Vous êtes sur la page 1sur 16

8

M%8?%ii g&Re medIatIon n

Modeling Natural Attenuation of Total BTEX and Benzene Plumes with Different Kinetics
by Monica P. Suarez and Hanadi 5. Rifai

Abstract
Natural attenuation has emerged as a potential alternative for remediation of sites contaminated with fuel hydrocarbons. This paper compares the results from modeling the natural attenuation of BTEX (benzene, toluene, ethylbenzene, and xylene) at a coastal site to the results from a benzene model at the same site. Field data for total BTEX and benzene were used to develop model parameters for the Bioplume 111 site model. A first-order kinetics expression was used for benzene and an instantaneous expression for BTEX. Modeling results showed shorter cleanup timeframes for benzene than for BTEX. Natural attenuation Cleanup times using BTEX and assimilative capacity are 47% to 90% higher than those for benzene alone. Cleanup times for benzene of -100 years were estimated from model predictions, whereas predicted cleanup times for BTEX varied between 150 and 200 years.

Introduction
Natural attenuation has been studied extensively. Most case studies report on BTEX (benzene, toluene, ethylbenzene, and xylene) as a whole (Borden et al. 1994; Breedveld et al. 1999; Brown et al. 1997; Cho et al. 1997; Davis et al. 1999; Doyle et al. 1994; Klens et al. 1999; Lahvis et al. 1999; McLinn 1999; Wilson et al. 1986; Wilson et al. 1995; Wilson et al. 1994b; Yang et al. 1997). In addition, models have been developed (e.g., Bioplume I11 [Rifai et al. 19971 and BioScreen [Newell et al. 19963) which are used to simulate total BTEX. This is mainly because electron acceptors are hard to Proportion among BTEX components. However, these models may be used to simulate natural attenuation of a single compound, e.g., benzene, using first-order decay rates for the component, which have been extracted from field data from a site. Certainly, the fate and transport of fuel hydrocarbons in ground water can be simulated using either a lumped comPound or individual components. Nonetheless, both approaches present drawbacks that should be recognized. For example, modeling BTEX as a whole cannot account for selective or competitive biodegradation of the hydrocarbons. In addition, any model using instantaneousreaction is limited to when the microbial biodegradation kinetics are fast relative to the rate of ground water flow that mixes electron acceptors with dissolved contaminants. Finally, the model

Copyright 0 2004 by the National Ground Water Association.

using a lumped compound does not account for differences in mobility and tendency to sorb of the various compounds, as they are not modeled separately. On the other hand, when modeling individual components using first-order kinetics, site-specific information may not be accounted for, such as the availability of electron acceptors. In addition, first-order decay rates, which are determined in the laboratory, do not readily transfer to field situations. The model, furthermore, does not assume any biodegradationof dissolved constituents in the source zone. This paper compares the results from modeling the natural attenuation of BTEX at a coastal site to results from a benzene model at the same site. Field data for total BTEX and benzene were used to develop model parameters for the Bioplume 111 model (Rifai et al. 1997). The calibrated models were used to estimate cleanup times for total BTEX and benzene, and to determine the confidence interval in model results. A first-order kinetics expression was used for benzene and an instantaneous expression for BTEX. Field data were collected which, through modeling results, showed shorter cleanup timeframes for benzene than for BTEX. For these results, the benzene cleanup goal was set to the maximum contaminant level (MCL) in drinking water (i.e., 0.005 mgL). The BTEX cleanup goal, in turn, was set to 5 mg/L, as TEX compounds have much higher MCLs than benzene (1 mg/L, 0.7 mg/L, and 10 mg/L, respectively). An important consideration is that the site does not exhibit contamination with MTBE (methyl tertiary butyl ether). An MTBE plume would have required analysis as a separate compound because its behavior is different from BTEX and benzene in ground water.
53

Ground Water Monitoring L Remediation 24, no. 31 Summer 20041 pages 53-68

Site Description
The study site is an industrial facility in the United States located in a coastal area. The site is bounded on the south by the ocean, on the east by a river that discharges into a bay, and on the north and west by refineries. The facility produced ethylene, diethylene, and ethylene derivatives. Manufacturing activities at the facility started in the 1950s and ended in 1985, and its operational units have been decommissioned and dismantled. During this time, crude and refined BTEX were the only BTEX-containing materials handled at the site. Surface topography at the site slopes to the south. The conceptualized geology underlying this site can be divided into two main units: (1) an upper zone with alluvial deposits containing sand, silty sand, silty clay, and gravel; and (2) on the bottom, a Tertiary Age limestone formation. The thickness of the sand layers varies between 2.1 and 5.5 m. This is underlain by a silty clay unit of unknown thickness that appears to be limiting the extent of contamination to the more permeable upper layer. The water table is present between 1.0 and 5.5 m below ground surface and, in general, ground water flow is southward towards the sea. Ground water coming from the central part of the north site boundary travels toward the bay with slight deviations from a southerly direction. Ground water at the northeastern part of the site moves in the southwest direction toward the inlet canal (Figure 1). Results from falling head and constant head tests conducted in 1979 showed that the hydraulic conductivities of the silty and clayey sands, and to 2.3 x sandy silts with fine gravels, ranged from 3.3 x m/sec. In addition, available data suggest that there are not significant seasonal variations in ground water flow direction for the northern part of the site. Data further show that horizontal gradients vary between 0.0008 and 0.002 m/m for the dry and rainy season, respectively. Using a gradient of 0.002 m/m, an average hydraulic conductivity (geometric d s e c , and an assumed effective porosity mean) of 1.9 x of 0.2, the average seepage velocity is -1.9 x 10-7 m/sec. Contamination by petroleum hydrocarbons has been detected in the north and northeastern parts of the site over an area of -60 acres. Several on-site and off-site sources have contributed to BTEX contamination of the underlying soil and ground water. In 1977, a rupture in a pipe (Figure 1) that transported BTX to the adjacent northern facility caused a spill of -300,000 gallons of BTX, 70% of which was benzene. Pumping operations resulted in the recovery of 270,000 gallons of the spilled hydrocarbons; however, a significant number of wells still exhibit very high concentrations of benzene. In addition, observations from site personnel indicate that hydrocarbons were discharged into a creek that flows into the ocean (Figure 1). Cloudy and white water running into the creek and strong aromatic odors were reported several times during the period 1965-1975. Lastly, several oil spills into the creek were reported between 1989 and 1995. In addition, some wells in the northeastern part of the facility have shown the presence of light nonaqueous phase liquids (LNAPLs). These LNAPLs were found floating above the piezometric surface with thicknesses varying from 0.02 to 0.7 rn. Compositional analysis of the observed NAPL showed xylene as the main component with the exception of wells near the pipe rupture. Finally, an extensive NAPL
5 4

Figure 1. Site map and sampling locations.

plume present at the adjacent facility to the north (thickness up to 1.5 m according to 1996 data) likely contributes to hydrocarbon contamination in ground water.

Extent of Contamination
Dissolved Contamination Dissolved organic contamination has been monitored at the site since 1979. Historical data include BTEX concentrations from the existing monitoring wells collected in 1979, 1982, and 1996. Data for this study were collected in November 1998, March 1999, November 1999, May 2000, and November 2000, from -40 wells at the site (Figure 1). Data in Figures 2 and 3 show the extent of the BTEX and benzene plumes, respectively, for each of the sampling events. In 1979, benzene concentrations within the studied area varied from 1780 mg/L to < 1 mg/L across the site. Maximum concentrations of benzene were observed in wells D-1 1 and D-7 near the northern site boundary (Figure 2). Overall, benzene concentrations have declined with time so that in November 2000, the maximum measured concentration was 560 mg/L in well D-11. However, in November 1998 and March 1999, an increase in the maximum benzene concentration was observed likely due to a decline in water table levels. A similar general decline has been observed for the BTEX plume, where the maximum concentration observed in 1979 was 1893 mg/L and 570 mg/L in November 2000. Mapping of the BTEX and benzene plumes over time (Figures 2 and 3) shows that despite the observed concentration

M.P. SuurczondHS. Rifuil Ground Water Monitoring 8. Remediation 24, no. 3: 53-68

NO
/

150 300

600m

Notes: the labels on the graphs h d i i t e m o n m w e n bcation used tocontour the data.

concentralis in In@

Figure 2. BTEX contamination.

increases mentioned previously, plume extent has remained relatively constant since 1996. While plume lengths declined from 550 to 305 m between 1979 and 1996, they have not changed substantially since then. Additionally, the benzene a d BTEX plume extents are similar, which suggests parallel behavior of benzene and "EX compounds.

Observed Patterns of Intrinsic Bioremediation In March 1999 and May 2000, a number of geochemical parameters were measured at selected wells including dissolved oxygen, temperature, conductivity, redox potential, pH, alkalinity, ferrous iron, nitrate, sulfate, chloride, and methane. Overall, the geochemical parameters measured at the site showed similar patterns of natural attenuation as those observed at other sites (Barker and Mayfield 1988; Metzinger and Capps 1997; Troy et al. 1995; Wiedemeier et
55

M.P. SuarezandH.5 Rifai/ Ground Water Monitoring L Remediation 24, no. 3: 53-68

Notes: h labels on he graphs indicate e monitoringwell location used to contour the data.

Concentrationsin m$

Firmre 3. Benzene contamination. al. 1995; Wilson et al. 1994a; Wilson et al. 1995). That is, highly contaminated areas had depleted electrcn acceptors (dissolved oxygen and nitrate) and a high concentration of byproducts (ferrous iron and methane). Sulfate concentrations, however, did not follow the expected trend observed at many other sites. In general, these concentrations were very high in the most contaminated area in the north part of the site, the origin of which is unknown. Figure 4 includes the results of the March 1999 geochemical analysis.
56

Conceptual Model and Model Setup


Input parameters used for the Bioplume 111 site model are based on available site data and a review of the pertinent literature. Where site-specific data were not available, reasonable assumptions for the types of materials comprising the aquifer were made according to widely accepted literature values. The site was modeled using a grid size of 30 rows by 25 columns. Each grid cell was 76.2 m long by 76.2 m wide.

M.P. SuorezondHS. Rifoi/ Ground Water Monitoring 8, Remediation 24, no. 3: 53-68

Ben

Concentrations in

Figure 4. Site geochemistry in March 1999: (a) dissolved oxygen, (b) nitrate, (c) ferrous iron, and (d) methane.

The conceptual model and calibration parameters are detailed here and presented in Table 1.

Geology As mentioned previously, the geology underneath the site is complex. For modeling purposes, the contaminated zone

Table 1
Site Parameters for Model Calibration Parameter
Porosity Hydraulic conductivity Longitudinal dispersivity Ratio transverse/ longitudinal dispersivity Retardation factor Background dissolved oxygen concentration Dissolved oxygen utilization factor Background nitrate concentration Nitrate utilization factor Background ferric iron concentration Iron utilization factor Background sulfate concentration Sulfate utilization factor Background carbon dioxide concentration Carbon dioxide utilization factor First-order decay rate Simulation period

BTEX Plume (InstantaneousReaction)


0.2 3.38 X I@ to 2.30 X 10 d s e c 7.6 m 0.1 a,
1

Benzene Plume (First-OrderKinetics)


0.2 3.38 x 10-7t02.30 x lc3 s e c d 7.6 m 0.1 a, 1

8 mg/L 3.13 8.3 mg/L 4.85 50 mg/L 21.85 100 mg/L 4.7 50 mg/L 2.15

18 years

0.Ooouday 18 years

M.P. SuorezondM. Rifoi/ Ground Water Monitoring & Remediation 24, no. 3: 53-68

57

was conceptualized and modeled as a shallow, continuous, unconfined aquifer comprised of silty clay and sandy silt across the site. The modeled aquifer had a variable thickness ranging from 2.1 to 5.5 m.
Ground Water Flow The closest bodies of surface water that influence site conditions to varying degrees include the ocean (Figure l), two canals that transport sea water through the site, and a creek binding the eastern side of the site. The canals had a significant impact on ground water levels in the southern part of the site since water was pumped from one of these, the inlet canal. Additionally, seasonal fluctuations and tidal influences further complicate the site hydrology. Because of limited data and little historical information, the site hydrology was conceptualized as follows. 0 Ground water discharges to the inlet and outlet channels were simulated by introducing pumping wells at the northernmost points of the modeled canals. Pumping rates were varied until observed water levels were matched in the vicinity of the conceptualized canals. 0 Measured hydraulic conductivity values were used to develop a hydraulic conductivity distribution across the site. The values ranged between 3.35 x lW7 and 2.32 x lo-' d s e c . The distribution was generated by kriging. 0 Flow conditions at the site were calibrated to August 1996 data (Figure l), using ground water conditions generally typical of site flow conditions. The model was run under steady-state flow conditions. BTEX Plume Model For modeling purposes, the BTEX plume was conceptualized as follows. BTEX biodegradation was modeled using instantaneous reaction kinetics. The base year was 1978 (after the BTEX pipeline mpture). An initial dissolved plume was assumed in order to account for contaminant conditions prior to 1978. Sources of contamination (LNAPL dissolution and partitioning from soil) were represented using injection wells. These were placed at the points were LNAPL had been detected or where known spills had been reported. There were 1 1 cells within the grid identified as sources (Figure 5). Source concentrations and rates were assumed variable in time. The injection rates varied between 2.8 x 1W and 1.1 x m3/sec,and concentrations varied between 100 and 1800 mg/L. The modeled injection rates were sufficiently low as not to impact the water balance in the system. Initial concentrations of electron acceptors were estimated from field data collected in 1999 (Suarez and Rifai 2002). Benzene Plume Model The benzene model differed from BTEX in two key ways. 1. To model biodegradation, a first-order reaction was used. A biodegradation rate of 0.OOOUday (half-life of 10
58

years) was used to model the benzene plume (Suarez and Rifai 2002). 2. Seven wells were used to model sources (Figure 5). These wells correspond to the locations where hot benzene spots had been observed.

Model Calibration
The Bioplume I11 model for the site was calibrated by altering hydraulic parameters and sources (i.e., injection wells) in a trial-and-error fashion. This was done until simulated heads and plumes approximated observed field conditions.
Ground Water Level Calibration The water table was calibrated by varying the pumping rates at the canals and varying hydraulic conductivity (within the reported range) at locations where this parameter had not been measured. Observed and simulated water contours were compared to determine the goodness of fit. In addition, water levels for 27 wells were used to compare actual and modeled heads, and to estimate the calibration error. The calculated value of the root mean squared error (RMSE) is 0.19 m, which corresponds to 8.6% of the head drop across the site.

BTEX Plume Calibration Due to the presence of multiple sources at the site, 11 injection wells were used to simulate BTEX spills into the ground water (Figure 5 ) . Rates and concentrations in these wells were varied to simulate changes in source strength observed from the field data. BTEX concentrations measured

Figure 5. Model grid and injection well locations.

M.P. Suorez ond H.S. Rifoi/ Ground Water Monitoring 8 Remediation 24, no. 3: 53-68

Figure 6. BTEX plume in 1996: (a) observed and (b) modeled.

in May 1979 were input as the initial condition. Injection rates were set up at a maximum value of 1.1 x m3/secto minimize mounding effects. In particular, the concentration of BTEX injected through wells 1 and 2 (close to monitoring wells D-11 and D-7) was varied within the period 1978-1999 based on the observed Concentrations. Additionally, injection rates were decreased with time to account for changing dissolution rates as a result of LNAPL volume loss (and subsequent surface area decrease), and for the effect of Volatilization. Through a trial-and-error procedure, the concentrations in the injection wells were calibrated SO that the total dissolved mass in 1996 (existing mass plus injected mass minus mass lost via biodegradation) matched the total dissolved BTEX mass observed in 1996. The calibration error was determined in the same fashion as that for simulated heads, an RMSE of 17 mg/L (3.0%) being obtained for the predicted concentrations. Figure 6 shows a comparison of the observed and modeled BTEX plumes in 1996, showing that the calibrated plume is in reasonable agreement with the measured plume. The simulated concentrations are slightly higher than those measured in 1996,but plume extent was in good agreement with that measured in the field. The model was further run up to 1998 and the simulated concentrationswere compared to those measured in November 1998. In order to match the observed concentrations, source concentrations were changed to reflect observed patterns of source decay with time. First, the sources identified with the numbers 9, 10, and 11 in Figure 5 were removed.

Second, sources 1.2, and 4 were moved -76 m down in the direction of the ground water. Finally, new injection wells were added on the north border near wells H-2 and D 4 (sources 7 and 3, respectively). This was necessary to match very high concentrations observed in 1998 in these wells. The RMSE calculated using the predicted and observed concentrations for 1998 was 15.4 mg/L (1.8%). Figure 7, which compares the observed and modeled plumes in 1998, shows that the simulated conditionsrelate well to the contamination found. To verify the appropriateness of the selected parameters, the model was later run up to 1999 and the simulated concentrations compared to those measured in April and September 1999. For the wells in which both sampling rounds were performed, the observed value was assumed to be the average of the two measurements.Model validation using the 1999 data yielded an RMSE of 23 mgL (1.9%). which is within an acceptable range for this type of simulation. In addition, 1998 and 1999 concentrationsalong the centerline of the plume were compared to observed values as illustrated in Figure 8. These comparisons show that the calibrated model adequately represents site conditions.
Benzene Plume Calibration Seven injection wells were used to simulate benzene sources into the ground water (Figure 5). Benzene data from May 1979 were input as the initial benzene plume. As with the BTEX simulation, changes in injection rates within the period 1978-1999 correspond to the changes observed from
59

M.P. SuarezandH.S. Rifod Ground Water Monitoring 8, Remediation 24, no. 3: 53-68

Figure 7. BTEX plume in 1998: (a) observed and (b) modeled.

IOOO,

+Simulated

value Measuredvalue

100

0
0

50

100

IS0

200

250

300

350

400

Dbtance from Well D-ll (m)

Measured value

100

200

300

406

Dis(8oc.e from Well Dl1 (m)

Figure 8. BTEX concentrations along plume centerline: (a) 1998 and (b) 1999.
60

field data. The 1996 plume and dissolved mass were matched in a trial-and-error fashion. A plot of the concentrations generated using the aforementioned sources of contamination vs. the observed concentrations in 1996 shows that the simulated concentrations match the 1996 field observed values relatively well (Figure 9). A calibration error (measured as RMS) equal to 20 mg/L (2.3%) was obtained for the predicted concentrations. Figure 9 shows that these concentrations are generally higher than those measured in 1996, but the shape of the plume is similar to that observed. The area of the simulated plume is, however, greater than the plume dimensions measured in the field. This is due to difficulties in matching small concentrations measured in the leading edge using a small biodegradation rate (Le., half-life of 10 years). The model was further run up to 1998 and compared to concentrations measured in November 1998. As a result of this exercise, two new benzene sources were added to the model in the northeastern area (BE6 and BE7). The RMSE calculated using the predicted and observed concentrations for 1998 was 19.6 mg/L (2.3%). A comparison of the observed and modeled plumes in 1998 (Figure 10) shows that the extent of the calibrated plume is in reasonable agreement with the dimensions of the measured plume. As with the BTEX plume, the model was validated by matching simulated concentrations with those measured in April and September 1999. For the wells in which both sampling rounds were performed, the observed value was assumed to be the average of those measured. Model validation using the 1999 data yielded an RMSE of 13.2 mg/L (1.7%), which is within an acceptable range for this simulation. In addition, 1998 and 1999 concentrations along the

M.P.SuorczondH.S. Rifoi/ Ground Water Monitoring & Remediation 24, no. 3: 53-68

Figure 9. Benzene plume in 1996: (a) observed and (b) modeled.

LEGEND

SCALE IN METERS

9 BENZENE CONCENTWITION (&) UNE OF ECUAL ETEX CONCENTRATION ( g ) mk


(04j

Figure - 10. Benzene plume in 1998: (a) observed and (b) modeled.
M.P. Suorez and H.S. Rifai/ Ground Water Monitoring 8 Remediation 24, no. 3: 53-68
61

centerline of the plume were compared to observed values as illustrated in Figure 11. These comparisons showed that the calibrated model adequately represents site conditions.

Sensitivity Analysis o f Calibrated Model


Much uncertainty is associated with ground water contamination depiction as the majority of aquifer characteristics and contaminant concentrations are spatially variable. While it is true that adequate site characterization is required for a natural attenuation assessment (and its associated costs are justified when compared to the more traditional approach of installing expensive cleanup systems when they are not necessary), sufficient data needed to adequately characterize uncertainties in these systems is very difficult and costly to collect. In natural attenuation assessments, the impact of these uncertainties on plume geometry and concentrations is important, particularly with regard to potential risks associated with the contamination. Various methods are used to assess this uncertainty, e.g., stochastic modeling. Sensitivity analysis is also commonly used. The latter determines the effect of model input parameters on model output. For the calibrated models, reasonable ranges of variation for each parameter were determined using this approach. The model was run assuming the limits of such ranges, varying each parameter individually. Table 2 includes the parameters eval-

1000,

100 0 0

loo

200

300

400

Mitance from Well D-11 (m)

9001
800

1
(a)

uated in the sensitivity analysis, as well as the ranges within which they were varied. The variations in model output due to variations in input parameters were quantified using two different criteria. These were plume length (assuming 5 pg/L as the minimum concentration) and concentrations in two locations, i.e., source area (well D 1 1) and midplume point (150 m downgradient from source well). The values obtained for these two criteria were compared to the values for the base case (1996 calibration plume) to determine percentage of variation (Table 3). With respect to plume length, data in Table 3 indicate that this criterion was most sensitive to changes in the biodegradation parameters. These were, respectively, assimilative capacity and biodegradation rate for the instantaneous and first-order models. The maximum simulated BTEX plume length was 686 m for a run assuming no biodegradation. In contrast, the maximum benzene plume length was 1067 m for the higher hydraulic conductivity value. Additionally, concentrations at two different locations were evaluated for the aforementioned scenarios. The percentage of variation of the simulated concentrations was then plotted in Figures 12 and 13. As can be seen in Figure 12, for the location within the source area (well D-1 I), the model was most sensitive to source definition (concentration and injection rates) and hydraulic conductivity if assuming instantaneous kinetics. Conversely, the first-order model was most sensitive to biodegradation rate and source definition. For the BTEX plume specifically, the maximum concentration in the source area (12 12 mg/L) was obtained when the source concentrations were multiplied by a factor of two. On the other hand, the minimum concentration (61 mg/L) was reached when the hydraulic conductivities were assigned the upper limit value. For the benzene plume, on the other hand, the maximum concentration in the source area (1 194 mg/L) was obtained for the higher injection rate and the minimum (4 mg/L) for the higher biodegradation rate. Data in Figure 13 show the sensitivity analysis for concentrations in a midplume location. In this case, the parameter that caused the most impact on concentrations was biodegradation for both the BTEX and benzene models. The data presented in Table 3 and Figures 12 and I3 also indicate that longitudinal and transverse dispersivities do not have a significant impact on model results.

2 700

a 600 8
P

Model Predictions
Using the calibrated and verified models, future site conditions were predicted by running the Bioplume I11 model up to the year 2206. Two different scenarios were considered. 1. The injection rates and concentrations were decreased with time using a line of best fit of source data between 1978 and 1999. 2. The model was run assuming 20% annual reduction in LNAPL volume. For each scenario, six criteria were evaluated: maximum concentration, average concentration across the site, distance to downgradient edge from well D-l 1, total dissolved mass, plume length, and time to reach the cleanup goal. This was defined as the time required to achieve concentrations lower

I 500
300
200
100

P pO

0
0

50

100

I50 200 250 300 Distance from WeU D-11 (m)

350

400

Figure 11. Benzene concentrations along plume centerline: (a) 1998 and (b) 1999.
62

M.P.Suarezand H.S. Rifail Ground Water Monitoring L Remediation 24, no. 3: 53-68

Table 2
Variation of Parameters for Sensitivity Analysis
Parameter
Hydraulic conductivity

Range
1 x lo-* to 2 x dsec

Explanation
Representative values for silts and sands (Wiedemeier et al. 1999). All the hydraulic conductivity values were multiplied by a factor such that the maximum and minimum values were never outside the reported range. These multipliers were calculated to be 0.009 and 6 ; thus the model was run first for the lower limit (all values X 0.009) and later for the higher limit. Lower limit assumes no dispersion. Higher limit set to 0.1 plume length (Spitz and Moreno 1996).
X

Longitudinal dispersivity

0.003 to 30.48 m
0 to 0.33 ax

the

Ratio transverse1 longitudinal dispersivity Retardation factor

Lower limit assumes no transverse dispersion. Transverse dispersivity can also be estimated as 0.3301~ (ASTM 1995; US.EPA 1986); this estimate was used as the upper limit. For representative values of fraction of organic carbon for sands and silts within the range O.OOO5 to 0.007 (Wiedemeier et al. 1999). published values of Kw for the BTEX compounds, and an assumed bulk density of 1.9 g/cmJ (Freeze and Cherry 1979). the coefficient of retardation for the BTEX compounds ranged between 1.2 and 17. The lower limit assumed no biodegradation (absence of electron acceptors). The upper limit was determined assuming the maximum background concentrations reported in the reviewed literature as cited in (Suarez 2000). That is 10,40,500, IOOO, and 500 mg/L for oxygen, nitrate, ferric iron, sulfate, and carbon dioxide, respectively. The maximum reported biodegradation capacity via iron reduction and methanogenesis was multiplied by a factor of five since byproduct concentrations do not give a good estimate of assimilative capacity. The lower limit assumed no benzene biodegradation. The upper limit was determined as the 90th percentile of biodegradation rates obtained from 38 field sites and reported in Rifai et al. (unpublished). The injection rate was decreased by one order of magnitude for the lower limit. The upper limit was determined finding the maximum rate that did not cause disturbance of the ground water contours.

1 to 17

Biodegradation capacity (BTEX simulation)

0 to 480 mg/L

First order biodegradation rate (Benzene simulation)

0 to 0.0 I2lday

Source injection rate

0.1 to 5 (multiplier)

Source concentration

0.5 to 2 (multiplier)

than 5 mg/L in 95% of the cells for the BTEX plume and

0.005 mg/L for the benzene plume. The results of the analyses are subsequently discussed.

imum BTEX concentrations by the year 2206 will be 5.2 and 244 mg/L, respectively.
Alternative 2-Decaying Sources Assuming 20% Annual LNAPL Removal

BTEX Plume Predictions


Alternative I -Decaying Sources Using Regression Values

Results from model simulations show that the plume will decrease substantially as it moves southward toward the sea. For instance, model results indicate that by the year 2106 the Plume will be -381 m long and will have a maximum concentration of 336 mg/L. Results from the model predictions show that for this scenario the remediation goal will not be ) reached until the year 2206 (Table 4 .The average and max-

The LNAPL removal was simulated by decreasing both injection rate and concentration by 20% per year. Table 4 includes the results of model predictions. It can be seen that by the year 2206 the average BTEX concentration across the site will be 4.0 mg/L with a maximum concentration of 214 mg/L. For this alternative, a 50% reduction in total dissolved mass is expected to occur between the years 2006 and 2206. Model predictions show that remediation goals for this scenario will be achieved by the year 2 156.

M.P. SuorezondH.5. Rifoi/ Ground Water Monitoring 8, Remediation 24, no. 3: 53-68

63

Benzene Plume Predictions Alternative 1-Decaying Sources Using Regression Values

Results from model simulations (Table 4) show that once the sources are depleted (year 2096), the plume will shrink rapidly. By the year 2106, the benzene plume will have reduced its length by 60% and concentrations across the site will be zero. Overall, model simulations using a first-order biodegradation rate showed that the leading edge of the plume will not travel farther from where it was in 1999. Remediation goals for this alternative will be achieved by the year 2 I06 and the benzene plume will be completely depleted by the year 2 156.
Alternative 2-Decoying Sources Assuming 20% Annual LNAPL Removal

3
4
I
Figure 12. Maximum variations in simulated concentrations at the source area: (a) instantaneous model and (b) first-order decay model. Additionally, by the year 2106 the average BTEX concentration across the site would be at most 28.2 mgL, while the average benzene concentration would be 14.7 mgL. Finally, plume length predictions have a confidence of f 132 and 175 m for BTEX and benzene, respectively.

It can be seen in Table 4 that by the year 2156 the average benzene concentration across the site will be 0.01 mg/L with a maximum concentration of 1.O m a . For this alternative, a 99% reduction in total dissolved mass is expected to occur between the years 2006 and 2106. As for the previous scenario, cleanup goals for the benzene plume will be reached by the year 2106 and the plume will be depleted by the year 2 156.

Uncertainty Analysis for Model Predictions


Alternative 1, which assumes decaying sources, was selected to perform the uncertainty analysis for model predictions. Based on the sensitivity analysis results for the calibrated model, five parameters were evaluated for estimating variations of the predictive model. These parameters included injection rate, source concentration, hydraulic conductivity, retardation factor, and biodegradation capacity. To quantify model simulation uncertainty, a modification of the two-point technique (Yen and Guymon 1990) was used. For this kind of analysis, reasonable ranges of variation for the different parameters were determined (Table 2). Each variable was then assigned two values that corresponded to the upper and lower limits of such ranges. The two-point technique calls for the model to be run for all the possible permutations of the uncertain variables. Consequently, a total of 32 (F)simulations were run to establish a statistical population for the different output criteria. The 1996 calibrated model was used as the starting point for model predictions, and four criteria were evaluated in the year 2 106. These criteria include average plume concentration, distance traveled by the plume, plume length, and dissolved mass. The results of the uncertainty analysis associated with model assumptions are presented in Figure 14. As can be seen, the first-order model resulted in lower values of average concentration and dissolved mass than the instantaneous model. Regarding plume length and traveled distance, both models resulted in similar median and maximum values. Output data were further analyzed to calculate confidence intervals with a significance level of 95%. The results, as summarized in Table 5 , indicate that the longest distances from well D-I1 that the plumes would travel by the year 2 106 are 720 and 7 19 rn for BTEX and benzene, respectively.
64

Summary and Conclusions


The Bioplume 111 model was used to predict the fate and transport of BTEX and benzene plumes at the study site using two different kinetic expressions for biodegradation (instantaneous and first-order rate). It can be concluded that natural attenuation models can be used to simulate BTEX plumes using instantaneous biodegradation and one-component plumes using first-order kinetics. The input parameters the Bioplume 111 model was observed to be most sensitive to when using instantaneous kinetics are source definition, hydraulic conductivity, and assimilative capacity. For a first-order kinetics simulation, on the other hand, the model is most sensitive to changes in biodegradation rate and source definition. In general, the sensitivity analysis showed that simulated plumes using a first-order model for the site were longer than those obtained using an instantaneous reaction model. The sensitivity analysis for plume length showed that biodegradation is a key parameter for both scenarios, whether for instantaneous or first-order kinetics. However, for the first-order kinetics, the Bioplume 111 model was most sensitive to hydraulic conductivity when looking at plume length. Model predictions showed that if no source is removed, cleanup goals will be achieved by the years 2206 and 2106 for the BTEX and benzene plumes, respectively. Removing

M.P. Suorezond H.S.Rifoi/ Ground Water Monitoring 8, Remediation 24, no. 3: 53-68

Table 3
Summary of Sensitivity
Max Plume Plume Concentration Length Value (m@) (mto5ppb)' Concentration(m&) Mid D-11 D-16 Plume variation In Plume Length
A Concentration
to Base 'ase D-16 Plume

Variable Base Case Hydraulic conductivity Dispersivity Ratio aT/at Retardation factor Biodegradationcapacity Source injection rate Source concentration Average

D-11

1
*

i B
m
2 _

6K 0.009K 100 0.01 0.33 0 12 . 1 7 0 480 01 . 5 2c 0.5C

588 281 846 531 640 593 603 600 358 619 450 250 1043 1212 313 595 560

547 457 229 305 533 305 305 305 229 686 152 305 533 305 305 361 686
I067 686

588 61 732 516 638 593 603 600 312 619 420 67 01 4 1212 313 548 560

41 225 0 0 1 9 0 0 0 0 69 0 0 2 1 0 0 25 2 1
105 15

195 I72 108 267 98 266 280 267 1 6 165 0 245


160

0%
-50%

31 1 263 188 108 11 7 46 116 105 107 19 0 87 36 364 0


81 292

-33% 17% -33% -33% -33% -50% -50% -67% -33% 17% -33% -33% -30%

-90% 24% -12% 9%


1%

449% -100% -100% -54%


-100% -100% -10% -100%

3% 2% -47% 5% -29% -89% 60% 106% 47% -7%

68% -100% -100% 49% -100%


-10%

-42%

-12% 45% 37% -50% 42% 44% 37% -92% -15% -100% 26% -18% 59% 35% 4 %

Base Case Hydraulic conductivity

E E

6K 0.009K 100 00 .1 0.33 0

355 974 716 794 768 775 681 I 74 1101 219 117
1444

1% 462
476 598

56%
0%

-65% -18% -15% 7%

400%

-29% 14% -5% 0% 0%

58% -57% 7% -3%


-1% 1%

Dispersivity Ratio *aL Retardation factor Biodegradation rate Source injection rate
Source concentration

762 6% 762 762 762 686 762 152 762 838 762 762 726

24 20 2 1

11%
0%

0
cr:

555
563 526 174 960 4 77 1194 1116 282 516

21
20 15

g
0

11% 11%
11%

-1% 1%
-6%

I .2
I7

0%
11%

-69% 71% -99% -86% 113%


99%

-5% -29%
248%
-m 1

-19% -67% 237%


-100%

II

0 0.012 01 . 5 2c O X

73 0 20 2 5 2 1 2 1 28

-78%
11%

22%
11%
11%

-5% 19%
0%
0%

-25% 170% 14%


-6%

1545 387 707

123
11 0

-50%
-8%

Average

123

6%

36%

15%

*Maximum distance fmm wl D-11traveled by the leading plume edge el

Table 4
Model Predictions
2006
Parameter
~

2056
Alt 2 Alt 1 Alt 2 Alt 1

2106
Alt 2 Alt 1

2156
Alt 2
Alt 1
244.0

2206

Alt 1 BTEX Benzene BTEX Benzene BTEX Benzene BTEX Benzene BTEX Benzene

Alt 2

Maximum concentration(m@) Average concentration( m a ) Distance from D- I I to downgradient edge (m)

672.0 1079.0 1. 19 1. 01

527.0 974.0
11.1 1. 03

458.0 223.0 1. 05 17 . 533 457 29.470 3540 381 533 55.6 3

367.0 1. 10 86 . 04 . 457 38 1 24.580 700 305 381 52.I 2

336.0 70 . 78 . 01 . 686 533 25,200 220 381 38I 45.6 0

281.O
I .o

300 2.
1.O

261.0
1

.o

21 . 40

6.4 0.2 610 457 20.420 5 229 76

62 . 0 762 0 20980
10 305

50 . 00 .1
686 457 16.760 5 229 0 4.I 0

5.2

40 .

T t l dissolved mass (kg) oa

38 1 838 28,910 24,090 451 914


56.5 1 6

38I 838 27,250 2.9 530 381 914 57.1 1 7

838
17.580

762 13.790

0 0

305 29 .

0 20.1 0

229 0

Concentration > 95% of cells (m@)

BTEX Benzene

25.5 0

c
M.P. Suarez and H.I Rifai/ Ground Water Monitoring & Remediation 24, no. 3: 53-68

65

Table 5
Uncertainty Estimation for Model Predictions

I
~

Criterion Average concentration (mg/L) Distance traveled by plume' (m) Plume length (m) Dissolved mass (kg)
~ ~ ~

Plume BTEX Benzene BTEX Benzene BTEX Benzene BTEX Benzene

Mean 17.4 9.0 552 526 485 466 42,823 19,890

Variance 1.005 x 103 2.873 X lo2 2.446 3.299


X

Standard Deviation 31.7 16.9 495 574 386 512 8.45 x 104 3.45 X 10"

Confidence Interval (or = 0.05) 17.4 f 10.8 8.9 f 5.8 552 f 168 526 f 1% 485 f 132 466f 175 4.28 x 104i2.88 x 104 1.99 x 1@* 1.18 x 1@

105

105

1.487 X 105 2.623 X l(r 7.145 x 109 1.193 X lo9

'Maximum dismce f o well D-11traveled by the leading plume edge rm

LNAPL at an annual rate of 20% will shorten the remediation time for the BTEX plume by 50 years, but it will not have an impact on the benzene remediation time. While BTEX and benzene plume dimensions are similar, cleanup times may
differ because of maximum concentrations, cleanup goals, and calculated decay rate on a component-by-component basis. Natural attenuation cleanup times using BTEX and assimilative capacity are 47% to 90% higher than those for benzene alone. Finally, the uncertainty analysis for model predictions showed similar plume lengths and distances traveled by the leading edge for both the benzene and BTEX plumes.

Acknowledgments
The authors would like to acknowledge the funding provided by Union Carbide Inc. and the Gulf Coast Hazardous Substances Research Center for the completion of this work.

References
ASTM (American Society for Testing and Material). 1995. Emergency standard guide for risk-based corrective action applied at petroleum release sites, ASTM E-1739. Philadelphia, Pennsylvania: ASTM. Barker, J.F., and C.I. Mayfield. 1988. The persistence of aromatic hydrocarbons in various ground water environments. In Proceedings of the National Water Well Association and American Petroleum Institute Conference on Petroleum Hydrocarbons and Organic Chemicals in Ground Water: Prevention, Detection and Restoration, November 9-1 1. Houston, Texas, 649-667. Dublin, Ohio: National Water Well Association. Borden, R.C., C.A. Gomez, and M.T. Becker. 1994. Natural bioremediation of a gasoline spill. In Hydrocarbon Bioremediation, ed. R.E.Hinchee. Boca Raton, Florida: Lewis Publishers. Breedveld, G.D., M.Sparrevik, J. Aadnanes, andP. Aagaard. 1999. Natural attenuation of jet fuel contaminated run-off water in the unsaturated zone. In Natural Attenuation of Chlorinated Solvents, Petroleum Hydrocarbons, and Other Organic Compounds, ed. B.C. Alleman and A. Leeson. Columbus, Ohio: Battelle Press. Brown, K., P. Sekerka, M. Thomas, T. Perina, L. Tyner, and B. Sommer. 1997. Natural attenuation of jet-fuel impacted groundwater. In In Situ and On-Site Bioremediation, vol. 1. ed. B.C. Alleman and A. Leeson. Columbus, Ohio: Battelle Press. Cho, J.S., J.T. Wilson, D.C. DiGiulio, J.A. Vardy, and W. Choi. 1997. Implementation of natural attenuation at a JP-4 jet fuel release after active remediation. Biodegradation 8,265-273. Davis, G.B.. C. Barber, T.R. Power, J. Thiemn. B.M. Patterson, J.L.Rayner, and Q. Wu. 1999. The variability and intrinsic remediation of a BTEX plume in anaerobic sulphate-rich groundwater. Journal of Contaminant Hydrology 36, no. 3-4: 265-290. Doyle, G., D. Graves, and K.Brown. 1994. Natural attenuation of jet fuel in ground water. In Proceedings of rhe U.S.Environmental Protection Agency Symposium on Natural Attenuation of Ground Water, August 30-September 1, Denver Colorado. Washington, D.C.: U.S. Environmental Protection Agency.

Figure 13. Maximum variations in simulated concentrations at a midplume location: (a) instantaneous model and (b) firstorder decay model.

66

M.P. Suarcz ond H.S. Rifoil Ground Water Monitoring 8, Remediation 24, no. 3: 53-68

1.OE+06

T
1.OE+OS 1.OE+O4
rb
0
Ll .

n
L

9
0

k h
r

1.OE+03

1.OE+02
I

9
&

!l.OE+OI
1.OE+OO
1.OE-0 1

1.OE-02 Average Concentration bgn) Plume length ( 4 Distance traveled (m> Dissolved M s as (kg)

Figure 14. Summary of uncertainty analysis for model predictions.

Freeze. R.A., and J.A. Cherry. 1979. Groundwater. E n g l e w d Cliffs, New Jersey: Prentice-Hall. Klens, J., G. Roberts, and D. Graves. 1999. Rapid qualification of sites for natural attenuation potential. In Natural Attenuation of Chlorinated Solvents, Petroleum Hydrocarbons, and Other Organic Compounds,ed. B.C. Alleman and A. Leeson. Columbus, Ohio: Battelle Press. Lahvis, M.A., A.L. Baehr, and R.J. Baker. 1999. Quantificationof aerobic biodegradation and volatilization rates of gasoline hydrocarbons near the water table under natural attenuation conditions. Water Resources Research 35, no. 3: 753-765. McLinn, E. 1999. Biodegradation of gasoline constituentsin a fractured aquitard: A field study. In Natural Attenuation of Chlorinated Solvents, Petroleum Hydrocarbons, and Other Organic Compounds,ed. B.C. Alleman and A. Leeson. Columbus,Ohio: Battelle Press. Metzinger. C.S., and M. Capps. 1997. Spatial variations in intrinsic bioremediation processes. In In Situ and On-Site Bioremediation, vol. 1, ed. B.C. Alleman and A. Leeson. Columbus, Ohio: Battelle Press. Newell, C.J., R.K. McLeod. and J.R. Gonzales. 1996. BIOSCREW: Natural Attenuation Decision Support System User's Manual, Version 1.3, EPMNWl-96J087. Ada, Oklahoma: Robert S.Kerr Environmental Research Center. Rifai, H.S., C.J. Newell. J.R. Gonzales. S. Dendrou, L. Kennedy, and J.T. Wilson. 1997. BIOPLUME 111: Natural Attenuation Decision Support System, Version 1.0, User's Manual. Prepared for the U.S. Ar Force Center for Environmental Exceli lence, Brooks AFB, San Antonio, Texas. Rifai, H.S.. M.P. Suarez, and C.J. Newell. Unpublished. Estimating first-order decay rates for petroleum hydrocarbon attenuation. spitz, K., and J. Moreno. 1996. A Practical Guide to Groundwater and Solute Transport Modeling. New York Wiley.

Suarez. M.P. 2000. Assessing intrinsic remediation of BTEX compounds at a coastal industrial facility. M S thesis. Civil and .. Environmental Engineering Department, University of Houston,Texas. Suarez, M.P., and H.S. Rifai. 2002. Evaluation of BTEX remediation by natural attenuation at a coastal facility. Ground Water Monitoring & Remediation 22, no. 1: 62-77. Troy, M.A., K.H.Baker, and D.S. Herson. 1995. Evaluating natural attenuation of petroleum hydrocarbon spills. In Intrinsic Bioremediation, ed. R.E. Hinchee, J.T. Wilson, and D.C. Downey. Columbus, Ohio: Battelle Press. U.S.EPA (Environmental Protection Agency). 1986. Background document for the ground-water screening procedure to support 40 CFR part 269: Land disposal, EPN530-SW-8647. Washington, D.C.: U.S.Environmental Protection Agency. Wiedemeier, T.H., H.S. Rifai, C.J. Newell. and J.T. Wilson. 1999. Natural Attenuation of Fuels and Chlorinated Solvents in the Subsurface. New York: John Wiley & Sons. Wiedemeier, T.H., M.A. Swanson, J.T. Wilson, D.H. Kampbell. R.N. Miller, and J.E. Hansen. 1995. Patterns of intrinsic bioremediation at two US. Air Force bases. In Intrinsic Bioremediation. ed. R.E. Hinchee, J.T. Wilson, and D.C. Downey. Columbus, Ohio: Battelle Press. Wilson, B.H., B.E. Bledsoe. J.M. Armstrong, and J.H. Sammons. 1986. Biological fate of hydrocarbons at an aviation gasoline spill site. In Proceedings of the National Water Well AssociatiodAmerican Petroleum Institute Conference on Petroleum Hydrocarbons and Organic Chemicals in Ground Water-Prevention, Detection and Restoration. November 12-14. Houston, Texas, 78-90. Dublin, Ohio: National Water Well Association. Wilson. B.H., J.T. Wilson, D.H. Kampbell. B.E. Bledsoe, and J.M. Armstrong. 1994a. Traverse City: Geochemistry and intrinsic bioremediation of BTX compounds. In Proceedings of the U.S. Environmental Protection Agency Syntposium on Natural
67

M.P.Suorezond H.S. RifoJ Ground Water Monitoring & Remediation 24, no. 3: 53-68

Attenuation of Ground Water, August 30-September 1, Denver, Colorado, 94-102. Washington, D.C.: U.S. Environmental Protection Agency. Wilson, J.J., G.Sewell, D. Caron, G.Doyle, and R.N. Miller. 1995. Intrinsic bioremediation of jet fuel contamination at George Air Force Base. In Intrinsic Bioremediation,ed. R.E. Hinchee, J.T. Wilson. and D.C. Downey. Columbus, Ohio: Battelle Press. Wilson, J.T., D.H. Kampbell, and J. Armstrong. 1994b. Natural bioreclamation of alkylbenzenes (BTEX) from a gasoline spill in methanogenic groundwater. In Hydrocarbon Bioremediation, ed. R.E. Hinchee. Boca Raton, Florida: Lewis Publishers. Yang, X.,H.Glasser, R. Stoelting, M. Barden, G. Mickelson, J. Delwiche, and G. Alvarez. 1997. Natural attenuation demonstration in Wisconsin. In In Situ and On-Site Bioremediation, vol. 1, ed. B.C. Alleman and A. Leeson. Columbus, Ohio: Battelle Press. Yen, C.C., and G.L. Guymon. 1990. An efficient deterministicprobabilistic approach to modeling regional groundwater flow, 1. Theory. Water Resources Research 26, no. 7: 1559-1567.

Biogra phicaI Sketches


Monica P Suarez is a researcher in civil and environmental . engineering at the University of Houston. She holds an M.S. in environmental engineeringfrom the University of Houston and has six years of experience in waste water treatment and sustainable development. Her current research focuses on understanding natural attenuation processes at the field scale. She may be reached at the University of Houston, 4800 Calhoun Rd., Room N107D, Houston, Tx 772044003; (713) 743-0753; f a x (713) 743-4260; monica.suarez @mail.uh.edu. Hanadi S. Rifait corresponding author, is an associate professor in civil and environmental engineering at the University of Houston. Her research effortsfocus on contaminantfate and transport modeling, and remediation and natural attenuation. She has coauthored two textbooks: Ground Water Contamination: Transport and Remediation, published by Prentice-Hall in 1994 and 1999, and Natural Attenuation of Fuels and Chlorinated Solvents in the Subsurface, published by McGraw Hill in 1999. She is the editor-in-chief of Bioremediation Journal and a member of the US. Environmental Protection Agency Science Advisory Board Environmental Engineering Committee (FYOO) Natural Attenuation Subcommittee, 2000. Raip may be reached at the University of Houston, 4800 Calhoun Rd., Room N107D, Houston, ITX 772044003; (713) 743 4271;far (713) 743 4260; rifai@uh.edu.

nternational Center for Ground Water & Public Policy


For public health and natural resources officials of the world

A /
\

info.ngwa.org/icgwpp
601 Dempsey Road, Westerville, Ohio 43081 U.S.A. 614 898.7791 or fax to 614 898.7786

68

M.P. Suum und H.S. Rifoi/ Ground Water Monitoring & Remediation 24, no. 3: 53-68

Vous aimerez peut-être aussi