1. Introduction
1.1. Purpose
2. Site Setting
4.1 3D Visualization Model 4-1
4.2 LNAPL Presence in North Olive Stratum 4-2
4.3 LNAPL Presence in Rand Stratum 4-3
4.4 LNAPL Presence in Main Silt, EPA, and Main Sand Strata 4-4
4.5 Comparison of Historical and Recent LIF Results 4-7
4.6 LNAPL Distribution Across the Hydrostratigraphic Units 4-9
4.7 Proposed Data Collection and Updates to the Comprehensive CSM
5.1 Soil Petrophysical Analysis 5-3
5.2 LNAPL Transmissivity Measurements 5-7
5.3 Historical LNAPL Recovery within the Main Sand Stratum 5-9
5.4 Historical LNAPL Recovery at the Premcor Facility 5-11
5.5 Proposed Data Collection and Updates to the Comprehensive CSM
7. REFERENCES
Tables
Figures
2. Volatile Petroleum Hydrocarbons Recovered Via Soil Vapor Extraction System
3. Fourth Quarter 2013 Groundwater Elevations, North Olive Hydrostratigraphic Unit
4. Fourth Quarter 2013 Groundwater Elevations, Rand Hydrostratigraphic Unit
5. Fourth Quarter 2013 Groundwater Elevations, EPA Hydrostratigraphic Unit
6. Fourth Quarter 2013 Groundwater Elevations, Main Silt Hydrostratigraphic Unit
7. Fourth Quarter 2013 Groundwater Elevations, Main Sand Hydrostratigraphic Unit
8. Collocated LNAPL Sample and LIF Boring Locations
9. Collocated Benzene Effective Solubility and Dissolved Phase Concentrations
10. LNAPL Viscosity in the Rand Stratum
11. LNAPL Viscosity in the Main Sand Stratum
12. Data Points Used to Generate the 3D Model
13. Boring Locations and 3D Lithology Data Display
14. LIF Locations and 3D LNAPL Data Display
15. LNAPL Distribution in the North Olive Stratum
16. LNAPL Distribution in the Rand Stratum
17. Fluid Level Elevations Within Select Wells Screened in the Rand Stratum
18. LNAPL Distribution in the EPA and Main Sand Strata
19. Fluid Level Elevations Within Select Wells Screened Within or Across the Main Silt Stratum
20. Fluid Level Elevations Within Select Wells Screened in the EPA Stratum
21A. Fluid Level Elevations Within Select Wells Screened in the Main Sand Stratum (Areas A, B1, and B2)
21B. Fluid Level Elevations Within Select Wells Screened in the Main Sand Stratum (Areas B3/B4 and C)
22. Maximum LNAPL Thickness in the Main Sand (2003-2005)
23. Maximum LNAPL Thickness in the Main Sand (2007-2009)
24. Maximum LNAPL Thickness in the Main Sand (2011-2013)
25. Lased Induced Fluorescence Boring Locations
26. LNAPL Distribution within the Hydrostratigraphic Units
27. Schematic Diagram of the Dual Optimal LNAPL Response Model
28. Historical LNAPL Soil Core Sample Locations
29. Fluid Saturations for Soil Cores Below the Water Table in the Main Sand
30. Mobile and Residual LNAPL Saturations Based on Lab Centrifuge
31. LNAPL Transmissivity Summary for 2005 High Vacuum Recovery Pilot Testing
32. LNAPL Transmissivity Summary for 2011 Multiphase Extraction Pilot Testing
33. Locations Considered as Part of Evaluating the Dual Optimal LNAPL Response Model
34. Historical LNAPL Recovery From the Main Sand Stratum
35. Schematic Diagram of Low Flow Dual Phase Extraction and Focused Pumping
Appendices
Appendix A
Appendix B
Appendix C
Appendix D
Appendix E
Appendix F
Appendix G
Download Full
Document Here
Conceptual Site Model, May 2014
Hartford Petroleum Release Site
Report
an
Issue
5.0 LNAPL RECOVERABILITY
The Dual Optimal LNAPL Response (DOLR) conceptual model was developed (H2A 2006) to explain the occurrence and potential recoverability of LNAPL under various hydraulic conditions. The DOLR model applies to the LNAPL present in the Main Sand, where the water table periodically transitions from unconfined to confined conditions. The DOLR model might also be applicable to shallower permeable strata such as the Rand and Main Silt, where LNAPL transitions between unconfined and confined conditions. However, as described in Section 4.0, LNAPL is detected infrequently in monitoring locations screened within these strata, suggesting that the fraction of total LNAPL that is potentially mobile and recoverable is relatively low. Therefore, the DOLR model is most useful in understanding historical LNAPL recovery within the Main Sand stratum, where the majority of LNAPL appears to be distributed as described in Section 4.0. This conceptual model can also be used to evaluate potential methods and select the optimal approach for attaining additional LNAPL recovery in the future.
Prior to describing the DOLR model, it is important to review general heuristics for LNAPL behavior and movement within unconsolidated sediments. LNAPL, when present, shares available pore space between sediment grains with water and air. In order for LNAPL to be mobile and recoverable it needs to be continuous or connected within the pore spaces. Within the saturated zone, where the pore spaces are primarily filled with water, LNAPL is generally present as less connected globules within the smaller pore spaces (2-phase conditions). That is, while some of the LNAPL might be connected and potentially capable of mobilizing to a well, much of it is often present as separate ganglia due to the majority of pore space being filled with water. Within the capillary fringe and vadose zone where water content is lower and air is also present (3-phase conditions), LNAPL tends to be more connected within the larger pore spaces. Put another way, LNAPL residual saturation can vary depending on whether 2-phase or 3-phase conditions are present (Charbeneau 2007). When LNAPL saturations are high and/or water saturations are low, LNAPL is better connected and therefore potentially mobile (i.e., the LNAPL is above the residual saturation). LNAPL preferentially moves within coarse-grained sediments such as sand and gravel (i.e., lower pore entry pressure), and is less able to migrate through fine-grained sediments such as silt and clay (assuming similar water content within the pore space).
With the above heuristics in mind, the first part of the DOLR model can be considered. Under confining conditions (created when groundwater within the Main Sand stratum intercepts and is forced against overlying finer-grained stratum), hydrostatic forces drive LNAPL into wells that behave essentially as pressure relief points. This is schematically depicted in the first panel on Figure 27. As the water table rises, some LNAPL in the smear zone also rises within connected pore spaces between the coarse-grained sediments and eventually contacts the bottom of the overlying fine-grained stratum. At that point, even though the piezometric surface continues to rise, the LNAPL remains trapped at the bottom of the confining stratum as it is unable to displace water from the smaller pore spaces present between the fine-grained sediment. Although the LNAPL is unable to move any further vertically, it is able to move laterally along the contact of the coarser Main Sand and overlying fine grained stratum. This potential for lateral movement is limited under these confined conditions because any portion of the pore space not occupied by LNAPL tends to be filled with water (2-phase conditions). Still, if a well is screened across the contact of the confining unit and the Main Sand, some fraction of LNAPL can move laterally into the well. The top elevation of the LNAPL in this well will be higher than the base of the confining stratum since it is under hydrostatic pressure resulting in an exaggerated LNAPL thickness. Such a condition could mean relatively high initial LNAPL recoverability from the well if mobile LNAPL can collect at the base of the confining layer and water in the well does not exert a significant backpressure. However, under this condition the “mass of available mobile LNAPL is minimal since much of the LNAPL mass is trapped underneath this high water table” (p. 59 of Appendix E within the Active LNAPL Recovery System 90% Design Report [Clayton 2006]). As LNAPL is removed from the formation adjacent to the well, LNAPL saturations may decrease as water saturations increase, resulting in reduced recoverability. Only if the LNAPL in the vicinity of the recovery well remains above residual saturations (i.e., has sufficient connectivity in this 2-phase condition) would recovery remain sustainable.
The second part of the DOLR model states that under unconfined conditions, LNAPL can vertically drain from the coarse sediments within the Main Sand as the water table falls below the confining strata. Under intermediate unconfined conditions (i.e., when the aquifer is unconfined but the water table is still relatively high), LNAPL thicknesses in wells can be relatively low because the confining pressure is no longer present and “much of the LNAPL is still submerged and entrapped under the water table” (p. 60 of Appendix E within the Active LNAPL Recovery System 90% Design Report [Clayton 2006]). LNAPL will subsequently accumulate above and below the water table, as depicted in the second panel on Figure 27. If the water table falls further, “much of the submerged residual LNAPL drains from the Main Sand, (and) larger volumes of mobile LNAPL are available to accumulate in wells” (p. 60 of Appendix E within the Active LNAPL Recovery System 90% Design Report [Clayton 2006]). The further the water table falls, the more LNAPL that drains and accumulates near the water table. This LNAPL is also able to move laterally within the Main Sand. If the screen interval within a well intersects the mobile LNAPL interval and the water table is sufficiently low for a sustained period of time, LNAPL can enter it and have an elevation that is consistent with the vertical interval of recoverable LNAPL in the formation (i.e., no exaggerated thickness). As shown on the third panel on Figure 27, “greater thicknesses [in wells] occur and relatively larger production capacities are observed” (p. 60 of Appendix E within the Active LNAPL Recovery System 90% Design Report [Clayton 2006]).
In summary, the DOLR model predicts that: (1) LNAPL thickness in wells will be high under confined conditions, with initially high LNAPL recovery rates, but a potentially lower mass of available mobile LNAPL within the zone of capture of the recovery well, (2) under intermediate unconfined conditions, LNAPL thicknesses may be smaller and recovery rates lower due to a significant portion of the LNAPL mass remaining submerged below the unconfined water table, and (3) high recovery rates may be attainable under the lowest water table conditions due to a larger mass of mobile LNAPL present under 3-phase conditions (i.e., unsubmerged) and therefore potentially recoverable.
At the Hartford Site, LNAPL recoverability has been previously assessed by several methods, such as: (1) soil coring, petrophysical analysis, and subsequent modeling; (2) LNAPL transmissivity estimates; and (3) LNAPL recovery pilot testing. These data, and information about LNAPL recovery at the adjacent Premcor facility, are reviewed in the remainder of this section. Of particular focus is the relationship of the hydrologic conditions (confined, intermediate unconfined, highly unconfined) and LNAPL recoverability estimates, and how these data support the DOLR model.
5.1 SOIL PETROPHYSICAL ANALYSIS
Soil coring and petrophysical analysis were conducted in 2005. The purpose of this sampling and analysis was “an evaluation of LNAPL recoverability within differing geological and hydrogeological settings” (Clayton 2006). The petrophysical data were used as inputs to two calculations related to LNAPL recoverability: estimates of the LNAPL specific thickness (Do) and LNAPL recovery modeling. The petrophysical analysis and calculations are reviewed herein in the context of LNAPL recoverability.
Six soil cores were collected from the smear zone in 2005 at the locations displayed on Figure 28. One location, boring HCSB-1, was installed along the eastern boundary of the Hartford Site along North Olive Street, and the other five were obtained within the interior portions of the smear zone to the north and west of boring HCSB-1. Core samples were collected within the Main Sand stratum from each boring. Core samples were also collected from the shallower strata (i.e., North Olive and Rand) from a subset of the borings. The soil cores were submitted to PTS Laboratories for petrophysical analysis. At the laboratory, individual plugs (i.e., subsamples) were extracted from each core and analyzed for moisture content, density, porosity, pore fluid saturations, grain size distributions, air-water drainage curves, and free product mobility by centrifuge. The results of these analyses are reported in Appendix C of the Active LNAPL Recovery System 90% Design Report (Clayton 2006).
A total of 73 plugs were analyzed for pore fluid saturations. LNAPL was detected in all but two of the plugs (HCSB-1 at 21.9 feet and HCSB-2 at 29.3 feet), with detected saturations ranging from 0.3% pore volume (HCSB-2 at 31.5 feet and HCSB-4 at 29.5 feet) to 40.2% pore volume (HCSB-1 at 31.6 feet). The total fluid saturations are displayed graphically for plugs obtained below the water table in the Main Sand on Figure 29. Water saturations were considerably higher than LNAPL saturations, with the exception of sample HCSB-1 at 27.3 feet. This suggests that in situ LNAPL saturations were generally low below the water table at the time of sampling. However, there is a limitation to this inference. As shown on Figure 29, between 14% and 40% of the plug pore volumes were reported as being absent of both LNAPL and water (shown as Sa on the figure). While it is possible that this indicates significant quantities of gases are trapped below the water table, it seems more likely that fluids were preferentially lost from the cores during collection in the field. If these fluids were primarily water, then this would not affect the inference that LNAPL saturations at many of the coring locations were relatively low. However, if the majority of fluids that drained from the cores during collection were primarily LNAPL, this would open the possibility that LNAPL was not adequately measured in the cores. LNAPL drainage from the cores during collection has been observed at other sites and is not considered a reason to reject the pore fluid saturation analytical results. However, any inferences based on measured saturations (especially LNAPL) should be considered with other lines of evidence. For instance, the measured LNAPL saturations might be considered low-end estimates of the true in situ values.
5.1.1 LNAPL SPECIFIC THICKNESS ESTIMATES
The LNAPL saturation results were subsequently used to estimate the LNAPL specific thickness (Do) across the full vertical interval of the smear zone at the five coring locations as described in Appendix D of the LNAPL Active Recovery System Conceptual Site Model (Clayton 2005). The Do calculations used site-specific data, including calibration of the van Genuchten “N” and “” to measured LNAPL saturations (provided in Section 3.1.4 of Appendix D of the LNAPL Active Recovery System Conceptual Site Model [Clayton 2005]). The calibrated Do calculations for the five soil cores were then used to generate Do correction factors that could be applied to measured LNAPL thicknesses in wells where soil coring data were not available. Using this correction, Do values were estimated for the Rand, EPA, and Main Sand strata for the October 2005 gauging event (Figures 3-4 through 3-6 of LNAPL Active Recovery System Conceptual Site Model [Clayton 2005]). The Do values were higher across a wider footprint in the Main Sand than the overlying strata, which follows from the Do calculation being directly based on LNAPL thicknesses gauged in wells.
The Do estimates are useful for evaluating the volume of LNAPL in a well, which tends to be greater than that within the adjacent formation (per unit lateral area) due to the formation volume also being occupied by soil, water, and to a lesser degree gases. Still, there are several limitations to the Do calculation that are worth noting:
5.1.2 LNAPL SATURATIONS
PTS Laboratories assessed the mobile versus residual LNAPL saturations for 12 plugs (9 from soil cores collected from boring HCSB-1 and 3 from boring HCSB-5) by centrifuging the plugs and measuring the volume of removed LNAPL. The reported residual LNAPL saturations can be considered low-end estimates because the force applied in the lab (i.e., 1,000 times gravity) are greater than what would be achieved in the field for a hydraulic recovery system. The results are displayed graphically on Figure 30.
Plugs collected from the North Olive and Rand strata were reported with lower mobile fractions versus those collected from the Main Sand. While the mobile versus residual LNAPL saturation data is considerably smaller than the routine well gauging data set (specifically, frequency of LNAPL occurrence in wells, an imperfect indicator) for the Hartford Site, both data sets suggest a relatively small portion of the pore space in the North Olive and Rand strata contain mobile LNAPL.
5.1.3 LNAPL RECOVERY MODELING
The petrophysical laboratory analyses were used to support LNAPL recovery modeling in the Active LNAPL Recovery System 90% Design Report (Clayton 2006). Four different technologies were modeled: skimming, vacuum-enhanced skimming, dual pump recovery, and dual phase extraction. The simulations, performed for 1, 6, and 10 year timeframes, suggested that LNAPL skimming would attain recovery rates that are relatively low versus the other technologies. For the selected input parameters, vacuum-enhanced skimming and dual pump recovery attained similar rates, although the following should be noted:
This second bullet is worth further consideration in the context of the modeled results from boring HCSB-1 located near groundwater monitoring well HMW-044C in Area A. The measured versus modeled saturation profiles, from Appendix F of the Active LNAPL Recovery System 90% Design Report (Clayton 2006), are presented herein in Appendix F. LNAPL was detected in plugs obtained from approximately 10 to 41.6 ft-bgs. The maximum measured LNAPL saturation of 40.2% pore volume was measured near the water table at 31.6 ft-bgs. The maximum LNAPL saturation below the water table was 8.3% at 39.7 ft-bgs. Based on the centrifuge data (Figure 30), LNAPL saturations at both of these depths are above the residual saturation and indicate potentially mobile LNAPL. However, the modeled interval of LNAPL presence shown on the right-hand graph presented in Appendix F is only for the interval located near the water table, from 30.5 to 33.4 ft-bgs. This modeled interval is based on the measured LNAPL thickness in well HMW-044C during gauging in September 2005. The model assumes that LNAPL is not present below this interval, and therefore does not simulate recovery of the potentially mobile LNAPL as deep as 39.7 ft-bgs. If groundwater extraction were conducted near boring HCSB-1 at a high enough rate to substantially lower the water table, potentially mobilized LNAPL within lower portions of the smear zone would not be simulated in this model.
The limitations to the LNAPL recovery modeling provided in the Active LNAPL Recovery System 90% Design Report (Clayton 2006), as well as inherent uncertainties in estimating LNAPL saturations via petrophysical analysis, suggest that additional pilot testing to assess LNAPL recoverability is warranted. MPE and LNAPL skimming have been tested at the Hartford Site over the past decade. Pilot testing of LNAPL recovery under higher groundwater extraction rates has been a data gap for LNAPL recoverability. The Final Light Non-Aqueous Phase Liquid Recovery Pilot Test Work Plan Addendum (Trihydro 2013a) proposes to lower the piezometric surface to expose currently submerged portions of the smear zone to determine whether the deeper LNAPL is potentially recoverable at rates that exceed vacuum enhanced skimming. This could provide an indicator of mobile LNAPL across a large footprint (i.e., the radius of capture of the recovery well) for lower water table conditions.
5.2 LNAPL TRANSMISSIVITY MEASUREMENTS
LNAPL transmissivity has been proposed as a metric to assess the potential for hydraulic recovery of LNAPL under the specific conditions tested (WSP 2012). LNAPL transmissivity can be estimated by analyzing data collected via multiple methods, such as LNAPL baildown tests, manual skimming, and continuous recovery (ASTM 2011). LNAPL transmissivities have been measured at the Hartford Site using various methods since 2004. These data are reviewed in this subsection in the context of LNAPL recoverability.
The dynamic nature of the piezometric surface within the Main Sand is one of the most important conditions affecting LNAPL transmissivity. As stated by WSP (2012), “. . . as the water table elevation at a site fluctuates, some of the mobile LNAPL at the water table may travel with the water table or it may be submerged (rising water table) or released (falling water table). As a result, the mobile and residual LNAPL saturations can decrease or increase.” The piezometric surface in the Main Sand rises and falls in response to the Mississippi River elevation and precipitation events, frequently transitioning between confined, semiconfined, and unconfined conditions. This phenomenon, and its relationship to measured LNAPL transmissivities, is considered herein.
As shown on Table 4, a total of 96 quantitative LNAPL transmissivity estimates have been completed for 26 different groundwater monitoring wells and multipurpose monitoring points between 2004 and 2012. The LNAPL transmissivity estimates provided in Table 4 are generally based on LNAPL baildown testing, with the exception of recharge data following MPE testing in 2005. No LNAPL transmissivity values are available between 2006 and 2008. Reported values span multiple orders of magnitude, from a low of 0.0010 square feet per day (ft2/day) in monitoring point MP-52C in August 2005 to a high of 94 ft2/day in monitoring point MP-50C in May 2005. The transmissivity estimates were conducted at a minimum piezometric surface elevation of approximately 396 ft-amsl and a maximum of
410.5 ft-amsl.
These LNAPL transmissivity estimates can be compared to the minimum endpoint (0.3 ft2/day) proposed by HWG for hydraulic recovery of LNAPL (WSP 2012). In 2004 and 2005, 69% (46 of 67) of the estimated LNAPL transmissivities were above this threshold, while in 2009 and 2010, 60% (15 of 25) of the values were above this threshold. In 2011 and 2012, none (0 of 4) of the values were above this threshold. Overall, these results might be an indication of decreasing LNAPL recoverability at the Hartford Site. That is, the quantity of LNAPL above residual saturations in the Main Sand has decreased as a result of LNAPL recovery by skimming, natural source zone depletion, and continued smearing of LNAPL due to piezometric fluctuations. However, this inference is limited by variations in the location, test method, and hydraulic conditions during testing. The repeatability of the results and influence of groundwater elevation changes on LNAPL transmissivity are examined more closely in the following bullets.
These results suggest that if LNAPL recoverability is to be assessed under ambient conditions, then LNAPL transmissivities should be estimated at multiple locations and points in time corresponding to the range of ambient fluctuations of the piezometric surface. Additionally, if the elevation of the piezometric surface has an influence on LNAPL transmissivities, then it follows that a lower water table induced by pumping could result in an increase in LNAPL transmissivity not observed under ambient conditions. This was suggested within Appendix D of the Light Non-Aqueous Phase Liquid Recovery Pilot Test Interim Report (WSP 2012) in which it is stated that: “…given a suitable physical setting, residual LNAPL that is submerged may become mobile during a given decrease in the water table thereby resulting in an increase in LNAPL transmissivity.” It is understood that WSP was likely referring to an ambient change in water table conditions and even suggested that it may be impractical to use hydraulic recovery to address submerged LNAPL. Still, it stands to reason that LNAPL transmissivities may increase under stressed conditions within the radius of influence of the pumping well.
5.3 HISTORICAL LNAPL RECOVERY WITHIN THE MAIN SAND STRATUM
For the most part, historical LNAPL recovery at the Hartford Site has been conducted under confined (the first panel on Figure 27) or intermediate unconfined (the second panel on Figure 27) conditions. The figures provided in Appendix G depict fluid level elevations and LNAPL recovery rates since 2005 within the 22 recovery wells, groundwater monitoring wells, and multipurpose monitoring points shown on Figure 33. These locations were selected because they were included as part of pilot testing of MPE (referred to as high-vacuum recovery) in 2005 (H2A 2006) and have subsequently undergone routine manual LNAPL skimming between 2010 and 2012. During manual skimming, field personnel would visit a location with more than 0.5 foot of LNAPL, conduct skimming, and then allow the LNAPL to recharge above this thickness before skimming again. This method would mean that drawdown in the monitoring well or monitoring point would only be maximized immediately after skimming, and would decrease over time until the next skimming event.
The LNAPL recovery rates depicted on the figures in Appendix G are the monthly volume recovered via manual skimming (i.e., units are gallons per month). These figures also display the elevation of the confining stratum based on the lithology recorded during installation of these wells and monitoring points. Confined conditions are inferred where the groundwater elevation (piezometric surface) is above the bottom of the fine grained stratum. Finally, the hydrographs display the vertical interval of the LNAPL smear zone based on nearby LIF borings. Note that the lithology and LIF data presented on the figures included within Appendix G are a subset of the data used to develop the 3D model described in Section 4.0.
These hydrographs indicate that the majority of manual LNAPL skimming has been conducted under confined conditions compared to intermediate unconfined conditions since 2005, as confined conditions occurred more frequently and lead to exaggerated LNAPL thicknesses targeted for skimming by the field personnel. The LNAPL recovery rates for skimming under confined conditions show significant variability at a given location that may be related to competing LNAPL recoverability characteristics as described by the DOLR model. The confining pressures may result in the groundwater monitoring well or monitoring point acting as a pressure relief point; however, elevated water saturations may mean only a minimal mass of mobile LNAPL is available within the radius of capture of the monitoring location.
When skimming is conducted under unconfined conditions, LNAPL recovery tends to remove smaller volumes of LNAPL. Inspection of the hydrographs provided in Appendix G, shows that between 2005 and 2012 when skimming was performed, unconfined conditions have been limited to the upper portions of the smear zone in the Main Sand stratum. This may be an indication of intermediate unconfined conditions described in the DOLR model where the majority of the smear zone remains submerged.
Overall, LNAPL recovery rates decreased between 2010 and 2012 as shown on Figure 34, which displays cumulative LNAPL recovery from the Main Sand stratum, as well as average recovery rates from individual wells and monitoring points. This decrease in LNAPL recovery rates can be inferred to represent a decrease in the LNAPL transmissivity near those wells and monitoring points where skimming has been conducted (providing a supporting line of evidence for the LNAPL transmissivity estimates described in Section 5.2). However, it is possible that other factors such as natural smear zone depletion, fluctuation of the piezometric surface resulting in additional smearing of LNAPL, and LNAPL recovery conducted at the adjacent facilities may have contributed to decreasing LNAPL recoverability within these monitoring locations. Since manual skimming targeted any monitoring point or groundwater monitoring well where LNAPL was measured above 0.5 feet, there are few locations where skimming was not performed. Therefore, it is difficult to assess the effect of manual skimming versus the other contributing factors that might have affected LNAPL recoverability in recent years.
Efforts to estimate the radius of capture for manual LNAPL skimming conducted in the groundwater monitoring wells or multipurpose monitoring points have not been performed via volumetric analyses, LNAPL tracer testing, or some other method. Charbeneau and Beckett (2007) suggest a radius of capture for LNAPL skimming between 10 and 30 feet. It is expected that the radius of capture for manual skimming would be on the low end of the suggested radius of capture as a result of the methodology used to recover LNAPL. As previously described, drawdown would be maximized immediately after skimming, and would decrease over time until the next skimming event. Since drawdown was lower during recharge, this probably meant a lower radius of capture than would have been achieved with a dedicated skimmer (i.e., consistently maximized drawdown). An evaluation of the radius of influence of manual skimming and its effects on future remedial efforts may be considered in the Comprehensive CSM. This evaluation will determine if future LNAPL recovery in the zone of typical piezometric surface fluctuations is expected to be low only adjacent to previously skimmed wells (as observed during pilot testing performed by WSP in 2011 and 2012), or if low recovery might also be expected across the rest of the smear zone footprint due to other contributing factors reducing LNAPL recoverability.
As shown on the figures provided in Appendix G, testing under all anticipated hydraulic conditions described in the DOLR model has not been performed at the Hartford Site. Specifically, the DOLR model predicts that when groundwater elevations are within the lower portions of the smear zone, LNAPL recovery rates may be higher (as more of the smear zone is under 3-phase conditions). As described in Section 2.4.4, water table elevations have not approached these lower portions of the smear zone since construction of Dam No. 27 (a.k.a. the Chain of Rocks Dam), between 1959 and 1963, down-stream of the Hartford Site. Therefore, a viable means to observe LNAPL recovery under low water table conditions may be to induce these conditions via extraction of groundwater at higher rates within focused portions of the Hartford Site. In this scenario, LNAPL recoverability could be observed in the vicinity of the groundwater extraction well. Such a pilot test has been approved by the USEPA as part of the Final Light Non- Aqueous Phase Liquid Recovery Pilot Test Work Plan Addendum (Trihydro 2013a). Area A along North Olive Street was selected as the pilot test area. Elevated LNAPL transmissivities have been estimated using data collected from monitoring wells within this portion of the Hartford Site (Clayton 2006) and previous pilot testing has been conducted in Area A in 2005 and 2011 through 2012, which allows comparative analysis of the various approaches. However, any comparison in the remedial approaches will not be perfect, as nearly a decade has passed since the original pilot tests were performed by HWG, which allows for additional factors to affect LNAPL recoverability. These include natural source zone depletion, continued smearing of LNAPL due to piezometric surface fluctuations, manual LNAPL skimming performed throughout the Area, and LNAPL recovery performed at the adjacent Premcor facility. Still, the proposed pilot test will allow for an assessment of the amount of LNAPL that could be mobilized under stressed conditions within the radius of capture of the production well. The results of this additional pilot test will help to resolve many of the remaining data gaps regarding LNAPL recoverability and will be integrated into the Comprehensive CSM.
5.4 HISTORICAL LNAPL RECOVERY AT THE PREMCOR FACILITY
LNAPL recovery conducted along the western property boundary of the adjacent Premcor facility can provide some insight into the applicability of the DOLR model to the Hartford Site. LNAPL recovery at the Premcor facility has been focused within the EPA and Main Sand strata. From 1994 to 2002, LNAPL was primarily removed using scavenger pumps deployed in select wells. Following this timeframe, LNAPL recovery at the Premcor facility has been conducted using a combination of techniques including automated LNAPL skimming, periodic manual skimming, LFDPE, and SVE. It should be noted that a significant volume of LNAPL has also been recovered from the groundwater production wells installed at the Premcor facility for gradient control purposes using a skimmer pump and intermittently applied vacuum.
Pumping has been conducted for the purpose of gradient control using various wells installed along the western boundary of the Premcor facility, with well RPW-01 being used as the primary well since 2005. Pumping has been performed to enhance inward hydraulic gradients in an attempt to prevent migration of petroleum-related constituents from the refinery to beneath the Hartford Site. In 2006, following approximately four months of groundwater extraction at 100 to 120 gpm from well RPW-01, LNAPL was observed in the pumping well (WSP 2011). A skimmer pump was installed in well RPW-01 in March 2006 to recover LNAPL, and beginning in January 2007, a vacuum was applied during low water table conditions. Approximately 118,000 gallons of LNAPL have been recovered from well RPW-01 since 2006, with the maximum LNAPL recovery observed in 2008 (groundwater elevations during that year varied from 398 to 411 ft-amsl within the Main Sand stratum, transitioning from unconfined to confined). An additional 80,000 gallons have been recovered via skimming in other wells installed along this portion of the Premcor facility. In recent years, the specific capacity of groundwater production well RPW-01 has decreased due to a combination of biofouling and silting of the gravel pack, and LNAPL recovery rates have diminished. Consistent with the DOLR model, pumping at the Premcor facility mobilized additional LNAPL towards the recovery wells (greater than what may otherwise be observed). Pumping which stimulates unconfined conditions within the Main Sand stratum locally beneath the western property boundary may have enhanced LNAPL recovery.
LFDPE has been conducted within 14 extraction wells screened in the EPA Stratum and shallow portions of the Main Sand along the western portions of the Premcor facility since December 2010. This method involves limited groundwater extraction at flow rates of approximately 5 to 10 gpm, combined with LNAPL removal with either submersible pumps or a vacuum stinger placed above the LNAPL-groundwater interface. Dual phase recovery is optimized when LNAPL is under confined conditions (depicted schematically in Figure 35). The recovery system uses a vacuum above the LNAPL/air interface to induce fluid movement to the well while a groundwater pump maintains the LNAPL/water interface in the well adjacent to the interval of mobile LNAPL within the formation. Approximately 122,000 gallons of LNAPL have been removed via LFDPE from beneath this portion of the Premcor facility. In addition, approximately, 100,000 gallons of LNAPL have been recovered via the mobile LFDPE system. In the case of the Premcor LFDPE system, the confining clay stratum is the B/C Clay, which is present beneath the western boundary of the refinery. The success of the LFDPE at the Premcor facility is reportedly attributed to installing recovery wells within localized high points within the overlying confining unit where LNAPL preferentially migrates. It is also possible that LFDPE has been successful at the Premcor facility, due to the presence of the D Clay, which has potentially acted as a barrier to downward LNAPL movement during historically low water table elevations. This might have allowed LNAPL to accumulate within the EPA stratum between the B/C Clay and the D Clay and yielded relatively high LNAPL saturations in this stratum. The EPA stratum and D-Clay are not present beneath the majority of the Hartford Site. Instead, the majority of the LNAPL appears to be located within the Main Sand, which does not have a shallow fine grained layer to act as a barrier to downward LNAPL movement. Therefore, at the Hartford site, the degree of historical smearing might have been greater, yielding lower recoverability under confined conditions (i.e., deeper submerged LNAPL). The effectiveness of the LFDPE will be considered further as part of the Comprehensive CSM.
WSP pilot tested LFDPE beneath Area A of the Hartford Site in January 2012, as described in Section 2.2.1. During the WSP pilot test, groundwater and LNAPL were confined. Groundwater extraction was tested up to a maximum rate of 25 gpm, in an attempt to define rates necessary to optimize LNAPL recovery using a vacuum stinger. Pilot testing of LFDPE failed to induce significant LNAPL recovery beneath this portion of the Hartford Site. Additional vacuum enhanced recovery performed in combination with focused pumping may be performed in Area A of the Hartford Site. The results of this approach will be integrated into the Comprehensive CSM.
5.5 PROPOSED DATA COLLECTION AND UPDATES TO THE COMPREHENSIVE CSM
In recent years, LNAPL recovery at the Hartford Site has been conducted via skimming activities primarily under confined conditions, with some recovery under intermediate unconfined conditions. Recovery rates have been observed within these groundwater monitoring wells and multipurpose monitoring points where manual skimming has been performed. The available data suggest that under these conditions LNAPL transmissivities are approaching the low practical recovery endpoints (0.3 to 0.8 ft2/day) suggested by HWG for recovery within the upper portions of the smear zone. An evaluation of the radius of influence of manual skimming, as well as other factors contributing to reduced recoverability, and the combined effect on future remedial efforts may be considered in the Comprehensive CSM. This evaluation will determine if future LNAPL recovery in the zone of typical piezometric surface fluctuations is expected to be low only adjacent to previously skimmed wells (as observed during pilot testing performed by WSP in 2011 and 2012), or if low recovery might also be expected across the rest of the smear zone footprint.
However, predictions of the DOLR model, as well as enhanced LNAPL mobility achieved via hydraulic controls implemented along the western boundary of the Premcor facility, suggest that additional LNAPL recovery may be possible beneath portions of the Hartford Site using a focused pumping approach. Focused pumping can induce a transition to unconfined conditions near the production well if performed during already low ambient water table conditions. This technique will be pilot tested in Area A at the Site, as described in the Final Light Non-Aqueous Phase Liquid Recovery Pilot Test Work Plan Addendum (Trihydro 2013a). The pilot testing will assess whether inducing unconfined conditions in the vicinity of a groundwater production well installed in Area A can increase LNAPL recovery rates within the Main Sand stratum. Additional enhanced vacuum recovery may also be performed as part of this additional pilot testing in Area A. The results from this pilot test will also be incorporated into the Comprehensive CSM.