GE Comments to the Modeling Peer Review Panel
on EPA’s Model Framework Design
Lenox, MA
April 25, 2001
April 25, 2001 2
GE’s Role in the Modeling Effort
Goal of the Modeling
– Develop a scientifically supportable mathematical representation of the system
– Objectively evaluate risk reduction associated with alternative remedial actions
GE’s Modeling-Related Responsibilities under the Consent Decree (CD) – GE will perform the Corrective Measures Study (CMS) for the Rest of River – GE will apply the EPA models to evaluate remedial alternatives
Introduction
April 25, 2001 3
GE’s Role in the Modeling Effort
Interactions with the EPA Modeling Team – Cooperative arrangement established in the CD to provide a forum to:
• share insights on the system • establish and maintain a common database • identify and fill important data gaps • provide comments on EPA work products
– EPA responsible for model development, calibration, and validation
GE Is Actively Engaged in Modeling the System – Effort initiated prior to signing the CD – Effort continues to provide insights into mechanisms controlling PCB
fate within the System – GE’s modeling efforts have informed GE’s comments on EPA’s MFD
Introduction
April 25, 2001 4
GE’s Major Concerns with EPA Modeling Approach
Model Domain Should Be Expanded to Include – Upper 2 mile reach, including plant site area – Downstream of Woods Pond into CT
EPA Modeling Approach Is Overly Complex Without Supporting Data – Focused on fine-scale phenomena
• not necessarily important at scale of remedial decision • poorly understood and not feasible to simulate
– Model linkages complicated by model incompatibilities • models not simulating same parameters (e.g. solids) • models not simulating same processes consistently
– Insufficient data to constrain some model complexities • lateral variations in sediment/PCB transport processes • changes in ecosystem parameters (e.g., fish biomass) • could lead to erroneous predictions
Additional Data Are Needed to Simulate Potentially Important Processes – Sediment bed load – PCB partitioning
Overview
April 25, 2001 5
Model Domain Should Be Expanded
Model Domain Should Include the Upper 2 Mile Reach
– Understanding of PCB loading from 2 mile reach important to understanding Rest of River PCB dynamics
• Approx. 20% of low flow water column load • 20 to 60% of high flow water column load
– Explicit modeling will provide data-based mass balance constraints on loadings originating from this reach of the River
• provides a means to estimate bed load – Active remediation is on-going or planned for this reach of the River
• work in upper ½ mile has found both PCB and coal-tar DNAPLs – Data collected after upstream remediation will allow evaluation of the
impacts of this remediation on Rest of River
Charge Question 3
April 25, 2001 6
Upper ½ Mile DNAPL Seeps
March 27, 2001 – DNAPL in Excavation Cell F3 Aroclor 1260 – 438,000 ppm
Charge Question 3
April 25, 2001 7
Model Domain Should be Expanded
Model Domain Should Include the Upper 2 Mile Reach
– EPA should take advantage of this unique calibration opportunity • models will be used to project the effects of remedial alternatives
for Rest of River • should apply them to simulate the observed impacts of actual
remediation in upper 2 miles – Comparison of model simulation of this post-remediation period with
actual post-remediation data would represent a robust test of the model
Charge Question 3
April 25, 2001 8
Model Domain Should be Expanded
Model Domain Should Include Reaches Downstream of WP – Biota consumption advisories in MA and CT – Expansion of model domain will allow objective downstream
predictions of the impact of: • natural recovery • rare floods • upstream remediation activities
– Extensive CT biota data • could provide a robust data set for model calibration
Charge Question 3
April 25, 2001 9
Cornwall, CT Fish PCB Concentrations
Charge Question 3
April 25, 2001 10
Housatonic River Is a Complex System
Variety of Hydrodynamic Regimes – main channel and backwaters – floodplains – impoundments
Diversity of Sediment Transport Processes – cohesive sediment transport – non-cohesive sediment transport – bed load – floodplain sediment deposition – bank erosion
PCB Fate – potential “third phase” – partitioning onto autochthonous solids – bed load PCB transport
Rea
ch 5
5A
5B
5C
WP
SackettBrook
RoaringBrook
West BranchConfluence
PittsfieldSTP
Charge Questions 1,2, and 3
April 25, 2001 11
EPA Modeling Approach Is Overly Complex
Impractical to Simultaneously Represent All the Following System Complexities – Floodplain interactions – Lateral gradients in current velocity and turbulent eddies in main
channel – Bank erosion – Bar and terrace formation and degradation
Lack of Sufficient Data to Constrain Some Processes
– Food web dynamics – Adding such complexity to model without supporting data will:
• increase uncertainty of model predictions • obscure model uncertainties • possibly lead to erroneous predictions
Charge Questions 1,2, and 3
April 25, 2001 12
EPA Modeling Approach Is Overly Complex
Need to Develop a Model at a Scale That – Is technically feasible given computational constraints – Is supported by available data – Possesses sufficient predictive capabilities to be applied to guide
remedial decisions
Charge Questions 1,2, and 3
April 25, 2001 13
Rea
ch 5
5A
5B
5C
WP
SackettBrook
RoaringBrook
West BranchConfluence
PittsfieldSTP
Model Grid Challenges
Narrow Meandering Channel Nested within Broad Floodplain
Complex Interactions between Main Channel and Floodplain during Floods (~1.5 yr return interval) – Complex flow routing – “Wetting and drying” of floodplains – Sediment and PCB transport
Charge Questions 1,2, and 3
April 25, 2001 14
Nested Curvilinear Grid Proposed by EPA
200 0 200 400 600 8001000 Feet
•Orthogonal curvilinear grid ambitious, but not practical for this system due to complex geometry
–extremely small elements –excessive computational burden
•Nested grid to represent floodplains is creative, but may not realistically represents channel/floodplain interactions
–momentum is not conserved –doesn’t represent sheet flow under high flow conditions
–sediment and PCB dynamics in flood plain and channel may not be realistically represented
Charge Questions 1,2, and 3
April 25, 2001 15
Model Grid Alternative
GE/QEA Has Simulated Hydrodynamics and Sediment Transport Using 20-meter Rectangular Grid Between Confluence and Woods Pond – 2,288 channel elements – 1,313 backwater elements – 8,981 floodplain elements
Similar Grid Envisioned for Upper 2 Miles, Including Plant Site,
While Coarser Grid Envisioned for Downstream to CT
Grid Attributes – Computationally feasible to perform 30 year projection runs – Should be sufficient to model remedial alternatives for the CMS
Charge Questions 1,2, and 3
April 25, 2001 16
Hydrodynamic Model Grid (20m x 20m)
Rea
ch 5
5A
5B
5C
WP
SackettBrook
RoaringBrook
West BranchConfluence
PittsfieldSTP
Charge Questions 1,2, and 3
April 25, 2001 17
Model Grid Alternative
A 20-meter Grid Necessitates Aggregation of Some Hydrodynamic and Sediment Transport Processes – Model cannot simulate lateral differences in bottom shear and sediment
transport – Individual point bar and terrace deposits not resolved
Small Scale Deposits Only Important If They Dominate PCB
Transport Process and Goal is to Remediate at That Scale – That is, if they contain higher PCB concentrations than surrounding
channel sediments such that aggregation of sediment transport processes misrepresents PCB transport
A 20-meter Grid Requires Appropriate Data to Constrain the Inherent Empiricisms of the Approach
Charge Questions 1,2, and 3
April 25, 2001 18
Sediment PCB Distribution
Charge Questions 1,2, and 3
April 25, 2001 19
Areal Distribution of Bar and Terrace Deposits
Source: BB&L 1994 Probing data
0
5000
10000
15000
20000
25000
30000
128129130131132133134135136137138River Mile
Are
a (m
2 )
0
5
10
15
20
Are
a (%
)
Bars and TerracesOther Channel Samples % Bars and Terraces
WBC PSTP
Charge Questions 1,2, and 3
April 25, 2001 20
Constraining the Inherent Empiricisms
Sediment Transport
– Data-based solids mass balance calculations during flood events • spatial patterns in erosion and deposition • generate understanding of system behavior
Charge Questions 1,2, and 3
April 25, 2001 21
Temporal Trends in TSS During May 1999 Flow Event
New Lenox Rd
Pror
ated
Col
tsvi
lle F
low
(c
fs)
TSS
(mg/
L)
Charge Questions 1,2, and 3
April 25, 2001 22
Constraining the Inherent Empiricisms
Solids Mass Balance Calculations – Develop hourly time series of TSS concentrations from data using
linear interpolation • upstream and downstream boundaries • tributaries
– Develop time series of flow rate from model inputs and predictions • inflows developed from upstream and tributary inputs • outflow determined from hydrodynamic model results
– Calculate hourly solids loads at inflow and outflow boundaries using TSS concentration and flow rate information
– Calculate sediment mass balances for different reaches of the River
Charge Questions 1,2, and 3
April 25, 2001 23
Sediment Mass Balance – POM – NLR Charge Questions 1,2, and 3
April 25, 2001 24
Sediment Mass Balance – NLR - WP Charge Questions 1,2, and 3
April 25, 2001 25
Constraining the Inherent Empiricisms
PCB Fate and Transport – Low flow
• diffusional processes • molecular, biologically enhanced
– High flow
• net effect of sediment erosion and deposition processes
Charge Questions 1,2, and 3
April 25, 2001 26
Examples of Water Column PCB Load Dynamics
0
100
200
300
400
500
600
PCB
Loa
d (g
/day
)
Flow Regime Low Flow High Flow
HeadwatersDam
Holmes Rd. Woods Pond
Effect of Flow Trapping in WP
Charge Questions 1,2, and 3
April 25, 2001 27
Comments on Specific Models Proposed by EPA
EPA Approach Complicated by Use of Two Models to Simulate PCB Transport and Fate – EFDC and AQUATOX
• different spatial and temporal domains and scales • different representations of PCBs (total vs. congener) • complex model linkages
– floodplain inundation – solids dynamics
– Complicates water, solids, and PCB mass balances
EPA Should Use Only EFDC to Simulate PCB Fate – AQUATOX incapable of effectively simulating floodplain PCBs
Charge Questions 2 and 4
April 25, 2001 28
Comments on Specific Models Proposed by EPA
AQUATOX Proposed to Simulate Ecosystem Dynamics – Simulating “population dynamics” in multiple species and trophic levels – Complex predator-prey relationships
Housatonic River Data Insufficient to Support the Development
and Calibration of Ecosystem Model – No data on biomass changes over time – Data from other sites inadequate
Artificial Ecosystem Dynamics Will Impact Fish Exposures and
PCB Bioaccumulation – Exposure related to prey population abundances – No constraints on prey population calculations
Charge Questions 1,2, and 4
April 25, 2001 29
Comments on Specific Models Proposed by EPA
AQUATOX Should Be Used Only to Simulate “Average” Fish
Exposure – Consider temporal trends in water column and sediment concentrations
predicted by EFDC – Bound the diet of each species based on site data, information from
other studies – Calibrate by adjusting diets within observed bounds
Charge Questions 1,2, and 4
April 25, 2001 30
Additional Data Required
Constrain Potentially Important Processes – Sediment bed load – PCB partitioning
EPA Recently Committed to Collecting New Data
– Water column partitioning data – Sediment pore water partition data – Sediment bed load measurements
Additional Data May Be Required to Further Constrain Inherent
Empiricisms of Approach – Finer longitudinal spacing in water column sampling program – Site-specific fish gut content survey
Charge Questions 5 and 6
April 25, 2001 31
Summary of GE Comments
EPA Should Expand Model Domain to Include – Upper 2 mile reach, including plant site area – Downstream into CT
EPA Should Modify Modeling Approach
– Hydrodynamics/sediment transport • develop model at a resolution that is computationally feasible and
realistically represents floodplain hydrodynamics – PCB fate
• conduct PCB fate modeling with a single code (i.e., EFDC) – PCB bioaccumulation
• abandon plans to simulate ecosystem dynamics
EPA Should Use Data Collected After 2 Mile Remediation to Calibrate the Models
EPA Should Collect Additional Data Required to Constrain
Empiricisms of Approach
Summary