By Joseph Robb and Ellen Moyer

This paper discusses the natural attenuation of benzene and methyl tertiary butyl ether (MTBE) at four Midwestern U.S. retail gasoline marketing outlets.  We chose four sites with many rounds of analytical data that for confidentiality purposes are described as Sites A, B, C and D.  This article has the following major objectives:

• Assess the natural attenuation of benzene and MTBE using datasets “typical” of gasoline station sites (i.e., data from a limited monitoring well network that are not collected with the primary goal of evaluating natural attenuation).
     

• Use the Mann-Kendall trend test and the coefficient of variation (CV) test to identify decreasing, stable or increasing concentration trends at individual wells and, by extension, identify decreasing, stable or increasing plumes.
     

• Use Sen’s nonparametric trend estimator to quantify the magnitude of concentration trends.
     

• Demonstrate how trend evaluations can also be used to help identify the need for additional source control measures or site characterization work.
     

• Identify additional data that would be useful to collect for natural attenuation evaluations at petroleum release sites.
     

• Develop practical recommendations for transitioning the focus of site characterization, if warranted, from plume delineation to evaluation of natural attenuation.

The term “natural attenuation” as used in this paper and defined by the United States Environmental Protection Agency (U.S. EPA, 1999) refers to all of the physical, chemical or biological processes that act without human intervention to reduce the mass, toxicity, mobility, volume or concentration of chemicals in soil or groundwater.  In situ natural attenuation processes include biodegradation, dispersion, dilution, adsorption, volatilization, and chemical reactions.  The three lines of evidence that may be used to demonstrate natural attenuation include:  1) analytical data showing plume stabilization or a decrease in contaminant mass over time, 2) geochemical data that show conditions are favorable for natural attenuation and 3) bench-scale microbiological laboratory data. 

This evaluation will rely entirely on the first and second lines of evidence to demonstrate natural attenuation. In order to demonstrate stable or decreasing contaminant plumes, time trend evaluations were performed to ascertain the behavior of benzene and MTBE at individual monitoring wells.  Trend evaluation results from individual monitoring wells were then interpreted within the hydrogeologic context of each site to describe site-wide trends in plume behavior. 

Trend Analysis Approach

The non-parametric Mann-Kendall test and Sen’s non-parametric trend estimator were combined with the CV test to evaluate time trends in benzene and MTBE data.  The Mann-Kendall test and Sen’s trend estimator are considered well suited to the dataset because they can be used on data that are non-parametric (do not have a specific distribution, such as normal or log normal) and the dataset can contain data at irregularly spaced intervals.  In addition, the dataset can contain elevated values compared to the average (outliers) or data reported as below the practical quantitation limit.  As with many statistical tests, the validity of the results is increased when the sample size is larger.

Essentially, the Mann-Kendall test is performed as follows: 

• The data are listed in the order in which they were collected. 
     

• Each data point is compared to the points that follow in time. 
     

• The number of times the data increase is compared to the number of times the data decrease.  The greater the number of increases or decreases, the more evidence there is for an upward or downward trend. 

Sen’s estimate of trend is calculated as follows:

• The data are listed in the order in which they were collected.
     

• Individual slope estimates are made between each data point and all successive data points.
    

• A median slope is calculated from the individual slope estimates.

The following guidelines were followed when applying the Mann-Kendall test and Sen’s estimate of trend to the dataset and interpreting the results:

• Trend analyses were performed for benzene and MTBE.
     

• A confidence level of 90% was deemed appropriate for natural attenuation trend evaluations.
     

• All non-detect values were assigned one half of the lowest detection limit in the time series data, as described by Gibbons (1994).  This approach removes the bias introduced by detection limits that change over time. 
     

• Mann-Kendall evaluations were not performed for time series data that consisted entirely of points near and below the detection limit, due to the inherent uncertainty of analytical data near the detection limit.  For example, a Mann-Kendall evaluation of the series 2 micrograms per liter (mg/L), 1.5 mg/L, 1 mg/L and 0.5 mg/L would treat each value as different and lower than the preceding value, and result in a determination of a downward trend, even though from a data interpretation perspective these values may essentially be the same number.
     

• Mann-Kendall trend evaluations were not performed on datasets that exhibited seasonality.
     

• Trend results for individual wells were grouped together to develop an interpretation of concentration trends across the entire site.
     

• Trend results for individual wells were interpreted considering the location of source area(s), groundwater flow directions and approximate dissolved plume locations.

Analytical data from site perimeter monitoring wells often have many consecutive rounds of non-detect data interspersed with several rounds that include detected compounds.  Since the non-detect values are all considered equal in the Mann-Kendall evaluation, the result is often a determination of “no trend.”  However the validity of the “no trend” result is drawn into question because the Mann-Kendall test does not take into account the variability in the dataset.  Since the identification of no trend is important for the evaluation of natural attenuation (i.e., identification of stable plumes or continuing sources), the CV was used to measure the variability within each time series dataset. 

CV = Standard Deviation / Arithmetic Mean

The CV should be less than or equal to 1 for a Mann-Kendall “no upward or downward trend” result to be considered a stable trend.  If the CV is greater than 1, the trend is considered unstable.  Thus, in order to show a stable plume, one would look for a preponderance of wells, especially at the downgradient edge of a plume, that show no trend in the Mann-Kendall analysis with a coefficient of variation less than or equal to one.  Stable benzene and MTBE concentrations near the source area indicate an ongoing source.

Sen’s estimator of slope was used to quantify downward trends in benzene and MTBE at individual monitoring wells.   The guidelines described above for the Mann-Kendall trend test were also applied for the Sen’s trend estimator.  Slope estimates, calculated in mg/L per year, for individual wells were grouped together to develop an interpretation of concentration trends across the entire site.  Slope estimates for benzene were qualitatively  compared to MTBE slope estimates at individual monitoring wells and across the entire site.  

Although present at each site, toluene, ethylbenzene and xylenes are not discussed in this paper.  Benzene and MTBE were chosen because benzene is often the risk driver at petroleum release sites and MTBE is often present in high concentrations, due to its relatively high solubility.  In addition, for the purposes of comparing MTBE and other gasoline constituents, the fate and transport characteristics of benzene are generally representative of other gasoline constituents such as toluene, ethylbenzene and xylenes.

Geochemical Data

The identification of significant downward trends provides good evidence that natural attenuation is occurring, but does not illuminate to what extent biological mechanisms are contributing to observed decreases in concentrations.  It is now recognized that biological degradation of benzene and MTBE can occur under a variety of terminal electron accepting processes (see the article by J. Thomson in this issue). Dissolved oxygen (DO) and oxidation-reduction potential (ORP) data were available for the four sites considered in this paper and were qualitatively evaluated to assess the importance of biological degradation.  Although the data were insufficient to precisely identify terminal electron accepting processes, general observations of trends in redox chemistry across the site were made.  In general, lower ORP values and depressed DO values in the source area, when compared to background, were considered good evidence of the consumption of oxygen for the biological degradation of benzene and MTBE. 

The natural attenuation of benzene and MTBE for four sites are evaluated below.  Hydrogeologic conditions, release history, trend evaluation results and geochemical conditions are discussed for each of the four sites. Site A data are presented and discussed in slightly greater detail than for Sites B, C and D, in order to demonstrate the methods use to evaluate natural attenuation.   At each of the four sites an unspecified amount of soil was removed in the early 1990s during underground storage tank (UST) upgrades or other subsurface investigation.  No further engineered soil or groundwater remediation has been performed at these sites.    

Site A

Site A was reported to regulatory authorities in January 1992 when light non-aqueous phase liquid (LNAPL) was identified in the vicinity of the former dispensing island, canopy footings and dispenser lines during excavation activities associated with UST upgrades.  However, no LNAPL has been detected in site monitoring wells. 

Hydrogeology

The depth to groundwater at Site A is 5 to 12 feet below ground surface (bgs).  The shallow unconfined aquifer consists primarily of unconsolidated glacial deposits described as reddish brown medium stiff clay, containing a trace of silt and fine sand.  The average hydraulic conductivity (K), estimated by rising head slug tests performed in site monitoring wells, is 3 x 10-4 centimeters per second (cm/sec), and the horizontal hydraulic gradient (i) across the site is approximately 0.09.  Based on the equation v = K i / n, where n is effective porosity estimated at 0.25, the average horizontal groundwater velocity (v) is estimated to be approximately 110 feet per year.  The horizontal groundwater flow direction is to the southwest.  The dimensions of Site A are 75 feet by 95 feet.  Thus, based on a release date of at least 9 years ago, and an estimated average groundwater velocity of 95 feet per year, dissolved contaminants have had ample opportunity to migrate off-site. 

Seasonality

Between January 1995 and July 1998, increases in groundwater elevations generally coincided with increases in dissolved benzene concentrations in source area monitoring wells MW-1, MW-2, MW-3 and MW-4.  Since Site A is paved and relatively impervious to infiltration, increases in dissolved benzene concentrations could be due to the seasonal rise of the water table into zones of residual soil contamination.  After January 1998, seasonal effects appear to play a smaller role in concentration trends, which indicates a potential depletion of residual soil contamination within the range of seasonal groundwater fluctuations. 

Trends 

Mann-Kendall trend analyses for MTBE were evaluated over ten sampling rounds from July 1997 to November 2000 and over seven sampling rounds from April 1998 to November 2000 for benzene.   Mann-Kendall trend analyses were not performed for benzene data collected prior to April 1998, or for MTBE data prior to July 1997, due to the presence of seasonal trends.   Mann-Kendall trend analyses indicate decreasing trends in benzene and MTBE concentrations at the 90% confidence level in source area monitoring wells MW-1, MW-2, and MW-3.  Source area monitoring well MW-4 exhibits a stable trend in MTBE and an unstable trend in benzene concentrations.

Decreasing concentrations of benzene and MTBE in wells MW-1, MW-2, and MW-3, are strong evidence of an attenuating source of benzene and MTBE.  Where benzene concentrations are highest, (MW-1, MW-2 and MW-3), Sen’s slope estimates show benzene is decreasing at rates between 377 mg/L and 620 mg/L per year.   At these same wells, MTBE is decreasing at rates between 222 mg/L and 2,024 mg/L per year. The highest estimate of MTBE decline (2,024 mg/L per year at MW-1) may be an overestimate due to a single elevated concentration value in October 1997 and lower values in subsequent sampling rounds.  In general, the slope estimates suggest benzene and MTBE concentrations are declining at approximately the same rate in source area wells.

Trend evaluations were not performed for volatile organic compound (VOC) data from perimeter wells MW-5, MW-6 and MW-7 because the data consisted primarily of concentration values near or below the detection limit (see discussion above).  These wells effectively delineate the extent of the dissolved VOC plume to the northeast (hydraulically upgradient) and northwest (cross-gradient).

Geochemical Conditions

DO and ORP values measured over seven sampling rounds from April 1998 to November 2000 did not show any trends.  However, average ORP and DO values over this time period indicate redox conditions are favorable for the biological degradation of dissolved benzene and MTBE. April 1998 to November 2000 average MTBE, benzene, DO and ORP data.  In general, there is a good correlation between the presence of MTBE and benzene, and low ORP values, relative to background values.  In addition, the DO concentration is higher in background well MW-5, and slightly lower in the source area, cross-gradient and downgradient wells.   These results suggest DO is being used as a terminal electron acceptor in the source area.

Site B

Site B was reported to regulatory authorities in August 1993 when photoionization detector field screening identified petroleum-impacted soil in the vicinity of the gasoline transfer lines.  No LNAPL was observed during the 1993 excavations, nor has LNAPL been observed in Site B monitoring wells. 

The shallow unconfined aquifer consists primarily of unconsolidated glacial deposits described as dark gray silt with some clay and a trace of fine gravel and coarse sand.  The hydraulic conductivity is estimated to be between 10-6 and 10-4 cm/sec and the horizontal hydraulic gradient across the site is approximately 0.05.  Assuming an effective porosity of 0.25, the average horizontal groundwater velocity is estimated to range from 0.2 to 20 feet per year.  The depth to groundwater ranges from 3 to 11 feet bgs, and the horizontal groundwater flow direction is to the southeast.  The distribution of contaminants at the site is consistent with the timing of the release and the estimated groundwater velocity.

Three groundwater monitoring wells have been installed at Site B. Approximately 20 rounds of data were available from April 1994 to November 2000.  Data prior to April 1997 appeared to exhibit seasonal trends and were therefore excluded from the trend evaluations.

MTBE exhibits a stable trend at monitoring wells MW-2 and MW-3 and an upward trend at MW-1.  Benzene is non-detect at MW-1, exhibits a stable trend at monitoring well MW-2 and an unstable trend at MW-3.   The elevated, stable benzene and MTBE concentrations at MW-2, located immediately downgradient of the gasoline-dispensing island, provide strong evidence of a continuing source of petroleum hydrocarbons.  The stable trend in MTBE at well MW-3 and upward trend at MW-1 indicate the plume of MTBE may be stable in the northeast, but is expanding in the western part of the site.  Elevated MTBE and benzene concentrations at well MW-2 suggest dissolved benzene and MTBE are migrating off site. 

Average DO and ORP values are lowest where benzene and MTBE concentrations are highest (MW-2), and higher average ORP and DO values were measured at wells MW-1 and MW-3, which contain little or no benzene and low to moderate concentrations of MTBE. The DO, ORP, benzene and MTBE results are consistent with the consumption of oxygen for the biological degradation of gasoline constituents.  

Overall, results for Site B indicate the following:

• Residual contamination is acting as a continuing source of petroleum hydrocarbons at Site B.  Additional source control may be warranted.
     

• Biodegradation may be limiting the migration of benzene and MTBE.

Site characterization and development of the natural attenuation argument would be facilitated by installation and sampling of additional monitoring wells.

Site C

Site C was reported to regulatory authorities in December 1994 when LNAPL was identified in a gasoline transfer line trench excavated for UST upgrades.  No LNAPL has been detected in site monitoring wells. 

The shallow unconfined aquifer consists primarily of unconsolidated glacial deposits described as gray silty clay with a trace of sand and gravel.  The average hydraulic conductivity, based on slug tests performed in site monitoring wells, is estimated to be   9x10-8 cm/sec and the horizontal hydraulic gradient across the site is approximately 0.02.  Assuming an effective porosity of 0.25, the average horizontal groundwater velocity is estimated be 0.01 feet per year.  The depth to groundwater is 5 to 8 feet bgs, and the horizontal groundwater flow direction is to the northeast.  The shallow depth to groundwater and the low permeability of the native aquifer material suggest shallow groundwater flow may be influenced by high permeability fill material in utility conduits.   Contaminant distribution and the estimated age of the release suggest actual groundwater flow rates may be several orders of magnitude greater than 0.01 feet per year.

Six groundwater monitoring wells have been installed at Site C.  Between 15 and 17 rounds of benzene and MTBE data were available from August 1995 to November 2000.  Concentration data from this time period did not appear to exhibit seasonality and were utilized in the trend evaluations.

MTBE detected in downgradient well MW-5 does not exhibit a downward trend for the time period from April 1995 to November 2000.  However, when the MTBE dataset at well MW-5 was divided in two halves, the Mann-Kendall analysis identified an upward trend from April 1995 to October 1997, and a downward trend from October 1997 to November 2000.  These results suggest the dissolved plume of MTBE at Site C has been shrinking since October 1997.  Recent MTBE concentrations near the property boundary at MW-5 are relatively low (50 mg/L, 75 mg/L), and are similar to MTBE maximum contaminant levels that have been established by various state regulatory agencies.

At monitoring wells MW-1 and MW-4, benzene appears to be degrading more rapidly than MTBE.  However, MTBE appears to be degrading more rapidly than benzene at monitoring well MW-6.  Downward benzene trends at downgradient wells MW-1, MW-4, and MW-6 indicate the source of benzene is decreasing in strength, and the dissolved plume has receded considerably. The disappearance of benzene at downgradient well MW-5 indicates the dissolved benzene plume no longer extends to the property boundary.   Downward trends in MTBE at source area monitoring wells MW-1, MW-3, MW-4 and MW-6 indicate the source of MTBE is also decreasing in strength.

ORP values measured during groundwater sampling activities suggest biodegradation is limiting the migration of dissolved benzene and MTBE.  There is a good correlation between low ORP values, relative to background values, and the presence of benzene and MTBE.  For example, monitoring well MW-1 typically has greater than 3,000 mg/L total benzene and MTBE, and has an average ORP of –38 millivolts (mV); while background well MW-2 typically has no benzene or MTBE and has an average ORP of 85 mV.  Average DO concentrations range from 2.6 to 3.8 milligrams per liter (mg/L) across the site and show little correlation to the presence of benzene or MTBE.

Site D

Site D was reported to regulatory authorities in July 1992 when a leak was detected in a premium grade gasoline transfer line.  No LNAPL has been detected in site monitoring wells. 

The shallow unconfined aquifer consists primarily of unconsolidated glacial deposits described as light brown sand and silt overlying brownish gray silty clay with a trace of sand.  The depth to the silty clay layer is approximately 5 feet bgs and the depth to groundwater is 4 to 6 feet bgs.  Thus, groundwater appears to migrate through the thin, more permeable sand and silt layer, and through the underlying silty clay material. The average hydraulic conductivity of the lower unit, based on slug tests performed in site monitoring wells, is estimated to be 5 x10-7 cm/sec and the horizontal hydraulic gradient across the site is approximately 0.01.  Assuming an effective porosity of 0.25, the average horizontal groundwater velocity in the lower unit is estimated be 0.02 [JR1] feet per year.  The average groundwater velocity in the upper sand and silt layer is expected to be several orders of magnitude greater, in the range of 0.3 to 30 feet per year.  The horizontal groundwater flow direction is to the southwest.  Based on the timing of the petroleum release, and the distribution of dissolved VOCs at Site D, it appears likely some benzene and MTBE migration is occurring in the more permeable sand and silt layer.

Four groundwater monitoring wells have been installed at Site D. Approximately 20 rounds of benzene and MTBE data were available from April 1994 to November 2000, however data prior to July 1997 exhibited seasonal trends.  Concentration data from after July 1997 appeared to exhibit minimal seasonal variation and were therefore utilized in the trend evaluations.

Concentrations of benzene compounds at all four site monitoring wells have decreased to near or below the analytical detection limit.  Thus, it appears that the plume of benzene has decreased in size, and no longer migrates off-site.  The detection of MTBE at downgradient monitoring well MW-4 suggest the plume of MTBE may still extend off-site, but the downward trend indicates the plume is receding.  The concentration of MTBE near the property boundary from the most recent round of sampling (November 2000) was relatively low (56 mg/L).  The rate at which MTBE is decreasing over time in individual monitoring wells at Site D are similar to those observed at Sites A and C.

In general, the background average DO concentration is high (5.5 mg/L at MW-1), while source area wells (MW-2 and MW-3) exhibit depressed average DO concentrations.  The average ORP at source area well MW-3 (-36 mV) is depressed relative to background well MW-1 (110 mV).  These trends provide good evidence for the aerobic biodegradation of dissolved petroleum hydrocarbons.  Downgradient well MW-4 has relatively high DO and ORP values, which are consistent with the depletion of dissolved petroleum hydrocarbons and the reoxygenation of the aquifer downgradient of the source area.  Based on the stoichiometric relationship describing the aerobic biodegradation of benzene and MTBE, 3.1 mg of oxygen are consumed during the biodegradation of 1 mg benzene and 2.75 mg of oxygen are consumed during the biodegradation of 1 mg MTBE. Based on the upgradient DO concentration (5.5 mg/L) and the source area benzene concentrations (< 1 mg/L), there is more than enough oxygen migrating into this site via groundwater migration for the aerobic biodegradation of the remaining dissolved benzene.  However, based on the stoichiometric relationship between DO and MTBE, and the concentrations of MTBE in the source area (2.2 mg/L), aerobic biodegradation of MTBE may be oxygen-limited.

Conclusions

Four petroleum retail marketing outlets with documented releases of petroleum hydrocarbons were evaluated to characterize the natural attenuation of benzene and MTBE.  Dissolved plumes were characterized as stable, expanding or shrinking based on the non-parametric Mann-Kendall trend evaluations. Datasets that did not exhibit upward or downward trends were further characterized as stable or non-stable trends with the coefficient of variation test.  The slopes of downward trends were quantified using Sen’s non-parametric indicator of median slope.

The trend evaluations provided a preliminary evaluation of the potential for natural attenuation to play a role in management of each of the four sites.  Trend evaluations at Sites A, C and D provided good evidence of decreasing sources and shrinking or stable benzene and MTBE plumes, while Site B showed evidence of a continuing source and steady state dissolved plumes of benzene and MTBE. Additional source control measures may be beneficial at Site B. 

At Site A, MTBE and benzene concentrations are declining at similar rates.  In addition, the downgradient extents of benzene and MTBE appear to be roughly equal at Site A.  The comparable behavior of the benzene and MTBE plumes at Site A suggests MTBE and benzene may be attenuating at similar rates.  At Sites C and D, the MTBE plume extends beyond the monitoring well network, but the MTBE plumes appear to be shrinking and concentrations at the property boundary are low.  The benzene plumes do not appear to extend past the property boundary at Sites C and D.  The benzene plumes appear to be attenuating more rapidly than the MTBE plumes at Sites C and D.

In general, average DO and ORP values were lowest where benzene and MTBE concentrations were highest, and higher average ORP and DO values were measured at locations with little or no benzene and low to moderate concentrations of MTBE. The DO, ORP, benzene and MTBE results are consistent with the consumption of oxygen for the biological degradation of gasoline constituents.

To reiterate, in situ natural attenuation processes include biodegradation, dispersion, dilution, adsorption, volatilization, and chemical reactions.  However, it should not be necessary to quantify the significance of individual natural attenuation processes in order to show natural attenuation mechanisms are at work.  Demonstration of decreasing contaminant concentrations across the site over time should be sufficient to demonstrate the efficacy of natural attenuation.

Recommendations

1. Monitoring well networks installed for purposes of plume delineation often are inadequate for purposes of thoroughly evaluating natural attenuation.  If preliminary investigations show natural attenuation will play an important role in site management, monitoring wells should be installed along the axis of the plume, extending from the source area to the downgradient edge of the dissolved plume.  Monitoring wells should be placed to provide three dimensional plume delineation.

2. All of the monitoring wells at the four sites considered in this paper were installed as water table wells, with screen intervals bridging the water table.  As a result, no information on vertical hydraulic gradients or vertical distribution of contaminants was available.  Downward hydraulic gradients could cause downward migration of the dissolved plume, which may not be detected with the conventional monitoring well network.   Installation of a shallow-deep well couplet at the downgradient property boundary could strengthen the case for natural attenuation by verifying that the plume is not merely diving beneath the monitoring well network.

3. Accurate measurement of DO in groundwater is useful for an understanding of biological degradation.

4. The role of biodegradation in reducing contaminant concentrations could be more fully understood by sampling and analyzing for alternate electron acceptors and other natural attenuation parameters including ferrous and ferric iron, sulfate, sulfide, nitrate/nitrite, carbon dioxide, methane and hydrogen.  Samples should be collected at several points in time, including a baseline round. Adequate detection limits and reliable results can usually be achieved at reasonable cost through the use of colorimetric field test kits for ferrous and ferric iron, sulfate, sulfide, nitrate and nitrite and carbon dioxide.

5. Demonstration of the production of breakdown products could bolster the case for natural attenuation. Since MTBE is known to biodegrade to tertiary-butyl alcohol (TBA) as an intermediate product, groundwater samples should also be analyzed for this compound.  TBA may also be present as an oxygenate or as a contaminant of the original MTBE.  Examination of MTBE/TBA ratios over time at individual monitoring wells could provide evidence of biodegradation.

6. Reliable estimates of hydraulic conductivity are useful to make reasonable estimates of biodegradation rates.  Residual drawdown tests as described by Driscoll (1995) can provide more reliable K estimates than slug tests because they test a larger portion of the aquifer.  Residual drawdown tests are not significantly more expensive than slug tests and are cheaper than full scale pump tests.

References

Driscoll, Fletcher G., 1995. Groundwater and Wells, Second Edition. U.S. Filter/Johnson Screens, St. Paul.

Gibbons, Robert D., 1994. Statistical Methods for Groundwater Monitoring.  John Wiley & Sons, Inc. New York.

U.S. EPA, 1999. Use of Monitored Natural Attenuation at Superfund, RCRA Corrective Action, and Underground Storage Tank Sites.  U.S. EPA Office of Solid Waste and Emergency Response Directive 9200.4-17P

Top