Development of a Coupled Reactor Model for Prediction of Organic

Aug 28, 2008 - AND. MORTON A. BARLAZ §. Risk and Environmental Modeling Program, Environmental. Health and Safety Division, RTI International, 3040...
0 downloads 0 Views 3MB Size
Environ. Sci. Technol. 2008, 42, 7444–7451

Development of a Coupled Reactor Model for Prediction of Organic Contaminant Fate in Landfills MICHAEL I. LOWRY,† S H A N N O N L . B A R T E L T - H U N T , * ,‡ STEPHEN M. BEAULIEU,† AND MORTON A. BARLAZ§ Risk and Environmental Modeling Program, Environmental Health and Safety Division, RTI International, 3040 Cornwallis Road, Research Triangle Park, North Carolina 27709, University of Nebraska-Lincoln, 203B Peter Kiewit Institute, Omaha, Nebraska 68182-0178, Department of Civil, Construction and Environmental Engineering, North Carolina State University, Box 7908, Raleigh, North Carolina 27695-7908

Received April 1, 2008. Revised manuscript received June 9, 2008. Accepted July 22, 2008.

Models describing the behavior of organic chemicals in landfills can be useful to predict their fate and transport and also to generate input data for estimates of exposure and risk. The landfill coupled-reactor (LFCR) model developed in this work simulates a landfill as a series of fully mixed reactors, each representing a daily volume of waste. The LFCR model is a numerical model allowing time-variable input parameters such as gas generation, and cover type and thickness. The model was applied to three volatile organic chemicals (acetone, toluene, benzene) as well as naphthalene and the chemical warfare agent sarin under three landfill conditions (conventional, arid, bioreactor). Sarin was rapidly hydrolyzed, whereas naphthalene was largely associated with the landfill solid phase in all scenarios. Although similar biodegradation rates were used for acetone and toluene, toluene was more persistent in the landfill due to its hydrophobicity. The cover soil moisture content had a significant impact on gaseous diffusive losses.

Introduction The occurrence of organic compounds in municipal solid waste (MSW) landfills is apparent from their concentrations in leachate and landfill gas (1-3). Trace organic compounds may enter landfills in household hazardous waste (HHW) (e.g., acetone from nail polish remover, toluene in paint remover), in wastewater treatment residuals, and via disposal of industrial wastes not regulated as hazardous wastes (e.g., solvent-contaminated industrial wipes). In California, HHW is estimated to comprise 0.2% (2000 mg kg-1) of landfilled waste (4). The fate and transport of trace organic compounds in landfills is complex given waste heterogeneity and the combined influences of sorption, volatilization, leaching, and biological and abiotic transformation. While landfills represent a potential reservoir for many organic contaminants, * Corresponding author phone: 402-554-3868; fax: 402-554-3288; e-mail: [email protected]. † Risk and Environmental Modeling Program, Environmental Health and Safety Division, RTI International. ‡ University of Nebraska-Lincoln. § North Carolina State University. 7444

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 19, 2008

there have been limited studies investigating organic contaminant fate. Predictive landfill models can be useful to assess the fate and transport of specific organics and provide input data for models that assess the potential risk associated with the disposal of specific compounds. This paper describes a landfill coupled-reactor (LFCR) model designed to support large-scale risk assessment, whereby the landfill represents a source for chemical release to the environment. In a typical risk assessment, fate and transport models (e.g., groundwater and air) predict contaminant concentrations at potential receptors, and exposure and risk models estimate the potential risk. Historically, many landfill contaminant transport models have focused on gas emissions through cover systems (5-7), mass loss in leachate (8), and/or leakage through liner systems (9). There have been relatively few models that integrate contaminant fate and transport in multiple phases within the entire landfill. Models that have been developed often require extensive and complex parametrization (10) or provide only qualitative assessments of fate routes (11). The model for organic chemicals in landfills (MOCLA) predicts the distribution of organic chemicals between leachate, gas, and solid waste fractions as well as the removal of organic chemicals via leachate and gas (12). MOCLA predicts contaminant fate based on a relatively small number of input parameters. However, MOCLA represents the entire landfill as a homogeneous, fully mixed reactor with all waste placed in the landfill at one time. In addition, MOCLA does not consider transient gas or liquid flow conditions as leachate and gas production rates are constant throughout the simulation. The LFCR model extends the basic MOCLA approach by simulating the landfill as a series of fully mixed reactors, each representing a daily volume of waste. In contrast to MOCLA, the LFCR model is a numerical model allowing transient conditions (e.g., gas and liquid flow rates). In this paper, the LFCR model was used to analyze the fate of three volatile organic solvents (acetone, benzene, and toluene) commonly present in leachate and gas. An example waste disposed in landfills containing these solvents is disposable wipes generated in various manufacturing industries (e.g., printing, automobile maintenance) (13). In addition, the model was used to evaluate compounds with varying properties, including naphthalene (nonvolatile, hydrophobic) and sarin (a chemical warfare agent that undergoes rapid hydrolysis). A comparison of LFCR and MOCLA model results is provided to illustrate differences in predicted behavior.

Modeling Approach and Data Development The LFCR model describes transport between daily cells in a landfill as transport between fully mixed reactors coupled in series (Figure 1). Theoretical retention time curves for fully mixed reactors provide a reasonably good fit with experimental landfill tracer data, suggesting that modeling a daily cell as a fully mixed reactor is appropriate (14). The model accounts for contaminant losses for each daily cell via diffusion both through uncovered waste and through cover and liner materials, advective losses in gas and leachate, and biotic and abiotic degradation. Diffusive transport occurs vertically between daily cells, to the atmosphere and to groundwater through the landfill liner system. Diffusive mass transport through daily, intermediate and final covers is considered. The model allows 10.1021/es800907j CCC: $40.75

 2008 American Chemical Society

Published on Web 08/28/2008

where Vcell and Ci are the daily cell i volume and total contaminant concentration, respectively, and Sa,i, Sw,i, Sλ,i, and SD,i are the vapor advection, aqueous advection, degradution, and diffusion sink terms, respectively. As described in the Supporting Information (SI), the total volumetric concentration (the dependent variable) includes the mass in solid, aqueous, and vapor phases. Retardation coefficients represent the equilibrium partitioning of chemical mass between phases and relate the total concentration to phasespecific concentrations. For example, the air-phase concentration is equal to the total concentration divided by the air-phase retardation coefficient. Each volume-normalized source/sink term is presented below along with the final form of the equation used in the model solution. The SI provides additional model formulation details. Each of the source/sink terms is expressed in terms of coefficients of the total concentration within the current (i), overlying (i + 1), and underlying daily cell (i - 1). The contaminant degradation term accounts for potential abiotic and biotic first-order degradation in the air, water, and solid phases:

(

)

εa(λa1 + λa2) εw(λw1 + λw2) Fb(λs1 + λs2) Sλ,i ) + + Ci (2) Vcell Ra Rw Rs where εa and εw are the air-and water-phase volumetric fractions (L3 L-3), respectively; F is the solid-phase bulk density (M/L-3); λa1, λw1, and λs1 are the first-order biotic contaminant degradation rates (T-1); λa2, λw2, and λs2 are the first-order abiotic contaminant degradation rates (T-1); Ra, Rw, and Rs are the retardation coefficients; and all other terms are as defined previously (subscripts a, w, and s refer to the air-, water-, and solid-phases, respectively). Diffusion through cover and liner systems is described using Fick’s law, whereby the diffusive flux is proportional to the concentration gradient across the layer and inversely proportional to the thickness. The total diffusion term considers diffusion between the underlying and overlying daily cells in the water and air phases:

FIGURE 1. Column of coupled reactors (horizontal fill pattern) with cover and liner layers (vertical fill pattern). Ci is concentration; Sλi is the degradation sink term; Qai is the air flow rate; Qw is the water flow rate; JQw and JQw are the flux in flowing air and water, respectively; and JD is the diffusive flux in air and water. the liner and final cover to be soil-only layers or composite layers consisting of a geomembrane and soil. Compositelayer diffusion depends on a permeation coefficient accounting for geomembrane sorption effects (15). A composite clay plus geomembrane liner, as required by U.S. regulations, is assumed for all simulations (16). The model also describes diffusive transport prior to placement of daily cover. The associated sink term is based on an analytical solution to the diffusion equation applied to the active daily cell for the user-specified fraction of a day with no cover. Advective transport occurs in gas generated by waste biodegradation and within leachate percolating through the landfill. Gas is generated at an exponentially decaying rate as in the LandGEM model (17). Aqueous advective transport occurs in leachate at user-specified infiltration rates that vary with placement of final cover. A user-specified liner leakage rate describes advective mass loss through liner leakage. The LFCR formulation starts with a mass balance for daily cell i: d(VcellCi) ) Sa,i + Sw,i + Sλ,i + SD,i dt

(1)

(

[(

) ( ) ] [( [(

) ] [( ) ( ) (

) )] ) ]

Da Da Da C C CLiHcellRa i+1 LiHcellRa i Li-1HcellRa i Da Dw Dw C + C C Li-1HcellRa i-1 LiHcellRw i+1 LiHcellRw i Dw Dw Ci C (3) Li-1HcellRw Li-1HcellRw i-1

SD,i ) Vcell

where Da and Dw are the air- and water-phase diffusion coefficients, respectively (L2 T-1); Li is the thickness of the overlying cover layer (L); Hcell is the daily cell height (L); and all other terms are as defined previously. In the case where there is no top cover, diffusive losses were estimated using an analytical solution as described in the SI. Vapor advection occurs in an upward vertical direction and the vapor advection term includes advection from the underlying cell (Ci-1 term) as well as advection out of the current cell (Ci term): i-1

Sa,i ) Vcell



i

qa,j

j)1

Ra

Ci-1 -

∑q j)1

Ra

a,j

Ci

(4)

where qa,j is the gas production rate per waste volume (L3 gas L-3 waste T-1) and all other terms are as defined previously. Aqueous advection occurs in a downward vertical direction. The aqueous advection term includes downward advection from the overlying daily cell (Ci+1 term) as well as advection out of the current cell.: VOL. 42, NO. 19, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7445

FIGURE 2. Simulated waste addition pattern, representing the landfill cell filled to the fourth lift, fourth row, third daily cell.

(

)

(

)

Sw,i N N ) C C Vcell HcellRw i+1 HcellRw i

(5)

where N is the net infiltration rate (LT-1); and all other terms are as defined previously. Using the above source/sink terms, the mass balance can be written as the change in total concentration as a function of the concentrations in the current (Ci), overlying (Ci+1), and underlying (Ci-1) daily cells. dCi ) k1Ci-1 + k2Ci + k3Ci+1 dt

(6)

where:

( )( [( ) ( ) i-1

k1 )

∑q j)1

Ra

a,j

+

)(

Da Dw + Li-1HcellRa Li-1HcellRw

)

(6a)

values. SI Table S2 provides the physical-chemical properties of the modeled compounds. To assess the impact of temperature on chemical fate, selected physical-chemical properties of the compounds were adjusted for temperature within the LFCR model as described in the SI. Simulations were run for a 62 year period, which corresponds to 32 years of operation plus a 30 year postclosure monitoring period as required by current U.S. regulations (16). Simulations were performed for (1) a base-case scenario representing “typical” conditions in the U.S., (2) an arid climate scenario, and (3) a bioreactor operation scenario in which the rate of refuse decomposition is enhanced. LFCR results for the 62 year simulation period were compared to MOCLA to evaluate the significance of the more sophisticated LFCR approach. MOCLA simulations were for a 62 year period using the same input data as presented in Table 1 and SI Table S2; however, constant gas generation and infiltration rates were used in MOCLA as required by the model. The constant gas generation rate was calculated in MOCLA using the gas decay rate and the potential methane generation capacity for each scenario (Table 1) as input data to LandGEM (17), and assumes that all waste is buried at the beginning of the simulation. Resulting gas generation rates were 1.9, 3.8, and 10.7 m3 gas m-3 yearr1 for the arid, base case, and bioreactor scenarios, respectively. While LFCR allows net infiltration to change after final cover installation, MOCLA requires a constant value. Therefore, the average infiltration values during the 62 year LFCR simulation were used in MOCLA (1.1 × 10-3 m year-1 for the arid scenario; 1.9 × 10-2 m year-1 for the base case, and 4.4 × 10-2 m year-1 for the bioreactor scenario).

i

∑q

Results and Discussion

a,j

N + + k2 ) Ra HcellRw εa(λa1 + λa2) εw(λw1 + λw2) Fb(λs1 + λs2) + + + Ra Rw Rs j)1

(

(

)(

) ( )( ) ( ) (

Da Da Dw + + LiHcellRa Li-1HcellRa LiHcellRw k3 )

(

)

Dw Li-1HcellRw

Da Dw N + + HcellRw LiHcellRa LiHcellRw

)

]

)

(6b) (6c)

Equation 6 is the form used within the numerical solution. For a given column of n daily cells, this formulation generates a system of n ordinary differential equations coupled through the transport between daily cells. The model was implemented in C++ using a Runge-Kutte numerical solution with a user-specified truncation error tolerance of 10-11, which ensured that mass balance error remained below 1% in all simulations. In a typical MSW landfill, waste is added in discrete units (daily cells and lifts) to minimize the active area. Operators cover the waste daily with native soil or alternative materials. The LFCR model simulates these practices through the waste addition algorithm illustrated in Figure 2. Input parameters to the waste addition algorithm include landfill volume, surface area, operating life, and the number of landfill cells (an active operational area including many daily cells). The SI provides additional detail about the waste addition algorithm. The algorithm generates adjacent, independent columns of daily cells, with each column generating mass loss to the atmosphere, leachate, and the underlying groundwater. The total landfill mass loss is the sum of mass loss from all columns. Table 1 presents input parameters describing the simulated landfills. The SI gives additional rationale for parameter 7446

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 19, 2008

Figure 3 presents the predicted fate routes for the target chemicals under the base case, arid, and bioreactor scenarios for the 62 year simulation period. SI Table S3 provides the values for the cumulative mass fractions lost. The leachate mass fraction represents the contaminant mass leaving the landfill with flowing water. It is assumed that 1% of leachate is released to the environment through liner leakage over the simulation period; the remaining leachate would be recovered by a leachate collection system (18). In the base case, acetone and sarin were predominantly lost via biodegradation and abiotic transformation, respectively. As acetone is soluble and biodegradable, its transformation is as expected. Similarly, the short half-life of sarin suggests rapid transformation. Although the biological half-life of toluene is similar to acetone, much more remains in the landfill due to its relatively greater partitioning to the refuse. The dominant fate pathway for benzene and toluene was gas-phase advection. In contrast, 98% of the initial naphthalene, which is less volatile than toluene and benzene, is predicted to remain in the landfill at the end after 62 years. The mass of acetone, toluene, and benzene remaining in the landfill after 62 years is lower in both the arid and bioreactor cases compared to the base case, although the dominant fate pathways for each contaminant under the arid and bioreactor scenarios are quite different. In the arid scenario, the dominant fate pathway for acetone, toluene and benzene was gaseous diffusion through the cover. The increase in mass lost via diffusion under the arid scenario can be attributed to the reduced moisture content of the cover materials. The moisture content for the base case was double that of the arid case (Table 1); however, the diffusive losses in the arid case exceeded the base case by more than an order of magnitude. This sensitivity to the moisture content can be explained by the tortuosity. The tortuosity adjusts the diffusivity to account for the tortuous transport paths within the porous medium (19). The arid case tortuosity is 0.047,

TABLE 1. LFCR Input Parameters Describing Landfill Characteristics parameter

units

arid

base case

bioreactor

source

4.5 × 24.4 1.1 × 107 32 4

4.5 × 24.4 1.1 × 107 32 4

4.5 × 24.4 1.1 × 107 32 4

assumed value 24 calculated value assumed value assumed value

0.4

0.4

0.4

25

m3 H2O m-3 total vol.

0.17

0.34

0.34

calculated value

m year-1

3.35 × 10-3

6.70 × 10-2

19.5 × 10-2

24

m year-1

1.5 × 10-4

4.1 × 10-4

4.1 × 10-4

liner leakage rate

m year-1

1.5 × 10-6

4.1 × 10-6

4.1 × 10-6

average waste temperature daily cell height landfill gas decay rate volumetric moisture content of waste organic carbon fraction in waste waste porosity potential methane generation capacity time before final cover installed on landfill cell waste bulk density (dry) working face angle to horizontal daily waste sublayer period time without cover on active daily cell thickness (daily cover) thickness (final cover and liner) geomembrane thickness thickness (interim cover)

°C m year-1

37 3.05 0.02

37 3.05 0.04

42 3.05 0.1

18 equivalent to leachate collection efficiency (18) 26, 27 assumed value 28

m3 H2O m-3 total vol.

0.12

0.15

0.33

calculated value

0.5

0.5

0.5

25

0.4

0.4

0.4

29

100

100

100

16

10

10

10

assumed value

landfill area landfill depth landfill volume landfill operating lifea landfill cell operating lifea porosity, daily/final cover and liner volumetric moisture content, daily/ and final cover and liner infiltration rate before final cover infiltration rate after final cover

m2

105

m m3 year year

L

kg-1

year kg

L-1

105

105

0.49

0.49

0.49

calculated value

deg

20

20

20

30

hr

1

1

1

assumed value

day

0.17

0.17

0.17

assumed value

m m m m

0.152 0.61 1.5 × 10-3 0.305

0.152 0.61 1.5 × 10-3 0.305

0.152 0.61 1.5 × 10-3 0.305

subtitle subtitle subtitle subtitle

d d d d

reg reg reg reg

(16) (16) (16) (16)

a The landfill operating life must be evenly divisible by the landfill cell life, where a landfill cell is a fraction of the landfill area. Landfills are constructed and filled sequentially in cells over their operating life.

FIGURE 3. Predicted fate of acetone, sarin, naphthalene, toluene, and benzene under three scenarios: (A) base case, (B) arid climate; and (C) bioreactor operation over 62 years. Aqueous degradation includes both biological and abiotic processes. whereas the base case tortuosity is 5.3 × 10-4. This strong dependence of the tortuosity on moisture content explains the dominance of diffusion in the arid case for acetone, toluene, and benzene versus base case predictions. SI Figure

S1 provides additional information on tortuosity and its sensitivity to moisture content. Under the bioreactor scenario, biodegradation was still the dominant fate route for acetone. Gas phase advection VOL. 42, NO. 19, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7447

FIGURE 4. Fate of benzene as a function of time under the base case scenario. increased relative to the base case and remained the dominant fate pathway for toluene and benzene. The increase in gas phase advection is attributed to increased gas production in the bioreactor scenario and the relatively high volatility and low solubility of toluene and benzene. The mass loss results for naphthalene and sarin were relatively insensitive to the operating conditions. Given the large degree of uncertainty, biodegradation rates were not varied between the three scenarios. Thus, biodegradation did not increase for the bioreactor scenario. Additional biodegradation of trace organics would generally be expected in bioreactors as the microbial population is stimulated. The additional water in a bioreactor would also result in increased dissolution and bioavailability. However, there are only limited data comparing the behavior of toluene and organic arsenic compounds in conventional and bioreactor landfills (20, 21). The LFCR model describes diffusion in the open cell prior to placement of daily, intermediate, or final cover and diffusion that occurs after cover placement. Diffusion losses from the open cell were highly variable, but only accounted for a maximum of 3% of the total mass loss under any scenario. Open cell diffusion was the dominant contaminant fate pathway for the active daily cell during the time that the waste was uncovered; however, the time was short relative to the total simulation, and the mass of uncovered refuse was small relative to the entire landfill. Figure 4 presents the predicted fate routes for benzene over time under the base case scenario. SI Table S4 provides the corresponding results for all modeled chemicals. With the exception of sarin, contaminant masses in the landfill increase through year 30, because waste deposition continues for 32 year. Vapor advective losses dominate for benzene and toluene while biodegradation dominates for acetone 7448

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 19, 2008

given its higher solubility. Hydrolysis of sarin occurs faster than the rate of disposal, so sarin does not accumulate in the landfill. SI Table S5 shows the effect of temperature for benzene, toluene, and acetone. Advective and diffusive gaseous losses increased for all compounds with the most dramatic effect for toluene, which has the highest KH (KH and diffusivities are adjusted for temperature within the model). Comparison of LFCR to MOCLA. The predicted fate routes for acetone and toluene under the base case scenario are compared between LFCR and MOCLA in Figure 5; SI Table S6 provides comparisons for all chemicals and scenarios. In all cases other than sarin, the mass remaining in the landfill at the end of the simulation (62 year) was greater for LFCR versus MOCLA. For toluene and benzene, the difference was significant with up to 39% more mass remaining for LFCR (toluene, bioreactor scenario). In MOCLA, all waste is in place at the beginning of the simulation, whereas LFCR represents typical landfill operations more realistically with waste disposed sequentially in daily cells over the operating life. Thus, mass loss mechanisms are active in MOCLA over a significantly longer time frame than LFCR, which results in greater mass loss from the landfill. For toluene, benzene, and naphthalene, vapor advection accounts for most of the increased mass loss in MOCLA. For acetone, aqueous degradation accounts for more mass loss in MOCLA versus LFCR. The predicted fate of sarin was insensitive to the model used, as its short abiotic half-life dominates. Based on these results, the use of a constant value of gas production and the initial placement of all waste in MOCLA may result in over prediction of mass loss via gas phase transport when compared to a model that uses a time-variable gas production rate and a more realistic waste disposal algorithm. The potential implications for risk assessment could be significant,

FIGURE 5. Comparison of predicted fates of acetone and toluene under the base case scenario using MOCLA and LFCR over a 62 year simulation period. with MOCLA predicting greater mass in the air and water leaving the landfill. In addition, the maximum contaminant concentration in MOCLA will always be at time zero. Model Limitations. The LFCR model makes it possible to represent changes in landfill waste volume, gas generation, infiltration, and cover type over time. There are nonetheless several simplifying assumptions that were considered appropriate relative to the primary model objective of predicting chemical behavior over long time scales (greater than decades) in support of large-scale risk analysis. It is recognized that such assumptions do not incorporate factors such as variation in gas flow fields and preferential moisture flow that may fluctuate daily. Anaerobic biodegradation rates in the landfill are likely the most uncertain parameter. In addition, the potential for aerobic biodegradation in the refuse near the landfill surface, and in landfill cover soils was not included although biodegradation in soil covers has been reported (22). By neglecting this mechanism, LFCR may overestimate gaseous losses. LFCR allows vertical but not lateral transport within the landfill. Fully describing multidimensional flow within a landfill, including heterogeneity and preferential flow is difficult given the limited understanding of landfill fluid flow and the challenges of characterizing waste heterogeneity. In

addition, the LFCR model assumes that the waste temperature is constant both spatially and temporally. The LFCR model assumes long-term average and homogeneous moisture conditions within the waste, liner, and cover layers. The model does not provide a dynamic water balance to simulate varying moisture content in response to short-term weather fluctuations. Without a dynamic water balance, water storage does not change, and infiltration rates are equal to leaching rates. This assumption deviates from observed behavior in some landfills, where the waste moisture content increases over time. Finally, the model assumes uniform as opposed to preferential liquid flow as well as instantaneous and reversible equilibrium between the solid and aqueous phases (sorption/desorption). The cumulative impact of these simplifications on trends in model predictions is unclear but likely to vary more over shorter time scales (daily to monthly) than over the lifetime of a landfill. The LFCR model structure and solution would allow some of the simplifying assumptions to be relaxed if additional information were available and if appropriate for the modeling objectives. Specifically, additional heterogeneous and time-variable conditions could be represented, because the numerical solution allows for spatially and temporally variable conditions. As an example, the refuse, cover, and VOL. 42, NO. 19, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7449

TABLE 2. Leachate and Gas Concentrations Predicted by LFCR and Range of Reported Valuesa LFG at 20 Yr

Leachate at 20 Yr

base case (37 °c) arid case (37 °C) bioreactor case (42 °c) literature reportsb

11.5 2.3

benzene 31.1 10.2

11.0

22.5

base case (37 °c) arid case (37 °C) bioreactor case (42 °c) literature reportsb

33.1 11.7

0.2 - 1630 toluene 66.8 35.8

24.4

63.6

10-287

1-12 300