4.4.0    Surfactant Solubilization (Cell 6)

The University of Oklahoma, in cooperation with the U.S. EPA, operated the surfactant solubilization cell.  The cell layout is shown in Figure 4.4.0.1.  The test cell was approximately 4.3 m by 3.5 m with four injection wells labeled 264x in the figure, three extraction wells labeled 265x and twelve multilevel samplers labeled 261x, 262x and 263x. Small diameter monitoring wells were installed to permit monitoring the head during the studies and are labeled 266x in the figure.  The injection wells were along the south side of the cell and the extraction wells were along the north side of the cell. The well labels match the location information in the database attached in the supporting data section. The elevations indicated on the figure in red are the elevation of the top of the device.  This particular cell had leakage along the west side of the cell.  The general location of leakage was determined by installing ten (10) monitoring wells around the perimeter of the test cell (three on each long side and two on each of the short sides).  A tracer was injected into the cell and all the monitoring wells continuously pumped to create small local zones of depression allowing the identification of the leaking region.  The method of sampling only identified the aereal region of leaking not the vertical extent of the leak so it was not possible to determine if the leak was a result of a bad joint or because there was not a good seal between the sheet pile and the clay aquitard.  Attempts to seal the cell by injecting a bentonite grout in the formation outside the test cell in the area associated with the leak were unsuccessful.  The injection strategy had to be modified to protect the environment and it was decided to use a hydraulic barrier and inject clean ground water in well 2644 next to the side of the cell that was leaking and remedial fluid in the remaining three injection wells.  The extraction wells were pumped normally.  The result should have been that the remediation would take place only in the eastern three-fourths of the cell.

Essentially seven studies were performed. Initially, the formation was sampled by coring the formation during the installation of wells and multilevel samplers. Descriptive logs of the wells cored are available in the database in the table labeled BoringLogs. The core samples were analyzed for a group of chemicals initially believed to be representative of the site and of regulatory concern.  The selected chemicals represented several classes of contaminants. Following the installation of the instrumentation, the formation was allowed to reequilibrate before any additional measurements were made.  Ground-water samples were collected under both static and dynamic conditions.  Injecting uncontaminated ground-water into the formation simulating a forced gradient flush or water flood created the dynamic flow conditions.  The static samples would be representative of the maximum concentration that would be anticipated in water samples from this location.  Ground-water sampling was followed by a pre-remediation tracer study.  The pre-remediation tracer study was designed to estimate the mass of NAPL in the formation.  Once the pre-remediation tracer study was completed, the remedial technology was implemented and its performance monitored.  At the termination of the remediation demonstration the three characterization steps were repeated in reverse order. Post-remediation core locations are shown in Figure 4.4.0.1  with the identification x69x.  A time line for the activities is shown in Table 4.4.0.2. An analysis of this data permits an evaluation of the technology.

The objective of the study was to solubilize the contaminants simultaneously avoiding the creation of macromolecules that could dramatically reduce the effective permeability of the formation.  From a theoretical standpoint, the number of pore volumes required to achieve an acceptable end point is dependent on the compound in question.  It should take fewer pore volumes to extract the more soluble compounds and more pore volumes to extract the less soluble compounds.  The actual number of pore volumes would depend on the micelle:water partition coefficient and the cosolvency effects of the surfactant molecule up to the critical micelle concentration (CMC).  The remedial fluid  selected contained 5% by wt Dowfaxtm.  It was injected at three liters per minute split equally into the three injection wells. A total of 52000 l of remedial fluid were injected.  The remaining injection well received clean water.   The extraction wells were pumped at a constant rate to maintain a constant head in the test zone.  During the test, the water table was elevated to permit contact of the remedial fluid to the contaminated aquifer material.  The number of pore volumes selected for injection was based on injecting a sufficient amount to determine how the system might perform under realistic field conditions.  It was not anticipated that cleanup goals would be achieved with the number of pore volumes injected.

 Table 4.4.0.1 Operating conditions for the experiments
Parameter
Activity
Pre-remediation Tracer Demonstration Post-Remediation Tracer
Average Saturated Thickness (m) 2.62 2.8 4.33
Average Head Across Cell (m) 0.22 0.33 0.06
Average Influent Flow rate (lpm) 2.57 2.61 5.68
Average Effluent Flow rate (lpm) 2.69 2.46 6.02

Table 4.4.0.2 Study sequence
Test Activity Fluid injected * Total flux rate (lpm)- volume (l) Duration
Core and install instrumentation Collect soil samples None    1/8/96 - 1/10/96
Ground-water sampling Collect water samples None    
Pre-remediation tracer Establish flow field Water    
  Inject tracer suite 1  br=863, dmp = 2045 methhep = 2640  2.7 lpm
473 (l)		 7/8/96 
  Maintain flow field Water  3 (lpm)  7/8/96 - 7/19/96
Remedial technology Establish flow field  Water  2.5 lpm  
  Inject remedial fluid  5 Wt % dowfax  2.5 lpm  7/23/96 - 8/2/96
  Remove surfactant Water  2.5 lpm  8/2/96 - 8/7/96
Post-remediation tracer Establish flow field Water    
  Inject tracer suite 2  br = 879 dmp = 591 methhep = 414  6 lpm 
348 (l)		  9/8/96
  Maintain flow field Water  6 lpm  9/8/96 - 9/18/96
Ground-water sampling Collect water samples None    
Core  Collect soil samples None    9/25/96 - 9/20/96

* br = bromide, dmp = 2,2-dimethyl-3-pentanol, methep = 6-methyl-2-pentanol

4.4.1    Core Analysis

The figures that describe the core analysis plot the concentration of the individual contaminants as a function of depth without regard for spatial location within the test cell.  Black data were collected prior to remediation and the blue data were the observations collected after remediation.  The vertical bars are the means for the values over the bar's depth zone;  95% confidence limits are plotted, where possible, with the limits represented by the small + at the center of the bar. The two magenta lines depict the vertical extent of the target zone. At the top of the figure, several different methods are reported to calculate the total mass in the target zone of the test zone and the surface area within the cell is printed. Prior to remediation, the contamination in the surfactant solubilization cell was fairly tightly clustered with very little variation in concentration from sample to sample. Figure 4.4.1.1  dichlorobenzene shows that there is a relatively uniform distribution of contamination in the aquifer. After remediation, there is a considerable amount of variability with frequent samples below quantification limits. As a general rule, it appears that there was better removal in the upper portion of the aquifer and possibly some accumulation near the bottom of the aquifer. This wide variability in concentration after remediation compared to pre-remediation suggests the remedial fluid had not perfused the entire cell or that, based on the limited solubility of the remedial fluid, insufficient remedial fluid had perfused the cell. The data have not been plotted as a function of travel path that might be used to evaluate this hypothesis. Since only a few pore volumes of remedial fluid were used, it is likely the performance would be improved significantly if more remedial fluid had been used or if the remedial fluid had been circulated.  Dichlorobenzene is one of the easier chemicals to remove, based on the octanol:water partition coefficient Table 4.1.1.1., and, with solubilization as the remediation approach, should have the best potential for remediation using this approach.

The pre-remediation pattern of  decane and undecane  is quite similar to the pattern of dichlorobenzene; observations of decane and undecane after remediation suggest very little removal.  None of the samples were below quantification limits.  Decane and undecane are the two most difficult chemicals to remove by solubilization based on their octanol:water partition coefficients.  It is not surprising that with only five pore volumes of remedial fluid utilized relatively poor performance was observed for these chemicals.

Ethylbenzene and dichlorobenzene behaved quite similarly.  Again, with ethylbenzene it does not appear that sufficient remedial fluid had been used to achieve a cleanup level comparable to the cleanup level that was achieved with the ethanol-pentanol mixture. If the same number of pore volumes of remedial fluid had been used, it is likely that a similar removal pattern would be observed.

A significantly larger number of samples were observed below the detection limit for o-xylene than the previous chemicals discussed. The o-xylene figure shows approximately 50% of the samples were below quantification limits after remediation and almost no samples were below the quantification limits before remediation.

There is a reasonably consistent pattern between the number of samples that are below the quantification limit after remediation and the water solubility of the compound. The higher the solubility the larger the number of samples that are below quantification limit after remediation. The most soluble chemical tracked, 1,1,1-trichloroethane, was, initially, a rather small fraction of the NAPL mass. The removed fraction during the study was less than would have been expected for this compound. This low removal fraction probably results from subtracting small numbers and the potential error in the initial numbers. . Results for the other chemicals tracked are shown in the figures naphthalene, ortho-xylene, trimethylbenzene,  and toluene.  The results shown are quite similar with very consistent distribution of the contamination prior to remediation and a more scattered result after remediation. the more soluble compounds were more effectively removed.  This is easily seen when one  compares of fraction remaining and Log Kow of the compounds.  It is reasonable to conclude that with more cosolvent the apparent removal effectiveness would have been improved. The results depicted should be considered only as one point on a performance curve.

Figure 4.4.1.1 Summary of Core data based on boxcar averaged results
Chemical Pre-remediation concentration (mg/kg) Post-remediation concentration (mg/kg) Fraction removed
dichlorobenzene  0.88 0.3  0.65 
1,1,1-trichlorethane 0.13 0.081 0.38
toluene 0.16 0.053 0.67
o-xylene 1.6 0.66 0.59
m-xylene      
naphthalene 3 1.2 0.60
trimethylbenzene 3.7 1.7 0.54
decane 34 34 0
undecane 77 91 0
ethylbenzene 0.33 0.14 0.58

4.4.2    Tracer Analysis

The tracer studies can be summarized in the following table.

Figure 4.4.2.1 Summary of tracer studies
Tracer Pre-remediation Post-remediation
Concentration (mg/l) Total Mass Injected (g) Mass recovered 

(g)*

 Concentration (mg/l) Total Mass Injected (g) Mass recovered 

(g)*
Bromide 863 408  425 879 306  650
2,2-dimethyl-3-pentanol 2045 968  1222 591 206  245
Hexanol 2640 1249        
6-methyl-2-heptanol       414 144  
Tracer Volume (l) 473 348
Injection Time (min) 234 360

* Based on zero moment of extrapolated data.

The first step in evaluating the tracer data is to determine if there was conservation of mass.  In the pre-remediation tracer study, 104% of the nonretarded tracer was recovered.  This is within expected analytical limits, suggesting that there is no apparent loss of the chemicals.  The clean water that was injected as a buffer on the west side of the cell was apparently successful in controlling leakage of chemicals from the cell. For the retarded tracer 2,2-dimethyl-3-pentanol (DMP) the mass recovered was about 26% more than what was reported to be injected.  The two analytical techniques are quite different and there appears to be some problem with the QA on the organic analysis since an increase in mass of 26% is more than should be expected due to normal analytical error. Looking at the zero moments in Tables 4.4.2.2 through 4.4.2.4, there are major differences in the flow field. The east side of the test cell (Table 4.4.2.2) received 50% more of the tracer mass than average and the west side received 50% less of the tracer mass. At the same time, the travel time on the east is about 30% slower than the rest of the cell.  If the formation were homogeneous with respect to water-filled porosity, one would anticipate a shorter travel time when more flow is passing through the formation.  One possible explanation for this confusing data is that it might have been possible, during the grouting for the leak, that some of the bentonite entered the test cell and filled some of the voids on the west side of the test cell lowering the porosity.  If there were some way of lowering the porosity without lowering the hydraulic conductivity, one could explain this confusing data. The purpose of the bentonite injection was to lower the hydraulic conductivity, not just fill the voids.  If the grouting created a blind spot in the formation forcing water through a smaller highly permeable portion of the formation, this could possibly explain this observation.  Before any conclusion is reached, the data from the multilevel samplers should be carefully analyzed.  The data is in the database attached but not presented here.  The post-remediation tracer was performed with a significantly different flux and head.  Therefore, the volume swept by the tracer during the post-remediation tracer study is nearly twice the volume swept by the pre-remediation tracer study.  This makes it difficult to compare the pre-remediation data to the post-remediation data. Care should be exercised in interpreting the results. An additional problem with the data is that more than twice as much bromide was recovered as was injected.  This is an unacceptable error in either the data analysis or reporting of some of the input data required in the analysis.  From a qualitative analysis of the breakthrough curves, it appears that there is an error in the reporting of the start time for the injection since two days were required before the tracer was observed in the extraction well.  None of the other studies showed this apparent error in the tracer data.

Table 4.4.2.2. Summary of Extraction Well 1 tracer analysis
Extraction Well 1
Pre-remediation
Post-remediation
  conservative (bromide) 2,2-dimethyl 
-3-pentanol		 NAPL Saturation conservative (bromide) 2,2-dimethyl 
-3-pentanol		 NAPL Saturation
Pulse duration (days)  0.181  0.181    0.250  0.250  
zero moment data  167.4  383.6    114.63  42.27  
zero moment extrapolated  169.8  470.3    115.455  43.64  
first moment / zero moment data (days)  2.197  2.937  0.047  1.268  2.516  0.147
first moment extrapolated / zero moment (days)  2.334  5.309  0.178  1.297  2.673  0.158
Convective Dispersive Model            0.208
dispersivity (m)  2.859  1.034  0.049  0.796  1.108  
est. initial conc. (mg/l)  852  1720    446  182  
mean time of travel (days)  1.100  1.501    0.865  2.206  
Stochastic model      0.044      0.215
variance in travel time  0.915  0.637    0.562  0.641  
travel time (days)  1.168  1.547    0.879 2.287  
est. initial conc. (mg/l)  813  1724    444 184  

Figure 4.4.2.1  Extraction Well 1 Pre-remediation 2,2-dimethyl-3-pentanol
Figure 4.4.2.1.b  Extraction Well 1 Pre-remediation 2,2-dimethyl-3-pentanol log
Figure 4.4.2.2   Extraction Well 1 Post-remediation 2,2-dimethyl-3-pentanol
Figure 4.4.2.2.b  Extraction Well 1 Post-remediation 2,2-dimethyl-3-pentanol log
Figure 4.4.2.3   Extraction Well 1 Post-remediation methylheptanol
Figure 4.4.2.3.b  Extraction Well 1 Post-remediation methylheptanol log

Table 4.4.2.3. Summary of Extraction Well 2 tracer analysis
Extraction Well 2
Pre-remediation
Post-remediation
  conservative (bromide) 2,2-dimethyl 
-3-pentanol		 NAPL Saturation conservative (bromide) 2,2-dimethyl 
-3-pentanol		 NAPL Saturation
Pulse duration (days)  0.181  0.181    0.250  0.250  
zero moment data  96.29  336.794    90.483  30.119  
zero moment extrapolated  97.61  338.285    90.926  32.234  
first moment / zero moment data (days)  1.307  2.822  0.167  1.528  2.794  0.121
first moment extrapolated / zero moment (days)  1.382  2.866  0.154  1.541  3.165  0.154
Convective Dispersive Model            0.086
dispersivity (m)  1.150  0.415  0.103  0.378  0.763  
est. initial conc. (mg/l)  473  1405    369  121  
mean time of travel (days)  0.789  1.394    1.332  2.181  
Stochastic model      0.099      0.088
variance in travel time  0.653  0.418    0.402  0.554  
travel time (days)  0.810  1.403    1.341  2.213  
est. initial conc. (mg/l)  472  1401    370  121  

Figure 4.4.2.4   Extraction Well 2 Pre-remediation 2,2-dimethyl-3-pentanol
Figure 4.4.2.4.b  Extraction Well 2 Pre-remediation 2,2-dimethyl-3-pentanol log
Figure 4.4.2.5   Extraction Well 2 Post-remediation 2,2-dimethyl-3-pentanol
Figure 4.4.2.5.b  Extraction Well 2 Post-remediation 2,2-dimethyl-3-pentanol log
Figure 4.4.2.6   Extraction Well 2 Post-remediation methylheptanol
Figure 4.4.2.6.b  Extraction Well 2 Post-remediation methylheptanol log

Table 4.4.2.4. Summary of Extraction Well 3 tracer analysis
Extraction Well 3
Pre-remediation
Post-remediation
  conservative (bromide) 2,2-dimethyl 
-3-pentanol		 NAPL Saturation conservative (bromide) 2,2-dimethyl 
-3-pentanol		 NAPL Saturation
Pulse duration (days)  0.181  0.181    0.250  0.250  
zero moment data  61.759  130.494    18.936  6.092  
zero moment extrapolated  62.643  132.407    19.802  8.388  
first moment / zero moment data (days)  1.174  2.197  0.127  1.641  1.931  0.026
first moment extrapolated / zero moment (days)  1.226  2.305  0.128  1.750  2.932  0.098
Convective Dispersive Model      0.165      0.049
dispersivity (m)  0.614  0.676    0.494  1.349  
est. initial conc. (mg/l)  314  722    102  37.2  
mean time of travel (days)  0.727  1.621    1.153  1.576  
Stochastic model      0.167      0.060
variance in travel time  0.503  0.522    1.000*  0.745  
travel time (days)  0.736  1.650    1.153*  1.668  
est. initial conc. (mg/l)  314  724    102*  38.7  

* Not converged
Figure 4.4.2.7   Extraction Well 3 Pre-remediation 2,2-dimethyl-3-pentanol
Figure 4.4.2.7.b  Extraction Well 3 Pre-remediation 2,2-dimethyl-3-pentanol log
Figure 4.4.2.8   Extraction Well 3 Post-remediation 2,2-dimethyl-3-pentaonl
Figure 4.4.2.8.b  Extraction Well 3 Post-remediation 2,2-dimethyl-3-pentanol log

4.4.3    Extraction Analysis

Two alternatives could be utilized to interpret the data in the estimation of overall performance.  If there were a complete analysis of the NAPL, one could estimate the initial mass of a particular contaminant based on tracer data estimates of the NAPL saturation and the mole fraction of the individual contaminant in the NAPL.  Samples of NAPL were collected from several locations around the site and analysis performed.  When a NAPL is sampled in this manner, the fraction collected is what is readily mobile.  It is necessary to assume that this NAPL is in equilibrium (i.e., having the same composition) with the rest of the NAPL in the immediate area.  If there is a "pitch" portion to the NAPL, this fraction may not be in equilibrium with the energetically free NAPL creating a potential bias to the results. The average head during the pre-remediation tracer study and the remediation study is approximately the same and the flux through the test cell is approximately the same.  Thus, the volume swept by the pre-remediation tracers is approximately the same as the volume swept by the remedial fluid.  If there were high confidence in the tracer data, this would be a robust approach. Using core data creates a different set of  biases.  The core samples that were analyzed represent only a subsample of the total core removed from the formation.  Much of the formation is occupied by cobbles, that were too large to be included in the analysis.  Since a significant portion of the formation was occupied by cobbles the estimated mass is likely higher than actually the presented value.  This is not a significant problem when comparing core data to core data but may be a significant problem when using core data as an estimate for the starting mass in the individual test cell.  In addition to this problem, the geometry of the test cell creates dead zones or at least zones that are difficult to access.  Thus, the remedial fluid may not have been in contact with all of the contamination.  In the analysis presented here, core data were used as a starting point.  With the conservation of mass observed for the DMP, there is some question regarding the reliability of this data.

The elution curves are shown in the following figures: dichlorobenzene,naphthalene, ortho-xylene, undecane and summarized in the table below.

Table 4.4.3.1 Comparison of mass eluted from the test cell
Chemical Initial mass *(g) Mass extracted (g) Fraction removed
dichlorobenzene 51 358 7.0
o-xylene 420 39 0.09
undecane 5900 295 0.05
naphthalene 240 794 3.3
       
       
* Initial mass estimate based on geostatistical analysis.

There appears to have been major problems in either the QA or methods development.  Dichlorobenzene and naphthalene were dramatically overestimated in the elution curve.  Most likely, compounds other than dichlorobenzene and naphthalene were reported as these compounds.  The undecane is consistent with what was observed from the core analysis, and the ortho-xylene is much less than what would have been expected based on core analysis.  It is difficult to understand why there is such a discrepancy between the core and extraction data for a solubilization process.