Disposal Safety SDMS DocID 535361

Incorporated November 20, 1992

Ms. Ginger Graham Hope's Committee for a Clean Environment

^•pcrfumi Records Center% Thomaston Miniature Works SI T E: (MiOo O^ica!111 Main St. BREAK: WlThomaston, Maine 04861 OTHER: __ttJUI

Subject: Comments on Proposed Vapor Extraction System

Dear Ms. Graham:

The following are our comments on the vapor extraction system that has been proposed in place of thermal treatment for contaminated soils at the Union Chemical Company Superfund site.

Notice - This report has been prepared solely for the guidance of the Hope's Committee for a Clean Environment in interpreting information available to it. Other users should satisfy themselves independently as to facts and conclusions contained herein. In particular, such users should refer to original sources of information rather than to this report. This report is not intended for use in any real estate or other transaction, and should not be used or relied upon for such purposes.

Scope of review - We reviewed the following documents prepared by Balsam Environmental Consultants Inc. and Vapex Environmental Technologies:

• Assessment of in situ soil aeration as a preferred source control remedy for the Union Chemical Company Superfund site, South Hope, Maine, June 27, 1991.

• Vapor extraction treatability study, April 3, 1992.

• Letter to John Gilbert, Balsam, from Michael C. Marley and Peter E. Nangeroni, Vapex, July 3, 1992.

• Letter to John M. O'Donnell, Balsam, from Michael C. Marley and Peter E. Nangeroni, Vapex, October 9, 1992.

We obtained further technical explanations from Balsam and Vapex at meetings in Hope on June 2 and Salem, NH, on November 16 and in a conference call on August 7.

1660 L Street NW, Suite 314 Washington, DC 20036 (202) 293-3993

Ms. Ginger Graham, November 20, 1992 Page 2

The issue addressed here is the accuracy and reliability of Vapex's predictions of the time required for vapor extraction to clean the soil to the levels specified in the Record of Decision. We have not calculated clean-up times ourselves, and we have not checked Vapex's calculations. Rather, we have limited ourselves to an examination of the assumptions that underlie Vapex's calculations and how well they are supported by data.

Homogeneity of the soil - The Vapex predictions rest on the assumption that gas drawn into the extraction wells will uniformly permeate the soil. If heterogeneities in the soil cause the gas flow to be concentrated in discrete conduits, the less permeable zones will be poorly flushed and the clean-up will take longer than predicted. No soil, of course, is perfectly uniform, and Vapex's design uses safety factors to compensate for a certain amount of heterogeneity. The main question we have evaluated is whether the subsurface conditions at the site are reasonably well approximated by the assumption of uniformity made in Vapex's calculations. Some other significant issues will be discussed briefly at the end of this letter.

Several kinds of information are available to answer this question. This information includes:

• Geological descriptions of the site.

• Laboratory tests in which gas was pumped through soil samples and removal of chemicals was measured.

• Laboratory measurements of soil permeability.

• Short-term pump tests in the field, in which the pressure was measured in the pumping well.

• Long-term pump tests with gas flows and pressures measured in observation holes.

• Chemical analyses of soil samples before and after the field pumping tests.

We will discuss each kind of information in turn.

Site geology - The unsaturated zone is composed of glacial till, overlain by fill which is composed of similar glacial till. Till can often be a heterogeneous material. The geological setting cannot be used to support a presumption of homogeneity (in contrast to the case of, for example, a beach sand). The site geology points in the direction of large heterogeneity, but this kind of information is less reliable than the direct measurements discussed below.

Ms. Ginger Graham, November 20, 1992 Page 3

Laboratory tests - Only one laboratory test sample contained a substantial amount of contamination at the beginning of the test. Most of this contamination was removed during the test. However, soil samples in laboratory tests are smaller than the scale of heterogeneities that are of greatest concern in the field. The result of the chemical measurement is positive evidence for the potential effectiveness of vapor extraction, but because it represents a single small sample it does not answer the question of heterogenity.

Laboratory permeability measurements - Permeabilities measured on laboratory samples (July 3 letter, Table 2) ranged over two orders of magnitude. This is a relatively large range, but the samples are so small that this variation may represent small-scale heterogeneity. It is possible that low-permeability zones measured in this test are small enough that pollutants would diffuse out of them into zones of gas flow. The laboratory permeability measurements suggest a large degree of heterogeneity, but little reliance can be placed on them.

Single-well pumping tests - Permeabilities measured in single-well pumping tests (July 3 letter, Table 1) are relatively uniform; all but one of the measurement points fall within a range of a factor of 4. The volume measured by these tests is larger than for the laboratory tests. This is a high degree of homogeneity. However, tests were performed at only a limited number of points, and the method of test analysis involves some assumptions whose accuracy is hard to determine. The single-well pumping tests indicate that the medium is relatively homogeneous, but they are not conclusive.

Multiple-well pumping tests - In the large-scale field tests, gas was pumped from central extraction holes and allowed to enter through a surrounding ring of induction holes. Gas flows were measured in all of these holes and pressures were recorded at numerous probe points. These tests can be interpreted as pumping tests with observation wells.

The pneumatic data from thefield tests were interpreted with the aid of several models, all of which incorporate the assumption of homogeneity. The relatively large amount of data collected can be compared with model results; a poor fit between data and model casts doubt on the assumption of homogeneity. There are two sets of data that can be compared with models: pressures in vapor probes and gas flows through induction wells.

The pressures in the vapor probes are reasonably symmetric around the extraction wells. (Figures in the Treatability Study seem to indicate otherwise, but these figures reflect data transcription errors by Vapex.) This is consistent with the soil being homogeneous. However, a considerable degree of heterogeneity could also be consistent with the observed pressures, so no strong conclusion can be drawn from the pressure data.

A more demanding test is the flows in the induction holes. Vapex's calculations before the test predicted that the total flow into the induction holes would be 70% of the gas

Ms. Ginger Graham, November 20, 1992 Page 4

extracted from the central pumping well. In the field, only 1% to 3% of the outflow came in through the induction wells. If one subtracts Vapex's estimate of leakage through the surface cap (which the intial calculations did not take into account), the induction well inflow is still only about 10% of the remaining flow in the system. At the November 16 meeting, Vapex explained this discrepancy as resulting from the one-foot well diameter and complete penetration assumed in the initial modeling; the induction wells actually were screened through only part of the unsaturated zone and had diameters of two inches.

We tested this explanation by calculating induction well inflows, with assumptions similar to those in Vapex's initial calculations, for both well diameters. We did the calculations in two ways: with the finite-difference numerical model FLOWPATH, and with an approximate analytical solution. The two calculations agreed with each other very well, and they showed that the difference in well sizes cannot explain the difference between prediction and observation. In the FLOWPATH calculations, reducing the well diameter from 1 ft to about 1 in. changed the inflow only from 76% to 62%. The analytical solution shows that the percentage inflow to induction wells is quite sensitive to where one places the simulation boundary, but relatively insensitive to induction well diameter. Inflows to induction wells can, given the model assumptions, be as low as 10% only if the wells are very near the boundary. Details of these calculations are attached to this letter.

Partial penetration also fails to explain the discrepancy between prediction and observation. The screened intervals of the induction wells are 5 to 6 ft long. This is half or more of the unsaturated-zone thickness. Thus partial penetration reduces induction-well inflow by less than a factor of 2.

The small flow into the induction wells may result from heterogeneity in the soil. Alternatively, it might reflect greater leakage through the soil surface than estimated by Vapex, or placement of induction wells too close to the edge of the cap. Whatever the reason, the small flows into the induction wells are inconsistent with the assumptions underlying the tests. Thus, the large-scale pumping tests fail to support the conclusion that the soils are uniform.

Chemical analyses fromfield test - No clear pattern of reduced contaminant concentrations emerges from comparison of pre- and post-test soil analyses. The simplest hypothesis consistent with these data is that soil concentrations are highly variable over small distances, and that measured differences among samples represent random fluctuations. Thus no conclusions can be drawn from this information.

Other issues - Heterogeneity is not the only issue. We understand that preliminary data indicate the presence of large concentrations of dimethylformamide in the ground water at the site. Vapor extraction will be ineffective in removing DMF from the soil. It has been argued that any DMF remaining in the soils would be washed out by infiltrating rainwater,

Ms. Ginger Graham, November 20, 1992 Page 5

but that is unlikely beneath the clay cap that the vapor extraction system would require, and even more unlikely beneath the concrete pads. If DMF is present in the unsaturated zone, it would pose an additional problem for vapor extraction.

Additionally, it is worth noting that cleaning up contaminated ground water is routinely found to take much more time than predicted before the start of remediation. Usually, pollutant concentrations decline rapidly at first but then stabilize. Vapex agreed (at the June 2 meeting) that this behavior is also typical of vapor extraction systems, and it was observed in the field tests. Vapex has tried to allow for this phenomenon by including safety factors in its predictions. But the phenomenon is not predicted by Vapex's models, so it is hard to know whether the safety factors are adequate.

Conclusions - The best test of the hypothesis that the soils are relatively uniform is the flow into the induction wells. (There are many other kinds of data, but they are all inconclusive.) The measured induction-well intakes differ entirely from the predictions, and Vapex has not given a reasonable explanation of the discrepancy. Either the measurements show that the soils are heterogeneous, or they show that the tests were not well designed. In either case, the large-scale pumping tests fail to confirm the conceptual model on which the predicted clean-up times are based.

Taking all of the above considerations into account, we conclude that the clean-up times calculated by Vapex are not reliable predictions.

Sincerely,

Benjamin Ross President

Attachment A

Numerical Simulation

We calculated the flow through induction wells with the numerical model FLOWPATH [Waterloo Hydrogeologic Software, 1992]. This is a steady-state two-dimensional node-centered finite-difference model. The FLOWPATH model is normally used to calculate ground-water flow in confined aquifers. However, the gas-flow potential <t> is related to source and sink terms by the same equation as applies to ground-water flow (with different coefficients). Consequently, a ground-water model can be applied to calculate the proportion of flow that passes through the induction wells and across the outer boundary.

The simulation results were compared with the simulation performed by Vapex, the field measurements, and the analytical solution described in Attachment B. The principal objective was to calculate what percentage of the total inflow enters the soil through the induction wells, in order to determine whether the difference between Vapex's initial calculations and the field measurements can be explained by the difference in well diameters.

Conceptual Model

We assume that the unsaturated zone is homogeneous, confined porous media with a constant thickness. The model simulates a two-dimensional square region 60 ft by 60 ft. The model domain is bounded on all sides by a constant-potential boundary at atmospheric pressure, corresponding to a fixed potential of zero. A single pumping well, positioned in the center of the domain, extracts air at a constant rate. The pumping well isringed by six induction wells positioned at equal intervals along the circumference of a circle with a radius of 10 ft. Because these induction wells are open to the atmosphere, they are at atmospheric pressure. These wells are represented by nodes with a fixed potential of zero. All wells are fully penetrating through the unsaturated zone.

Simulations

Two simulations were performed. The only difference between the two was the grid spacing near the well. The theory of finite differences shows that a node set at fixed potential is equivalent to a rectangular hole whose sides are equal to the grid spacing. In the first case, one-foot grid spacing was used. This is the same size that Vapex used in its simulation. In the second case, cells with approximately one inch sides were used, which is smaller than the wells used in the field experiment and therefore provides an upper bound on the sensitivity of inflow to the well diameter.

1

The finite difference grids used to simulate the twelve-inch and one-inch well cases are attached. In both grids, the boundaries and the nodes corresponding to the induction wells were assigned a fixed potential of zero. In both cases the soil horizon was assigned a hydraulic conductivity of 0.86 m/d, and the pumping rate for the extraction well was specified as 1000 m3/d. Both of these quantities are arbitrary because the objective of the study was to determine the percentage of air entering the system through the induction wells.

Results

The calculated head distributions for the twelve inch and one inch well cases are also attached. The simulation of twelve inch wells predicts that 76% of the air enters through the induction well. This matches the Vapex simulation quite closely. The simulation of the one inch wells predicts that 62% of the air enters through the induction wells. This shows that wellbore size effects are relatively small. The discrepancy between Vapex's prediction that 70% of flow would enter through the induction holes and the analysis of field results indicating that induction-hole inflow is only a tenth of inflow through the boundaries cannot be explained by the wellbore size.

Reference

FLOWPATH: Steady-state two-dimensional horizontal aquifer-simulation model, Waterloo Hydrogeologic Software, 1992.

2

48.0

Simulation Domain and Boundary Conditions

60.0'

36.0'

24.0­

12.0

0.0

0.0 12.0 24.0 36.0 48.0 60.0

FLOWPATH

Copyri 4» 1989-199

by WHS

# Rows :

61

# Cols :

61

# Wells:

1

Units

Cm]

File :

HOPE

60.0^ Hydraulic Head Distribution

FLOWPATH

48.0 -

Copyrigh

1989-199

by WHS

36.0

14.0 ­

Steady

State

Flow

Min :

12.0 -2.60E+0;

Max :

0.00E+0C

0.0

0.01 1

12.0 1 1 24.0

1 1 ' 3 6 . 0

1 1 48.0 60.0

Inc :

1.30E+0:

Units

Cm]

File :

HOPE

Simulation Domain and Boundary Conditions 60.0 vjin LB u u i i i i aJ FLOWPATH

48.0

C o p y r i c M ^

1989-199

by WHS

36.0

24.0

# Rows :

79

# Cols :

89

# Wells:

1

1 2 . 0 ­

0.0 ­

0 .0 12.0 24.0 36 .0 48 .0 60 .0

Units

Cm]

File :

H0PE2

Hydraulic Head Distribution 60.0 -v FLOWPATH

Copyrigh

1989-199 48.0

by WHS

36.0 ­

Steady

State

24.0­ Flow

Min :

-3.96E+0 12.0

Max :

0.00E+0C

Inc : 0.0

0.0 12.0 36.0 48.0 60.0 2.00E+0

Units :

Cm]

File :

H0PE2

Attachment B

Analytical Approximation

The equation governing the air pressure in unsaturated zone under radial coordinated can be approximated as [Baehr and Hult, 1991]

J24> (1) dr2 r dr

where is defined as P2 - P^ 2. If the boundary conditions are defined as

4> = 0, r = r. (2) d$ -QRTy

r = r, dr nHarj k*

where the notation is the same as used by Baehr and Hult, except that our definition of <f> differs by the constant P 2. The analytical solution to Equations (1) and (2) is obtained by a t m

Baehr and Hult [1991] as follows:

QRT\x mi (3) • = r

The value of <t> at any position can be obtained by superposition of the six induction wells and the extraction well. The potential due to each induction well can be approximated by the potential of the same well if the boundary were a circle centered at that well. This is a good approximation unless the well is close to the true system boundary.

Because the induction wells are open to the atmosphere, <f> in any induction well should equal zero. Denoting q as the mass flux in the induction well, rd as the distance between wells, and r as the radius of the induction well, superposition then gives w

/r£-Qf (rJ> 2qf 2qf

(rJ> = 0 \ r 4 ) \ r < ) (4)

/ = ln̂ > r

Simplification of the above equation yields the portion of air flow coming from the induction wells:

1

6q (5) Q

)

The portion of air flow coming from outside the induction well region can be calculated as

Q-6q _ 1

Q (6) 1.0 Xjli

Figure 1 shows the portion of air flow coming from outside the region as a function of Rf (defined as Tf/rJ and Rh (rjr^. For a system with rd = 10 feet, r = 1 inch, and w

rd = 30 feet, we conclude that about 30% of the air flow comes from the outside region and 70% comes from the induction wells. On the other hand, if we have a system with rd = 10 feet, r = 6 inches, and rd = 30 feet, we found that about 15% of the air flows w

from outside the region and 85% comes from the induction wells. This agrees well with the numerical simulations described in attachment A, confirming the acceptability of the approximation.

For all values of Rb shown on the figure, the inflow through the induction wells is much greater than the inflow through the region boundary. This conclusion is independent of the induction well diameter. The model will calculate small induction well inflows only if the induction wells are close to the outer constant-head boundary of the system.

Reference

Baehr, A. L. , and M. F. Hult, Evaluation of Unsaturated Zone Air Permeability Through Pneumatic Tests, Water Resources Research, Vol. 27, pp. 2605-2617, 1991.

Figure 1. Portion of air flow coming from outside the region (A value 1.0 means 100%). Where R{ (rf/r^ is the ratio of the far boundary distance to the distance between the extraction well and the air induction well and Rh {rjr^ is the ratio of the radius of the induction well to the distance between the induction well and the extraction well).

3

December 14, 1992 Mr. Ed Hathaway U.S. Environmental Protection Agency HPS-1 90 Canal Street Boston, MA 02203

Mr. David Wright Maine Dept. of Environmental Protection State House Station 17 Augusta, Maine 04333

Re: Dr. Ross's Comments Concerning the Proposed Vapor Extraction System

Union Chemical Site

Dear Messrs. Hathaway and Wright:

Enclosed please find Dr. Ross's comments to the Hope's Committee for a Clean Environment, (Technical Assistance Grant) TAG Group on the vapor extraction system, proposed for the Union Chemical Site.

I will discuss the comments with you tomorrow.

Regards,

RCS:vll

cc: file

EPA14/UC16/RCS

P.O. Box 310 • Mont Vernon, N.H. 03057 • (603) 673-0004 • FAX (603) 672-0004