DSHW-2008-001969 - 0901a0688013cf3e

RECEIVED Launch Systems Group
P.O. Box 707
Brigham City, UT 84302
v^.atk.com JUN 02 2(^38
UTAH DlVISlOiM Uh
29 May 2008 SOLID & HAZARDOUS WASTE
8200-FY09-013 0^.Dl$i1
Mr. Dennis R. Downs, Executive Secretary
State of Utah
Department of Environmental Quality
Division of Solid and Hazardous Waste
288 N.1460 W.
P.O. Box 144880
Salt Lake City, Utah 84114-4880
Subject: Response to April 4, 2008 Division Comments on the Groundwater Flow and Contaminant
Transport Model Report, ATK Launch Systems Promontory Facility, EPA ID #UTD009081357
Dear Mr. Downs
Your office provided comments in an April 4, 2008 letter on the ATK Launch Systems, Promontory
Groundwater Flow and Contaminant Transport Model Report. Attached, please find responses to
these comments.
If you have questions regarding these comments, please contact Paul Hancock at [phone].
David P. Gosen, P.E., Direcior
Environmental Services
Cominents on the Groundwater Flow and Contaminant Transport Model (Revised)
ATK Launch Systems - Promontory Facility, January 2008
1) Section 3.2, Flow modeling Software, p. 3-3, and Section 4.5, Grid, p.4-6: The text states
that a uniform grid (one node every 200 feet) was selected for the entire modeling
domain. While this approach can be justified for the flow model, the transport model
could have benefited from either a grid refinement or a domain resizing, especially in
zones where calibration problems were encountered (e.g., areas of fracture flow, see p.4-
8). Please explain why a grid refinement for the transport portion ofthe model was not
contemplated.
Response: The dispersivity in most layers (layers 3, 4, 5, and 6) is set to 60 feet, except
the layer 1 (which is the confining clay). The Peclet number in these layers was 3.33 in
the model run and the Courant Number is set to 0.75. We used upstream finite difference
numerical method, thus, according to accuracy criteria, the Peclet number should be
smaller than 2/(l-C) as much as possible (here, C is courant number). In this model run,
2/(1-C) = 5.77, we can say it is satisfied for modeling accuracy with relatively larger grid
spacing. We agree that smaller grid spacing can make the transport model simulation
better; however, 200 foot grid spacing has already met the accuracy requirement and
shortens the computer time necessary for each model run. Therefore, the 200 foot grid
spacing was kept for the transport portion ofthe model.
2) Section 3.5.3, Recharge, p. 3-6: The text states that a recharge value ofthree inches/year
into the uppermost, active layer was determined during the calibration runs. Was this
value arrived by trial and error, or by a more numerically, objective "goal search"
routine, such as implemented in PEST? Please elaborate.
Response: The primary method to determine the recharge value to the uppermost active
layer was by trial and error. According to Bolke et al. (1972), average annual
precipitation averages 12-20 inches per year. Ofthe 184,000 acre-feet of rainfall that falls
on the Blue Creek Valley drainage annually, approximately 14,000 acre-feet will
recharge the ground-water. This equates to 8% ofthe total rainfall. The Blue Creek
Valley receives approximately 12 to 20 inches of rainfall annually. This equates to
between 1 and 1.6 inches of recharge (8% ofthe annual rainfall) annually. Using three
inches/year is considered a conservative estimate and appropriate for the model.
3) Section 3.6.1, Calibration Process, p.3-9: The text menfions that the target maximum
head change criterion of 0.1 foot (as stated in the original work plan) was not achieved;
however, multiple simulation runs demonstrated that the flow model solution was stable
and therefore acceptable. Please explain how that was done in practice, e.g., was the
' same (or approximately the same) solufion obtained with different numerical solvers
(SIP, SOR, PCG, AMG, etc.), and were different damping factors employed?
Response: WHS was used as the numerical solver. The primary difficuhy encountered
during calibrafion was trying to achieve convergence after adding the perched aquifer.
The head differences at the boundary ofthe perched aquifer change rapidly and therefore
made it more difficult to achieve convergence. An inherent problem with the Visual
Modflow software is that it does not support rapid head change between nodes. The flow
model solution is proven to be stable because we could not find much difference when
we ran the flow model with different initial conditions under steady state condifions. That
is the common way to test ifthe application of such a numerical method is stable or not.
4) Section 3.6.3, Calibration Results, p. 3-13: The text states that the greatest head
differences between the predicted, steady state, and measured potentiometric surfaces
were encountered in the perched aquifer, with a maximum drawdown of 27 feet in the
final iteration. It appears to us that more data from that area (i.e., one or two monitoring
wells) might be desirable (especially to the east of J-2, the east of J-3, and the northeast
of J-4), as contaminant plumes for both TCE and perchlorate still seem to be undefined
(i.e., not delineated satisfactorily).
Response: We added J-2, TCC2 and J-4 from the north to south as observation wells in
the eastem perched aquifer. Here are the TCE and Perchlorate concentrations compared
between the observed and model calculated concentrations:
Concentration Comparison of Contaminant plumes for TCE and Perchlorate
Well Name Observed value Calculated value
TCE (mg/L)
J-2
TCC2
J-4
0.0
0.42
2.54
5.77
2.85
2.14
Perchlorate
J-2
TCC2
J-4
0.0
0.851
0.014
0.0
0.0
1.98
5)
From the above data, we can see this region has very complex geologic and
hydrogeologic background with several normal faults passing through this very narrow
perched aquifer, so the concentrations varied largely between each well. The model
simulation basically can represent TCE and Perchlorate concentrations at downgradient
receptors but has difficulty within the interior ofthe plume specifically near the perched
aquifer. It is acknowledged that additional data would be useful to further define the TCE
and perchlorate plumes. For this modeling effort, however, the limited data available
were utilized to the extent possible in defining contributions from the perched aquifer.
Section 4.2, Transport Parameters, p 4-2: The text states that, because monitoring
reportedly demonstrated that perchlorate has degraded in the manufacturing area and
downgradient ofthe leach field, a decay rate (first-order) of 0.00065 1/day was
determined for modeling purposes. However, later in the text it is stated that because
Visual Modflow does not allow the decay rate to be varied with time, a decay rate was
not used (instead, source concentrations were varied over time). Please clarify.
Response: Section 4.2 indicates that the decay rate (0.000651 1/day) will reduce with
time but Visual Modflow does not allow this parameter to be changed with time. This
decay rate and this value is so low so that the whole decay rate for Perchlorate is set as
zero for the conservative consideration.
6) Secfion 4.3, Sources and Sinks, p. 4-3: There appears to be a typographical error in the
text. Please clarify the sentence "..., this difference in constant concentration values
between resulted in the unreasonable situation of removing .. ."
Response: The document shows that if we add Building Ml 15/M174 source as a
constant concentration source, it would cause the plume of contamination in the nearby
Buming Ground to move toward Building Ml 15 and to not migrate downgradient, (i.e.,
southward) as expected. Modflow can only accept one type of flux concentration
boundary (that is recharge concentration/Evapotranspiration concentration), and it only
works for the first layer.
7) Figures: Calibration, "goodness-of-fit" charts appear to be missing. Please attach the
relevant calibration charts of observed vs. final, calibrated heads for the modeling domain
layers (both for the flow model, as well as for the transport model).
Response: These charts have been included with this response.
Calculated vs. Observed Head : Steady state
a
A
Layer #1
Layer #2
Layer #3
Layer #4
95% confidence interval
95% interval
r-
St- "
TO •*
ra to-
:D
O
h-
'3-
to-1
r^
CM
/\/-'' ,••''
9\/^-'''
^y/'
, '. 1 . . 1 1
//,
/p>
•••;^
//
1 ' ' ' ' i ' ' 4234.7 4334.7 4434.7
Observed Head (ft)
4534.7
Num. of Data Points : 87
Max. Residual: -44.924 (ft) at A-3/A
Min. Residual: 0.031 (ft) at G-2/1
Residual Mean : -2.824 (ft)
Abs. Residual Mean : 4.874 (ft)
Standard Error ofthe Estimate : 0.942 (ft)
Root Mean Squared : 9.177 (ft)
Normalized RMS : 2.672 ( % )
Correlation Coefficient: 0.997
Calculated vs. Observed Concentration : Time = 50000 days
• Layer #2 : ConcOOl
• Layer#3 : ConcOOl
A Layer#4 : ConcOOl
95% confidence interval
95% inten/al
-0.228 4.772
Observed Concentration (mg/L)
9.772
Num. of Data Points : 81
Max. Residual: 5.774 (mg/L) at J-2/1
Min. Residual: 0 (mg/L) at BC-3/1
Residual Mean : 0.475 (mg/L)
Abs. Residual Mean : 0.949 (mg/L)
Standard Error of the Estimate : 0.174 (mg/L)
Root Mean Squared : 1.629 (mg/L)
Normalized RMS : 27.615 ( % )
Correlation Coefficient: 0.426
TCE Calibration Chart for Promontory Transport Model
Calculated vs. Observed Concentration : Time = 50000 days
• Layer#2 : ConcOOl
• Layer #3 : ConcOOl
A Layer #4 : ConcOOl
95% confidence interval
95% interval
-11.7 188.3 388.3
Observed Concenlration (mg/L)
588.3
Num. of Data Points : 81
Max. Residual: 82.604 (mg/L) at A-5/CONCENTRATION
Min. Residual: 0 (mg/L) at P-1/CONCENTRATION
Residual Mean : 9.54 (mg/L)
Abs. Residual Mean : 11.855 (mg/L)
Standard Error of the Estimate : 2.354 (mg/L)
Root Mean Squared : 23.118 (mg/L)
Normalized RMS : 41.655 ( % )
Correlation Coefficient: 0.528
'erclilorate Calibration Chart for Pronionlory Transport Model
Calculated vs. Observed Concentration : Time = 10950 days
• Lay^r #2 : TCA
• Layer #3 : TCA
A Layer #4 : TCA
95% confidence interval
95% inten/al
-1.99 48.01
Observed Concentration (mg/L)
98.01
Num. of Data Points ; 49
Max. Residual: -2.478 (mg/L) at A-8/CONCENTRATION
Min. Residual: 0 (mg/L) at BC-2/CONCENTRATlON
Residual Mean : -0.098 (mg/L)
Abs. Residual Mean : 0.182 (mg/L)
Standard Error ofthe Estimate : 0.067 (mg/L)
Root Mean Squared : 0.474 (mg/L)
Normalized RMS : 18.722 ( % )
Correlation Coefficient: 0.353
TCA Calibration Chart for Promontory Transport Model
Caicuiated vs. Observed Concentration : Time = 10950 days
• Layer #2 : DCE
f Layer #3 : DCE
Layer#4:DCE
95% confidence interval
95% interval
lri
=J CO.-
C
o
1
§ . c o O s
ro 3
^ CO
ro oo-
o ^-
"~M
7/' M'
//y ..••''
""
1 . . ••!•• • 1 1 . 1 •• • 1
1.88 3.88
Observed Concentration (mg/L)
5.88
Num. of Data Points : 49
Max. Residual: -0.73 (mg/L) at B-1/CONCENTRATION
Min. Residual: 0 (mg/L) at H-5/CONCENTRATION
Residual Mean : -0.063 (mg/L)
Abs. Residual Mean : 0.067 (mg/L)
Standard Error of the Estimate : 0.023 (mg/L)
Root Mean Squared : 0.17 (mg/L)
Normalized RMS : 23.09 ( % )
Correlation Coefficient: 0.377
DCB Calibration Chart for Proinontoiy Transport Model