Multibox Model of PCB Fate in San Francisco Bay

1
Multibox Model of PCB Fate in San Francisco Bay

• Presented to the CFWG
• January 15, 2008

2
Outline

• Brief Model Overview
• Model Confidence
• Hindcast Model Validation
• Uncertainty Analysis
• Model Sensitivity
• Uncertainty of Important Input Parameters
• Response to Extreme Scenarios
• Forecast Results
• Source and Loss Pathways
• Information Gaps
• Next Steps

3
Model Overview
4
Model Overview
5
Validation Salinity
6
Validation SSC
Dashed Model Solid Observations
7
Preliminary (Uncalibrated) Hindcast Results
Magnitude and spatial distribution of results
can be improved.
8
Targeted Calibration

• Considering Tetra Techs testing and Workgroup
  comments, do certain model parameters need
  revising?
• ? Magnitude of WY 2000 tributary loads.
• ? Temporal trend of historic PCB loads.
• Spatial distribution of tributary loads
• Removed this tweak per CFWG recommendation

9
Historic PCB Loads
Breivik et al, 2002
10
Hindcast Results After Targeted Calibration
Still room for improvement. Workgroup
recommended spatially explicit Koc.
11
Objective Calibration of Partitioning (Koc)

• RMP data insufficient for regional Koc
• Filter size overestimates dissolved fraction
• Need objective means of determining Koc
• Should consider more than just sediment PCBs.
• ? Model Bias Estimator
• Iteratively run model with small perturbations of
  regional Koc and search for model bias closest to
  1

12
Model Bias Estimator

• Estimate model bias for each Bay segment (j) and
  each matrix (i).
• For each segment (j) determine mass-weighted bias
  for all matrices (i).
• Determine overall (Bay-wide) model bias.

13
Hindcast Results After Calibration
Error Bars EMAP RMP Standard Deviation of
Samples Model Aggregate Uncertainty
14
Hindcast Results After Calibration
Net Erosional
Net Depositional
15
Uncertainty Analysis

• 10,000 runs made as part of Tetra Tech testing
  re-analyzed to determine aggregate uncertainty
  of model results given uncertainty in input
  parameters.
• 3 sediment results (SSC, net sedimentation,
  sediment export)
• 10 PCB results (wct, sed, mass in water, mass in
  sed, burial, deposition, degradation, erosion,
  outflow, volatilization)

16
Uncertainty Analysis Results

• Uncertainty, expressed as std. dev., is function
  of mean.
• Use this relation to project uncertainty onto
  forecast

17
Uncertainty Analysis Results
18
Base Forecast Setup

• Initialized with water and sediment
  concentrations from RMP 2006 data and profiles
  from end of hindcast
• Future loads attenuate from current levels with
  50yr half-life
• Relationships of PCBs to SSC _at_ Mallard Island
  remain
• Delta outflow includes potential climate change
• Sedimentation rates/patterns continue
• Uncertainty determined by hindcast

19
Forecast Setup Sedimentation

20
Model Sensitivity
O Model Output P Model Input Parameter
21
Model Sensitivity
High
Relative Sensitivity
Low
22
Uncertainty of Important Input Parameters
Vertical Profile
Profiles that increase subsurface mass change
forecast predictions considerably.
23
Model Sensitivity
High
Relative Sensitivity
Low
24
Extreme Scenarios Delta Outflow
25
Extreme Scenarios Instantaneous Inputs
Double Check of South Bay 200 kg PCBs mixed into
2.3x1013kg sediment 8.8 ng/g increase
26
Model Confidence Summary
2005 survey suggests South Bay depositional.
Survey to which model calibrated (1984?)
suggested erosional. - D. Schoellhamer,
personal communication
27
Base Forecast Recovery Due to Natural
Attenuation
Net Erosional Controlled by Subsurface Mass
Net Depositional Controlled by Attenuation
Degradation
28
Loading Scenarios Local Tributary Loads
29
Loading Scenarios Delta Loads
30
Loading Scenarios Wastewater
31
Loading Scenarios No External Loads
32
PCB Source and Loss Pathways
Active Sediments refers to top 5 cm
Erosion Inputs Total External Inputs
33
Information Gaps

• Sediment cores
• Hindcast depositional cores more important
• Forecast erosional cores more important
• Better info on attenuation, degradation, Koc
• Possible to get better info?
• Estimate outflow of sediment PCBs
• USGS et al. have study planned for 2008

34
Possible next steps or PCB modeling

• Is South Bay depositional?
• Congener specific model
• Smaller spatial scales hot spots or sub-
  embayments (e.g., LSB)
• 3D hydrodynamics sediment transport
• Apply multibox to multiple contaminants
  (screening tool)