid these complications, the performance data presented in the following table were obtained by using core data to estimate initial contaminant mass.  These data are plotted in the linked elution comparison curve.

Table 5.2.3.1 Fraction Removed by the Technologies Based on Initial Masses Derived from Core Data (Based on  Geostatistical Interpretation)
Chemical Co-sol. solub. Co-sol. mob. Surf solub. Surf. mob. Micro emul. Macro
mol.		 Sparg
Vent		 In-well aeration Steam
dichlorobenzene 1.03 0.34  5.1  0.51    0.97  6.52    
1,1,1-trichloroethane                  
toluene            0.75      
ortho-xylene      0.09 1.86    0.01      
meta-xylene              0.45    
naphthalene    0.49  3.3  3.6  0.88  0.61  0.05    
trimethylbenzene            0.26  0.03    
decane 0.66      0.23  0.37  0.01  0.05    
undecane    0.35  0.05  0.12  0.66  0.01  0.1    
ethylbenzene            0.66      

Most of the cells in the table are empty, indicating inadequate elution data or incompatibility with core data.  Frequently, the mass removed is reported to be much more than the initial mass estimated in the test cell.  Other times the mass reported as removed is insignificant.  It is obvious that the data presented is inadequate for making  general conclusions as to the performance of the different technologies.  Potential reasons for this include: sampling errors, inadequate analytical method as a result of interfering compounds, or co-elution.  In addition to errors associated with measuring elution curves, any error in the estimation of initial mass will contribute uncertainty to the evaluation of technology performance.
 

5.2.4 Test Cells Evaluated with Tracer Data


The test cells were designed to have four injection wells and three extraction wells, with twelve multilevel samplers in the flow domain.  This configuration allows a line drive flow field to be established.  By carefully selecting a suite of tracers which contain both nonreactive tracers that behave like water in the flow domain and reactive tracers which partition into the resident NAPL in a known and predictable manner, the swept volume and NAPL mass can be estimated.  The theory is that the retarded tracers will separate chromatographically from the non-retarded tracers and that the separation can be correlated to NAPL content.  Thus, it is possible to estimate the mass of NAPL present as a ratio of the time of travel of the two tracers.  At the present time, tracers provide the only direct method of integrating the mass of NAPL between two points (Annable et al., 1998).   Several assumptions must be made when interpreting the data.  For example, the partition coefficient is known and is not spatially or temporally variable.  In this study, it was possible to obtain small amounts of NAPL from several of the wells.  Partition coefficients were measured both at room temperature and at 10oC.  The partition coefficients were significantly temperature dependent.  The results presented here are based on partition coefficients for 2,2-dimethyl-3-pentanol (dmp) measured at 10oC (7.44).  Studies were performed measuring the effect of the remedial fluid on the partition coefficient but not all of the remedial fluids were tested.  Cores collected after the remedial activities retained a blackened appearance, indicating the presence of residual NAPL.  This occurred even in zones in which core analysis suggested the contaminants had been effectively extracted.  Thus, it is possible that the residual NAPL in these zones was composed predominantly of high molecular weight "pitch."  If this is the case, tracer partitioning coefficients for this material could be substantially different from those measured on the original NAPL.  This would complicate interpretation of tracer data.

Reliable estimates of NAPL saturation from tracer data require good resolution of tracer breakthrough curves.  The tailing portion of these curves often contain crucial information needed to accurately assess NAPL content, but these tails are difficult to resolve because of the very low tracer concentrations.  Breakthrough curve extrapolation is often used to minimize the errors associated with early truncation of these curves.  For the data presented here, tracer concentrations were extrapolated to 1 mg/l by assuming the last 25% of the samples could be fit to an exponential function. The following table gives estimates of NAPL saturations before and after each demonstration and removal fractions computed as the differences of these estimates.
 

Table 5.2.4.1 Fraction Removed by the Technologies Based on Mass Estimated by Moment Analysis of the Extrapolated Tracer Data
Test Cell Initial saturation based on moment analysis of extrapolated data Final saturation based on moment analysis of extrapolated data Fraction removed
Cosolvent solubilization 0.053 0.028 0.47
Cosolvent mobilization 0.084 0.023 0.73
Surfactant solubilization 0.153 0.137 0.10
Surfactant mobilization 0.029 0.031 (0.069)
Micro emulsion 0.093 0.038 0.59
Macro molecular 0.087 0.562 (5.46)
Sparging/venting      
In-well aeration 0.191 0.313 (0.64)
Steam 0.069 0.146 (1.12)

Much of the data in the table above is reasonable.  For instance, the co-solvent solubilization , cosolvent mobilization and micro emulsion data suggest mass removals in the range of 50%.  This compares to a removal fraction for the target chemicals of about 90% based on core data.   For four of the technologies, tracer results indicate increases in NAPL content during the course of the demonstration.  Possible explanations for these apparent increases in NAPL saturations include:  1) inadequate resolution of breakthrough curves,  2)redistribution of NAPL into the treatment zone from adjacent formation, 3) greater accessability of the NAPL to the tracers or to reactive mineral surfaces, 4) changes in partitioning coefficients as a result of changes in NAPL composition, residual remedial fluid, or formation temperature, 5) changes in the swept volume, and 6) partitioning of tracers into biomass formed as a result of increased biological activity during or after the remedial activity, or into residual remedial adjuvants.  Although there are large discrepancies in performances shown here and those based on core data,  there are similarities in the relative ranking of the technologies.  For both performance assessment approaches, Microemulsification, cosolvent mobilization, and cosolvent solubilization are among the top performers while in-well aeration is at the bottom of the list.

5.3 DISCUSSION AND CONCLUSIONS


The evaluation of the performance of a remedial technology in the field is not a trivial activity.  Even with carefully designed studies, it is difficult to accurately determine the initial mass of the contaminants in the environment or even the mass of contaminants removed from a contaminated environment.  It appears that it is necessary to generate several lines of evidence evaluating the performance with different independent methods before confidence can be attained in the suitability of a particular remedial approach.  From the data summarized here, one could conclude that core data does provide a good estimate of the performance if sufficient samples are collected.  Each analysis technique, however, yields different results.  The variation based on analysis methodology is generally within a factor of two.  Of the methods used, it is recommended that a box car average gives a reasonably accurate result and is easy to implement for these small scale studies. Extreme care must be observed in interpreting both the elution curves and tracer data.  It was difficult to make good conclusions when the data is inconsistent.  The tracer data appears sound when it is performed carefully.

Combining the data from the core analysis and tracer analysis one would conclude that co-solvent solubilization, co-solvent mobilization and micro emulsion technologies as implemented perform well at this site.  Any of the three methods should be capable of removing significant amounts of NAPL mass and thereby, reducing the environmental risk. Surfactant mobilization  and possibly steam stripping and macro molecular solubilization appear to be effective when considering only the core data alone. There were apparent problems with the tracer data for these two technologies and thus not as many lines of evidence supporting their applicability.  The core data for the surfactant mobilization looks particularly promising and, thus, surfactant mobilization should not be excluded from consideration.

It needs to be reiterated that the core data and tracer evaluations as collected in this study gives only snapshots of the performance.  Thus, only one point on a performance curve has been generated.  Different implementations of the technologies or different endpoints would yield different results.  Simplified design procedures, which take into consideration key parameters, are still needed for these technologies. The design procedures need to permit evaluating design considerations including: the effect of variable remedial fluid composition, effects of heterogeneity, and precursor reuse to permit optimization remedial design. The attached data set can be used to help in evaluating the validity of design procedures as they are developed.