PHASE 2 REPORT - REVIEW COPY
        FURTHER SITE CHARACTERIZATION AND ANALYSIS
VOLUME 2D - REVISED BASELINE MODELING REPORT
         HUDSON RIVER PCBs REASSESSMENT RI/FS
                       JANUARY 2000
                            For

               U.S. Environmental Protection Agency
                          Region 2
                            and
                  U.S. Army Corps of Engineers
                      Kansas City District
                    Volume 2D- Book 1 of 4
                   Fate and Transport Models
                    TAMS Consultants, Inc.
                       Limno-Tech, Inc.
                  Menzie-Cura & Associates, Inc.
                       Tetra Tech, Inc.

-------
                      UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
                                             REGION 2
                                          290 BROADWAY
                                     NEW YORK, NY 1 0007-1 866
   January 25, 2000

   To All Interested Parties:

   The U.S. Environmental Protection Agency (EPA) is pleased to release the Revised Baseline
   Modeling Report for the Hudson River PCBs Superfund site. This report presents results and
   findings from the application of mathematical models for PCB transport and fate and
   bioaccumulation in the Upper Hudson River.  This report provides predictions under baseline
   conditions, that is, without any remediation measures for the PCB -contaminated sediments in the
   Upper Hudson River.

   The Revised Baseline Modeling Report incorporates changes made to the models based on
   comments received during the public comment period on the May 1999 Baseline Modeling Report
   (BMR) and from additional analyses that were conducted to refine the models for predicting future
   PCB levels in sediment, water and fish. EPA is also releasing a responsiveness summary for the
   BMR which provides readers with responses to significant comments, or directs them to the
   appropriate section of the Revised Baseline Modeling Report where the comment has been
   incorporated.

   The Revised Baseline Modeling Report supercedes the May 1999 Baseline Modeling Report. EPA
   will evaluate if the revised modeling results would change the overall conclusions of the Human
   Health and Ecological Risk Assessments, and will update the risk and hazard values in the
   responsiveness summaries for the risk assessments, as necessary. The results of models used to
   calculate the loads leaving the Upper Hudson, which were subsequently utilized in the calculation of
   risks and hazards for the Mid- and Lower Hudson River, are presented  in the Revised Baseline
   Modeling Report.

   Because the Revised Baseline Modeling Report was prepared in response to public comment (with
   some additional analyses that were outlined in the May  1999 BMR), there is no public comment
   period on this document. Please remember that we will, of course, accept public comment on all
   aspects of the Reassessment during the comment period on the Proposed Plan.

   The Revised Baseline Modeling Report is being peer reviewed by a panel of independent experts.
   The peer reviewers will discuss their comments on the Revised Baseline Modeling Report at a
   meeting that will be held on March 27 and  28, 2000 at the Sheraton Saratoga Springs Hotel and
   Conference Center.  Observers are welcome and there will be limited time for observer comment.

   If you need additional information regarding the Revised Baseline Modeling Report, please contact
   Ann Rychlenski, the Community Relations Coordinator for this site, at (212)  637-3672.

   Sincerely yours,
J, Richard L. Caspe, Director
7) Emergency and Remedial Response Division
                                 Internet Address (URL) • http://www.epa.gov
              Recycled/Recyclable • Printed with Vegetable OH Based Inks on Recycled Paper (Minimum 25% Postconsumer)

-------
               PHASE 2 REPORT - REVIEW COPY
        FURTHER SITE CHARACTERIZATION AND ANALYSIS
VOLUME 2D - REVISED BASELINE MODELING REPORT
         HUDSON RIVER PCBs REASSESSMENT RI/FS
                       JANUARY 2000
                           For

              U.S. Environmental Protection Agency
                         Region 2
                           and
                 U.S. Army Corps of Engineers
                     Kansas City District
                    Volume 2D- Book 1 of 4
                   Fate and Transport Models
                    TAMS Consultants, Inc.
                      Limno-Tech, Inc.
                 Menzie-Cura & Associates, Inc.
                       Tetra Tech, Inc.

-------
Table of Contents

-------
                    PHASE 2 REPORT - REVIEW COPY
            FURTHER SITE CHARACTERIZATION AND ANALYSIS
            Volume 2D - REVISED BASELINE MODELING REPORT
                HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4
                               CONTENTS
                                                                    "Jaee
LIST OF TABLES	viii
LIST OF FIGURES	xii

GLOSSARY	xxvii

EXECUTIVE SUMMARY	ES-1

1. INTRODUCTION	1
  1.1 PURPOSE OF REPORT	l
  1.2 REPORT FORMAT AND ORGANIZATION	2
  1.3 PROJECT BACKGROUND	3
    1.3.1 Site Description	3
    1.3.2 Site History	4
  1.4 MODELING GOALS AND OBJECTIVES	5
2. MODELING APPROACH	7
  2.1 INTRODUCTION	7
  2.2 CONCEPTUAL APPROACH	7
  2.3 HYDRODYNAMIC MODEL	9
  2.4 DEPTH OF SCOUR MODEL	9
  2.5 MASS BALANCE MODEL	10
  2.6 MASS BALANCE MODEL APPLICATIONS	11
  2.7 MASS BALANCE MODEL CALIBRATION	12
  2.8 HUDSON RIVER DATABASE	13
3. THOMPSON ISLAND POOL HYDRODYNAMIC MODEL	15
  3.1 OVERVIEW	15
  3.2 HYDRODYNAMIC MODELING APPROACH	16
   3.2.1 Governing Equations	16
   3.2.2 Computational Sequence and Linkages	18
  3.3 AVAILABLE DATA	18
   3.3.1 Model Grid	19
   3.3.2 Manning's V	19
   3.3.3 Boundary Conditions	19
  3.4 HYDRODYNAMIC MODEL CALIBRATION	20
  3.5 HYDRODYNAMIC MODEL VALIDATION	21
   3.5.1 Rating Curve Velocity Measurements	21
   3.5.2 FEMA Flood Studies	22
   3.5.3 100-Year Peak Flow Model Results	22
                                                       Limno-Tech, Inc.

-------
                      PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4
                                 CONTENTS
  3.6 HYDRODYNAMIC MODEL SENSITIVITY ANALYSES	22
    3.6.1 Manning's 'n'	23
    3.6.2 Turbulent Exchange Coefficient	23
  3.7 CONVERSION OF VERTICALLY-AVERAGED VELOCITY TO BOTTOM SHEAR STRESS	23
  3.8 DISCUSSION OF RESULTS	26

4. THOMPSON ISLAND POOL DEPTH OF SCOUR MODEL	27
  4.1 OVERVIEW	27
  4.2 DOSM MODEL DEVELOPMENT	28
   4.2.1 Conceptual Approach	28
   4.2.2 Formulation for Cohesive Sediments	29
     4.2.2.1 Background	29
     4.2.2.2 Basic Equations	29
     4.2.2.3 Reparameterization to a Probabilistic Model	30
     4.2.2.4 Calculation of PCB Erosion	31
   4.2.3 Formulation for Non-cohesive Sediments	32
     4.2.3.1 Background	32
     4.2.3.2 Equations	32
   4.2.4 Time Scale of Erosion Estimates	32
  4.3 DOSM PARAMETERIZATION	33
   4.3.1 Data	33
     4.3.1.1 Distribution of Types of Bottom Sediment	33
     4.3.1.2 Resuspension Experiments	33
     4.3.1.3 Non-Cohesive Particle Size Distributions	34
     4.3.1.4 1984 Cohesive Sediment PCB Concentration	35
   4.3.2 Parameterization for Cohesive Sediments	35
   4.3.3 Parameterization for Non-cohesive Sediments	36
  4.4 DOSM APPLICATION	37
   4.4.1 Application Framework	37
   4.4.2 Probabilistic Model Application to High Resolution Coring Sites	37
   4.4.3 Poolwide Model Application	38
     4.4.3.1 Cohesive Sediments	38
     4.4.3.2 Non-Cohesive Sediments	39
  4.5 DOSM FINDINGS	40
5. FATE AND TRANSPORT MASS BALANCE MODEL DEVELOPMENT          41
  5.1 INTRODUCTION	41
  5.2 MODEL APPROACH	41
   5.2.1 Introduction	41
                                                            Limno-Tech, Inc.

-------
                      PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4
                                 CONTENTS
                                                                         Page

   5.2.2 Conceptual Framework	42
   5.2.3 Governing Equations	43
  5.3 WATER TRANSPORT	44
  5.4 SOLIDS DYNAMICS	46
   5.4.1 Solids Gross Settling	46
   5.4.2 Cohesive Sediment Flow-Driven Resuspension	46
   5.4.3 Non-Cohesive Sediment Resuspension	47
   5.4.4 Sediment Bed Particle Mixing	48
   5.4.5 Scour and Burial	48
  5.5 PCB DYNAMICS	50
   5.5.1 Equilibrium Sorption	50
   5.5.2 Air-Water Exchange	53
   5.5.3 Dechlorination	55
   5.5.4 Sediment-Water Mass Transfer of PCBs	55
  5.6 MODEL SPATIAL SEGMENTATION	56
   5.6.1 Water Column Segments	56
   5.6.2 Sediment Segments	57
  5.7 MODEL IMPLEMENTATION	59
6. DATA DEVELOPMENT FOR MODEL APPLICATIONS	61
  6.1 INTRODUCTION	61
  6.2 AVAILABLE DATA	61
  6.3 MODEL APPLICATION DATASETS	63
   6.3.1 Sediment Datasets	63
   6.3.2 Water Column Data	65
   6.3.3 Conversion of PCB Data in Historical Calibration Datasets	66
     6.3.3.1 USGS Water Column Data	66
     6.3.3.2 1976-1978 NYSDEC Sediment Data	67
     6.3.3.3 1984 NYSDEC Sediment Data	67
     6.3.3.4 GE Water Column and Sediment Data	67
     6.3.3.5 USEPA Water Column and Sediment Data	67
   6.3.4 Data conversion for Total PCB and Congeners	68
  6.4 FLOW BALANCE	68
   6.4.1 Overview	68
   6.4.2 Flow Data	69
   6.4.3 Flow Estimation Methods	70
   6.4.4 Results of Flow Balance	73
     6.4.4.1 Validation of the Flow Estimation Approach	73
     6.4.4.2 Application of Estimated Flows in Modeling	74
                                                            Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4
                                  CONTENTS
                                                                            Page

     6.4.4.3 Summary of Flow Balance	74
  6.5 MAINSTEM AND TRIBUTARY SOLIDS LOADS	75
   6.5.1 Overview	75
   6.5.2 Solids Data	75
   6.5.3 Methods for Estimating Solids Loads	77
     6.5.3.1 Mainstem Solids Loads	77
     6.5.3.2 Tributary Solids Loads	81
     6.5.3.3 Development of Long-term Solids Balance	83
   6.5.4 Results	86
   6.5.5 Summary of Solids Load Estimates	87
  6.6 MAINSTEM AND TRIBUTARY PCB LOADS	87
   6.6.1 Overview	87
   6.6.2 PCB Data	88
     6.6.2.1 Data Availability for Estimating PCB Loads	88
     6.6.2.2 Thompson Island Dam West Shore Station Bias Correction	89
     6.6.2.3 Data Development for Computing PCB Loads	89
     6.6.2.4 Overview	90
     6.6.2.5 Mainstem Tri+ Loads  1977-1997	91
     6.6.2.6 Tributary Tri+ Loads 1977-1997	93
     6.6.2.7 Tri+ Load Results  1977-1997	93
     6.6.2.8 Mainstem and Tributary Total PCB and Congener Loads 1991-1997	95
   6.6.3 Total PCB and Congener Load Results 1991-1997	96
   6.6.4 Summary of PCB Load Estimates	96
  6.7 SEDIMENT INITIAL CONDITIONS	97
   6.7.1 Overview	97
   6.7.2 Sediment Specific Weight	97
   6.7.3 1977 Tri+ Initial Conditions	98
     6.7.3.1 1977 NYSDEC Sediment Data	98
     6.7.3.2 Methods	99
     6.7.3.3 1977 Initial Condition Results	100
   6.7.4 1991 Initial conditions and model calibration targets	100
     6.7.4.1 Data	100
     6.7.4.2 Methods	101
     6.7.4.3 1991 Initial Condition Results	101
   6.7.5 Summary	101
  6.8 WATER AND AIR TEMPERATURES	102
  6.9 PARTITIONING	103
   6.9.1 Overview	103
   6.9.2 Partition Coefficients	105
                                       iv                      Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4
                                  CONTENTS
     6.9.2.1 Water Column Organic Carbon Concentrations	106
     6.9.2.2 Water Column DOC	107
     6.9.2.3 Water Column/oc	107
     6.9.2.4 Sediment Organic Carbon Concentrations	108
     6.9.2.5 Porewater DOC	108
     6.9.2.6 Sediment foc	109
     6.9.2.7 Distribution of PCBs in sediment and water	 109
     6.9.2.8 Partitioning Summary	110
  6.10 VOLATILIZATION	110
    6.10.1 Overview	110
    6.10.2 Volatilization Mass Transfer	111
     6.10.2.1 Henry's Constant and Molecular Weight	Ill
     6.10.2.2 Film Transfer Coefficients	112
     6.10.2.3 Atmospheric PCB Concentrations	112
    6.10.3 Gas Exchange at Dams	113
  6.11 SEDIMENT PARTICLE MIXING	114
  6.12 DECHLORINATION	115
  6.13 SEDIMENT-WATER MASS TRANSFER	115
    6.13.1 Overview	115
    6.13.2 Calculation of kf for Tri+	118
     6.13.2.1 Data	118
     6.13.2.2 Approach	119
     6.13.2.3 kf Results	120
     6.13.2.4 Implementation in HUDTOX	121
    6.13.3 Analysis of congener and total PCB mass transfer coefficients	121
    6.13.4 Estimation of Particulate and Pore water Mass Transfer Rates	123
7. MASS BALANCE MODEL CALIBRATION	125
  7.1 OVERVIEW	125
  7.2 CALIBRATION STRATEGY	126
  7.3 SOLIDS DYNAMICS	127
    7.3.1 Calibration Approach	127
    7.3.2 Solids Calibration Results	129
     7.3.2.1 Burial Rates	129
     7.3.2.2 High and Low-flow Solids Loads	129
     7.3.2.3 Water Column Solids Concentrations	130
     7.3.2.4 Spring 1994 High Flow Event Solids Mass Balance	131
     7.3.2.5 Further Model-Data Comparisons	131
    7.3.3 Components Analysis for Solids	132
                                                              Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4
                                  CONTENTS
    7.3.4 Solids Calibration Summary	133
  7.4 HISTORIC ALTRI+CALIBRATION	134
    7.4.1 Calibration Approach	134
    7.4.2 Tri+Calibration Results	135
     7.4.2.1 Long-Term Sediment Tri+ Concentrations	135
     7.4.2.2 Longitudinal and Vertical Sediment Profiles	136
     7.4.2.3 Water Column Tri+ Concentrations	137
     7.4.2.4 High and Low-flow Tri+ Loads	137
     7.4.2.5 Further Model-Data Comparisons	138
    7.4.3 Components Analysis for Tri+	139
    7.4.4 Comparison to Low Resolution Sediment Coring Report (LRC) Results	140
    7.4.5 Tri+ Calibration Summary	141
  7.5 SENSITIVITY ANALYSES	142
    7.5.1 Solids loadings	143
     7.5.1.1 Solids loads at Fort Edward	143
     7.5.1.2 External Tributary Solids Loads	144
     7.5.1.3 Tributary Solids Loads Based on the Original Rating Curves	144
    7.5.2 Partition Coefficients	144
    7.5.3 Sediment-Water Mass Transfer Rates	145
     7.5.3.1 Variation of Sediment-water Transfer Rate	145
     7.5.3.2 Differences in Sediment Water Transfer between Cohesive and Non-Cohesive
     Areas	146
    7.5.4 Burial Rates in Cohesive Sediments	146
    7.5.5 Particle Mixing in Sediments	147
    7.5.6 Sediment Initial Conditions	147
    7.5.7 Henry's Law Constant	148
  7.6 1991-1997 HINDCAST APPLICATIONS	148
    7.6.1 Overview	148
    7.6.2 Approach	149
    7.6.3 Results	149
    7.6.4 Hindcast Applications Summary	 151
  7.7 CALIBRATION FINDINGS AND CONCLUSIONS	151
8. FORECAST SIMULATIONS FOR NO ACTION	155
  8.1 OVERVIEW	155
  8.2 No ACTION FORECAST SIMULATION DESIGN	156
    8.2.1 Hydrograph	157
    8.2.2 Solids Loads	157
    8.2.3 PCB Loads	157
                                        vi                     Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4
                                  CONTENTS
                                                                          Page

    8.2.4 Initial Conditions for the Forecast	158
    8.2.5 Specification Of Other Model Inputs	159
  8.3 No ACTION FORECAST RESULTS	159
    8.3.1 Forecast Results: Surface Sediment PCB Concentrations	159
    8.3.2 Forecast Results: Water Column PCB Concentrations	161
    8.3.3 Forecast Results: PCB Loads to the Lower Hudson River	162
  8.4 100-YEAR PEAK FLOW SIMULATION DESIGN	162
    8.4.1 Specification of the  100-year Flood Hydrograph and Loadings	162
  8.5 100-YEAR PEAK FLOW SIMULATION RESULTS	163
  8.6 SENSITIVITY ANALYSIS	164
    8.6.1 Sensitivity to Specification of Forecast Hydrograph	164
    8.6.2 Sensitivity to Solids Loads at Fort Edward	165
    8.6.3 Sensitivity to Tributary Solids Loads	165
    8.6.4 Sensitivity to Particle Mixing	166
    8.6.5 Sensitivity to Sediment Initial Conditions	166
  8.7 EXPOSURE CONCENTRATIONS FOR AUGUST 1999 AND DECEMBER 1999 RISK ASSESSMENTS
  	167
  8.8 PRINCIPAL FINDINGS AND CONCLUSIONS	167
    8.8.1 No Action Forecast	168
    8.8.2 100-Year Peak Flow Simulation	169
9. HUDTOX VALIDATION	m
  9.1 OVERVIEW	171
  9.2 VALIDATION APPROACH	171
    9.2.1 Validation Results	172
    9.2.2 Validation Summary	172

REFERENCES	173


Note:        Book 3 and Book 4 Tables of Contents are located in the respective books.
                                      vii                    Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4

                               LIST OF TABLES

TABLE                 TITLE

3-1         Comparison of Manning's 'n' from Previous Studies
3-2         Modeled Hudson River Flows at the Upstream Boundary of Thompson Island
            Pool
3-3         Comparison of Model Results with Rating Curve Data
3-4         Effect of Manning's 'n' on Model Results for 100-Year Flow Event
3-5         Effect of Turbulent Exchange Coefficients on Model Results
4-1         Summary of Inputs for Depth of Scour Model at Each High Resolution Core
4-2         Predicted Depth of Scour Range for 100 Year Flood at Each High Resolution
            Core Location
4-3         Thompson Island Pool Cohesive Sediment Expected Values of Solids Erosion and
            Mean Depth of Scour for 100-Year Flood, from Monte Carlo Analysis
5-1 a        HUDTOX Water Column Segment Geometry in Thompson Island Pool (2-
            dimensional segmentation)
5-1 b        HUDTOX Water Column Segment Geometry Below Thompson Island Pool (1-
            dimensional segmentation)
5-2 a        HUDTOX Sediment Segment Geometry in Thompson Island Pool for Surficial
            Sediment Segments (2-dimensional segmentation)
5-2 b        HUDTOX Sediment Segment Geometry Downstream of Thompson Island Pool
            for Surficial Sediment Segments (1-dimensional segmentation)
6-1         Sediment Data Sets Used in Development and Application of the HUDTOX
            Model
6-2         USGS Gage Information For Gages Used In Flow Estimation
6-3         Drainage Areas and Reference Tributaries Used to Estimate Daily Tributary
            Flows
6-4         Mean Seasonal USGS Flows For Select Flow Gauges in the Study Area for the
            Period 3/1/77 to 6/30/92
6-5         Seasonal Tributary Flow Adjustment Factors applied to Tributaries between Fort
            Edward and Stillwater, and between Stillwater and Waterford
6-6         Hudson River Flows Yearly Averages Estimated and USGS Gage Data
                                   viii                        Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4
                                LIST OF TABLES
TABLE                  TITLE
6-7         Summary of Available Solids Data for Mainstem Stations; Number of Samples
            and Source of Suspended Solids Sample Data by Station
6-8         Summary of Available Solids Data for Tributaries; Number of Samples and
            Source of Suspended Solids Sample Data by Station
6-9         Reference Tributaries for Unmonitored Tributaries
6-10        Tributary Solids Rating Curve Equations for Data-Based Rating Curves and
            Adjusted Rating Curves for the Long-Term Solids Balance
6-11        Cumulative Mainstem Solids (SS) Loads and Yields
6-12        Cumulative Solids Loads and Corresponding Yields by Reach (10/1/77 - 9/30/97)
6-13        Solids (TSS) Trapping Efficiencies by Reach Estimated by QEA Using SEDZL
            and Applied to Estimate Tributary TSS Loads in HUDTOX
6-14        Comparison of LTI and Literature-Based Annual Average Sediment Yield
            Estimates by Watershed
6-15        Number of Tri+ PCB Data Available by Source and Year at Each Hudson River
            Mainstem Sampling Station
6-16        Number of Days With Available PCB Data for Monitored Tributaries (Batten
            Kill, Hoosic River, Mohawk River)
6-17        Number of PCB Data Available for Each Congener and Total PCB by Source and
            Year at Each Hudson River Mainstem Sampling Station
6-18        Criteria and Factors Used in Adjustment of Thompson Island Dam West Shore
            PCB Data Bias
6-19        Tri+ and Total PCB Concentration Statistics for Monitored Tributaries
6-20        Comparison of Annual Tri+ PCB Load Estimates at Hudson River Mainstem
            Station Presented in the DEER and Calculated in this Report
6-21        Estimated Average Annual Load at Fort Edward by PCB Type from 1991 -1997
6-22        Cohesive/non-cohesive Sample Classification Criteria Applied to 1977 NYSDEC
            Data to Compute HUDTOX Sediment Tri+ Initial Conditions
6-23        Sample Count and Averaging Groups for Specifying 1977 Sediment Initial
            Conditions for HUDTOX from the NYSDEC Data
                                       ix                      Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4



TABLE

6-24

6-25

6-26

6-27


6-28

6-29

6-30
6-31
6-32

6-33

6-34

6-35

6-36

6-37

6-38
                   LIST OF TABLES
            TITLE
Averaging Groups for Specifying Sediment Initial Conditions from the 1991 GE
Composite Sampling Data
Pool-Wide Average Surficial Sediment Concentrations for Each PCB State
Variable
3-Phase Partition Coefficients Estimated from Phase 2 Water Column Data and
GE Sediment Data
Mass Fraction of Total PCB Represented by Tri+, BZ#1, BZ#4, and BZ#8 at
Mainstem Hudson River Stations Determined from GE and USEPA Phase 2 (P2)
Data
Estimated Partition Coefficients (KPOc, KDOC) for Total PCB by Source and
Agency at Mainstem Hudson River Stations
Estimated Partition Coefficients (KPOc, KDOc) for Total PCB at Mainstem Hudson
River Stations and Averaged Over Study Reach
Statistical Summary of Dissolved Organic Carbon (DOC) Water Column Data
Mean DOC Concentrations by Reach in Upper Hudson River
Mean Sediment f0c Values Specified from GE 1991 Composite Data for River
Mile intervals in HUDTOX
Illustration of Typical Low and High Flow Partitioning Behavior During Cold
Weather and Warm Weather Periods
Henry's Law  Constants Developed Experimentally by Brunner, et. al. (1990) for
Selected Congeners
Congener Distribution of Total PCB by Mass Fraction at Mainstem Hudson River
Stations Using 1993 USEPA Phase 2 Data (Number of observations)
Congener Distribution of Total PCB by Mass Fraction at Mainstem Hudson River
Stations Using 1991-1998 GE Data (Number of observations)
Estimated Henry's Law Constants (HLC) for Total and Tri+ PCB by Source and
Agency at Mainstem Hudson River Stations
Estimated Henry's Law Constants (HLC) for Total PCB at Mainstem Hudson
River Stations and Averaged Over Study Reach
                                                              Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4
                                LIST OF TABLES
TABLE                  TITLE
6-39         Estimated Molecular Weight for Total and Tri+ PCB by Source and Agency at
             Mainstem Hudson River Stations
6-40         Estimated Molecular Weight for Total PCB at Mainstem Hudson River Stations
             and Averaged Over Study Reach
6-41         Estimated Henry's Law Constants and Molecular Weight by PCB Type
6-42         Coefficients Used to Estimate Depth and Velocity as a Function of Cross-Section
             Average Flow in HUDTOX for Calculation of Liquid-Phase (KL) Air-Water
             Transfer Rates
6-43         Annual Average Bulk Sediment Concentrations by PCB Type
6-44         Annual Average Pore Water Concentrations by PCB Type
6-45         Estimated Sediment Properties in Thompson Island Pool Based on Area
             Weighting by  Sediment Type
6-46         Annual Time Series of Sediment-Water Mass Transfer Rate for Tri+ PCBs
6-47         Correlation of Particulate-mediated Sediment-Water Mass Transfer Coefficient
             with Suspended Solids Concentration, Fort Edward Flow, and Water Temperature
6-48         Annual Time Series of Pore Water and Particulate Mass Transfer Coefficients by
             PCB Type
7-1          HUDTOX Solids Model Calibration Parameter Values
7-2          HUDTOX Cohesive Sediment Resuspension and Armoring Parameters
7-3          HUDTOX Fraction Organic Carbon and Dissolved Organic Carbon
             Parameterization by Reach
7-4          HUDTOX PCB Model Calibration Parameter Values
7-5          Tri+ Mass Loads (1977-1997) at Mainstem Stations for Sensitivity Analyses
8-1          Sequencing of Annual Hydrographs to Develop 70-year Forecast Hydrograph
8-2          Surface Sediment Tri+ Initial Conditions for the No Action and 100-Year Event
             Simulations
8-3          Effect of the 100-Year Flood Event on the Non-cohesive (N) and Cohesive (C)
             Sediment Bed in Upper Hudson River Reaches between Fort Edward and Federal
             Dam (Year 1 - 3/28 to 4/13)
                                       xi                     Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4
                              LIST OF FIGURES
FIGURE                TITLE

1 -1         Hudson River Watershed
1-2         Upper Hudson River Watershed
1-3         Thompson Island Pool
2-1         Upper Hudson River Modeling Framework
2-2         Upper Hudson River Modeling Framework with Model Inputs
3-1         Thompson Island Pool Study Area
3-2         Thompson Island Pool RMA-2V Model Mesh
3-3         Thompson Island Pool Velocity Vectors for 100-Year Flow Event
3-4         Shear Stress Computed from Vertically Averaged Velocity
3-5         Thompson Island Pool Bottom Shear Stress for 100-Year Flow Event
4-1         Erosion versus Shear Stress in Cohesive Sediments
4-2         Armoring Depth versus Shear Stress
4-3 a,b       Likelihood of PCB Scour for Selected Phase 2 High Resolution Sediment Cores
4-4         Cumulative Percent versus Mean Depth of Scour for Cohesive Sediment in
            Thompson Island Pool
4-5         Cumulative Percent versus Total Solids Scoured from Cohesive Sediment in
            Thompson Island Pool
5-1         Conceptual Framework for the HUDTOX PCB Model
5-2         Illustration of Sediment Scour in the HUDTOX Model
5-3         Illustration of Sediment Burial in the HUDTOX Model
5-4 a,b       HUDTOX Model Water Column Segmentation Grid for Upper Hudson River,
            Parts A and B
5-4 c,d       HUDTOX Model Water Column Segmentation Grid for Upper Hudson River,
            Parts C  and D
5-5         Thompson Island Pool Study Area
5-6         Schematic of HUDTOX Water Column Segmentation Grid
5-7         HUDTOX Water Column Segment Depths by River Mile

                                   xii                       Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4
                               LIST OF FIGURES
FIGURE                 TITLE
5-8         Percent Cohesive Area Represented in HUDTOX Sediment by River Mile
6-1         Upper Hudson River Basin USGS Flow Gage Stations Used in HUDTOX
            Modeling
6-2         Log Pearson Flood Frequency Analysis for Fort Edward gage, Hudson River, NY
            Analysis
6-3         USGS Flow Time Series at Fort Edward from 1/1/77 - 9/30/97
6-4         Comparison of LTI-Estimated Flow (DAR-based, seasonally & high-flow
            adjusted) and the USGS-Reported Flow
6-5   •      Estimated Daily Average Mainstem and Tributary Flows for the Upper Hudson
            River between Fort Edward and Federal Dam (1/1/77-9/30/97)
6-6         Relative Percent Flow Contribution from Fort Edward and Tributaries between
            Fort Edward and Waterford
6-7         1993 - 1997 Estimated versus USGS-Reported Daily Average Flow at Stillwater
            and Waterford
6-8         1993 - 1997 Estimated versus USGS-Reported Daily Average Flow Time Series
            at Stillwater and Waterford
6-9         Upper Hudson River Basin Primary Mainstem and Tributary Sampling Locations
            for Solids Used in HUDTOX Modeling
6-10        Monitored and Unmonitored Subwatesheds for Solids Between Fort Edward and
            Waterford
6-11        GE versus USGS TSS Data at Fort Edward for High and Low Flow Data Pairs
            from 4/1/91 to 9/15/97
6-12        Observed Total Suspended Solids (TSS) versus Flow, 1977-1997 and TSS Rating
            Curves for this Period at Fort Edward, Stillwater and Waterford
6-13        Comparison of Total Suspended Solids (TSS) High-Flow Rating Curves for Fort
            Edward, 1977-1997, Using MVUE (Cohn et al. 1989) and Non-linear Regression
            Analysis.
6-14        Comparison of 1977-1990 and 1991-1997 Total Suspended Solids (TSS) Rating
            Curves at Fort Edward versus the 1977-1997 Rating Curve
                                      xiii                     Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4


FIGURE
6-15

6-16

6-17

6-18

6-19

6-20 a,b

6-21
6-22

6-23
6-24

6-25

6-26
6-27

6-28

6-29

6-30

6-31
                   LIST OF FIGURES
             TITLE
Tributary TSS Rating Curves: Based on Data and Adjusted to Achieve Solids
Balance
Mainstem and Tributary Suspended Solids Watershed Loads and Yields based on
HUDTOX Suspended Solids Loading Estimates (10/1/77-9/30/97)
Relative Percent Solids Contribution from Fort Edward and Tributaries between
Fort Edward and Waterford
Distribution of TSS Load Over Flow Range at Fort Edward, Stillwater, and
Waterford from 1977-1997
Upper Hudson River Basin Primary Mainstem and Tributary Sampling Locations
for PCB Data Used in HUDTOX Modeling
Distribution of Available Tri+ PCBs Concentration Data by Flow Intervals for
Mainstem Hudson River Sampling Stations (January 1977-May 1998)
Tri+ PCB concentrations and Load versus Flow at Fort Edward for Selected Years
Tri+ PCB Concentrations and Loads versus Total Suspended Solids (TSS)
Concentration at Fort Edward for Selected Years
Interpolated Daily Tri+ PCB Concentration and Flow at Fort Edward, 1977-1997
Examples of Apparent Tri+ Pulse Loading Events at Fort Edward in 1983 and
1994
Estimated Annual Tri+ Load at Mainstem Hudson River Sampling Stations
Compared to DEIR Estimates
Estimated Annual Tri+ Load at Hudson River Mainstem Sampling Stations
Distribution of Tri+ Load Over Flow Range at Fort Edward, Stillwater, and
Waterford from 1977-1997
Distribution of Tri+ Load Gain Across  Thompson Island Pool (TIP) Over Flow
Range for 1993-1997
Relative Contribution of Estimated External Tri+ PCB Loads to the Upper
Hudson River by Source, 1977-1997
Ratio of Congener BZ#4 to Total PCBs at Fort Edward, 1991-1997, GE and
Phase2 Data
Estimated Annual Total and Congener  PCB Loads at Fort Edward
                                       XIV
                                                              Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4

                               LIST OF FIGURES

FIGURE                 TITLE

6-32         1977 Sediment Tri+ PCB Initial Conditions Computed from the NYSDEC Data,
             Fort Edward to Federal Dam
6-33         1977 Sediment Tri+ PCB Initial Conditions Computed from the NYSDEC Data,
             Thompson Island Pool
6-34 a,b      1977 Sediment Tri+ Initial Conditions Computed from 1977 NYSDEC Data:
             Vertical Profiles
6-35         Comparison of Measures Total PCB & Tri+ PCB Data to 1991 Model Initial
             Conditions in the Top Layer (0-5 cm) of Cohesive and Non-cohesive Sediments
6-36         Comparison of Measured BZ#4 (#10) & BZ#52 Data to Model Initial Conditions
             in the Top Layer (0 to 5 cm) of cohesive and Non-cohesive Sediment
6-37         Comparison of Measured BZ#28 and BZ#90+101 Data to 1991 Model Initial
             Conditions in the Top Layer (0 to 5 cm) of Cohesive and Non-cohesive Sediments
6-38         Comparison of Measured BZ#138 Data to Model Initial Conditions in the Top
             Layer (0 to 5 cm) of Cohesive and Non-cohesive Sediments
6-39         Ratio of Average BZ#4 1991 Concentrations to Average BZ#52 1991
             Concentrations by Sediment Depth
6-40         Monthly Average Water Temperature Functions Applied in HUDTOX and
             Observed Water Temperatures
6-41         Comparison of Monthly Mean Temperatures at Mainstem Upper Hudson River
             Stations
6-42         Estimated Partition Coefficients for Total PCB by Station and by Source
6-43         Observed Dissolved Organic Carbon (DOC) Concentrations versus Normalized
             Flow between Fort Edward and Federal Dam
6-44         Observed Dissolved Organic Carbon (DOC) Data versus River Mile between Fort
             Edward and Federal Dam
6-45         River-wide Fraction Organic Carbon (foe) Function Based on a Power Function Fit
             to foe Data for Mainstem Hudson River Stations
6-46         Specified  Sediment Dissolved Organic Carbon (DOC) Concentrations in
             HUDTOX
                                       xv                     Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4

                               LIST OF FIGURES

FIGURE                 TITLE

6-47        foc versus River Mile from the 1991 GE Composite Sampling and Values
            Specified for Cohesive and Non-cohesive Sediment in HUDTOX
6-48        Estimated Henry's Law Constant for Selected Congeners Determined
            Experimentally by Brunner, et. al (1990)
6-49        Estimated Henry's Law Constants for Tri+ and Total PCB by Station and Data
            Source
6-50        Estimated Molecular Weight for Tri+ and Total PCB by Station and Data Source
6-51        Specification of Historical Atmospheric Gas-Phase PCB Boundary Concentrations
            for the 1977-1997 HUDTOX Calibration Period
6-52 a-c      Vertical Profiles of PCB3+ within Finely Segmented Sediment Cores Collected
            from the Upper Hudson River (from QEA, 1999)
6-53        Comparison of Same-Day Suspended Solids (TSS) Concentration Data at Fort
            Edward and Thompson Island Dam when TSS Concentration is less Than 10
            mg/L and Fort Edward Flow is less Than 10,000 cfs (1993-1997)
6-54        Temporal Patterns in Water Column Tri+ PCB Concentration at Fort Edward and
            Thompson Island Dam, Tri+ PCB Loading Increase Across Thompson Island
            Pool, and Calculated Effective Sediment-Water Mass Transfer Rates Across
            Thompson Island Pool
6-55        Computed Effective Mass Transfer Rate for Tri+ PCBs in Thompson Island Pool,
            1993-1997
6-56        Scatter Plots of Estimated Sediment-Water Mass Transfer Rate: Congeners versus
            Total PCB
6-57        Comparison of Estimated Site-Specific Water Column and Sediment Koc Values
            for Congeners as Determined in the DEFR
6-58        Average Observed versus Porewater and Particulate Predicted Relative Load Gain
            at Thompson Island Dam by Season, 1991-1997
6-59        Comparison of Congener Specific Apparent Sediment-Water Mass Transfer Rates
            by Date
6-60        Comparison of Fit using Ratio of Pore Water to Particulate Mass Transfer
            Coefficients to Average Observed Predicted Relative Load Gain at Thompson
            Island Dam by Season, 1991-1997
                                       xvi                     Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4
                               LIST OF FIGURES
FIGURE                 TITLE
7-1         Computed Annual Average Burial Rates, 1977-1997
7-2         Comparison between Model Estimated and Data Estimated In-River Solids
            Loadings Stratified by Fort Edward Flow at 10,000 cfs (1/1/77-9/30/97)
7-3 a,b      Comparison Between Computed and Observed Solids Concentrations at
            Mainstem Sampling Stations
7-4         Comparison Between Computed and Observed Total Suspended Solids
            Concentrations (TSS) for the Spring 1983 High Flow Event
7-5         Comparison Between Computed and Observed Total Suspended Solids
            Concentrations (TSS) for the Spring 1993 High Flow Event
7-6         Comparison Between Computed and Observed Total Suspended Solids
            Concentrations (TSS) for the Spring 1994 High Flow Event
7-7         Comparison Between Computed and Observed Total Suspended Solids
            Concentrations (TSS) for the Spring 1997 High Flow Event
7-8         Comparison Between Computed and Observed Suspended Solids Concentrations
            for Fort Edward Flows less than 10,000 cfs
7-9         Comparison Between Computed and Observed Suspended Solids Concentrations
            for Fort Edward Flows greater than 10,000 cfs
7-10        Comparison Between Computed and Observed Probability Distributions for Total
            Suspended Solids (TSS) for Fort Edward Flows less than 10,000 cfs
7-11         Comparison Between Computed and Observed Probability Distributions for Total
            Suspended Solids (TSS) for Fort Edward Flows greater than 10,000 cfs
7-12        Computed Cumulative Sediment Bed Elevation Change in Thompson Island Pool,
            1977-1997
7-13        Computed Annual Average Solids Burial Rates, 1977-1997
7-14        Computed Solids Mass Balance Components Analysis for 1977-1997
7-15 a       Comparison between Computed and Observed (Surficial and Depth-Composited)
            Sediment Tri-i- Concentrations for Thompson Island Pool
7-15 b       Comparison between Computed and Observed (Surficial and Depth-Composited)
            Sediment Tri+ Concentrations for Schuylerville Reach
                                      xvii                     Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                 HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4

                               LIST OF FIGURES

FIGURE                 TITLE

7-15 c       Comparison between Computed and Observed (Surficial and Depth-Composited)
            Sediment Tri+ Concentrations for Stillwater Reach
7-15 d       Comparison between Computed and Observed (Surficial and Depth-Composited)
            Sediment Tri+ Concentrations for Waterford Reach
7-15 e       Comparison between Computed and Observed (Surficial and Depth-Composited)
            Sediment Tri+ Concentrations for Federal Dam Reach
7-16        Comparison Between Computed and Observed Depth-Averaged Sediment Tri+
            Concentrations for Thompson Island Pool in 1984
7-17        Comparison Between Computed and Observed Depth-Averaged (0-5 cm)
            Sediment Tri+ Concentrations from Fort Edward to Federal Dam in 1991
7-18        Comparison Between Computed and Observed Depth-Averaged (5-10 cm)
            Sediment Tri+ Concentrations from Fort Edward to Federal Dam in 1991
7-19        Comparison Between Computed and Observed Depth-Averaged (10-26 cm)
            Sediment Tri+ Concentrations from Fort Edward to Federal Dam in 1991
7-20 a,b      Comparison between Computed and Observed Sediment Tri+ Concentrations at
            Mainstem Stations
7-20 c       Comparison between Computed and Observed Tri+ Concentrations at Thompson
            Island Dam, 1991-1997
7-21        Comparison of Same Day Tri+ Concentration Data by Source at Fort Edward,
            Stillwater, and Waterford
7-22        Comparison between Model Estimated and Data Estimated In-River Tri+
            Loadings from 1977-1997 Stratified by Fort Edward Flow at 10,000 cfs
7-23        Comparison between Computed and Observed Tri+ Concentrations for the Spring
            1983 High Flow Event
7-24        Comparison between Computed and Observed Tri+ Concentrations for the Spring
            1993 High Flow Event
7-25        Comparison between Computed and Observed Tri+ Concentrations for the Spring
            1994 High Flow Event
7-26        Comparison between Computed and Observed Tri+ Concentrations for the Spring
            1997 High Flow Event
                                      xviii                    Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
              FURTHER SITE CHARACTERIZATION AND ANALYSIS
              Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4
                               LIST OF FIGURES
FIGURE                 TITLE
7-27         Comparison between Computed and Observed Tri+ Concentrations for Fort
             Edward Flow Less Than 10,000 cfs
7-28         Comparison between Computed and Observed Tri+ Concentrations for at Fort
             Edward Flow Greater Than 10,000 cfs
7-29         Comparison Between Computed and Observed Probability Distributions for Tri+
             at Fort Edward Flow Less Than 10,000 cfs
7-30         Comparison Between Computed and Observed Probability Distributions for Tri+
             at Fort Edward Flow Greater Than 10,000 cfs
7-31         Computed Tri+ PCB Mass Balance Components Analysis for 1977-1997
7-32         Computed Cumulative Contribution Tri+ Load Gain between Mainstem Hudson
             River Sampling Stations from 1991 to 1997
7-33         Sediment Responses in Thompson Island Pool to Alternate Solids Loads at Fort
             Edward
7-34         Sediment Responses in Waterford to Alternative Solids Loads at Fort Edward
7-35         Sediment Responses in Thompson Island Pool to Changes in Tributary Solids
            Loadings
7-36         Sediment Responses at Waterford to Changes in Tributary Solids Loadings
7-37         Water Column Responses to Changes in Tributary Solids Loadings
7-38         Sediment Responses in Thompson Island Pool to Changes in Tributary Solids
             Loads due to Specification of Rating Curves
7-39         Sediment Responses at Waterford to Changes in Tributary Solids Loads Due to
             Specification of Rating Curves
7-40         Water Column Responses to Changes in Tributary Solids Loadings Due to
             Specification of Rating Curve
7-41         Sediment Responses in Thompson Island Pool to Changes in Partitioning
7-42         Sediment Responses in Waterford Reach to Changes in Partitioning
7-43         Water Column Responses to Changes in Partitioning
7-44         Time Series for Effective Mass Transfer Rate in HUDTOX
                                       xix                     Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
              FURTHER SITE CHARACTERIZATION AND ANALYSIS
              Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4
                               LIST OF FIGURES
FIGURE                 TITLE
7-45         Sediment Responses in Thompson Island Pool to Changes in Sediment-Water
             Mass Transfer Rate
7-46         Sediment Responses in Waterford Reach to Changes in Sediment-Water Mass
             Transfer Rate
7-47         Water Column Responses to Changes in Sediment-Water Mass Transfer Rate
7-48         Sediment Responses in Thompson Island Pool to Changes in Cohesive and Non-
             cohesive Specific Sediment to Water Effective Mass Transfer Rates
7-49         Sediment Responses in Waterford to Changes in Cohesive and Non-cohesive
             Specific Sediment to Water Effective Mass Transfer Rates
7-50         Responses of Burial Rates in Cohesive Sediments to Changes in Gross Settling
             Velocities
7-51         Responses of Burial Rates in Non-Cohesive Sediments to Changes in Gross
             Settling Velocities
7-52         Sediment Responses in Thompson Island Pool to Changes in Gross Settling
             Velocities
7-53         Sediment Responses in Waterford Reach to Changes in Gross Settling Velocities
7-54         Water Column Responses to Changes in Gross Settling Velocities
7-55         Sediment Responses in Schuylerville Reach to Enhanced Mixing (top 6 cm) in
             Non-cohesive Sediments
7-56         Sediment Responses in Stillwater Reach to Enhanced Mixing (top 6 cm) in Non-
             cohesive Sediments
7-57         Sediment Responses at Waterford to Enhanced Mixing (top 6 cm) in Non-
             cohesive Sediments
7-58         Sediment Responses in Federal Dam Reach to Enhanced Mixing (top 6 cm) in
             Non-cohesive Sediments
7-59         Sediment Responses in Thompson Island Pool to Changes in Sediment Initial
             Conditions
7-60         Sediment Responses in Waterford to Changes in Sediment Initial Conditions
7-61         Water Column Responses to Changes in Sediment Initial Conditions
7-62         Water Column Responses to Changes in Henry's Law Constant
                                       xx                     Limno-Tech, Inc.

-------
                       PHASE 2 REPORT - REVIEW COPY
             FURTHER SITE CHARACTERIZATION AND ANALYSIS
             Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4



FIGURE

7-63

7-64

7-65

7-66 a



7-66 b


7-67 a

7-67 b

7-67 c

7-67 d

7-67 e

7-67 f

7-68 a

7-68 b

7-68 c
                   LIST OF FIGURES
            TITLE
Predicted versus Observed BZ#4, BZ#28 and BZ#52 Concentrations Using
Historical Calibration Model Parameters
Comparison between Computed Surficial Sediment Tri+, BZ#28, BZ#52 and
BZ#4 Concentrations for Thompson Island Pool
Predicted versus Observed BZ#4, BZ#28 and BZ#52 Concentrations Using
Sediment-Specific Partitioning (from GE Data)
Predicted versus Observed BZ#4, BZ#28 and BZ#52 Concentrations Using
Sediment-Specific Partitioning (from GE Data) and Particulate and Porewater
Sediment-Water Mass Transfer Pathways
Predicted versus Observed BZ[#90+101], BZ#138 and Total PCB Concentrations
Using Sediment-Specific Partitioning (from GE Data) and Particulate and
Dissolved Sediment-Water Mass Transfer Pathways
Predicted versus Observed BZ#4 Concentrations below Thompson Island Dam,
1991-1993
Predicted versus Observed BZ#28 Concentrations below Thompson Island Dam,
1991-1993
Predicted versus Observed BZ#52 Concentrations below Thompson Island Dam,
1991-1993
Predicted versus Observed BZ#[90+101] Concentrations below Thompson Island
Dam, 1991-1993
Predicted versus Observed BZ#138 Concentrations below Thompson Island Dam,
1991-1993
Predicted versus Observed Total PCB Concentrations below Thompson Island
Dam, 1991-1993
Comparison of Model versus Observed Congener Concentrations Ratios:
Thompson Island Pool, September 25, 1996 Float Study
Comparison of Model versus Observed Congener Concentrations Ratios:
Thompson Island Pool, September 26, 1996 Float Study
Comparison of Model versus Observed Congener Concentrations Ratios:
Thompson Island Pool, June 4, 1997 Float Study
                                       XXI
                                                  Limno-Tech, Inc.

-------
                        PHASE 2 REPORT - REVIEW COPY
              FURTHER SITE CHARACTERIZATION AND ANALYSIS
              Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4

                                LIST OF FIGURES

FIGURE                 TITLE

7-68 d        Comparison of Model versus Observed Congener Concentrations Ratios:
             Thompson Island Pool, June 17, 1997 Float Study
7-69         Model versus Observed Down-river [BZ#28]/[BX#52] Ratios by Season,  1991 -
             1997
7-70         Model versus Observed Down-river [BZ#28]/[BZ#52]Ratios Stratified by Fort
             Edward Flow (<10,000 cfs and >10,000 cfs), 1991-1997
8-1           70-Year Hydrograph for the No Action Forecast Simulation: 1998-2067
8-2           Observed Total PCB and Tri+ PCB Concentrations at Fort Edward During 1997
             and  1998
8-3           Data-Based Estimate of Annual Total and Tri+ PCB Load by Year at Fort
             Edward, 1991-1997
8-4 a         Forecast Sediment Tri+ Concentrations for Thompson Island Pool with Constant
             Upstream Tri+ Concentrations at 10 ng/L, 30 ng/L, and 0 ng/L, 1998-2067
8-4 b         Forecast Sediment Tri+ Concentrations for the Schuylerville Reach with Constant
             Upstream Tri+ Concentrations at 10 ng/L, 30 ng/L, and 0 ng/L, 1998-2067
8-4 c         Forecast Sediment Tri+ Concentrations for the Stillwater Reach with Constant
             Upstream Tri+ Concentrations at 10 ng/L, 30 ng/L, and 0 ng/L, 1998-2067
8-4 d         Forecast Sediment Tri+ Concentrations for the Waterford Reach with Constant
             Upstream Tri+ Concentrations at 10 ng/L, 30 ng/L, and 0 ng/L, 1998-2067
8-4 e         Forecast Sediment Tri+ Concentrations for the Federal Dam Reach with Constant
             Upstream Tri+ Concentrations at 10 ng/L, 30 ng/L, and 0 ng/L, 1998-2067
8-5 a         Predicted Sediment Tri+ Concentrations for Thompson Island Pool with
             Forecasted Constant Upstream Tri+ Concentration at 10 ng/L
8-5 b         Predicted Sediment Tri+ Concentrations for Schuylerville Reach with Forecasted
             Constant Upstream Tri+ Concentration at 10 ng/L
8-5 c         Predicted Sediment Tri+ Concentrations for Stillwater Reach with Forecasted
             Constant Upstream Tri+ Concentration at 10 ng/L
8-5 d         Predicted Sediment Tri+ Concentrations for Waterford Reach with Forecasted
             Constant Upstream Tri+ Concentration at 10 ng/L
                                        xxii                     Limno-Tech, Inc.

-------
                        PHASE 2 REPORT - REVIEW COPY
              FURTHER SITE CHARACTERIZATION AND ANALYSIS
              Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4

                                LIST OF FIGURES

FIGURE                  TITLE

8-5 e        Predicted Sediment Tri+ Concentrations for Federal Dam Reach with Forecasted
             Constant Upstream Tri+ Concentration at 10 ng/L
8-6 a        Forecast Average Annual Tri+ Concentrations at Thompson Island Dam and
             Schuylerville with Constant Upstream Concentrations of 10 ng/1, 30 ng/L, and 0
             ng/L, Tri+ at Fort Edward, 1998-2067
8-6 b        Forecast Average Annual Tri+ Concentrations at Stillwater and Waterford with
             Constant Upstream Concentrations of 10 ng/1, 30 ng/L, and 0 ng/L, Tri+ at Fort
             Edward, 1998-2067
8-7 a        Forecast Average Summer Tri+ Concentrations at Thompson Island Dam and
             Schuylerville with Constant Upstream Concentrations of 10 ng/L, 30 ng/L, and 0
             ng/L Tri+ at Fort Edward, 1998-2067
8-7 b        Forecast Average Summer Tri+ Concentrations at Stillwater and Waterford with
             Constant Upstream Concentrations of 10 ng/L, 30 ng/L, and 0 ng/L Tri+ at Fort
             Edward, 1998-2067
8-8 a        Predicted Average Annual Water Column Tri+ Concentrations at Thompson
             Island Dam and Schuylerville with Forecasted Constant Upstream Tri+
             Concentration at 10 ng/L, 1977-2067
8-8 b        Predicted Average Annual Water Column Tri+ Concentrations at Stillwater and
             Waterford with Forecasted Constant Upstream Tri+ Concentration at 10
             ng/L, 1977-2067
8-9 a        No-action Forecast Annual Tri+ Load to the Lower Hudson River with Constant
             Upstream Concentrations of 10 ng/L,  30 ng/L, and 0 ng/L Tri+ at Fort Edward,
             1998-2067
8-9 b        No-action Forecast Cumulative Annual Tri+ Load to the Lower Hudson River
             with Constant Upstream Concentrations of 10 ng/L, 30 ng/L, and 0 ng/L Tri+ at
             Fort Edward, 1998-2067
8-10         Adjustment of the Fort Edward Hydrograph to Include the 100 Year Flow (47,330
             cfs)
8-11         Predicted 100 Year Event (3/28 to 4/13) Impact on Tri+ PCB Levels at Thompson
             Island Dam (West)
8-12         Predicted 100 Year Event (3/28 to 4/13) Impact on Tri-t- PCB Levels at Federal
             Dam
                                       xxiii                     Limno-Tech, Inc.

-------
                        PHASE 2 REPORT - REVIEW COPY
              FURTHER SITE CHARACTERIZATION AND ANALYSIS
              Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4

                                LIST OF FIGURES

FIGURE                  TITLE

8-13         Cumulative Net Increase of Tri+ PCB Mass Loading at Various Locations in the
             Upper Hudson River Due to the 100 Year Flood Event (versus the No Action
             Scenario)
8-14 a        Forecast Sediment Tri+ Concentrations for Thompson Island Pool for Alternative
             Hydrographs (Constant Upstream Tri + Concentration of 10 ng/L) at Fort Edward
8-14 b        Forecast Sediment Tri+ Concentrations for Schuylerville Reach for Alternative
             Hydrographs (Constant Upstream Tri + Concentration of 10 ng/L) at Fort Edward
8-14 c        Forecast Sediment Tri+ Concentrations for the Stillwater Reach for Alternative
             Hydrographs (Constant Upstream Tri + Concentration of 10 ng/L) at Fort Edward
8-14 d        Forecast Sediment Tri+ Concentrations for the Waterford Reach for Alternative
             Hydrographs (Constant Upstream Tri + Concentration of 10 ng/L) at Fort Edward
8-14 e        Forecast Sediment Tri+ Concentrations for the Federal Dam Reach for Alternative
             Hydrographs (Constant Upstream Tri + Concentration of 10 ng/L) at Fort Edward
8-15 a        Forecast Annual Average Tri+ Concentrations at Thompson Island Dam and
             Schuylerville for Alternative Hydrographs (Constant Upstream Tri +
             Concentration of 10 ng/L at Fort Edward),  1998-2067
8-15 b        Forecast Annual Average Tri+ Concentrations at Stillwater and Waterford for
             Alternative Hydrographs (Constant Upstream Tri + Concentration of 10 ng/L at
             Fort Edward), 1998-2067
8-16         Sensitivity of Thompson Island Pool Surface Sediment Tri+ Concentrations to an
             Alternative Total Suspended Solids Load at Fort Edward, 1998-2047
8-17 a        Sensitivity of Thompson Island Pool Surface Sediment Tri+ Concentrations to
             Changes in External Tributary Solids Loadings, 1998-2067
8-17 b        Sensitivity of Thompson Island Dam to Schuylerville Surface Sediment Tri+
             Concentrations to Changes in External Tributary Solids Loadings, 1998-2067
8-17 c        Sensitivity of Schuylerville to Stillwater Surface Sediment Tri+ Concentrations to
             Changes in External Tributary Solids Loadings, 1998-2067
8-17 d        Sensitivity of Stillwater to Waterford Surface Sediment Tri-f Concentrations to
             Changes in External Tributary Solids Loadings, 1998-2067
8-17 e        Sensitivity of Waterford to Federal Dam Surface Sediment Tri+ Concentrations to
             Changes in External Tributary Solids Loadings, 1998-2067
                                        xxiv                     Limno-Tech, Inc.

-------
                        PHASE 2 REPORT - REVIEW COPY
              FURTHER SITE CHARACTERIZATION AND ANALYSIS
              Volume 2D - REVISED BASELINE MODELING REPORT
                  HUDSON RIVER PCBs REASSESSEMENT RI/FS

BOOK 1 of 4

                                 LIST OF FIGURES

FIGURE                  TITLE

8-18 a       Sensitivity of Thompson Island Pool Surface Sediment Tri+ Concentrations to
             Enhanced Mixing (top 6cm) in Non-Cohesive Sediments, 1998-2067
8-18 b       Sensitivity of Thompson Island Dam to Schuylerville Surface Sediment Tri+
             Concentrations to Enhanced Mixing (top 6cm) in Non-Cohesive Sediments, 1998-
             2067
8-18 c       Sensitivity of Schuylerville to Stillwater Surface Sediment Tri+ Concentrations to
             Enhanced Mixing (top 6cm) in Non-Cohesive Sediments, 1998-2067
8-18 d       Sensitivity of Stillwater to Waterford Surface Sediment Tri+ Concentrations to
             Enhanced Mixing (top 6cm) in Non-Cohesive Sediments, 1998-2067
8-18 e       Sensitivity of Waterford to Federal Dam Surface Sediment Tri+ Concentrations to
             Enhanced Mixing (top 6cm) in Non-Cohesive Sediments, 1998-2067
8-19         Sensitivity of Tri+ Concentrations at Stillwater to Enhanced Mixing (top 6 cm) in
             Non-cohesive Sediments, 1998-2067
8-20 a       Sensitivity of Thompson Island Pool Surface Sediment Tri+ Concentrations to
             Specification of Sediment Initial Conditions, 1998-2067
8-20 b       Sensitivity of Schuylerville Reach Surface Sediment Tri+ Concentrations to
             Specification of Sediment Initial Conditions, 1998-2067
8-20 c       Sensitivity of Stillwater Reach Surface Sediment Tri+ Concentrations to
             Specification of Sediment Initial Conditions, 1998-2067
8-20 d       Sensitivity of Waterford Reach Surface Sediment Tri+ Concentrations to
             Specification of Sediment Initial Conditions, 1998-2067
8-20 e       Sensitivity of Federal Dam Reach Surface Sediment Tri+ Concentrations to
             Specification of Sediment Initial Conditions, 1998-2067
8-21 a       Sensitivity of Forecasted Average Annual Tri+ Concentrations to Specification of
             Initial Conditions at Thompson Island Dam and Schuylerville, 1998-2067
8-21 b       Sensitivity of Forecasted Average Annual Tri+ Concentrations to Specification of
             Initial Conditions at Stillwater and Waterford, 1998-2067
9-1          HUDTOX Validation: Comparison of Predicted and Observed Thompson Island
             Dam Tri+ Concentrations
9-2          HUDTOX Validation: Comparison of Predicted and Observed Schuylerville Tri+
             Concentrations
                                        xxv                     Limno-Tech, Inc.

-------
                     PHASE 2 REPORT - REVIEW COPY
            FURTHER SITE CHARACTERIZATION AND ANALYSIS
            Volume 2D - REVISED BASELINE MODELING REPORT
                HUDSON RIVER PCBs REASSESSEMENT RI/FS
BOOK 1 of 4



FIGURE

9-3

9-4

9-5
                 LIST OF FIGURES
           TITLE
HUDTOX Validation: Predicted versus Observed Tri+ Concentrations at
Thompson Island Dam and Schuylerville
Monthly Average Scatter Plots of Observed Data and Model Output at Thompson
Island Dam, 1998-1999
Monthly Average Scatter Plots of Observed Data and Model Output at
Schuylerville, 1998-1999
                                   XXVI
                                                        Limno-Tech, Inc.

-------
                                    GLOSSARY
BAF
Bayesian updating

BMR
BSRE
BURE
CEAM
CD-ROM
cfs
cm
Corp.
DAR
deg.C
DEIR
DOC
DOSM
e.g.
etal.
FA
FEMA
foe
fps
g
GBTOX
GE
CIS
GLI
HEC-2

HOC
HUDTOX
i.e.
IADN
kg
LDEO
Likelihood profile

LRC
m/s
mg/1
mi2
MT
MVUE
Biota Accumulation Factor
calibration procedure based on conditional probability in Bayes Rule
(optimizes predicted distribution based on observed distribution).
Baseline Modeling Report
Beale's Stratified Ratio Estimator
Beale's Unstratified Ratio Estimator
Center for Exposure Assessment Modeling
Compact Disc - Read Only Memory
Cubic feet per second
Centimeter
Corporation
Drainage Area Ratio
Degree Celsius
Data Evaluation and Interpretation Report
Dissolved Organic Carbon
Depth of Scour Model
For example
and others
Flow Average (Phase 2 Water Column Monitoring Program)
Federal Emergency Management Agency
Fraction organic carbon
Feet per second
Gram
Green Bay Toxic Chemical Model
General Electric
Geographic Information System
Great Lake Initiative
US Army Corps of Engineers, Hydraulic Engineering Center,
Surface Water Profile Model
Hydrophobic Organic Chemicals
Hudson River Toxic Chemical Model
That is
Integrated Atmospheric Deposition Network
Kilogram
Lamont-Doherty Earth Observatory
maximum likelihood estimation technique to determine parameters of
prior and posterior distributions
Low Resolution Sediment Coring Report
Meters per second
Milligrams per liter
Square miles
Metric Ton
Minimum Variance Unbiased Estimator
                                     XXVII
                                             Limno-Tech, Inc.

-------
NAPL
NPDES
ng/m3
ng/1
NGVD
NOAA
NWS
NYSDEC
NYSDOH
NYSDOT
OC
PCBs
PMCR
Posterior distribution

Prior distribution
RBMR
RMA-2V
ROD
RPI
SS
TDD
TIN
TIP
TSCA
TSF  (tsf)
TSS
ug/g  (ppm)
ug/L
USAGE
USEPA
USGS
WASPS
TOXI5
WY
Non-aqueous Phase Liquid
National Pollutant Discharge Elimination System
Nanograms per cubic meter
Nanograms per liter
National Geodetic Vertical Datum
National Oceanic and Atmospheric Administration
National Weather Service
New York State Department of Environmental Conservation
New York State Department of Health
New York State Department of Transportation
Organic Carbon
Polychlorinated Biphenyls
Preliminary Model Calibration Report
optimized input distribution based on Bayesian updating calibration
procedure; revised prior distribution
empirical or likelihood-function-based probability distribution initially
specified in FISHRAND before implementing any calibration procedure;
"best guess"
Revised Baseline Modeling Report
Thompson Island Pool Hydrodynamic Model
Record of Decision
Rensselaer Polytechnic Institute
Suspended Solids
Thompson Island Dam
Triangulated Irregular Network
Thompson Island Pool
Toxic Substances Control Act
Temperature slope factor
Total Suspended Solids
Micrograms per gram (parts per million)
Micrograms per liter
United States Army Corps of Engineers
United States Environmental Protection Agency
United States Geological Survey
(USEPA) Water Quality Analysis Simulation Program, Version 4
Toxic Chemical Module in WASPS
Water year
                                        XXVlll
                                                                 Limno-Tech, Inc.

-------
Executive Summary

-------
                           EXECUTIVE SUMMARY
                  REVISED BASELINE MODELING REPORT
                                 JANUARY 2000

This report presents results and findings from the application of mathematical models for
PCB physical/chemical transport and fate, as well as PCB bioaccumulation in the Upper
Hudson River.  The modeling effort for the Hudson River PCBs site Reassessment has
been designed to predict future levels of PCBs in Upper Hudson River sediment, water
and fish.  This report provides predictions under baseline conditions, that is, without
remediation of PCB-contaminated sediment in the Upper Hudson River (equivalent to a
No Action scenario).  The predicted sediment, water and fish PCB concentrations from
the models are used as inputs in the Human Health and Ecological Risk Assessments.
Subsequently, the models will be used in the  Feasibility Study  (the Phase 3 Report) to
help evaluate and compare the effectiveness of various remedial scenarios.

The Revised Baseline Modeling Report (RBMR or Revised BMR) incorporates changes
to the  May  1999 Baseline Modeling Report (BMR)  based on public comments and
additional analyses, and supercedes the May 1999 report. The Revised BMR consists of
four books. Books 1 and 2 are on the transport and fate models, with Book 1 containing
the report text and Book 2 containing the corresponding tables,  figures and plates.
Similarly, Books 3 and 4 are on the bioaccumulation  models, with Book 3 containing the
report  text and Book  4 containing  the  corresponding  tables,  figures  and plates.
Predictions of future PCB concentrations in sediment and water from the transport and
fate  models  are  used   as  input  values for the  bioaccumulation  models.    The
bioaccumulation models forecast PCB concentrations  in various fish  species  based on
these inputs.

MODELING OBJECTIVES

The overall goal of the modeling is to develop scientifically credible models capable of
answering the following principal questions:

   •   When will PCB levels in fish populations recover to levels meeting human health
       and ecological risk criteria under continued No Action?

   •   Can remedies other than No Action significantly shorten the time required to
       achieve acceptable risk levels?

   •   Are there contaminated sediments now  buried that  are likely  to become
       "reactivated" following a  major flood, possibly resulting in an  increase in
       contamination of the fish population?

The work presented in this Revised BMR  provides information relevant to the first and
third questions. Forecasts regarding the potential impacts of various remedial scenarios,
thus addressing the second question, will be presented in the Feasibility Study (the Phase
3 Report).
                                         ES-l                             USEPA

-------
 MODEL DEVELOPMENT

 A large  body of information  from site-specific  field measurements (documented in
 Hudson River Database Release 4.1), laboratory experiments and the scientific literature
 was synthesized within the models to develop the PCB transport and fate and the PCB
 bioaccumulation models. Data from numerous sources were utilized including USEPA,
 the New York State Department of Environmental Conservation, the National Oceanic
 and  Atmospheric  Administration, the US  Geological Survey and the General Electric
 Company.

 The  proposed modeling approach and preliminary demonstrations of model outputs were
 made available for public review in the Preliminary Model Calibration Report (PMCR),
 which was issued in October 1996. The modeling framework of the PMCR was revised
 based on a peer review and public comment,  as well as the incorporation of additional
 data.  The baseline  modeling effort and results  were documented in the Baseline
 Modeling Report (BMR) issued in May 1999.  USEPA decided to revise the BMR to
 reflect changes to the models based on public comment and additional analyses that were
 conducted.  The Revised BMR includes model refinements, additional  years of data,
 longer model forecasts, validation  to an  independent dataset,  and  additional model
 sensitivity analyses. This Revised BMR supercedes the May 1999 BMR.

Transport and Fate Models

HUDTOX - The  backbone of the modeling  effort is the Upper Hudson River  Toxic
Chemical Model (HUDTOX).  HUDTOX was developed to simulate PCB transport and
fate  for 40 miles  of the Upper Hudson River from Fort Edward to  Troy, New  York.
HUDTOX is a transport and fate model, which is based on the principle of conservation
of mass. The fate and transport model simulates PCBs in the water column and sediment
bed, but  not in fish.  It balances inputs, outputs and internal sources and sinks for the
Upper Hudson River.   Mass balances are constructed first for water, then  solids and
bottom sediment,  and finally PCBs.  External inputs  of water,  solids loads and PCB
loads, plus values for many  internal model coefficients,  were specified  from field
observations.  Once inputs are specified, the  remaining internal model  parameters are
calibrated so that concentrations computed by the model agree with field observations.
Model calculations of forecasted PCB concentrations in  water and sediment  from
HUDTOX  are used  as  inputs for  the forecasts  of the bioaccumulation models (as
described in Books 3 and 4).

Depth of Scour Model (DOSM) - The Depth of Scour Model was principally developed
to provide spatially-refined information on sediment erosion depths in response to high-
flow events such as a  100-year peak flow.  The DOSM is a two-dimensional, sediment
erosion model that was applied to the Thompson Island Pool.  The Thompson Island Pool
is characterized by high levels of PCBs in the cohesive sediments. DOSM is linked with
a hydrodynamic model that predicts the velocity  and shear stress (force of the  water
acting on the sediment surface) during high flows.  There is also a linkage between the
                                         ES-2                             USEPA

-------
 DOSM and  HUDTOX.   Relationships between  river flow  and cohesive sediment
 resuspension were developed using the DOSM for a range of flows below the 100-year
 peak flow.  These relationships were used in the HUDTOX model for representing flow-
 dependent resuspension.

 Bioaccumulation Models

 Three  separate  bioaccumulation  models  were developed  in a  sequential  manner,
 beginning with a simple,  data-driven empirical approach (Bivariate  BAF Analysis),
 followed by a  probabilistic  food  chain  model,  and  ending with  a time-varying,
 mechanistic approach (FISHRAND).  The three approaches are complementary, with
 each progressively more complex model building on the results of the preceding, simpler
 effort.  All three bioaccumulation models are presented in  the Revised BMR; however,
 the FISHRAND model is the final bioaccumulation model  that is used to predict future
 fish PCB body burdens.

 Bivariate BAF Analysis - The Bivariate BAF (Bioaccumulation Factor) Analysis is a
 simple empirical approach that draws on the wealth of historical PCB data for the Hudson
 River to relate PCB levels in water and sediments (two  variables, or "bivariate") to
 observed PCB levels in fish. This analysis  is useful  in understanding the relative
 importance of water and sediment sources on particular species of fish. As this empirical
 approach does not  describe causal relationships,  the analysis has limited predictive
 capabilities and accordingly was not used for forecasts.

 Empirical Probabilistic Food Chain Model  - The Empirical Probabilistic Food Chain
 Model  is a more sophisticated representation of the steady-state relationships between
 fish body burdens and PCB exposure concentrations in  water and sediments.  The model
 combines information from available PCB exposure measurements with knowledge about
 the ecology of different fish species and the food chain relationships among  larger fish,
 smaller fish,  and  invertebrates in the  water column and sediments. The Probabilistic
 Model  provides  information  on the  expected range of uncertainty and  variability
 associated with the estimates of average fish body burdens.

 (FISHRAND) Mechanistic Time-Varying Model - The FISHRAND model is based on
 the peer-reviewed uptake model developed by Gobas  (1993  and 1995) and  provides a
mechanistic, process-based, time-varying representation of PCB bioaccumulation. This
is the same form of the model that was used  to develop criteria under the Great Lakes
Initiative (USEPA, 1995). The FISHRAND model  incorporates distributions instead of
point estimates for input parameters, and calculates distributions of fish body burdens
from which particular point estimates can be obtained, for example, the median, average,
or 95th percentile.  FISHRAND was used to predict the future fish PCB body burdens for
the Human Health and Ecological Risk Assessments.
                                         ES-3                             USEPA

-------
MODEL CALIBRATION

The principal HUDTOX application was a long-term historical calibration for a 21-year
period from 1977 through 1997.  Consistent with the Reassessment principal questions,
emphasis  was placed on  calibration of the  model to long-term trends in sediment and
water column PCB concentrations. However, a short-term hindcast calibration test was
also conducted from  1991 to 1997 to establish model performance for certain individual
PCB congeners.

Model applications included mass balances for seven different PCB forms: total PCBs,
Tri+, and five  individual PCB congeners  (BZ#4, BZ#28, BZ#52,  BZ#[90+101] and
BZ#138).  Total PCBs represents the sum of all measured PCB congeners and represents
the  entire  PCB  mass.    Tri+  represents  the sum   of  the  trichloro-  through
decachlorobiphenyl   homologue  groups.  Use  of Tri+  as the  historical  calibration
parameter allows for the comparison of data that were analyzed by  congener-specific
methods with data analyzed by packed-column methods (that did not separate the various
PCBs as well and did not measure many of the mono- and dichlorobiphenyls). Therefore,
use of the operationally defined Tri-i- term allows for a consistent basis for comparison
over the entire period for which  historical data were  available.  Tri+ is also  a good
representation of the PCBs that bioaccumulate in fish.

The five PCB congeners were selected for model calibration based primarily on their
physical-chemical properties  and  frequencies  of  detection in  environmental  samples
across different  media. These individual congener simulations help provide  a better
understanding of the  environmental processes controlling PCB dynamics in the river by
testing the model with PCBs with widely varying properties.  BZ#4 is  a  dichloro
congener that represents a final product of PCB dechlorination in the sediments. BZ#28 is
a trichloro congener  that  has similar physical-chemical properties to Tri+.  BZ#52 is a
tetrachloro congener  that was selected because of its resistance to degradation and based
on its presence in Aroclor 1242, the main Aroclor used by General Electric at the Hudson
River  capacitor  plants.   BZ#[90+101]  (a pentachloro  congener)  and  BZ#138  (a
hexachloro congener) represent higher-chlorinated congeners that strongly partition  to
solids in the river and bioaccumulate in fish.

The HUDTOX model calibration strategy can be considered minimal and conservative.
It is minimal in that external inputs and internal model parameters were  determined
independently to the fullest extent possible from site-specific data and only a minimal
number of parameters were adjusted during model calibration.  It is conservative in that
parameters determined through model  calibration were  held spatially and temporally
constant unless there was supporting information  to the  contrary.  Consistent with the
Reassessment principal questions, emphasis was placed on calibration to long-term trends
in sediment and water column  PCB concentrations,  not  short transient changes  or
localized variations.

The 21-year historical calibration for Tri+ served as the main development vehicle for the
PCB fate and transport model used in the Reassessment. This calibration was successful
in reproducing observed long-term trends in  water and sediment PCB concentrations over
                                          ES-4                              USEPA

-------
the 21-year period.   This was primarily demonstrated through comparisons between
model results and available data for long-term Tri+ surface sediment concentrations, in-
river solids and Tri+ mass transport at low and high flows, and water column solids and
Tri+ concentrations.  Many  different metrics were  used  collectively in a "weight  of
evidence" approach to demonstrate model reliability.

The calibration of the FISHRAND model was conducted by a process known as Bayesian
updating.  This approach optimizes the agreement between predicted distributions of fish
concentrations from the FISHRAND model as compared to empirical distributions based
on  the data by adjusting  three input distributions (percent lipid in fish, total  organic
carbon in sediment, and the octanol-water partition coefficient  or KOW).  Initial  input
distributions (referred to as prior distributions) are specified based on site-specific data
and values from the published scientific literature.  The model is run and calculates  the
likelihood of obtaining an  output distribution that matches observed measurements given
the input distribution. The prior input distributions are then adjusted (within constraints
of the data) and these adjusted distributions are referred to as posterior distributions. The
focus of the calibration was on the wet weight concentrations (as opposed to the lipid-
normalized concentrations)  because the wet  weight  concentrations  are  generally  of
primary interest  to USEPA and other regulators, the lipid content of any given fish is
difficult to predict, and  the model predicts fish body  burdens on a wet weight basis and
then lipid-normalizes.  It was determined that, overall, the FISHRAND model predicts
wet weight Tri+ PCB  fish  body burdens  to within  a factor  of two, and typically
significantly less than that.

MODEL VALIDATION

Model validation is the comparison of model output to observed data for a dataset that
was not included in the calibration of the model.  A HUDTOX model validation was
conducted to compare  predicted  and observed  water  column concentrations for Tri+
using a dataset  acquired  in  1998 for the Upper  Hudson River by General  Electric.
Results indicated good agreement at both Thompson  Island Dam and Schuylerville over
an entire year, spanning a range of environmental conditions in the river. The validation
was judged successful and it enhances the credibility of the model as a predictive tool.

Several  approaches were used to validate the FISHRAND model.  One method was to
calibrate FISHRAND for one river mile, and then to run the model for a different river
mile.  Satisfactory agreement for both river miles implied model validity across locations
in the Hudson River.  In  addition, a calibration was conducted using  only part of  the
available dataset, and then the model results were compared with the remaining portion
of the dataset. The posterior distributions obtained using  only the partial dataset were
compared to  the posterior distributions  obtained using the full dataset.   Finally,  the
partial-data calibrated model  was run for the forecast period and these results  compared
to the full-data calibrated model results.  Good agreement across all three metrics implied
confidence in the performance of the model.
                                          ES-5                              USEPA

-------
 MODEL FORECAST

 In the Revised BMR, the HUDTOX model was run for a 70-year forecast period from
 1998 through 2067  for Tri-K  The  forecast  period was lengthened from the  21-year
 forecast in the May 1999 BMR for two reasons. First, the fish body burdens attained for
 the 21-year forecast presented risks and hazards above levels of concern as documented
 in the risk assessments (i.e., the  21-year forecast was too short to predict when PCB
 concentrations  in fish would decrease below levels  of concern).   Second, the  70-year
 forecast period was selected in order to provide exposure concentrations that can  be used
 directly in the Monte Carlo analysis  in the Human Health Risk Assessment.  Tri+ was
 simulated because it reflects PCB congeners that bioaccumulate in fish and hence are key
 to the risk assessment.

 In order to conduct forecast simulations with  the HUDTOX model, it was necessary to
 specify future conditions in the Upper Hudson  River for flows, solids loads, and upstream
 Tri+ loads. These model inputs are not easily predicted (similar to  predicting the future
 weather),  but reasonable  estimates were  made based on  historical observations  and
 current information regarding PCB loading trends.

 The baseline  forecast simulation was run for an assumed constant Tri-t- concentration of
 10 ng/L at the  model's upstream  boundary at Fort Edward.  This level represented  the
 annual average Tri+ concentration that was observed in 1997 and assumes that there will
 be no future  load increases or reductions from upstream sources.   In particular, it also
 assumes that  the PCB migration from the GE Hudson Falls Plant site would not increase
 or decrease and that there  would not be any  type of event similar to the releases  that
 occurred with the partial failure of the Allen Mill gate structure in 1991.  Recognizing the
 uncertainty in this upstream load,  model sensitivity runs were conducted for an assumed
Tri+ concentration of zero (0 ng/L) to represent a lower bound on future loads due to  the
 implementation of remedial measures upstream, and for an assumed concentration of 30
 ng/L to reflect increased loads similar to observations in 1998.

Results from  70-year  forecast   simulations  contain  inherent   uncertainty   due  to
uncertainties  in estimating  future flow and  solids loading conditions.   Furthermore,
various model  input  assumptions, while  less influential  in  21-year simulations,  can
become more important in 70-year forecast simulations. This uncertainty can be assessed
 and accounted for in USEPA's decision making by evaluating  predictions across a range
 of alternate scenarios for these inputs.  For this reason, model sensitivity runs were  also
conducted for three additional hydrologic conditions:  plus/minus 50 percent changes in
future tributary solids loads, a different assumption for the depth of particle mixing in the
 surface sediments, and different starting concentrations for Tri+ in the sediments.

 Risk-based target levels for fish PCB  body burdens have not yet been established. In the
Feasibility Study,  site-specific target  levels to  be protective  of human health  and  the
environment  will be developed from the risk  assessments.  However, it is beneficial at
this time to compare forecasted fish PCB levels against example target levels as a matter
of perspective.  The target levels used for this analysis provide several concentrations
spanning two orders-of-magnitude. Again, these are not endorsements of these values for
                                          ES-6                             USEPA

-------
decision making.  Appropriate values will be developed in the Feasibility Study for the
site.

MAJOR FINDINGS

The primary objective of the modeling effort is to construct a scientifically credible tool
to help in the understanding of PCS transport and fate and bioaccumulation in the Upper
Hudson River, and to use that tool for making forecasts of what will happen in the future.
As  such, one of the major findings was that it was possible to construct models that
simulate conditions that match the observed data reasonably well.  Consequently, the
model predictions can be reliably used to evaluate future ecological and human health
risks and to assess the relative time it takes for the river to recover under various remedial
scenarios.

There are numerous general observations about the river that are apparent from the mass
balance exercises.   Some important observations that impact the understanding of the
system include:

    •   The river is net depositional for solids in Thompson Island Pool, and apparently
       also in downstream reaches;

    •   Solids loads are dominated by tributary inputs;

    •   PCB (Tri+) loads to the water column are dominated by sediment to water mass
       transfer under non-scouring flow conditions; and,

    •   Water column and PCB (Tri+)  surface sediment concentrations are gradually
       declining due to reduced input loads and natural attenuation.

Beyond the general observations  above,  the model forecasts  provide  the following
findings regarding PCBs in the Upper Hudson River. It should be noted that the findings
below are made based on the evaluation of Tri+, and that some of the findings may differ
for other mixtures of PCBs, such as total PCBs or individual congeners.

1.   PCB (Tri+) concentrations in the surface sediment are forecasted to decline at annual
    rates of approximately 7 to 9 percent over the next two decades, consistent with long-
    term historical trends.

2.   PCB (Tri+) loads from upstream of the  model boundary at Fort Edward control the
    long-term responses of PCB  (Tri+) concentrations in the water column and surface
    sediments, and accordingly, body burdens in fish.

    •   For the first two to three decades of the model forecast, depending on location, the
       in-place PCB  (Tri+) reservoir in  the sediments and sediment-water transfer
       processes control responses of surface sediment concentrations.

    •   Water  column  PCB  (Tri+)  concentrations are  increasingly controlled by the
       upstream boundary at Fort Edward over the long term. The rate at which water
                                          ES-7                              USEPA

-------
       column concentrations  approach  an asymptote  depends  upon  the  assumed
       magnitude of the upstream boundary load and location within the river.

 3.  Forecasted surface sediment PCB (Tri+)  concentrations in  several localized areas in
    the Stillwater reach and the Thompson  Island Pool increase after 40 to 50 years,
    despite exponential-type decreases up to that time. These computed increases are due
    to relatively small annual erosion rates that eventually,  over an extended length of
    time, expose PCB  concentrations that were previously at depth.

    •  The relative magnitudes of computed increases in surface sediment PCB (Tri+)
       concentrations are small within the  context of long-term  trends in historical
       concentrations.

    •  The occurrence, magnitude and timing of these computed increases are dependent
       on forecast assumptions.

    •  It is reasonable to assume that localized erosion occurs within the river, but at
       scales smaller than the spatial scale of the model.  Therefore, the model may not
       accurately reflect the areal extent of such erosion or its timing.

4.  Results of the 100-year peak flow show that a flood of this magnitude would result in
    only a small additional increase in sediment erosion beyond what might be expected
    for a reasonable range of annual peak flows.

    • The small sediment scour depths produced by the 100-year peak flow  result in
      only very small increases in surface sediment PCB (Tri+) concentrations.  These
      increases decline to values in the base forecast simulation (without the  100-year
      peak flow) in approximately four years.

    • Increases in  water column PCB  (Tri+) concentrations in response to  a  100-year
      peak flow are very short-lived (on the order of weeks)  and decline rapidly after
      occurrence of the event.

    • The 100-year event causes an increase of less than 30 kg (70 Ibs) in cumulative
      PCB (Tri-f) mass loading across the Thompson Island Dam by the end of the first
      year of the forecast.  This increase represents approximately  13 percent of the
      average annual PCB (Tri+) mass loading across Thompson Island Dam during the
       1990's.

5.  The  FISHRAND model  results  for the 70-year forecasts  show  that predicted wet
    weight  PCB  (Tri+) fish  body  burdens  asymptotically  approach  steady-state
    concentrations.   These concentrations are species-specific, depending on the relative
    influence  of sediment versus water sources, and  reflect the  upstream boundary
    assumption. That is, the asymptotic value is lowest for the 0 ng/L upstream boundary
    condition  and approximately an order of magnitude higher  for the 10 ng/L upstream
    boundary  condition.  Under the 30 ng/L upstream boundary  condition, the asymptotic
    value is approximately a factor of five higher than the 10 ng/L result.
                                          ES-8                              USEPA

-------
6.  F1SHRAND model  results show that  PCB (Tri+) uptake in fish  is predominantly
    attributable to dietary sources, with a smaller contribution from direct water uptake.
    Analysis of relative  sediment and water contributions within the food chain yielded
    the following results.  Brown bullhead are most sensitive to changes in sediment
    concentration and not  very sensitive to changes in water concentration; largemouth
    bass are more sensitive to sediment concentrations than to water concentrations, but
    water plays a larger  role than for brown bullhead; yellow perch are driven primarily
    by the water; white perch show greater sensitivity to sediment; and  pumpkinseed and
    spottail shiner are sensitive to small changes in water concentration.

7.  The time it takes to  attain acceptable target levels in fish tissue is greatly dependent
    upon the target level selected.  Target levels will be selected as part of the Feasibility
    Study for the site.

SUMMATION

The modeling effort for the Reassessment has provided USEPA with  valuable insights
regarding  factors that control transport and fate and bioaccumulation of  PCBs in the
Upper Hudson River. Forecasted responses of water column and surface sediment PCB
(Tri+) concentrations in the Upper Hudson  River, as calculated  by HUDTOX, are
sensitive to changes in  hydrology, solids loadings, sediment particle mixing depth and
sediment initial conditions.  Forecasted  responses of fish  body  burdens using the
FISHRAND model  are sensitive to changes in lipid content of fish, total  organic carbon
in sediment, and the octanol-water partitioning coefficient (KoW).

The models are  useful tools for forecasting  future sediment, water  and fish PCB
concentrations.  The forecasts can  be reliably used to  evaluate future  ecological and
human health risks and  to assess the relative time it takes for the river to recover under
various remedial scenarios.
                                          ES-9                             USEPA

-------
THIS PAGE INTENTIONALLY LEFT BLANK

-------
Chapter 1

-------
1. INTRODUCTION

1.1 PURPOSE OF REPORT

This volume is the fourth in a series of reports describing the results of the Phase 2 investigation
of Hudson River sediment polychlorinated biphenyls (PCB) contamination.  This investigation is
being conducted under the  direction of the U.S. Environmental Protection Agency (USEPA).
This investigation is part of  a three phase remedial investigation and feasibility study intended to
reassess the 1984 No Action decision of the USEPA concerning sediments contaminated with
PCBs in the Upper  Hudson River.  Figure 1-1  contains a location map for  the Hudson River
watershed.  For purposes of the Reassessment, the  area of the Upper Hudson River considered
for remediation is defined as the river bed between the Fenimore Bridge at Hudson Falls (just
south of Glens Falls) and Federal Dam at Troy, New York (Figure 1-2).

In December 1990, USEPA issued a Scope of Work for reassessing the No Action decision for
the Hudson River PCB Site.  The scope of work identified three phases:

       •     Phase 1 - Interim Characterization and Evaluation

       •     Phase 2 - Further Site Characterization and Analysis

       •     Phase 3 - Feasibility Study

The Phase  1 Report (USEPA, 1991) is Volume 1 of the Reassessment documentation and was
issued by USEPA in August 1991.  It contains a compendium of background material, discussion
of findings and preliminary assessment of risks.

The Final Phase 2 Work Plan and Sampling Plan (USEPA, 1992) detailed the following main
data collection tasks  to be completed during Phase 2:

       •      High- and low-resolution sediment coring;

       •      Geophysical surveying and confirmatory sampling;

       •      Water column   sampling  (including  transects  and  flow-averaged
             composites); and,

       •      Ecological field program.

The data available from the Phase 2 investigation and other historical datasets  are documented in
the Database Report (Volume  2A in  the Phase  2 series  of  reports;  (USEPA,   1995) and
accompanying CD-ROM database.  This database  provides the validated data  for the Phase 2
investigation.  This  Revised Baseline  Modeling Report (RBMR  or Revised BMR) utilized the
Hudson River Database, Release 4.1b, which was updated in Fall 1998 (USEPA,  1998b).
                                                                  Limno-Tech, Inc.

-------
This Revised Baseline Modeling Report is Volume 2D of the Reassessment documentation. It
presents results and findings from application of mathematical models for PCB transport and
fate, and PCB bioaccumulation in the Upper Hudson River.

There were two modeling reports preceding this RBMR  in the Reassessment documentation.
The Preliminary Model Calibration Report (USEPA,  1996) was issued  for public review in
October 1996.  The purpose of the PMCR was to document the conceptual approaches, databases
and preliminary calibration results for the transport and fate, and bioaccumulation models. The
PMCR  did not contain results for any forecast simulations with the preliminary models.  The
modeling approaches in the PMCR were  reviewed  by an independent peer review panel in
September 1998. The modeling  approaches were revised in response to comments from the peer
review panel and from the public. The Baseline Modeling Report (USEPA, 1999c) was issued
for public  review in May  1999.  The BMR contained  model refinements recommended by
reviewers,  results  from  a long-term  historical  calibration  of the  transport  and  fate and
bioaccumulation models, and results from forecast simulations designed to estimate long-term
responses to continued No Action and impacts due to a  100-year peak flow.  USEPA decided to
revise the BMR to reflect  changes in the  models based on public comment  and additional
analyses that were conducted. The Revised BMR supercedes the May 1999 BMR.

The purpose of this Revised Baseline Modeling Report is to document:

       •     Additional model  refinements;

       •     Sensitivity of the historical calibration;

       •     Model validation to an independent dataset;

       •     Longer (70-year) model forecasts for continued No Action; and,

       •     Sensitivity of forecast simulations for continued No Action.

1.2 REPORT FORMAT AND ORGANIZATION

The information gathered and the findings of this  phase are presented here in a format that is
focused on answering questions critical to the Reassessment, rather than report results  strictly
according to Work Plan tasks.  In particular, results are presented in a way that facilitates input to
other aspects of the projects.

This report is presented in four books.  Books 1 and 2 contain results and findings from the PCB
transport and fate models. Book 1 contains the report text and Book 2 contains all tables, figures
and plates for the transport and fate models.  Books 3  and 4 contain results and findings from the
PCB bioaccumulation models.  Book 3 contains the report text  and Book 4 contains all tables,
figures and plates for the bioaccumulation models.

Books  1 and 2 contain results and findings for applications of PCB transport and fate models to
existing  historical  data, and  for forecast  simulations designed to estimate  both long-term
                                                                   Limno-Tech, Inc.

-------
responses to continued No Action and impacts due to a 100-year peak flow.  Books  1 and  2
chapters are as follows:

       •       Chapter 1 herein provides the report introduction;

       •       Chapter  2  presents  the  overall conceptual  approach used  for the
              mathematical models and the relationships among individual models;

       •       Chapter 3 presents the hydrodynamic model used for Thompson  Island
              Pool (TIP);

       •       Chapter 4 presents the Depth of Scour Model (DOSM)  used to estimate
              masses  of solids and PCBs eroded from  cohesive  and non-cohesive
              sediment areas in Thompson Island Pool (TIP) in response to peak flows;

       •       Chapter 5 presents the development of the Hudson River Toxic Chemical
              Model (HUDTOX) including conceptual framework, governing equations
              and spatial-temporal scales;

       •       Chapter 6 presents results from  data synthesis tasks necessary to provide
              model inputs and to support processing and interpretation  of model output;

       •       Chapter 7 presents results and findings from calibration of the HUDTOX
              model to historical data, including data collected as part of the USEPA
              Phase 2 investigation;

       •      Chapter 8 presents results and findings from  forecast simulations with the
              HUDTOX model designed to estimate long-term, responses to continued
              No Action and impacts due to a 100-year peak flow; and,

       •       Chapter 9 presents results from a model validation simulation  using an
              independent dataset acquired in 1998 by General Electric.

1.3 PROJECT BACKGROUND

1.3.1 Site Description

The Hudson River PCBs Superfund site encompasses the Hudson River from Hudson Falls (river
mile [RM] 198) to the Battery in New York Harbor (RM 0), a river distance of nearly 200 miles.
Because of different physical and hydrologic  regimes, approximately  40  miles  of the Upper
Hudson River, from Hudson Falls to  Federal Dam (RM 153.9), is distinguished from the Lower
Hudson River  below Federal Dam. Emphasis was placed on Thompson Island Pool (TIP), a 6-
mile portion of the river between Fort  Edward and Thompson Island Dam (TDD) (Figure  1-3),
because a substantial amount of PCB-contaminated sediment is contained in this location.
                                                                  Limno-Tech, Inc.

-------
 1.3.2 Site History

 Over a 30-year period ending in 1977, two General Electric (GE) facilities, one in Fort Edward
 and  the other in Hudson  Falls, NY, used PCBs  in the manufacture of electrical capacitors.
 Various sources have estimated that between 209,000 and 1,300,000 pounds (95,000 to 590,000
 kilograms [kg]) of PCBs were discharged between 1957 and 1975 from these two GE facilities
 (Sofaer, 1976; Limburg, 1984). Discharges resulted from washing PCB-containing capacitors
 and  PCB spills.  Untreated washings are believed to  have been discharged  directly into the
 Hudson from about 1951 through 1973 (Brown et al., 1984).  No records exist on which to base
 estimates of discharges from the beginning of PCB capacitor manufacturing operations in  1946
 to 1956; however, discharges during this period are believed to be less than in subsequent years.
 Discharges after 1956 have been estimated at about 30 pounds (14 kg) per day or about 11,000
 pounds (5,000 kg) per year (Bopp, 1979, citing  1976 litigation; Limburg, 1984, citing Sofaer,
 1976). In 1977, manufacture and sale of PCBs within the U.S. was stopped under provisions of
 the Toxic Substances Control Act (TSCA).  PCB use ceased at the GE facilities in 1975 and only
 minor discharges (about 0.5 kg/day or less [Brown et al., 1984; Bopp, 1979]) are believed to  have
 occurred during facility shutdown and cleanup  operations  through mid-1977  when  active
 discharges ceased.  GE  had been  granted a National Pollutant Discharge Elimination System
 (NPDES) permit allowing up to 30 Ibs/day to be discharged during this period (Sanders, 1989).
 According to scientists at GE, at least 80 percent of the total PCBs discharged are believed to
 have been Aroclor 1242, with lesser amounts of Aroclors 1254, 1221 and 1016 (USEPA, 1997).

 A significant portion of the PCBs discharged  to the river adhered to suspended particulates and
 subsequently accumulated  downstream  in bottom  sediments as they  settled in the  impounded
 pool behind the former Fort Edward Dam (RM 194.8), as well as in other impoundments farther
 downstream.  Because of the proximity to the  GE discharges, sediments behind the Fort Edward
 Dam were probably among the most contaminated to be found in the Hudson, although this was
 not well known in  the  1970s.  The Fort Edward Dam was removed in 1973 because of  its
 deteriorating condition.  During subsequent spring floods, the highly contaminated sediments
 trapped behind the Fort  Edward Dam were scoured and transported  downstream.  Substantial
portions of these sediments were stored in relatively quiescent areas of the river.  These areas,
which were  surveyed  by New   York State  Department  of  Environmental  Conservation
(NYSDEC)  in  1976 to  1978 and  1984,  have been  described as  PCB  "hotspots".  Exposed
sediments from the  former pool remaining behind  the dam site, called the "remnant deposits",
have been the subject of several remedial efforts.

PCB releases from the GE Hudson Falls Plant site near the Bakers Falls Dam have also occurred
through migration of PCB oil through bedrock.  The extent and magnitude of these releases are
not well quantified. This release through bedrock  continued until at least 1996, when remedial
activities by GE brought the leakage under better control. Despite some evidence for its existence
prior to 1991 based on U.S. Geological Survey (USGS) data, this leakage was not identified until
the partial failure of the abandoned  Allen Mill gate  structure near the GE Hudson Falls plant site
 in 1991.   This failure caused a large  release  of what were probably  PCB-bearing oils and
 sediments that had accumulated within the structure.  This failure also served to augment  PCB
 migration from the bedrock beneath the plant to the river until remedial measures by GE over the
                                                                   Limno-Tech, Inc.

-------
period 1993 to 1997 greatly reduced the release rate.  A more in-depth discussion of PCB sources
is contained in the Data Evaluation and Interpretation Report (DEIR) (USEPA, 1997).

1.4 MODELING GOALS AND OBJECTIVES

The goal of the PCB transport, fate and bioaccumulation modeling was to assist in answering the
following principal Reassessment questions:

    1.  When will PCB levels in fish populations recover to levels meeting human health
       and ecological risk criteria under continued No Action?

    2.  Can remedies other than No Action  significantly shorten the time required to
       achieve acceptable risk levels?

    3.  Are  there  contaminated sediments  now  buried that  are likely  to  become
       "reactivated" following  a major  flood, possibly resulting in an  increase in
       contamination of the fish population?

The approach to the PCB transport and fate modeling  was to develop and field  validate  a
scientifically credible  mass  balance model that  was  capable  of  predicting future PCB
concentrations in the  water and sediments.   The  model would be used  for evaluating and
comparing the impacts of continued No Action,  major flood events  and various remedial
scenarios.  The model  also provides water column  and sediment PCB exposures for the PCB
bioaccumulation model and the ecological and human health risk assessments.

The specific objectives of the  transport and  fate modeling  work in this RBMR were  the
following:

       •      Develop a mass balance model for PCB levels in the water column  and
              bedded sediments in the Upper Hudson River;

       •      Calibrate the mass balance model to available historical data, including
              data collected as part of the Phase 2 investigation;

       •      Conduct forecast  simulations with the calibrated mass balance model to
              estimate long-term responses to continued No Action and impacts due to a
              100-year peak flow; and

       •      Estimate short-term, fine-scale erosion of solids and PCBs in Thompson
              Island Pool in response to a 100-year peak flow.

Through these objectives, the modeling  work in this Revised Baseline Modeling Report is
directed at answering  Reassessment questions pertaining to continued No Action (Question  1
above) and impacts of a major flood (Question 3).  During Phase  3, the Feasibility Study, the
models will be used for evaluation and comparison of the impacts of various remedial scenarios
(Question 2).
                                                                   Limno-Tech, Inc.

-------
THIS PAGE INTENTIONALLY LEFT BLANK
                                   Lirnno-Tech, Inc.

-------
Chapter 2

-------
 2. MODELING APPROACH

 2.1 INTRODUCTION

 Mass balance models were developed for transport and fate of PCBs in the water column and
 bedded sediments,  and for PCB  bioaccumulation in fish.  The report herein (Books 1 and 2)
 focuses  only  on the  PCB transport and  fate model,  whereas the bioaccumulation model is
 described in Books 3 and 4. The spatial domain of these models was the Upper Hudson River
 between Fort  Edward and Federal Dam at Troy (Figure  1-2).  However, special emphasis was
 placed on Thompson Island Pond (TIP), a 6-mile portion of the river between Fort Edward and
 Thompson  Island Dam  (TID)  (Figure  1-3),  because  this reach  contains the highest  PCB
 concentrations and a disproportionately high PCB mass reservoir relative to downstream reaches.

 The following major sections are contained in Chapter 2:

      •      Section  2.2 presents the  overall modeling  framework used  in this
             Reassessment;

      •      Section 2.3 describes the hydrodynamic model developed for Thompson
             Island Pool which was linked to  the Depth  of Scour Model and  the
             HUDTOX model;

      •      Section 2.4 describes the Depth of Scour Model (DOSM) for Thompson
             Island Pool;

      •      Section  2.5  describes  the  Hudson  River  Toxic Chemical   Model
             (HUDTOX) that was developed and applied to the Upper Hudson River
             between Fort Edward and Federal Dam at Troy;

      •      Section  2.6 describes  the various  applications  conducted  with  the
             HUDTOX model; and,

      •      Section 2.7 presents an  overview of  the  database  used  for  model
             development and applications.

2.2 CONCEPTUAL APPROACH

The conceptual approach for the PCB transport and fate models of the Upper Hudson River was
driven by the principal Reassessment questions:

 1.  When will PCB levels in fish populations recover to levels meeting human health and
   ecological  risk criteria under continued No Action?

2.  Can remedies other than No Action significantly shorten the time required to achieve
   acceptable risk levels?
                                                                 Limno-Tech, Inc.

-------
 3.  Are there contaminated sediments now buried that are likely to become "reactivated"
    following a major flood, possibly resulting in an increase in contamination of the fish
    population?

 Answers to the first two questions required reliable representation of long-term trends in water
 column and sediment PCB exposure concentrations to fish populations in the Upper Hudson
 River.  To accomplish this  objective, a mass balance  model,  HUDTOX,  was developed to
 simulate water, solids and PCBs over the long-term historical period and  a  long-term forecast
 period.  Inputs, outputs, and internal sources and sinks were balanced on a daily time scale in
 order to simulate long term conditions for the entire Upper Hudson River from Fort Edward to
 Troy, New York.

 An answer to  the  third  question  required reliable representation  of flow-driven  sediment
 resuspension from highly contaminated areas, especially PCB "hotspots" associated with fine-
 grain, cohesive sediments.  To accomplish this objective, a two-step approach was used. First, a
 fine-scale hydrodynamic and sediment scour model, DOSM, was used to estimate flow-driven
 resuspension of sediments and associated PCBs in Thompson Island  Pool, the most heavily
 contaminated portion of the river, in response to a 100-year peak flow.  Second, the PCB mass
 balance model,  HUDTOX, was used  to estimate water column and sediment responses in the
 entire Upper Hudson River to the same 100-year peak flow.  The hydrodynamic and resuspension
 models provided an estimate of the likelihood that high PCB concentrations  now buried in the
 sediments would become re-exposed due to flow-driven scour of the sediment  bed. Results from
 the PCB mass balance model (HUDTOX) provided estimates of the resultant water column and
 sediment  concentration responses  due  to  flow-driven  scour and  subsequent transport and
 redistribution of contaminated sediments.

The operational  framework for  the Reassessment models  is illustrated in Figure 2-1, which
depicts the principal individual modeling components and their inter-relationships.  In  Figure 2-
2, the specific information input to each of these models is also presented.  The hydrodynamic
 model, the DOSM and HUDTOX comprise the transport and fate models. The Thompson Island
Pool models consist of a coupled hydrodynamic and resuspension model (DOSM) for sediments.
HUDTOX is the mass balance model that represents water,  solids and PCBs in the entire Upper
Hudson River,  including  Thompson  Island Pool.  There  is a linkage module that  serves to
process output from  the hydrodynamic  model and the DOSM for use in HUDTOX.  These
transport  and fate models  are  described in  the  following sections.  The  Bivariate  Biota
Accumulation  Factor (BAF) Model and  the bioaccumulation models  (FISHPATH  and
FISHRAND) quantify linkages between PCB water column and sediment concentrations and fish
body burdens. These models are the subject of Books 3  and 4, and not described herein.

Different models with different attributes were developed to most effectively answer the three
Reassessment questions  while  balancing the  issues  of  complexity, computational burden,
supporting data and system characteristics.  The Thompson Island Pool Hydrodynamic and Depth
of Scour Models were more refined and complex, as necessary to answer the issue of episodic
scour.  Thompson Island Pool, although only 6  miles (15 percent) of the entire model domain,
contains almost half  of the PCB mass reservoir in the Upper Hudson  (Tofflemire and Quinn,
 1979) and the highest PCB concentrations.  Hence, the Pool has been the  focus of remedial
                                                                  Limno-Tech. Inc.

-------
considerations, and more specialized modeling was warranted.  The same framework was not
applied to the remainder of the river because substantially fewer supporting data were available,
and the additional computational burden and model complexity were not warranted.

2.3 HYDRODYNAMIC MODEL

The hydrodynamic model used for Thompson Island Pool was the U.S. Army Corps of Engineers
RMA-2V. This model is two-dimensional and vertically-averaged.  It was applied to Thompson
Island Pool to provide velocity information for bottom shear stress calculations at the sediment-
water  interface using DOSM.   It also  provided  flow routing,  water depth, and  velocity
information to the HUDTOX model for Thompson Island Pool only.  The hydrodynamic model
includes explicit representation  of the existing river geometry as  well  as  the flood  plains to
account for overbank flow during flood events.

The hydrodynamic model was not directly integrated with the HUDTOX  model.  Hydrodynamic
model  results were spatially and temporally processed using a linkage module that transformed
water velocities into flows that  were routed among the HUDTOX model spatial segments in
Thompson Island Pool. Water velocities were also transformed into applied shear stresses at the
sediment-water interface for use in the DOSM.  The hydrodynamic model was run to steady state
for a range of different river flows, including the 100-year peak flow.

2.4 DEPTH OF SCOUR MODEL

The DOSM  is a two-dimensional, GIS-based  model of sediment erosion that was applied to
Thompson Island  Pool.  It is a  specialized tool for providing spatially-refined information on
sediment erodibility in response to high  flows,  including a catastrophic flood.   It calculates
sediment bed scour  based on flow-induced shear stress and  site-specific measurements  of
sediment properties  and resuspension  behavior.  Information on applied shear stresses at the
sediment-water interface was  calculated based on output from the hydrodynamic model.

The DOSM was developed principally to answer questions related to the likelihood that flood-
induced erosion of bottom sediment would reactivate buried PCB.  It was first used as a stand-
alone tool to provide mass estimates of solids and PCBs eroded,  and depth of sediment bed
scour, in response to a 100-year peak flow.  A constant 100-year flow was simulated with the
hydrodynamic model as a worst case scenario.  This simulation produced a map of bottom shear
stress throughout Thompson Island Pool for the 100-year flow. This shear stress map  was used
to compute estimates of depth of scour throughout the entire cohesive sediment bed in Thompson
Island Pool.  Based on various uncertainties in model inputs, DOSM also  calculates a probability
distribution for scour depth addressing the question of "likelihood".  The relationship between
cohesive sediment resuspension  and applied shear stress was  based on a formulation from the
published literature and parameterized  using site-specific measurements  from Thompson Island
Pool.  The DOSM shear stress map was also used to compute an  upper bound estimate of depth
of scour that could be expected for the non-cohesive sediment area in Thompson Island Pool.

The DOSM was also used to develop  relationships between  river flow  and  cohesive  sediment
resuspension for use in the HUDTOX model. The hydrodynamic model  was run for a range of
                                                                  Limno-Tcch, Inc.

-------
flow conditions spanning typical summer flows to the 100-year flow.  The DOSM was used to
estimate cohesive sediment resuspension for each of these flow conditions, thus producing a
family of resuspension-flow relationships.   These relationships were  used as input  to  the
HUDTOX model to represent cohesive sediment resuspension across all  flow conditions in the
Thompson Island Pool portion of the River.

2.5 MASS BALANCE MODEL

HUDTOX is the principal transport and fate modeling tool in this Reassessment.  HUDTOX is a
time-variable, three-dimensional model that includes three types of mass balances: (1) a water
balance; (2) a solids balance; and (3) a PCB mass balance.  A water balance is necessary because
PCB dynamics are influenced by  river flow and mixing rates.  A solids balance is necessary
because PCB dynamics are influenced by  the tendency  of PCBs  to  sorb (attach) to both
suspended and bedded solids in the river.  Finally, a PCB  mass balance is necessary to account
for all inputs, outputs, and internal  sources and sinks of PCBs in the river.  HUDTOX has a fully-
integrated  representation of solids and PCB  concentrations in the water column and bedded
sediments.

The spatial scales of the HUDTOX model application were determined by the Reassessment
questions and available site-specific data.  HUDTOX was  applied to the entire Upper Hudson
River from Fort Edward to Federal Dam at Troy.  Because  a substantial amount  of PCB-
contaminated sediment is contained in Thompson Island Pool, this portion  of HUDTOX included
greater spatial resolution than the  portion downstream of Thompson Island Dam.  In the Pool,
HUDTOX is two-dimensional in  the  water column  and  three-dimensional  in the sediments.
Between Thompson Island  Dam and Federal Dam, HUDTOX is one-dimensional in the water
column and two-dimensional in the sediments.

With respect to temporal  scale, the HUDTOX model was developed to  represent  long-term
average water column and sediment PCB exposure concentrations.  It  was not developed  to
represent short-term behavior associated with high flow events.  The reason is that PCB body
burdens in fish  are driven primarily by long-term average exposure concentrations,  not short-
term, event-scale exposures. The model does, however, represent differences between low-flow
and high-flow sediment resuspension  processes, and  differences between  cohesive and non-
cohesive sediment areas.  In this sense the model was designed to capture both mean low-flow
and mean high-flow solids and PCB dynamics.

In HUDTOX, hydraulic routing downstream of Thompson  Island Dam was one-dimensional and
was specified using USGS flow gage data at Fort Edward and estimated flows for downstream
tributaries.  In Thompson Island  Pool, the two-dimensional flow routing  was defined by  the
hydrodynamic model.

Sediment scour in  HUDTOX was determined through  use of output from  DOSM.   The
hydrodynamic model results were  used to calculate the bottom shear stress required for DOSM.
Output from the DOSM was linked to HUDTOX in the form of relationships between flow and
cohesive sediment resuspension.  This  linkage ensured internal consistency in representation of
flow-dependent  resuspension between these two  models for cohesive sediment areas.   In
                                          10                      Limno-Tech, Inc.

-------
Thompson Island Pool, the hydrodynamic, DOSM and HUDTOX models were linked in terms of
flow routing, depth, velocity, applied shear stress and cohesive sediment resuspension. Neither
the hydrodynamic model nor the DOSM was applied to the portion of the river below Thompson
Island  Dam.   Average relationships for cohesive  sediment resuspension developed  from the
DOSM in Thompson Island Pool were used in this portion of the river.

2.6 MASS BALANCE MODEL APPLICATIONS

The HUDTOX mass balance model was applied in a structured sequence as follows:

       •      Historical  calibration  for Tri+  (sum of trichloro through decachloro
              homologue groups) for a 21-year period from 1977 to 1997;

       •      Hindcast applications for total PCB and five congeners for 1991 to 1997;

       •      Independent model validation for 1998;

       •      70-year model forecasts from 1998 to 2067; and,

       •      Sensitivity  analysis   for  the historical  calibration  and the  forecast
              simulation periods.

Model  applications  included a total of seven different PCB forms:  total PCBs, Tri+, and five
congeners, BZ#4, BZ#28, BZ#52, BZ#[90+101] and BZ#138.  Total PCBs represents the sum of
all measured  PCB congeners  and is  the only PCB form that completely represents total PCB
mass.  A limitation to the use of total PCBs is that  data were available for only the period from
1991 to 1997. To extend the period of time for the  HUDTOX historical calibration,  Tri+ was
used as a surrogate for total PCBs and served as the  principal calibration and forecast model state
variable. Tri+ represents the sum of only trichloro  through decachloro homologue groups.  Due
to differences in analytical methods  among individual datasets,  Tri+ was the only internally-
consistent PCB form that could be operationally defined to approximate total PCBs over the
entire period  from 1977  to 1997 (USEPA, 1998a).  Tri+  was also an appropriate choice for
calibration and forecast simulations because it represents the principal distribution of PCB
congeners that bioaccumulate in fish.

The  historical calibration  was the  principal development  vehicle for the  model, which was
focused on representing long-term PCB trends in water and sediment for a 21-year period. Tri-i-
was the principal focus of the calibration because comparable measurements were available for
the entire 21-year period.  However, a subsequent 7-year hindcast application of the  model to
total PCB and five congeners provided a test of the  historical calibration to Tri+. The calibrated
model  was then subjected to validation using an independent set of water column  PCB data for
1998.   Following  successful  validation  of the  model,  70-year  forecast  simulations were
developed. The  forecasts were intended to assess the long-term system responses to continued
No Action and impacts due to a 100-year peak flow.  Additionally, model performance over the
historical calibration and forecast periods was assessed through sensitivity analyses.
                                                                   Limno-Tech. Inc.

-------
The congener simulations were  conducted to gain  better understanding of the environmental
processes controlling PCB dynamics in the river and to  strengthen and support the long-term
historical calibration.   The  five congeners were selected based  primarily on their physical-
chemical  properties and frequencies of detection in environmental samples across different
media.  BZ#4 is a dichloro congener that represents a final product of PCB dechlorination in the
sediments (USEPA, 1997).  BZ#28 is  a trichloro congener that has similar physical-chemical
properties to total PCBs.  BZ#52 is a tetrachloro congener that was selected as a normalizing
parameter for congener patterns based on its presence in Aroclor 1242, the main Aroclor used by
GE, and on its  resistance to  degradation or dechlorination in the environment (USEPA, 1997).
BZ#[90+101] (a pentachloro congener) and BZ#138 (a hexachloro congener) represent higher-
chlorinated congeners that are more strongly associated with suspended and bedded solids in the
river.

2.7 MASS BALANCE MODEL CALIBRATION

The calibration strategy can  be described as minimal and conservative. It was minimal in the
sense that external inputs and internal model parameters were determined independently to the
fullest extent possible  from site-specific data and  only  a minimal number were  determined
through model calibration. It was conservative in the sense that parameters determined through
model calibration were held spatially  and temporally constant  unless there was supporting
information to the contrary.  Consistent with the  Reassessment questions, emphasis was placed
on calibration to long-term trends in sediment and water column PCB concentrations, not short
transient changes or localized variations.

The following factors were found to be the most important in controlling long-term trends in
sediment and water column Tri+ concentrations in the Upper Hudson River:

       •     Hydrology;

       •     External solids loads;

       •     External Tri+ loads;

       •     Tri+ partitioning;

       •     Sediment-water mass transfer under non-scouring flow conditions;

       •     Solids burial rates; and,

       •     Particle mixing depth in the sediments.

The first three of these factors are external inputs defined largely by data, and the last four factors
are internal processes within  the river defined by data, scientific literature and model calibration.
Long-term solids burial rates were the  principal  factor controlling long-term Tri+ responses in
the river.  Partitioning controls the distribution of Tri+ mass between sorbed and truly dissolved
phases, thus influencing sediment-water and water-air mass transfer  rates,  and bioavailability to
fish.  Sediment-water mass  transfer under non-scouring  flow conditions  was found to be the


                                            12                       Limno-Tech. Inc.

-------
principal source of Tri+ inputs to the water column.  Particle mixing depth strongly influenced
long-term responses and the vertical distribution of Tri+ in the sediments. With the exception of
solids  burial rates  and particle  mixed depth, all model inputs and parameter values were
determined using site-specific data and were not adjusted during the model calibration.

Most of the effort during the HUDTOX model calibration consisted of determining solids burial
rates.  Solids burial rates were determined for the 21-year historical calibration for four major
reaches,  including  Thompson Island Pool  and  three  downstream reaches.   The  principal
calibration constraints on solids burial rates were the following:

       •     Measured burial rates from dated sediment cores;

       •     Computed burial rates from a sediment transport model;

       •     Tri+ surface sediment concentration trends; and,

       •     In-river solids and Tri+ mass transport at high and low flows.

The  historical calibration was conducted  by  applying simultaneous, mutual constraints on the
coupled solids and Tri+ mass balances.  Operationally, the approach consisted of adjusting four
model  parameters: gross settling velocities  into  cohesive and non-cohesive sediment areas;
resuspension rates from non-cohesive sediment areas; depth of particle mixing in the  sediment
bed;  and magnitude of sediment particle mixing.

2.8 HUDSON RIVER DATABASE

All modeling work in this report utilized the extensive database that was created to support this
Reassessment.  The  Database Report (USEPA, 1995) and accompanying CD-ROM  database
provides  the validated  data  for the  Phase 2 investigation.  This Revised Baseline Modeling
Report (RBMR) utilized the Hudson  River Database, Release 4.1b, which was updated in fall
1998 (USEPA,  1998b). This database contains information  from a large variety of  different
sources, including:

       •      New York State Department of Environmental Conservation (NYSDEC)

       •      New York State Department of Health (NYSDOH)

       •      New York State Department of Transportation (NYSDOT)

       •      General Electric Company (GE)

       •      Lamont-Doherty Earth Observatory (LDEO)

       •      Rensselaer Polytechnic Institute (RPI)

       •      U.S. Geological Survey (USGS)
                                          13                       Limno-Tech, Inc.

-------
       •     National Oceanic and Atmospheric Administration (NOAA)

       •     National Weather Service (NWS)

       •     U.S. Environmental Protection Agency (USEPA).

To supplement the database in Release 4.1b, a portion of the 1997 USGS flow, suspended solids
and PCB data were obtained directly from the USGS in Albany, New York.  Where necessary
and appropriate, information from the scientific literature and various technical reports was also
used in this modeling work. These sources are cited in the report text.
                                          14                       Limno-Tech, Inc.

-------
Chapter 3

-------
3. THOMPSON ISLAND POOL HYDRODYNAMIC MODEL

3.1 OVERVIEW

The six-mile long Thompson Island Pool is a special area of focus in the Reassessment because it
contains a disproportionate amount of the PCB mass (nearly half) in the 40-mile long portion of
Upper Hudson River.  Additionally, the highest PCB concentrations occur in the Pool.  These
factors have made the Pool a focus area for possible remediation. The Pool is also the most
extensively sampled reach of the Upper Hudson. As a result of the special focus on the Pool and
the greater data availability, a fine scale, two-dimensional hydrodynarnic model was applied for
the Pool to provide input to the PCB fate and transport model (Chapter 5) and the Depth of Scour
Model (Chapter 4).  The Depth of Scour Model uses fine scale velocity information from the
hydrodynarnic model to compute scour of sediments, especially under high flow conditions.

The Thompson Island Pool is defined as the reach  of the Hudson River  upstream from the
Thompson Island Dam at RM 188.5 and downstream  from the former Fort Edward Dam,  as
shown in Figure 3-1. The purpose of  the hydrodynarnic modeling effort for Thompson Island
Pool was to provide information on bottom shear stresses at the sediment-water interface for the
DOSM and HUDTOX models.   Additionally, the model provided  flow routing,  depth and
velocity information for the  two-dimensional portion  of the HUDTOX model in Thompson
Island Pool.

The hydrodynamic  model was used to calculate two-dimensional, vertically-averaged velocity
fields for a range of different river flows, including the 100-year peak flow in the Hudson River,
(estimated to be 47,330 cfs by Butcher, 2000a).  The computation of a two-dimensional,
vertically averaged velocity field is necessary to account for the lateral variability of the flow and
resultant bed shear.  The bed shear is used to compute the mass of cohesive sediments eroded in
the Depth of Scour Model (DOSM).  Because sediment properties and PCB concentrations are
not uniformly distributed, the bottom shear stresses must be determined for each element used in
the river model to correctly estimate Poolwide resuspension of PCBs.

The hydrodynamic model was applied for a range of steady flow conditions in the Thompson
Island Pool:  Transient effects due to storage and drainage were not included in the simulations
because the historical flow record at Fort Edward shows that the Hudson River high flow events
occur over several days, which gives the Pool enough time  to establish approximate steady  state
conditions. This means simulation  of transient water  storage and drainage could be reasonably
omitted from calculations of bottom shear at peak  flow conditions. Additionally, the Depth of
Scour Model (DOSM) presented in Chapter 4 requires only simulation of the peak flow hydraulic
conditions to estimate solids resuspension losses from cohesive sediment bed areas during flood
events. The credibility of the numerical simulation results was established by applying the model
to events where the flow in the river had  been measured. The model  was run for the 100-year
peak flow to provide the velocity field used by the DOSM.
                                          15                      Limno-Tech, Inc.

-------
 The following major sections are included in Chapter 3:

       3.2    Hydrodynamic Modeling Approach

       3.3    Available Data

       3.4    Hydrodynamic Model Calibration

       3.5    Hydrodynamic Model Validation

       3.6    Hydrodynamic Model Sensitivity Analyses

       3.7    Conversion of Vertically Averaged Velocity to Bottom Shear Stress

       3.8    Discussion of Results

3.2 HYDRODYNAMIC MODELING APPROACH

The hydrodynamic model used to compute the flow is the US Army Corps of Engineers RMA-
2V.  RMA-2V uses the finite element method to compute vertically-averaged velocities and
water surface elevations in the flow field. The model  has been extensively studied and applied
widely (Berger, 1990; Lin and Richards, 1993; McAnally et. al., 1984; and Richards, 1990). The
selection of a two-dimensional, vertically averaged model and the density of the grid mesh were
largely determined by the resolution needed to adequately define the flow field variations and
river bathymetry, and hence, shear stress variation. The shear stress exerted on the river bottom
is parameterized by the magnitude of the vertically averaged velocity and the depth of flow, as is
described in Section 3.7

A  short  summary  of the modeling procedure is as follows: A finite element grid  was first
constructed for the Thompson Island Pool section of the  river and floodplain. RMA-2V uses a
finite element procedure to  solve the governing equations that describe the vertically-averaged
velocities and water surface elevation.  The boundary conditions consist of a specified upstream
flow, the water elevation downstream and the resistance to flow.  The downstream boundary was
obtained from a rating  curve developed for the stage-discharge gage near the Thompson Island
Dam, and the resistance to flow is parameterized by Manning's 'n'.

3.2.1 Governing Equations

The RMA-2V model formulation is based on the conservation of mass and momentum equations
in  order  to simulate water elevation and two-dimensional velocity.  A brief description of the
model equations and framework is provided here.  A more rigorous presentation of the model is
available in the user's manual.

The two governing equations for continuity of mass  and momentum  focus on three  state
variables, water elevation (h) and downstream and cross-stream velocity (u and  v).  To solve for
these three variables,   three equations  are needed. Bottom stress is  computed  based  on the
vertically-averaged velocity using an additional equation. The equations are presented below.
                                           16                       Limno-Tech, Inc.

-------
    1.  Continuity
                      dt    dx     83;
    2.  Linear Momentum
       a.     x-direction (longitudinal) momentum
                      du    du    du             0               _         _          ._ „.
                         + u    +v   =_g  v     o;_c      _  Ł       +Ł          (3.2)
                      d/    dx    dy          dr          /i   p I     dx       dy
       b.     y-direction (transverse) momentum
v— = -
                                             ^
                            dx    dy         3y          /i   P
                                                 0    ^           r,        „           ,„ „,
                                                 °  -Cfq- + —\  Ł„—=- + Ł„„— r-     (3-3)
    3.  Bottom Friction Coefficient
       (English Units)
                       f   (1.486)2/zl"3)

       (Metric Units)
where:
       h      =      water depth [L]
       u      =      vertically-averaged flow velocity in the x-direction
                     (longitudinal) [L/T]
       v      =      vertically-averaged flow velocity in the y-direction (lateral)
                     [L/T]
       x      =      distance in the longitudinal direction [L]
       y      =      distance in the lateral direction [L]
       t      =      time [T]
       g      =      acceleration due to gravity [L/T2]
       a0     =      bottom elevation [L]
       Cf     =      bottom friction coefficient [dimensionless]
       n      =      Manning's 'n' channel roughness coefficient
                     [T/LI/3]
       E.C.V    =      normal turbulent exchange coefficient in the x direction [M/(LT)]
                                             17                        Limno-Tech, Inc.

-------
       Exy    =     tangential turbulent exchange coefficient in the x direction [M/(LT)]
       Eyy    =     normal turbulent exchange coefficient in the y direction [M/(LT)]
       Eyx    =     tangential turbulent exchange coefficient in the y
                    direction [M/(LT)]
       p      =     water density [M/L3]
       q      =     velocity magnitude = («2+v2)l/2 [L/T].

The Coriolis apparent force and the force imposed by wind stress have  been  neglected here
because these forces are small compared to forces induced by gravitation and friction.

3.2.2 Computational Sequence and Linkages

The hydrodynamic model for the Thompson Island Pool was not incorporated directly in either
the HUDTOX model or the Depth of Scour Model because its calculations could be performed
independently. As a result, output from the hydrodynamic model needed to be linked to the other
models.

The RMA-2V model was first calibrated  to the measured hydraulic data for the river, with
Manning's V as the primary calibration parameter. River data, such as river stage-discharge
relations for the upstream (Lock 7) gaging station, were used to calibrate the model. Other data,
such as velocity measurements made by the USGS during high flow events, were also used to
validate the model results.

The specific steps used in the modeling procedure to provide information to the other models are
as follows:

1 .   The flow field, velocity and depth for each node were calculated using  the RMA-2V
    model for a range of flow conditions, and bottom shear velocities («*) were computed
    from depth and vertically-averaged velocity.

2.   The Depth of Scour Model calculates the bottom shear  stress from the bottom shear
    velocities using the relation:
3. Intersegment  flows  between  the larger HUDTOX  segments  were  defined by
   integrating  velocity  field results  from  the  hydrodynamic  model  at  the  various
   corresponding nodes.

3.3 AVAILABLE DATA

The  hydrodynamic model  RMA-2V requires  specific  input  data  describing  the  hydraulic
conditions of the system chosen for simulation. These input data consist of the grid used for the
computation, Manning's  'n'  to  parameterize the bottom friction,  the forcing functions  or
upstream boundary conditions,  and  the  downstream  and side-channel boundary conditions.
These are described below.
                                           18                       Limno-Tech, Inc.

-------
 3.3.1 Model Grid

 The RMA-2V model uses a six-node triangular element scheme to describe the physiography of
 the TIP system. The model grid consists of approximately 6,000 nodes defining 3,000 elements.
 Each node is defined by an x-y coordinate and its corresponding elevation. The depth associated
 with  each  grid node for the  main channel is based on the  bathymetric survey performed  by
 General Electric in 1991 (O'Brien & Gere,  1993b). Figure 3-2 shows the finite element grid used
 in the model calibration.  The finite element grid in the floodplain was constructed  using
 elevations  taken from  the  USGS topographic maps.  As seen in Figure 3-2, the  grid in  the
 floodplain is much coarser than in the Thompson Island Pool channels.  This is justified because
 velocities in the floodplain are much smaller than in the Pool channels and do  not vary as much.
 The nodes of the finite  element grid in the main channel are located approximately every 50 feet
 across the River (laterally) and approximately 300 feet along the channel (longitudinally).

 During the course of model calibrations and runs, it  was necessary to refine the grid so that the
 water mass was conserved at  the various transects  corresponding with  HUDTOX  segment
 boundaries. Conservation was achieved within a few  percentage points for each transect. This
 level  of accuracy was  sufficient to allow  post-processing of the  RMA-2V results to meet  the
 mass balance of water  requirements for the HUDTOX model without significantly affecting the
 routing of advective flows through segments in the Thompson Island Pool. The refining of the
 grid consisted of eliminating isolated nodes along the sides of the flow and smoothing the bottom
 elevations.   These changes were minor and had little impact on the calculated overall  velocity
 field.

 3.3.2 Manning's 'n'

 The  input  parameter,  Manning's  'n',  expresses  the river's hydraulic  resistance to  flow.
 Conceptually, resistance to flow reflects the character of the sediments and the  nature of the flow
 pathways.  This parameter is commonly a calibration parameter, because its value cannot  be
 determined accurately  from a measurement of the physical dimensions of the river or from a
 description of the sediment type.  Two site-specific hydraulic flow modeling studies,  Zimmie
 (1985) and FEMA  (1982), had been conducted  previously;  the  Manning's 'n' values can  be
 expected to be near  the values used in these studies.  Table 3-1 contains the Manning 'n' values
 used in these two studies.

 For this study, the values of Zimmie were used initially and subsequently calibrated to best fit the
 recorded observations of the river, especially those at high flow. The sensitivity of the model to
 changes in this parameter is discussed below in Section 3.6.1.

 3.3.3  Boundary Conditions

The principal input  to  the model is the upstream boundary condition, the incoming flow.  The
 model was run for the  eight different flows at Fort Edward shown in Table 3-2.  The first four
 flows are of interest because the concentration of suspended sediment in  the river was sampled
 when they occurred. The fifth flow is of interest because it is the highest flow recorded in TIP
 after the Fort Edward dam was removed in 1973.  The final three flows are of interest  because
                                           19                       Limno-Tech, Inc.

-------
 they represent high flow events with a specified return period. The model results for these eight
 flow  simulations were used  in the DOSM  to develop  relationships between river  flow and
 cohesive sediment resuspension.

 Other boundary conditions of the model consist of the side-channel boundary condition and the
 downstream water elevations.  The side-channel boundary condition is the requirement that the
 velocity normal to the sides of the channel be zero. This is implicitly performed in the  RMA-2V
 model.  The downstream boundary condition consists of specifying the water surface elevation at
 the most downstream transect, which is the Thompson Island Dam. The downstream  boundary
 must  be specified as an elevation in order to incorporate the backwater effects of the dam into the
 model.

 The downstream  boundary surface elevation was taken from the rating curve for  USGS Gage
 118, which is located just above Thompson Island Dam. The rating curve was developed from a
 regression analysis performed on the discharge-water level data accumulated during the 11 year
 period of 1983 to 1993 (USEPA,  1997).  Examination  of this  rating curve  showed that the
 regression is good for flows up to 30,000 cfs; however, the third-order polynomial developed in
 the regression fails to accurately predict increasing river elevations for flows above 30,000 cfs.
 Refined extrapolation using best engineering judgment and a theoretical rating curve  (Zimmie,
 1985) was used to determine the water levels at Thompson Island Dam above these flows.

 3.4 HYDRODYNAMIC MODEL CALIBRATION

 The hydrodynamic model calibration approach consisted of specifying an appropriate  value for
 the turbulent exchange coefficients based on literature values and then varying the Manning's 'n'
 so  that  computed river levels agree with elevations from the  upstream rating  curve.  The
 agreement with the upstream rating curve was assessed for each flow input at the most upstream
 transect of the grid. Note that only one  value of Manning's 'n' was used for the entire length of
 the main channel, because there are no physical data on which to base a variation of Manning's
 'n'. The upstream rating curve used for comparing to model output during calibration was USGS
 Gage  119, near Lock Number 7, which is near the southern tip of Rogers Island (Figure  3-1).

 Because the calculation of velocity is of primary interest for larger flows on the Hudson River,
 the calibration first focused on the flow of 30,000 cfs, which is the highest flow for which the
rating curves for both USGS Gage 119  (upstream) and USGS  Gage 118 (downstream) are
 substantiated.  The Manning's 'n' values were calibrated for 30,000 cfs and were then used in the
 model to predict water elevations for lesser flows.  These predicted water elevations were then
compared with the elevations from the Gage 119 elevations.

The turbulent exchange coefficients were set to 4,790 Pa-sec (100 lb-sec/ft2) which is within the
range  of longitudinal turbulent dispersion (Kjj) values measured  in a variety of rivers (Fischer
et.al.,  1979).  The measured dispersion  numbers  can be directly translated  into  turbulent
 momentum exchange coefficients, since for most turbulent flows  the turbulent Prandtl number
 (Ejj/Kjj) equals 1.0 (Tennekes and Lumley, 1972).
                                           20                 .      Limno-Tech. Inc.

-------
As described  above,  the  model was primarily calibrated for the flow of 30,000 cfs.   The
Manning's 'n' values for the final calibration were 0.020 for the main channel and 0.060 for the
floodplain. The model computed the same river water surface elevation as was observed at Gage
119 using these Manning's 'n' calibration values.  Table  3-3 shows this result, along with the
comparison of model output vs. rating curve water levels for lesser flows. The elevations in the
table are listed in feet relative to the National Geodetic Vertical Datum (NGVD).

Comparing the last two columns in Table 3-3 shows that the model's results are slightly higher
than the rating curve for the smaller flows, implying that the calibrated Manning's 'n' might be
somewhat low for the lower-flow cases. It is possible that the rating curve used in the calibration
was biased at either low or high  flow, making calibration  difficult across the entire flow range.
Nevertheless, it was judged that a higher value could not be justified, given the model's close fit
for 30,000 cfs, (a higher Manning's 'n' would unacceptably increase the model's prediction of
the upstream water surface in that case).

The excellent model fit at the calibration  flow of 30,000 cfs, along with good results from two
validation exercises described below, provide confidence  in using the model to simulate  high-
flow events.

3.5 HYDRODYNAMIC MODEL VALIDATION

There were two additional and independent sources of information used to verify the calibration
results.  The  first source is the Hudson River velocity measurements  made  in the Thompson
Island Pool by the  USGS.  The second  source  is the flood study conducted  by FEMA.  A
comparison of model results with these sources of information is discussed below.

3.5.1 Rating Curve Velocity Measurements

The USGS periodically measures the flow in the Hudson River in the Thompson Island Pool to
develop and update the river's rating curves. For the rating curve located at Scott Paper, which is
upstream  of  Rogers Island, the flow  is measured by measuring the depth and velocity at
numerous points over the cross-section of the river at Rogers Island. These data are taken at the
bridges  over the Hudson River on both sides of Rogers Island. The model's simulated velocities
can be compared to these measured velocities as a check on the accuracy of the model.

The model was run for  the discharge (29,800 cfs) that was measured  on April  18, 1993. The
velocities computed by the model for locations  along  the  cross-section of  the  river were
approximately equal to  or slightly  lower than  measured.  For  example, the  river velocities
measured in the middle of the channel by the USGS were approximately 4.3 feet per second
(fps), while the model computed  velocities of approximately  4.1  fps.   These  values  are
sufficiently close for validation. It should be noted that since the velocities were measured from a
bridge, it is to be expected that  the measured velocities are slightly higher than the computed
ones, since the bridge piers will cause a localized acceleration in the flow. Constraints on model
resolution inhibit the ability to capture these localized effects on the flow.
                                           21                        Limno-Tcch, Inc.

-------
3.5.2 FEMA Flood Studies

The Federal Emergency Management Agency (FEMA) regularly conducts studies to predict the
flood elevations in rivers for flows of various return periods.  The results of the study conducted
by  FEMA in 1984 for the Upper Hudson River were used as an  additional  validation  of the
credibility of the model.  The 100-year flow used by FEMA (52,400 cfs) is greater than the 100-
year flow used in this study (47,330 cfs) so that a direct comparison  of 100-year flood elevations
was not initially possible. Estimates of the 100-year flow magnitude are different due to  use of
different datasets and estimation methods.  However, the model was also run for the  100-year
FEMA flow of 52,400 cfs, and the model predicted a river elevation at Fort Edward of 130.4 ft.
NGVD  (National Geodetic Vertical Datum, formerly Sea Level Datum of 1929).  The FEMA
flood study using the HEC-2 program predicted a river elevation  of 130.7 ft. NGVD.  These
results are comparable considering that the two models reflect a slightly  different representation
of the river hydraulics.

The RMA-2V model developed here was also run for 52,400 cfs with a  Manning's 'n' of 0.030
for  the main channel and 0.075 for the floodplain (approximately the same as the FEMA study).
This resulted in a predicted river elevation of 131.7 ft.  Most importantly, the river velocities do
not vary appreciably for the various representations. Therefore, the model results are judged to
be comparable to those produced from the FEMA flood study.

3.5.3 100-Year Peak Flow Model Results

The model was used to  simulate the  100-year  peak flow of 47,330 cfs.   The predicted river
elevation at the downstream tip of Rogers Island was 128.6 ft. This elevation is slightly lower
than the extrapolated rating curve's elevation of 129.1, but is reasonably close.

The vertically-averaged velocity field produced by RMA-2V for the 100-year peak flow is shown
in Figure 3-3. The velocity magnitudes are reflected by the length of the vectors in accordance
with the scale provided near the bottom of the figure. The vectors in the  floodplain that have no
visible tail indicate slow moving water in the overbank area. A vector was printed where the
water depth was greater than zero, even if the velocity was small, to indicate  the extent  of the
flow.

The RMA-2V velocity field was  used to compute the shear stresses in the DOSM within the
normal river banks of the Thompson Island Pool,  not in the  floodplain.  Floodplain simulation
was only  included to ensure an appropriate representation of the  in-river, vertically-averaged
velocity field.

3.6  HYDRODYNAMIC MODEL SENSITIVITY ANALYSES

The sensitivity of the model to the principal inputs was evaluated by varying the finite element
grid size,  the Manning's 'n', and the turbulent exchange  coefficient. The model's sensitivity to
the  grid size was checked by running the model for a flow of 40,000 cfs with a finite element grid
having approximately two times the number of elements as the baseline finite element grid.  The
results obtained with the  larger grid resolution were essentially the same  as the smaller grid and,
                                           22                       Limno-Tech. Inc.

-------
therefore, it was concluded that the finite element grid used here was of sufficient resolution to
simulate the river flow.  The sensitivity of the model to the Manning's 'n' and the turbulent
exchange coefficient was measured by the effect on the predicted water elevations for the 100-
year peak flow at the downstream tip of Rogers Island (Gage 119).  The sensitivity  results are
presented in the following discussion.

3.6.1 Manning's 'n'

The Manning's 'n' was varied  from 0.015 to 0.035 for the main channel and 0.040 to 0.080 for
the floodplain. These values of 'n' are  consistent with what has been previously used for this
reach of the Hudson  (Zimmie, 1985; FEMA,  1982), and with literature values (Chow,  1959;
Hicks and Mason, 1998). The model was run for the 100-year peak flow of 47,330 cfs; the results
are contained  in Table 3-4.  These results indicate that  changes  in Manning's  'n' do not
significantly affect results from the calibrated model.  It  is also evident that the main channel
Manning's 'n'  generally  affects the results much more than the floodplain Manning's 'n',  as
would be expected because most of the flow occurs in the  main channel. The model irisensitivity
to Manning's 'n' is due to the fact that the flows are large  and the system is  strongly forced. The
accurate prediction of stages and velocities in this flow regime depends  more on  having  an
accurate representation of the depth of the main  channel and the flood  plains.

3.6.2 Turbulent Exchange Coefficient

The four turbulent exchange coefficients, Exx,, Exy, Eyx, and Eyy were  all set to a value of  4,790
Pa-sec (100 lb-sec/ft2) in the  baseline run.     Table 3-5 shows  the  effects of varying these
turbulent exchange coefficient values on the water surface elevation at Rogers Island.

It  can be concluded that variations  in  turbulent viscosities do not affect  the  river elevation
dramatically, especially evidenced by the small increase in the river elevation for each doubling
of  the  coefficients.   The model predicts higher  elevations  for  higher  turbulent  exchange
coefficients in much the  same way that  it would predict  higher elevations  with a  larger
Manning's 'n'. Both  the Manning's 'n' and the turbulent  exchange coefficients parameterize
energy loss in  the system. This means that if higher turbulent exchange coefficients were used in
the calibration,  then  a  lower Manning's 'n'  would be  required to  obtain an equally good
agreement with the observed rating curve.  Given  these results, it was judged that a turbulent
exchange coefficient  of 100 lb-sec/ft2  was reasonable  and  that further calibration was not
required.

3.7 CONVERSION OF VERTICALLY-AVERAGED VELOCITY TO BOTTOM SHEAR STRESS

Conversion of the vertically-averaged River velocities, as obtained from the RMA-2V model, to
bottom  shear  stresses is required  to compute  resuspension of  Thompson Island  Pool bed
sediments in the DOSM and HUDTOX models. Several formulations were investigated. One of
these formulations computes shear stress directly from the  vertically-averaged velocity, while the
other three provide computed values of bottom shear velocity, u*, for use  in computing shear
stress as i = p (u* }2.  The four methods,  with a short description of each, are  presented below.
                                           23                       Limno-Tech, Inc.

-------
 1.  Smooth wall log velocity profile

This conversion method (Thomas and McNally,  1990;  Schlichting,  1979) derives from the
assumption that the vertical velocity profile at any point  in the river conforms to the "smooth
wall log velocity profile". The following equation describes this velocity profile:
                     „•        ^   v   )                                            <3'6)

where:
       u      =     vertically-averaged velocity [L/T]
       u*     =     shear velocity [L/T]
       d      =     depth of flow [L]
       v      =     kinematic viscosity [L2/T].
The applicability of this relation to the Upper Hudson River is suspect, because it is known that
the bottom of the river is not hydraulically smooth.

2. Gailani Method

This empirical method was used by Gailani (Gailani et al., 1991) for the Lower Fox River, as
follows:

                    ^- = 0.003 u2                                                   (3-7)
                     Po
where:
                                             f\
       tb     =     bottom shear stress [M/L/T ].

       Po     =      reference density [M/L3].
3. Rough wall log velocity profile

                     — = 6.25 + 2.51n(d/k)                                           (3-8)
w
  here:
       u      =     vertically averaged velocity [L/T],
       u*     =     shear velocity (friction velocity) [L/T],
       d      =     depth of flow [L],
       k      =     equivalent Nikuradse roughness [L].

This relation (Thomas and McNally, 1990) describes the velocity profile for a rough wall river
flow, which is typically the condition for river flows. The only free parameter for this equation is
k, the roughness factor.  This parameter can be estimated from the Manning's roughness (Chow,
1960): for 'n' = 0.02, k was determined to be 0.04 feet.
                                           24                       Limno-Tech, Inc.

-------
 4.  Manning shear stress equation

 The fourth formulation, the Manning shear stress equation was selected for use in the
 Depth of Scour Model. It involves a combination of the cross-section average velocity
 and bottom shear stress equations (Thomas and McNally, 1990).  Specifically, the bed
 shear velocity is expressed as:

                                                (English Units)                      (3-9)
                          (1.486)^
                                  ,/6
                     u* =                       (Metric Units)                      (3-10)
                            d

The channel average velocity is defined from the one-dimensional Manning equation, which is
given below, as:
                     u =      /?2/3 s"2                 (English Units)               (3-11)
                           n
                     u = -R2n S1'2                     (Metric Units)                (3-12)
                        n

The definition of the cross-sectional average shear stress (TO), can be written as,

                     T0 =  wRS = p gRS                                             (3-13)
where:
       u      =      channel averaged velocity [L/T],
       n      =      Manning's 'n' [T/L1/3] ,
       g      =      acceleration due to gravity [L/T2],
       p      =      density of fluid [M/L3]
       w      =      weight of the water (pg) [M/L2/T2],
       R      =      hydraulic radius [L],
       5      =      the slope of the river [dimensionless].
The definition of the friction velocity u* can be combined with Equation 3-13 to yield;

                                                                                   (3-14)
For flow in a wide open channel, the wetted perimeter is approximated by the depth (R = d).
Combining this assumption with Equations 3-12 and 3-14 will yield Equation 3-10.
                                           25                        Limno-Tcch, Inc.

-------
 Results comparing the model calculations using the  four different methods are  presented in
 Figure 3-4, which shows the variation of shear stress with the average vertical velocity among
 methods.  In Figure 3-4, the depth used to calculate the conversion for methods 1,2 and 4 was 10
 feet.  As  seen in Figure 3-4, Method 1, the  smooth wall velocity profile, and Method 2, the
 Gailani method, yield the smaller shear stresses, especially at higher flows. Methods 3 and 4, the
 rough wall and Manning's methods respectively, yield appreciably higher values  for stress at
 high  velocity  flows. Method 4 (Manning's) was chosen to estimate shear stress because it is
 consistent  with the RMA-2V approach,  and it provides the most critical (highest)  estimates of
 bottom shear stress for the DOSM.

 The shear  stress field for the Thompson Island Pool 100-year peak flow, as computed by the
 Manning method using the velocity field  shown in Figure 3-3, is plotted in Figure 3-5. Maximum
 stresses are observed in the flood plain, which is to be expected since the depths of the flow are
 smaller and the Manning's 'n' is 0.06, compared to 0.02 in the main channel.

 3.8 DISCUSSION OF RESULTS

 The calibrated RMA-2V model is a good representation of Thompson Island Pool hydraulics for
 various  flow regimes.  This conclusion is based on  the good  agreement found between model
 output for water levels and rating curve results  at Lock 7, and the good agreement between model
 output for velocities and those measured by the USGS. The model's ability to simulate flows
 well above the calibration flow, 30,000 cfs, is supported by the reasonable agreement between
 the 100-year peak flow predictions by this  model and the FEMA model, and also by the  lack of
 sensitivity of high-flow results to changes in internal model parameters.

The sensitivity analyses show that the RMA-2V model is not appreciably sensitive to changes in
 the calibration parameters.  However, the analysis of the conversion of the flow  field output
 (vertically-averaged velocity and depth) to river- bed shear stress shows that shear stress can vary
 significantly at high flow, depending on the conversion method used. The lower bound estimate
 for the smooth wall profile is not applicable.  However, the other three methods are potentially
valid and provide similar results. The most conservative method, that method which predicts the
largest shear stress given the magnitude  of  the  vertically-averaged velocity, was chosen  to
provide  shear  stress to the DOSM. However, overall differences among the three  methods are
approximately less than 30 percent.
                                           26                       Limno-Tech, Inc.

-------
Chapter 4

-------
4. THOMPSON ISLAND POOL DEPTH OF SCOUR MODEL

4.1 OVERVIEW

The Depth of Scour Model (DOSM) is a two-dimensional model of sediment erosion depth that
was applied to Thompson Island  Pool.  This model  was  developed  as  a  stand-alone tool
specifically to address one of the three principal study questions:

       •     Are  there contaminated  sediments now buried that are likely to become
             "reactivated" following a major flood, possibly resulting in an increase in
             contamination of the fish population?

The DOSM formulations were also integrated into the HUDTOX mass balance model, providing
consistency  between these two  models  in  Thompson Island Pool for  cohesive  sediment
resuspension.

The DOSM has different formulations for cohesive and non-cohesive  sediment scour. Cohesive
sediment  scour is calculated  on  the  basis  of site-specific measurements  of resuspension
properties for Thompson Island Pool cohesive sediments. Non-cohesive sediment scour depth is
computed via formulations available in the scientific literature, using sediment physical property
data for the Pool.  Scour depth calculations in each sediment type for  the  Thompson Island Pool
are based on the depth of scour equations, linked to steady-state hydrodynamic model predictions
of overlying water velocity (Chapter 3).

The cohesive scour depth calculations are probabilistic estimates. The 5th to  95th percentile
estimates were used to define the "likelihood" that buried contaminated sediments are reactivated
in a flood. However, the mean estimate at each location is used in HUDTOX transport and fate
simulations.   For non-cohesive scour depths,  only a  singular theoretical estimate is available.
Both cohesive and non-cohesive estimates assume that the flow condition  of interest persists long
enough to achieve  the estimated scour depth.   For cohesive sediments, laboratory observations
show that maximum scour will occur within approximately one hour for all flow conditions. The
time to maximum scour  cannot  be  determined for non-cohesive  sediment in  the DOSM
framework and hence, the model calculations are viewed  as an upper bound.

The DOSM is  used to answer the principal  study  question presented above by estimating
probable ranges of sediment scour depth expected in cohesive sediment areas  with the occurrence
of a 100-year flood event in Thompson Island Pool. These depth of scour  ranges are compared to
vertical PCB sediment  concentration profiles  at five specific locations.   Additionally, these
ranges  are used to estimate PCB mass and sediment eroded from cohesive sediments through a
poolwide application of DOSM. Cohesive sediment areas are of special  interest relative  to the
non-cohesive sediment areas due to their higher levels of contamination. Observed PCB hotspots
generally  coincide  with  cohesive  sediment   areas  and  exhibit  the   highest buried PCB
concentrations.

The DOSM  cohesive sediment resuspension algorithms are  also used in the HUDTOX mass
balance model (described in Chapter 5).  Although the DOSM does not account for transport or
                                          27                      Limno-Tech, Inc.

-------
redeposition of scoured sediment and PCBs, the HUDTOX  mass balance model does.  The
HUDTOX  model incorporates cohesive  sediment  resuspension algorithms obtained through
application of the DOSM model to a range of flow conditions. Median values were used from
the probabilistic DOSM model application.  Equations were developed to relate the predicted
cohesive sediment mass scoured to flow at Fort Edward for each HUDTOX cohesive sediment
segment in Thompson Island Pool. Through this approach, consistency between the DOSM and
HUDTOX  models is achieved for cohesive sediment resuspension.   In contrast  to cohesive
sediments,  time  to maximum scour  for non-cohesive sediments is uncertain.  Therefore the
DOSM non-cohesive sediment scour depths were viewed as an upper bound estimate, with actual
resuspension values calibrated in HUDTOX.

The findings from the Depth of Scour Model application show that the expected impact of a 100-
year flood  on surface sediment PCB concentrations in  the Thompson Island Pool is small
because computed scour does not expose higher concentrations of PCB in the sediments. Results
suggest that the 100-year event is not an important  concern from the standpoint of a potential
remediation decision for PCBs in the  Thompson Island Pool.  Specific findings are presented at
the end of this chapter. These findings were corroborated by HUDTOX simulation of long-term
response to a 100-year peak flow, presented in Chapter 8.

The following major sections are included in Chapter 4:

      4.2    DOSM Model Development

      4.3    DOSM Parameterization

      4.4    DOSM Application

      4.5    Major DOSM Findings

Section 4.4 presents  results of the DOSM estimates regarding depth of scour,  likelihood  of
"reactivating" buried PCBs at the five high resolution core locations, and an estimate of the mass
of PCB eroded from cohesive sediment areas due to a 100-year peak flow.

4.2 DOSM MODEL DEVELOPMENT

4.2.1  Conceptual Approach

Two categories of information are necessary to compute the depth of erosion and total mass of
solids eroded from bedded sediments  for a high-flow event. First, the hydrodynamic conditions
at the sediment-water interface  need  to be specified.   The primary forcing  function for
entrainment of  bottom sediments into the flowing water is  the shear stress exerted at the
sediment-water  interface by flowing water.  The Thompson Island Pool Hydrodynamic Model
yields estimates of vertically-averaged flow velocities at a fine spatial resolution.  Bottom shear
stresses  are computed from the velocities  by a  simple  formula  (Section  3.7).   Second, the
physical-chemical properties of the bedded sediments greatly influence the magnitude and rate of
entrainment of sediments for a given event, and the resulting depth of scour.  These are specified
from data.
                                          28                       Limno-Tech. Inc.

-------
Entrainment mechanisms can be classified into two distinct categories based on sediment bed
properties.  The main parameters affecting the entrainment of non-cohesive sediments include
grain size and shape (and their distributions), the applied shear stress, bed roughness, and specific
weight. Bed sediments that are primarily fine grained and/or possess a high clay content exhibit
interparticle effects that are cohesive in nature.  The resulting entrainment properties are very
different from non-cohesive sediments.   Since the toxic contaminants of interest (PCBs) are
associated preferentially  with  fine  grained sediments, this distinction  is  of considerable
importance. Each approach is described separately below.

4.2.2 Formulation for Cohesive Sediments

4.2.2.1  Background

Particle diameter has a  significantly lower influence  on  the entrainment characteristics  of
cohesive sediments compared to electrochemical influences.  Relatively small amounts of clay in
the sediment-water mixture  can result in critical shear stresses far larger than those in non-
cohesive  materials of similar  size  distribution (Raudkivi,  1990).  Previous  studies on  the
entrainment of cohesive sediments hypothesize that the scour magnitude is primarily influenced
by the excess applied shear stress (i.e., the difference  between the applied shear stress and the
critical  shear stress of the surficial  sediments), and  the state of consolidation (or age after
deposition) of the bed sediments (Partheniades, 1965; Mehta et al., 1989; Xu, 1991).  The mass
of material resuspended can be expressed in the  following functional form:

                M = f(T - TC; age, other sediment properties)                             (4-1)

where M  is the mass of material resuspended, T is the applied shear stress, and Tc is the bed
critical shear stress.  The function f has been expressed in a variety of different forms, including
linear, (e.g. Partheniades, 1965), exponential, (e.g. Parchure and Mehta, 1985), and the power
relationship, (e.g. Lick et al.,  1995; Gailani et al., 1991).

4.2.2.2  Basic Equations

Lick et al. (1995),  proposed an erosion equation based on statistical analysis of laboratory and
field  data.  This work forms the basis  of DOSM  calculations for cohesive sediments and is
expressed as follows:

                                                                                     (4-2,
where:
       e = the net total amount of material resuspended (g/cm2);
       T = the applied shear stress (dynes/cm2);
       Tc. = the bed critical shear stress (dynes/cm2);
       Id = the time after deposition (days); and
       ao, n, and m = empirical constants.
                                            29                        Limno-Tech, Inc.

-------
The empirical constants ao, n, and m are obtained through fitting of Equation 4-2 to experimental
data. The critical shear stress, Tc, is also determined through experimentation. The depth of scour
can then be calculated as:
                                                                                      (4-3)


where Zscour is the depth of scour (cm), and Qjuik is the dry bulk sediment density (g/cm ). These
equations have been applied to site-specific data for several rivers (Fox, Detroit, and Buffalo) by
McNeil (1994).

4.2.2.3  Reparameterization to a Probabilistic Model

The reassessment study asks if buried contaminated sediments are "likely" to become reactivated
following a major flood.  To address this issue of likelihood, the resuspension formulation was
adapted to provide probabilistic calculations of scour, as described below.

If the value of TC is assumed to have been defined from resuspension experiments, while the other
parameters are  unknown, then Equation 4-2 can be reduced from five parameters to two using a
dimensionless shear stress parameter, i7:
                               m                                                     (4-4)

where:

       T/      =      (T-TC) / TC,
       A      =      ao/tdn
Equation 4-4 can be linearized as follows:
                     ln(Ł)=ln(A)+mxln(T')                                        (4-5)

Therefore, a linear regression  may be performed to fit a straight line to data for erosion vs.
dimensionless shear stress in "log-log"  space.   The slope obtained from this regression will
correspond to the exponent "m" from Lick's equation, while the intercept will correspond to the
logarithm of the lumped term  ao/td".  Characterization of the distribution of errors around this
regression will allow estimation of the uncertainty in erosion predictions due to uncertainty in
measured resuspension properties.

Given a  regression  line  with  normally  distributed  residuals,  prediction  limits  for new
observations (for a given value  of the independent variable) fall on a Student-t distribution (Neter
et.  al.,  1990).  For large sample  sizes, the Student-t distribution  is approximately normal.
Predicted values  for  new observations are  therefore calculated as percentiles of  normal
distributions, in log-log space.  The resulting predicted distribution in ordinary space (again, for
given values of shear stress) is log-normal, and is  calculated according to Equation 4-6.
                                            30                        Limno-Tech, Inc.

-------
                     e =expA + mxlnT'+w                                        (4-6)

where:
                             MSEx
and

       T7      =     (T-TC) / tc,
       exp    =     exponentiation operator
       Z      =     a value of the standard normal distribution variable
       MSE   =     mean square error of regression
       ns     =     number of data used in the regression
       Xavg    =     mean of the natural log dimensionless shear stresses
       Xi      =     a particular natural log dimensionless shear stress value.
Division of the erosion by the bulk density gives the depth of scour in centimeters, as shown in
Equation 4-3.

4.2.2.4  Calculation of PCB Erosion

Equations 4-3 and 4-6 define a probabilistic model for predicting bottom cohesive sediment mass
erosion and depth of scour as a function of shear stress and sediment physical properties.  The
model is probabilistic in that it presents a range  of depth of scour estimates  with associated
probabilities based on the variability in  experimental resuspension measurements. For a given
scour depth, an estimate of the PCB erosion from cohesive sediments can then be estimated as a
function of sediment PCB concentration using Equation 4-7.

                     D_  S*CFCB                                                  (4.?)
                         (lOOOmg ]

                         I   *   J
where:
       P      =     quantity of PCBs eroded from cohesive sediments (g)
       5      =     mass of solids eroded from cohesive sediments (kg)
       CPCR    =     average cohesive sediment surficial PCB concentration (mg/kg).
In a stand-alone application of DOSM used to evaluate  the impact of a 100 year peak flow,
average surficial sediment PCB concentrations were  used to provide a conservative screening
estimate of eroded PCB.  However, in long-term forecasts with HUDTOX, model simulations of
PCB in individual PCB layers was used, rather than surficial averages.
                                           31                        Limno-Tech. Inc.

-------
 4.2.3 Formulation for Non-cohesive Sediments

 4.2.3.1 Background

 Net erosion of non-cohesive sediments occurs when the sediment transport capacity of the flow
 exceeds the  actual sediment burden being carried by the  flow.  A flow will have transport
 capacity for  a particular particle diameter (size class) when the shear stress applied to those
 particles by the flow exceeds the critical  shear stress of the particle size class. The transport
 capacity of the flow is inversely related to the particle size; hence, differential scouring takes
 place, with the smaller particles  being removed in greater proportion than the larger particles.
 The particle size distribution of the bed surface then shifts progressively towards larger particles.
 If sufficient large particles are present that cannot be transported under the flow conditions, the
 bed surface will  come to consist primarily of the larger particles, with the smaller particles
 underneath sheltered from scour. This layer of coarse particles, called the armor layer, may
 persist until higher flows and their associated shear stresses  erode it, causing  further coarsening
 and the establishment of a new armor layer. The armor layer can be degraded  by vertical mixing
 with the parent bed material and replenishment of fine material via deposition from the  water
 column.

 4.2.3.2  Equations

 Borah (1989) gives equations for the depth of scour that will occur before the establishment of an
 armor layer. His formulation assumes a well-mixed surface  layer with constant particle specific
 gravity, but different particle sizes.   After a scour event and armoring, the result is  a single
 surface layer of the smallest non-transportable particle size.  The formulation  may be viewed as
 conservative because the potential for finer particles to be trapped (hiding) in the armor layer is
 ignored. This means that the mass of sediment scoured to achieve armoring may be high because
 the fine particles  that may be trapped are assumed to be scoured in order to  achieve armoring.
 An active layer thickness is defined as:

                     T=,  D\                                                       (4-8)
where T is the thickness of the active layer (cm); Da is the smallest armor size  (cm); <}> is the
porosity of the bed  material;  and Pa is the fraction of all the armor sizes present in the bed
material. Da is computed using a modified version of the Shields Curve (Shields,  1936; van den
Berg and van Gelder, 1993). The scour depth is then computed as:
                     E = T-Da                                                      (4-9)

where E is the scour depth (in cm). These equations have been applied and the results validated
for laboratory (Little and Mayer, 1972) and field (Karim and Kennedy, 1982) data.

4.2.4  Time Scale of Erosion Estimates

The cohesive sediment scour calculations  result in a mass estimate at the peak flow for an event
assuming that the event peak shear stress is established essentially instantaneously. Experiments
                                            32                       Limno-Tech, Inc.

-------
by Lick et al. (1995) indicate that this mass is eroded over the time scale of approximately one
hour. The non-cohesive computations provide a mass estimate corresponding to scour down to
the armoring depth.  However, the time required to reach armoring depth cannot be directly
calculated with the available models.  Model predictions for non-cohesive sediments should
therefore be considered "upper bound" estimates, as they are based upon the assumption that the
flood event is of sufficient duration to allow erosion to proceed all the way down to the armoring
depth.  This upper bound estimate is suitable for determining  the  likelihood that the buried
contamination can be "reactivated", but it is not suitable for direct use in HUDTOX.  Hence, non-
cohesive sediment resuspension rates in HUDTOX were calibrated.

4.3 DOSM PARAMETERIZATION

4.3.1 Data

4.3.1.1 Distribution of Types of Bottom Sediment

The bedded sediments  in Thompson Island Pool were  differentiated as cohesive and non-
cohesive based on side-scan  sonar profiles of fine and coarse sediments (Flood,  1993).   The
analysis of sonar and sediment data suggested that the results of the 500 kHz digital image (i.e.
mean digital number, or DN) can be successfully correlated to mean grain size.  It was found that
DN values less than about 40 generally correspond to finer grain sizes (mean size less than about
4 phi) while DN values greater than about 60 generally correspond to coarser sediments (coarse
sand, gravel). For the purpose of characterizing the sonar images, sediment type is  described as
"finer" for DN less than 40, or as "coarse" or "coarser" for DN greater than 60.

The sonar maps were qualitatively divided into several categories including "coarse", "coarser",
"finer", "island",  and "rocky".  These maps were  digitized into a GIS coverage by TAMS
Consultants, Inc.  No sediments described as "coarse"  were listed for Thompson  Island Pool.
The two sediment categories considered for this analysis to be significant sources of potentially
erodible materials (due to magnitude of area and/or substrate type) were "coarser" - representing
non-cohesive sediments - and "finer" - representing  cohesive  sediments.  The  area  of non-
cohesive sediments in  Thompson Island Pool is approximately three times  that of  cohesive
sediments.

4.3.1.2 Resuspension Experiments

Data  used to parameterize the DOSM  for  cohesive Thompson Island Pool sediments  were
obtained from resuspension experiments described in a report by HydroQual (1995).  This report
contained two different sets of experimental data.

The  first dataset came from  an annular  flume study,  wherein sediments from three  different
locations  in Thompson  Island Pool were transported to  a laboratory at the University  of
California at Santa  Barbara and subjected to two  types of experiments involving  shear stress.
Multiple  shear stress tests were conducted by filling the flume  with sediment, allowing it to
compact  for 1,  3, or  14  days  with the flume at rest, and  running (i.e.,  rotating) the  flume at
successively higher levels of shear stress, with steady state suspended sediment concentrations
                                           33                       Limno-Tech, Inc.

-------
achieved (as indicated by concentration measurements at 30 minute intervals) before each shear
stress increase.  A continuous flow test was conducted by filling the flume with sediment and
running it continuously for 47 days at a shear stress of about one dyne/cm2, except that on several
days the shear stress was increased to 5 dynes/cm2 for two hours. Also, one multiple shear stress
test similar to those described above was conducted.

The  purpose of these experiments was to investigate the  effects of bed compaction and to
estimate the  value of the  critical shear stress,  within the framework  of the Lick equation,
Equation 4-2.  Based upon these laboratory flume experiments, HydroQual  (1995) concluded
that: 1) the critical shear stress was approximately 1.0 dyne/cm2, 2) the maximum time since
deposition (id) was 7 days (i.e., after 7 days no further significant bed compaction takes place),
and 3) the exponent, n, for t^ was 0.5.

Although laboratory-derived values were used in the DOSM, there are environmental factors that
were not accounted for in the laboratory experiments.  Critical shear stresses from resuspension
can vary seasonally due a number  of factors. These include disturbance of sediments by benthic
organisms, generation of  gases  from  decomposition of organic  matter, and  uprooting of
macrophytes.  Also, the bed surface in the river may be much more varied than the  planar
surfaces achieved in the laboratory annular flume experiments. Nonetheless, these laboratory
data were the best available information.

The second set of sediment resuspension measurements described in HydroQual (1995) consisted
of field studies  using a portable resuspension  device, commonly called  a shaker.   Surficial
sediment cores were collected at 20 cohesive sediment locations in Thompson Island Pool and 8
locations downstream; each location had one (Thompson Island Pool) or  two (downstream) sets
of three cores each.  Each  core was subjected to a shear stress in the shaker and the  resulting
resuspension potential was determined.  The field study produced  107 resuspension potential-
shear stress data pairs for the Hudson River, with 60 measurements specific to Thompson Island
Pool.   The shear stresses  used in the field study ranged from 5 to  11  dynes/cm2.  Observed
sediment erosion rates in Thompson Island Pool ranged from 0.06 to 28.84 mg/cm2.

From the Thompson Island Pool-specific data, HydroQual (1995) assumed a Thompson Island
Poolwide constant value of 3 for m, and back-calculated the core-specific  values for ao necessary
to produce the observed erosion.  The methodology used to determine the  value for m was not
provided.  HydroQual reported a mean value and standard deviation for ao  of 0.071 (in units of
mg- dayl/2/cm2) and 0.062, respectively, excluding certain results deemed to  be outliers.

4.3.1,3 Non-Cohesive Particle Size Distributions

The Borah formulation described above (Equations 4-8 and  4-9)  requires  sediment data on
particle size distribution,  particle  density,  and  wet bulk  density  (to  calculate  porosity).
Unfortunately, a large percentage  of the cores had missing or incomplete data for one or more
properties. This obstacle was overcome in two ways: 1) missing data on particle density and
bulk density were replaced by random deviates from the distributions found for the existing data,
and 2) particle size distributions, which  were occasionally incomplete on the large-particle end,
were extrapolated by plotting the data for each core as In(size) vs. In(fraction) and extending the
                                           34                       Limno-Tech, Inc.

-------
 curves smoothly (this was done for 81 cores with data to extrapolate).  The distribution used for
 particle density was normal with a mean of 2.438 g/cm3 and a standard deviation of 0.262.  The
 distribution used for wet bulk density was normal with a mean of 1.452 g/cm3 and a standard
 deviation of 0.212; random deviates  greater than 1.8 or less  than 1.04 were  rejected on the
 grounds of  physical improbability  and  were replaced  with new  deviates.    Particle  size
 distributions were extrapolated as far as size fraction 2.7 percent or size 20 mm.

 The data synthesis procedures (extrapolations and data substitutions) contribute to uncertainty.
 However, it was judged that  more uncertainty would result from ignoring  the sample datasets
 entirely where one parameter was missing.

 4.3.1.4  1984 Cohesive Sediment PCB Concentration

 The DOSM was used separately from HUDTOX to develop a conservative estimate of PCB mass
 and associated sediment eroded from cohesive sediment areas in response to a 100-year peak
 flow.  This was accomplished by using the 1984  NYSDEC  sediment PCB data from  grab
 samples and coarsely-segmented sediment cores.  Surface  sediment core sections for these data
 were on average about 10 inches thick.  The mean cohesive sediment surface sediment PCB
 concentration is approximately 32.5 mg/kg (USEPA,  1998a).  This value was applied to the
 DOSM  cohesive sediment elements, which are the same as the hydrodynamic model elements in
 Figure 3-2.  The depth of scour in each element at the 100-year flow was converted to PCB mass
 eroded using this concentration.

 Use of the 1984 median surface concentration results in a conservatively high estimate of PCB
 mass likely to be eroded under a future 100-year event for the following reasons:

    1.  The surface concentrations at the present have decreased significantly from those
       observed in 1984; and

    2.  The coarse vertical segmentation of the 1984 core samples likely resulted in an
       over-estimate of mean surface concentrations  because  peak concentrations are
       buried.

It should also be noted that the 1984 NYSDEC data do not represent total PCB and therefore do
not provide an estimate of total PCB mass eroded.  These data more closely represent the sum of
the tri- and higher-chlorinated congeners, which is discussed in detail in Section 6.3.3.

Results  of the mass erosion estimates from cohesive sediment are presented below in Section
4.4.3.1

4.3.2 Parameterization for Cohesive Sediments

There are several assumptions inherent in the application of Equations 4-3 and 4-6 to the shaker
data for parameterization of the DOSM. These include:

       •      The value for critical shear stress obtained from the annular flume study is
              constant and applies throughout Thompson Island Pool;
                                           35                        Limno-Tech, Inc.

-------
       •      The sediment cores used in the resuspension studies represent an unbiased
              random sample of Thompson Island Pool cohesive sediments;

       •      The experimental shear stress values are exact;

       •      The statistical model is valid for extrapolation to higher values of shear
              stress than were used experimentally; and,

       •      The bulk density, at a specific location, used for converting erosion to
              depth of scour can be represented as a single number.

                                                     fIS                          rtft
All statistical analyses were conducted using SYSTAT  Version  6.0  for Windows  (SPSS,
1996), and only data from Thompson Island Pool were considered. A linear regression of natural
log erosion (in mg/cm2) vs. natural log T/ produced an intercept  (A)  value of -3.829 and a slope
(m) value of 2.906 (Figure 4-1).  Of 60 Thompson Island Pool data points, two outliers were
deleted; 58 data points were used.  The outliers were identified solely on the basis that their
Studentized residuals were too large (absolute value greater than 3.0). The regression R-squared
value was 0.541, and p-values for both the regression constant and the slope were  <0.00001.  An
analysis of the residuals strongly indicated that they could be assumed to be normally distributed.
It was concluded on the basis  of these and other statistical indications that the use of linear
regression was supported by the data.

The value of 2.906 obtained for m is similar to the value of 3  reported by HydroQual (1995).
Assuming from the flume studies that the maximum time since deposition (td) was 7 days, and
the exponent, n, for tj was 0.5, the lumped term corresponds to a value of ao of 0.0575.  This
value is well  within one standard deviation of the value reported by HydroQual (1995), (Section
4.3.1.2).

4.3.3  Parameterization for Non-cohesive Sediments

The Borah formulation described previously was used to develop a relationship between depth of
scour and shear stress for the various size fractions in each core sample. The data points were
plotted on a  log-log plot.  One linear relationship was found for shear  stresses  below  about 5
dynes/cm2, and another for shear stresses above 5 dynes/cm2 (Figure 4-2).

Data for determining  particle size distributions are not available throughout Thompson Island
Pool, but shear stresses are available on a fine scale. A predictive relationship between armoring
depth and shear stress was sought.  Assuming that the core particle size distributions are typical
of particle size distributions  throughout Thompson Island Pool, the relationships between
armoring depth and shear stress discussed above can be considered predictive, even where the
particle size distribution is unknown.  Therefore, a linear regression was performed to fit the 355
data points above 5 dynes/cm2 (shear  stresses lower than 5 would not be, of course, as significant
in producing erosion) to Equation 4-10.

                     \n(Depth, cm) — A + m x \n(ShearStress, dynes I cm2)              (4-10)
                                            36                        Limno-Tech, Inc.

-------
 A constant (A) value of -1.6335 and a slope (m) value of 1.2407 were found.  The R-squared
 value was 0.5, and the p-values were less than 0.00001. The spread around the regression line is
 considerable, encompassing approximately two orders of magnitude.  This is not unexpected,
 since a similarly large spread was observed for the cohesive sediment correlation.  The graph of
 armoring depth vs. shear stress, with the regression line shown, is provided in Figure 4-2.

 4.4 DOSM APPLICATION

 4.4.1 Application Framework

 An ARC/INFO-based Geographical Information  System (GIS) (ESRI,  1997) was utilized to
 associate sediment and  hydrodynamic  properties  with geographic  locations  and  areas in
 Thompson Island Pool. Computations made use of shear stresses estimated at the nodal locations
 where flow field information  was available from  the Thompson Island Pool Hydrodynamic
 Model (Chapter 3). The sediments were spatially differentiated into cohesive and non-cohesive
 areas, as  described in Section 4.3.1, with separate analyses conducted for each sediment type.

 It is important to note that the DOSM, as a stand-alone model, has not been designed to simulate
 the subsequent transport and redeposition of eroded sediments.  It evaluates only the mass of
 bottom sediments potentially mobilized at a specified peak flow.  The  HUDTOX mass balance
 model includes a dynamic representation of solids and PCB  transport  and fate in the  water
 column and bedded sediments.

 The DOSM  was used to develop relationships  between river flow  and  cohesive  sediment
 resuspension  in Thompson Island Pool that were  subsequently used  in the HUDTOX  model to
 compute  flow-dependent cohesive sediment resuspension.  Details of the development of these
 equations are presented in Section 5.2.3.2.  The relationship between the DOSM and HUDTOX
 ensures  internal consistency in  representation of flow-dependent  resuspension for  cohesive
 sediments between these two  models. Use of the  non-cohesive  sediment  scour equations to
 determine non-cohesive resuspension rates  in HUDTOX was not possible due to limitations of
 the theoretical formulations.  As discussed in Section 4.2.4, the non-cohesive armoring equation
 only represents the maximum  potential  scour and  the actual armoring  depth depends on  the
 dynamic  characteristics of the flood hydrograph.

 4.4.2  Probabilistic Model Application to High Resolution Coring Sites

 As discussed  above, a Monte Carlo approach was used to assess probability of sediment scour
 depths based  on the variability  in site-specific measurements of cohesive  sediment resuspension
 properties.  This was done specifically for cohesive  sediment locations  where USEPA collected
 high-resolution sediment core PCB  profiles, and  the  range  of probable scour depths was
compared to  these profiles to  assess  the likelihood that higher PCB concentrations would be
 uncovered in  response to scour under a 100-year flow event.  Probabilistic calculations were not
conducted for non-cohesive  sediments because the method already provides an upper-bound
calculation.
                                          37                       Limno-Tech. Inc.

-------
 As part of the Phase 2 monitoring program, sediment cores were taken at five locations in areas
 containing cohesive  sediments  in  Thompson  Island  Pool,  and analyzed at a  high vertical
 resolution.  These sediment cores  exhibited fairly high  long-term sediment  burial rates  and
 showed  peak PCB concentrations  in  excess  of 2,000 ug/g (dry weight).    Core collection
 locations were specifically established in highly depositional areas of the River.  Although these
 five locations are not necessarily representative of PCB profiles in cohesive sediments in the
 entire  Pool, they  were used because each site  contained detailed measurements of sediment
 physical-chemical properties that were required for a finely resolved analysis of resuspension
 potential. Location-specific inputs consisted of predicted shear stress at each coring location  and
 sediment bulk density measured for each core.   Table 4-1 lists location-specific input data for
 each of the five cores.  For depths greater than 2 cm, core average values of dry bulk density were
 used for calculating depths of scour.

 Table 4-2 contains summary results  for each of the five sediment core locations.  The predicted
 median depths of scour for the five  locations, shown in the second column of Table 4-2, range
 from less than 0.08 (HR-19) to almost 4 cm (HR-25). The third and fourth columns in Table 4-2
 show the range of predicted scour depths encompassing the middle 90 percent of expected values
 (i.e. 5th to 95th percentile) for each core location. By comparing the depth of scour estimates in
 Table 4-2 with the input data in Table 4-1, one can see that bottom shear stress is a very strong
 determinant of erodibility in these cohesive sediments.

 The median predicted depth of  scour provides information  on quantities of solids that  can
 potentially resuspend during an  event;  however, this  information alone  does  not  define  the
 quantity  of PCBs  that can potentially resuspend.  The last column in Table 4-2 contains  the
 observed depth of the total PCB  peak at each of the five core locations.  By comparing median
 predicted depths  of scour and observed depths of PCB  peaks,  a more complete picture of
 potential PCB erodibility emerges.  These results are depicted graphically in Figure  4-3, which
 show the total PCB (as originally measured) profiles with depth for each of the five  sediment
 cores, along with the 5th, 50th and 95th percentile predicted depth of scour for each of the five core
 locations. Results indicate that Core HR-25 is likely to experience scour of sufficient magnitude
 to substantially erode the PCB peak  at that location. However, even if erosion occurs at the 95th
 percentile depth, PCB peaks at the  other four locations are predicted to be unsecured (i.e.  the
 PCB peaks are likely to stay intact after a 100-year peak flow event).

4.4.3 Poolwide Model Application

4.4.3.1 Cohesive Sediments

Equations 4-3 and 4-6 can conveniently be used to  estimate the total mass of solids remobilized
 from cohesive sediments  throughout Thompson Island Pool, and the mean depth of scour in
cohesive sediments, by means of a Monte Carlo  Analysis.  The cohesive  sediment areas of
Thompson Island  Pool were subdivided into polygons of constant shear stress and  dry bulk
density by  intersecting coverages for these properties in the  GIS system  discussed in Section
4.4.1.  The Monte Carlo technique  was  employed to calculate the depth of scour and the mass
 scour by  randomly varying parameters in the resuspension equation according to variability in the
 site-specific resuspension  measurements.   Poolwide results  for mass scour were obtained by
                                           38                        Limno-Tcch, Inc.

-------
summing the results at all locations, while an area-weighted average was calculated as the mean
depth of scour. The calculation was repeated many times to get a valid statistical distribution of
results.

Monte  Carlo  calculations  were  performed  with  the  Crystal  Ball®  computer  program
(Decisioneering, Inc., 1996).  Depth and mass of scour were computed together, with 3,000
repetitions conducted; a sensitivity analysis of the number of repetitions demonstrated that 3,000
repetitions were adequate to produce consistent results.  The results were plotted as cumulative
percent  vs. mean depth of scour or mass of scour, respectively.  Expected values for mean depth
and mass of scour were estimated by the mean of the Monte Carlo trials and are shown in Table
4-3.

Figure 4-4 shows the results  for mean depth of scour.  Most of the predictions fall into the range
of about 0.3 to 0.4 cm.  There is, therefore, a high probability that a future 100-year peak flow
would result in a mean depth of scour of between 0.3 and 0.4 cm.  Figure 4-5 shows the results
for total solids scoured.  Most of these predictions fall into  the range of about 1,500,000 to
2,000,000 kg.  There is,  therefore, a high probability that a future 100-year peak flow would
result in a mass scour of between 1,500,000 and 2,000,000 kg.

The PCB concentration in Thompson Island Pool surficial  sediments was estimated to be 32.5
mg/kg (USEPA, 1998a).  Using this concentration value  in Equation 4-6 with the above estimate
of 1,500,000 to 2,000,000 kg  of solids erosion provides an approximate range of gross PCB
erosion  of 49 to 65 kilograms.   This is  a conservative estimate due to the use of the 1984 PCB
data, as  discussed in Section 4.3.1.4. The range of solids scoured represents uncertainty due to
variability in sediment properties.  This range could be applied to a more recent estimate of
surface  sediment concentrations to get  a  more refined estimate of  the range of expected PCB
mass scoured.

4.4.3.2  Non-Cohesive Sediments

Equation 4-9 was applied using estimated shear stresses in non-cohesive sediment areas. For the
100-year peak flow, the mean,  non-area-weighted Thompson Island Pool non-cohesive sediment
armoring depth is 13.1 cm.  Therefore, 13.1 cm is an estimate of  the expected average upper
bound erosion from non-cohesive sediment areas in Thompson Island Pool resulting from a 100-
year peak flow. Upper bound  estimates of erosion at specific non-cohesive sediment locations
throughout Thompson Island Pool ranged from 1.5 to 42 cm.  This estimate of erosion in non-
cohesive sediment areas  is fundamentally different from,  and not  directly comparable to, the
above estimates of erosion in cohesive sediment areas.  Those cohesive estimates are predictive
of the actual  erosion expected  to occur under the specified conditions, including an uncertainty
band for the prediction. It is reasonably certain that the actual erosion would be less than the non-
cohesive sediment erosion estimate, perhaps much less. Given  the difference in the nature of the
estimates, it is not surprising that the 13.1  cm upper bound on the average erosion from non-
cohesive sediment areas of Thompson Island Pool substantially exceeds the 0.317  cm expected
value  of the mean depth of scour from cohesive sediment areas of Thompson Island Pool. If this
upper bound  scour depth were achieved in  non-cohesive sediments, some areas might result in
increased surface sediment PCB concentrations based on observations of the PCB distribution in
                                           39                       Limno-Tech, Inc.

-------
the various sediment PCB datasets.  For example, the 1977 NYSDEC sediment core data and
1991  GE composite  core data show that  higher PCB concentrations exist below  the surface
sediment layer in non-cohesive sediments.  This can be observed from inspection of Figures 6-34
and 7-21 through 7-23, which present data used in the development of sediment initial conditions
and model calibration datasets.

4.5 DOSM FINDINGS

The Depth of Scour Model (DOSM) was  developed for Thompson Island Pool specifically to
address the likelihood of a 100-year flood event uncovering buried high concentrations of PCBs
due to erosion of surface sediments.  Two separate applications of  the DOSM model were
conducted to address this question.  These applications found that:

   5. A probabilistic calculation of 100-year peak flow scour depths at the five USEPA
      high resolution sediment coring locations in Thompson Island Pool; and,

6.  A conservatively high estimate of total PCB mass eroded during a 100-year peak flow
   from cohesive sediments in the Pool.

These results lead to the following major findings:

      •     Predicted scour depths under 100-year peak flow conditions are small  and
             will  not result  in  significant remobilization of buried  sediments in
             Thompson Island Pool cohesive sediments;

      •     Non-cohesive scour depths  could only be computed as an upper bound
             because  the  time to armoring  is very  uncertain  and could not  be
             determined in the DOSM framework.  If this upper bound were  achieved,
             scour in non-cohesive  sediment  areas may result in  increased surface
             sediment PCB concentrations in some areas;

      •     Even at the 95th percentile of scour depth, the 100-year peak flow does not
             cause  scour  to  elevated PCB  concentrations  at the  high-resolution
             sediment core locations; and,

      •     Based on a conservative estimate of the mass of PCBs resuspended under
             a 100-year peak flow, the  100-year flow will result in only a slightly larger
             amount of PCBs resuspended than may be expected during typical annual
             high flow events.
                                          40                      Limno-Tech. Inc.

-------
Chapter 5

-------
5. FATE AND TRANSPORT MASS BALANCE MODEL DEVELOPMENT

5.1 INTRODUCTION

Chapter 5 describes the development of the Hudson River Toxic Chemical Model (HUDTOX),
the principal transport  and fate modeling tool in this  Reassessment. This chapter presents the
conceptual framework and  the governing equations  for model  state variables and process
mechanisms as well  as details on  the computer hardware and software operating environment.
The following major  sections are included in Chapter 5:

       5.2    General  Model Approach

       5.3    Water Transport

       5.4    Solids Dynamics

       5.5    PCB Dynamics

       5.6    Model Spatial Segmentation

       5.7    Model Implementation

5.2 MODEL APPROACH

5.2.1 Introduction

HUDTOX is the principal transport and fate modeling tool in this Reassessment. HUDTOX is a
time-variable, three-dimensional mass  balance model.  It is a fully-integrated representation of
solids and PCB concentrations  in the water column and bedded sediments.  HUDTOX was
applied to the entire Upper Hudson River from Fort Edward to Federal Dam at Troy. Because a
disproportionate amount of PCB-contaminated sediments is contained in Thompson Island Pool
(TIP), and because there is substantially more data available for the Pool HUDTOX included
greater spatial  resolution  for the Thompson Island  Pool than for the river downstream of
Thompson Island Dam (TID).  In  the Pool, HUDTOX is two-dimensional in the water column
and three-dimensional in the sediments. Between Thompson Island Dam and Federal Dam, it is
one-dimensional in the  water column and three-dimensional in the sediments.

The principal model  application was a  long-term historical calibration for a 21-year period from
1977 to 1997 for Tri+ PCBs. Short-term hindcast applications were also conducted from 1991 to
1997 in order to test the long-term historical calibration for several PCB forms (5 congeners, and
total PCBs) exhibiting  a range of physical-chemical properties (e.g., sorption to solids, Henry's
Law constants, molecular weights, etc.). The  calibrated model was  also  used to conduct  a
validation simulation with an independent dataset acquired in 1998 for the Upper Hudson River.
Calibration parameters  were not changed in this validation exercise. The calibrated model was
then used to conduct forecast simulations for 70-year periods beginning in 1998.  These forecast
simulations were intended to estimate  long-term system responses to continued No Action and
impacts due to a 100-year peak flow.


                                         41                      Limno-Tech, Inc.

-------
5.2.2 Conceptual Framework

Three different mass balances are represented in HUDTOX: (1) a water balance; (2) a solids
balance; and (3) PCB mass balances. A water balance is necessary because PCB dynamics are
influenced by  river flow rates and mixing rates. A solids balance  is necessary because PCB
dynamics  are  influenced by the tendency of PCBs to sorb, or attach,  to both suspended and
bedded solids  in the river.  Finally, a PCB mass balance itself is necessary to account for all
sources, losses and internal transformations of PCBs in the river.

HUDTOX represents PCBs  in both the water column and bedded  sediments.  PCBs in each
medium are comprised of three phases:

       •       Truly dissolved;

       •       Bound to dissolved organic carbon (DOC); and,

       •       Sorbed to total solids.

Organic carbon is the  principal  sorbent compartment for  hydrophobic organic chemicals  in
aquatic systems. A time-dependent mass balance was developed for the suspended and bedded
solids, and organic carbon fractions were assigned to these solids based on  data.   Dissolved
organic carbon (DOC) was not simulated in the mass balance.  Instead, concentrations were held
constant in the sediment bed and the water column.  These concentrations were developed from
site-specific data and specified as model inputs.

HUDTOX computes time-dependent mass balances  for two  state variables: solids and  PCBs
(total PCBs, Tri+, and congeners BZ#4, BZ#28, BZ#52, BZ#[101+90], or BZ#138, depending on
the particular application). It assumes that within each model spatial segment a local equilibrium
exists among the three  different PCB phases. It computes the PCB distribution among these
phases by applying an  organic  carbon-based  partition coefficient  to  the organic  carbon
concentration of each sorbent (dissolved and particulate organic carbon). This  local equilibrium
assumption allows the mass balance model to compute only a single PCB state variable while
still representing the specific process kinetics operating on each PCB phase.  For example, only
the solids-sorbed PCBs will settle; therefore, the settling velocity determined through the solids
mass balance is applied to only the solids-bound phase of PCBs within each spatial segment. On
the other hand, only truly dissolved PCBs can exchange across the air-water interface; hence, that
process is applied to only dissolved phase PCBs in water column segments at the air-water
interface.

Figure 5-1 contains  a conceptual diagram  for  HUDTOX  that illustrates PCBs in the water
column and  surface sediment spatial segments.   This diagram displays the three phases into
which PCBs can be partitioned, as well as the model processes which are applied to either the
whole PCB form or to an individual PCB phase.  Thus, each arrow into or out of a given control
volume (or spatial segment) represents a distinct  source or sink flux process that operates on the
PCB state variable and forms its full mass balance equation for that segment. The simultaneous
                                          42                       Limno-Tech, Inc..

-------
solution of those mass balance equations permits quantification of the relationship between
external inputs and within-system concentrations of PCBs over space and time.

5.2.3 Governing Equations

This section presents a summary of the state  variables and processes  in the  HUDTOX mass
balance model.  The  HUDTOX model is a  modified version  of the USEPA WASP toxic
chemical model WASP5/TOXI5.  The equations and framework are essentially the same except
for two major enhancements; one relates to handling of sediment bed segments under erosion and
scour, and the second relates to sediment scour formulations.

The  HUDTOX model code was originally developed using  an  earlier version of the WASP
model  (WASP4/TOXI4) which was later updated by EPA to reflect coding changes and various
enhancements.  The primary source for documentation of the updated WASP5/TOXI5 model is
Ambrose et al. (1993).  This document can be obtained via the Internet by downloading it from
the  USEPA  Center for  Exposure  Assessment  Modeling   (CEAM)  web  site located  at
"http://www.epa.gov/epa_ceam/wwwhtml/ceamhome.htm."   The  HUDTOX  model description
presented  in  this section  is  a summarized  version  of  the  WASP5/TOXI5 documentation
contained in Ambrose et al. (1993). Details are presented for those processes in HUDTOX that
were modified from the WASP5/TOXI5 model. Unless specifically noted, the HUDTOX model
processes are identical to those in the WASP5/TOXI5 model.

The  mass balance for the HUDTOX model accounts for all user-specified material entering and
leaving the  system by external  loading,  advective  and dispersive  transport,  settling and
resuspension,   and  physical,  chemical, and  biological  transformations.    The  generalized
HUDTOX mass balance (partial differential) equation for an infmitesimally small  fluid volume
in three-dimensions is:
where:
                                                                               (5-1)
             C =          concentration of the water quality constituent state variable,
                          mg/L (g/m3) [M/L3]
             t =           time, days [T]
             Ux, Uy, Uz =  longitudinal, lateral, and vertical advective velocities, m/day [L/T]
             Ex, Ey, E2 =   longitudinal,   lateral,   and   vertical   diffusion   (dispersion)
                          coefficients, m'/day [L2/T]
                                          43                       Limno-Tech, Inc.

-------
              SL =  direct and diffuse loading rate, g/m3-day [M/L3/T]
              SB =  boundary loading rate (including upstream, downstream, sediment,
                    and atmospheric), g/m3/day [M/L3/T]
              SK =  total kinetic transformation rate; positive indicates a source,
                    negative indicates a sink, g/m3/day [M/L3/T].
 By expanding the infinitesimally small control volumes into larger adjoining "segments" and
 specifying transport, loading,  and transformation parameters,  HUDTOX implements a finite-
 difference form of Equation 5-1 to solve for the concentration of each water quality state variable
 over time. A one-dimensional simplification of Equation 5-1  may be expressed by assuming
 vertical (z-domain) and lateral (y-domain) homogeneity:
                      Term 1	    Term 2     Term 3

    |-(AC)=   |-f-UxAC + ExA—1+A(SL + SB) + ASK                        (5-2)
    dt          dxl              dx I
where:
              A =   cross-sectional area, m2 [L2]
This equation represents the three major classes of water quality processes:

       •     Transport (term 1);

       •     External loading (term 2); and,

       •     Transformation (term 3).

These processes, which describe the fate of each HUDTOX solids and PCB model state variable,
are discussed in the following paragraphs. The finite-difference derivation of the general WASP
mass balance equations and the specific solution technique implemented to solve these equations
are described in Ambrose et al. (1993).

5.3 'WATER TRANSPORT

The physical transport of water column solids and PCBs in HUDTOX is governed principally by
advective  flow  and  dispersive  mixing in the water column.    Each  are described below.
Advective water column flows are important because they control the downstream transport of
dissolved and paniculate pollutants in many water bodies.  In addition, changes in velocity and
depth resulting from variable flows can affect such kinetic processes as reaeration, volatilization,
and photolysis.   HUDTOX tracks each separate inflow specified by the  user from its point of
origin and through each segment until it exits the model network. For each inflow, the user must
supply a continuity or unit flow response function (i.e., flow routing) and a time function. For the
HUDTOX  model,  the flow routing information is based upon the RMA-2V results for  the
Thompson  Island Pool two-dimensional water column segmentation grid. The advective  flows
are simply  routed  directly through  the one-dimensional  HUDTOX  segmentation existing
downstream of the  Thompson Island Pool.  Representation  of short-term transient effects due to


                                          44                       Limno-Tech, Inc.

-------
storage and drainage were deemed unimportant considering the goal of the HUDTOX model,
which was to describe long-term PCB concentration trends.

The flow continuity function describes how various flow inputs are routed throughout the model
network.  The time function describes the temporal  variability of the inflow.  The actual flow
between segments that results  from a given inflow is the product of the time function and the
continuity function. If several inflow functions are specified between any segment pair, then the
total flow between segments is computed as the sum of the individual flow functions. In this
manner, the effect of several tributaries joining, density currents, and wind-induced flow patterns
can be described in a simple manner.

Hydraulic  relating describing  depth and velocity to stream  flow  are based on formulations
developed by Leopold and Maddox (1953) which describe empirical observations of the velocity
and  depth  to stream  flow relationship.  These relationships,  which are used  for determining
chemical air-water mass transfer rates (gas phase absorption and volatilization), are described in
Ambrose et al. (1993). For the Thompson Island Pool portion of the Upper Hudson  River, the
HUDTOX  model  coefficients  describing this relationship  were  developed from the RMA-2V
hydrodynamic model  described in Chapter 3.  The relationship for downstream reaches was
developed using correlations between surface water elevations and flow (USEPA, 1997).  Note
that these relationships are only used to affect chemical gain or loss within a water column model
segment (through  volatilization);  they do  not  affect water volume or advective or  dispersive
transport of chemicals between model segments.

Dispersive  water  column  exchanges significantly  influence  the transport of  dissolved and
particulate pollutants by mixing between water of different concentrations.  In rivers, longitudinal
dispersion  can be an  important process in diluting  peak concentrations  that may result from
dynamic (unsteady)  loads or spills.   Natural or  artificial  tracers  such as  dyes, salinity,
conductivity or heat (temperature) are often used to calibrate dispersion coefficients for a model
network.

The dispersive exchange between HUDTOX segments i and j at time t is given by:


            =   E^t)'A"(Cj-Q)                                                 (5-3)
        dt        Lcij

where:
             Mj =  mass of constituent (state variable) in  segment i, g [M]
             C =   total constituent (state variable) concentration, mg/L (g/m3) [M/L^]
             Eij(t) =dispersion coefficient time function for exchange "ij", m2/day [L2/T]
             Ay =  interfacial area shared by segments i and j, m2 [L2]
             Lcy = characteristic mixing length between segments i and j,  m [L] .
                                           45                        Limno-Tech. Inc.

-------
The exchange coefficient may also be expressed as a mass transfer velocity by dividing the
dispersion coefficient by the characteristic mixing length:

                E-(t)
        vij(0=  -^                                                             (5-4)
                 Lcij

where:
              vjj(t) = mass transfer rate for exchange "ij", m/day [L/T] .

5.4 SOLIDS DYNAMICS

HUDTOX calculates sediment and PCB concentrations for every segment in a model grid that
includes surface water,  surficial sediment bed, and underlying sediment  bed layers.   During
simulation, solids are treated as a conservative constituent that is advected and dispersed through
water column segments, settles to and resuspends from surficial sediment segments, and moves
through the subsurface bed through burial/scour of the surficial bed or through particle mixing.
Due to large  uncertainties in externally  contributed solids loads  (see Chapter 6), internal
production and decay of water column biotic solids through primary production was not included
in the HUDTOX model calibration application to the Upper Hudson River.  The contributions of
primary production and decay of solids is dwarfed by the high upstream and tributary solids
loads.

5.4.1  Solids Gross Settling

HUDTOX differs from WASP5/TOXI5 with respect to gross settling of suspended solids from
the  water column to the sediment bed in  order to  capture effective differences  in settling
characteristics between cohesive and non-cohesive sediment areas. Constant settling velocities
(not dependent on flow or other factors) are specified in HUDTOX with different rates specified
for cohesive and non-cohesive  sediment areas.  Settling velocities are constant for each sediment
type throughout the river under all  flow conditions.   This approach was  developed to  be
consistent with the model calibration strategy presented in Chapter 7.  The differences between
cohesive and non-cohesive settling velocities arise as a result of lower resuspension  and higher
deposition for cohesive  sediment  areas  relative  to non-cohesive  sediment areas.    Lower
resuspension occurs in cohesive sediment areas due  to lower flow velocities and shear stress in
these  areas.  Due  to the  lower flow  velocities in cohesive sediment areas, higher deposition
occurs relative to non-cohesive areas.

The rationale  and approach for specification  of the cohesive and non-cohesive  settling rates
specified for HUDTOX in this  RBMR calibration is presented in Chapter 7 of this report.

5.4.2  Cohesive Sediment Flow-Driven Resuspension

The algorithm for flow-driven  resuspension of cohesive sediments used in the  DOSM (Equation
4-5) was  incorporated into the  HUDTOX model.  Total sediment  erosion  (e , mg/cm2)  is
incrementally  applied to  the rising side of the flood hydrograph. Non-linear  correlations were
developed relating the DOSM-predicted sediment erosion in each segment as a function of flow
                                           46                        Limno-Tcch, Inc.

-------
measured at Fort Edward. Equations of the following form were fit to DOSM results correlating
the mass of cohesive sediment erosion to flow:

                    e = ai + a2xQ?3                                               (5-5)

where:
              e =   cohesive sediment erosion, mg/cm2
              Qx  =  advective flow in 1000's of cfs
              tti =  empirical constant fit to DOSM results, mg/cm2
              v.2 =  empirical constant fit to DOSM results, mg/cm2/1000 cfs
              Ota =  empirical constant fit to DOSM results, dimensionless.
The cohesive sediment erosion is converted to an effective resuspension rate (vrH, m/day) in the
HUDTOX model  over each model time step during the  rising  side of a flood hydrograph.
Computational time steps in the model vary between approximately 5 and 25  minutes.  This
approach is consistent with observations by Lick et al. (1995) that most resuspendable material is
mobilized in approximately one hour.

In the HUDTOX  model,  resuspension occurring  over previous  model time steps within  an
increasing hydrograph is tracked such that total cumulative erosion equals the amount computed
using the maximum shear stress during that  event.  This mass flux  tracking occurs in  an
incremental fashion.  The amount eroded during any given model  time step being dependent on
the change in flow and the cohesive sediment solids concentration (dry bulk density).  The total
amount of sediment erosion is limited by the maximum predicted erosion (Łmax) associated with
the peak flow. Flow-driven resuspension  effectively stops depleting the existing sediment bed
once the peak flow is reached and that cohesive sediment armoring  is assumed to have occurred.

HUDTOX also includes a recovery period (trec, days) in which the maximum erosion constraint
prevents subsequent near-term smaller floods from eroding bedded cohesive sediments that have
reached an armored condition. Any solids depositing on the sediment bed during the recovery
period are allowed to erode based on Equation 5-5, but only to the extent that freshly deposited
material is available. Once the recovery  period is  ended,  no  differentiation is  made between
freshly deposited solids and older "bedded" solids in the surficial  sediment layer. At this point,
all the surficial sediments are again subject to scour based on Equation 5-5 (i.e., armoring of the
cohesive sediment bed ceases).  Model applications employing the Lick resuspension formulation
for cohesive sediments have generally used a recovery or "sediment  aging" timeframe of 7 days
for the surficial sediment layer  (Gailani et al., 1991; Ziegler et al.,  1994). Lick et al. (1995) cites
a range of 1  to 28  days for the consolidation process (i.e., the consolidation or "aging" of fresh
sediments to a condition consistent with sediments lying below this recently deposited material)
to take place based on experiments using sediments from various freshwater river systems.

5.4.3  Non-Cohesive Sediment Resuspension

In contrast to cohesive sediments, the  DOSM model  only  provides  a single,  upper-bound
estimate of scour for non-cohesive sediments.  The single estimate is characterized as an upper-
bound because it occurs when armoring is achieved and is a function of deposition and time over
                                           47                       Limno-Tech, Inc.

-------
which peak shear stresses are experienced (Chapter 4). Consequently,  the non-cohesive scour
formulation from DOSM was not incorporated into the HUDTOX model.

The representation of the flow-driven resuspension of non-cohesive sediments in HUDTOX was
developed for the calibration strategy presented in Chapter 7. This strategy attempts to describe
mean high and low-flow solids dynamics.  It is represented in HUDTOX model by specification
of  a constant high-flow resuspension velocity operative  during scouring conditions.   Non-
cohesive sediment scour is considered insignificant below specified flow thresholds, which are
spatially variable in the model.  Thus, non-cohesive resuspension rates switch between zero and
the specified high flow resuspension rate.

No attempt  has been made in HUDTOX to simulate armoring conditions  in the non-cohesive
sediments. The calibration approach relies on accurately capturing the long-term behavior of the
system based on solids burial rates, surface sediment PCB concentration trends, and in-river mass
transport of solids and PCBs, rather than description of event dynamics which vary over small
time scales.   This is consistent with the overall objective of this modeling study presented in
Chapter 1. Chapter 7 presents the calibration approach and a discussion of the model parameters
specified to  simulate non-cohesive  resuspension during high  flow  scouring conditions. These
parameters include: the transition flow between scouring and non-scouring conditions for  each
reach of the river, and the calibrated non-cohesive resuspension rate for scouring flow conditions.

5.4.4 Sediment Bed Particle Mixing

Bioturbation and other physical processes can result in vertical mixing of solids (and  sorbed
chemicals) within the bedded sediment.  Particle mixing rates tend to be site-specific and can
vary seasonally due to  temperature  influences on biological activity (e.g. McCall and Tevesz,
1982). Sediment bed particle mixing is an important model consideration.

Sediment mixing processes  are represented in HUDTOX  by  effective particle  diffusion
coefficients. The resulting particle diffusion  transfer between sediment layers induces a  flux of
sorbed contaminants between sediment layers in the model.  The direction of flux is determined
by the concentration gradient between layers. The form of the particle mixing equation is  similar
to that represented by Equation 5-3, but with the concentration gradient expressed in terms of the
solids concentrations (and sorbed  chemical concentrations) in the sediment layers across which
the  flux takes place. Particle mixing rates are not subjected to  temperature  influences in the
model.

The parameterization of particle mixing in HUDTOX requires specification of the depth over
which particle mixing occurs and the effective particle diffusion rates between sediment layers.
Specification particle mixing depths and particle diffusion rates is  presented in  Chapter 6 and
further developed in Chapter 7.

5.4.5 Scour and Burial

The HUDTOX  model  uses an  improved sediment  bed handling  approach  from  that in
WASP5/TOXI5.  The  HUDTOX approach  maintains and  allows  the  formation of  a  distinct
                                           48                       Limno-Tech, Inc.

-------
vertical chemical profile through the bedded sediments.  This modified sediment bed handling
routine is a better representation of transport of PCB mass through the sediment bed because it
maintains the integrity of the deeply buried sediment layers as burial or scour occurs.   The
standard  WASP5/TOXI5   model  can exhibit significant   numerical  dispersion over  long
simulation periods, leading to a "smearing" of vertical contaminant profiles.

To insure the  maintenance  and formation  of  a  distinct  vertical profile,  the following
modifications were made to WASP5/TOXI5: 1) use of a quasi-Lagrangian sediment bed handling
routine; and, 2) use of an archival stack of deep  sediment layers as a dynamic boundary condition
to track PCB mass beneath  the computational grid.  The  following paragraphs describe the
implementation  of this  alternative  bed  handling  through  a  set  of modifications  to the
WASP5/TOXI5 scour and burial processes.

The revised sediment bed handling routine maintains the integrity of the deeply buried sediment
profile as sedimentation and erosion  occur. With the revised framework, the surficial sediment
layer volume varies over time due to deposition  and resuspension. Thus variation continues until
either erosion or burial is triggered based on the volume (or equivalently, the thickness) reaching
a specified minimum  or maximum level. For burial,  the trigger is based on a doubling of the
surficial sediment thickness. Erosion is triggered by depletion (or near depletion) of the original
surficial sediment volume. Essentially, the HUDTOX model bed handling implements a quasi-
Lagrangian  (or  floating  frame  of reference) approach  to  burial  and  scour  versus the
WASP/TOXI5-based quasi-Eulerian (fixed frame of reference) approach.

Figures 5-2 and 5-3 illustrate the manner in which HUDTOX implements respectively scour and
burial of surface sediment segments.  In the HUDTOX bed handling framework, burial results in
no numerical mixing of chemicals to deeper sediments, because the surface sediment segment is
simply split into two and renumbering of the segments is triggered whenever its volume doubles.
Erosion of the surface sediments still provides a degree of mixing between the surface sediments
and the immediate segment below.  The degree of this mixing is dependent on the amount of
sediment remaining in a surface segment once it has been effectively depleted.  However, no
additional mixing occurs through the deeper sediment segments as a result of the HUDTOX bed
handling procedure. These deeper segments are subject to renumbering when erosion occurs, but
they still  maintain their original  pre-erosion  characteristics.  As described in the previous
discussion  of particle mixing,  HUDTOX allows  user-specified vertical mixing (via  particle
mixing and/or diffusion) through the active sediment bed to represent the effects of bioturbation
and other processes (e.g., ice scour, propagation of bed load waves,  prop wash, etc.) which serve
to maintain partially mixed conditions through some depth of the sediment bed.   Thus, the
degree of vertical "smearing" of sediment concentration profiles is user-controlled, rather than
being dependent  upon the sedimentation time  utilized in the standard WASP5/TOXI5 model
framework.

In order to provide long-term tracking of sediment layer PCB concentration, and allow possible
future exposure of deeply buried PCBs, a second modification of the WASP5/TOXI5 framework
maintains an "archival" stack  of  deep sediment  layers  beneath  the existing simulated bed
segments. A  user-defined reserve  stack of deep sediment layers can be specified to underlie the
existing simulated bed segments with distinct  stacks for each surface sediment segment.  In
                                           49                       Limno-Tech. Inc.

-------
essence, the archive stacks provide a dynamic boundary condition for the bottom sediments. The
stacks are not part of the computational grid, except to the extent that layers are moved between
the stack and the model  grid  to compensate for burial or erosion  of the  surface sediment
segments. The process of constituent decay is not represented in the archive stack.

When erosion results in a  surface sediment  segment being depleted, then "renumbering" of the
segments is triggered as previously described.  Additionally, the top layer of the archival stack is
then incorporated within the computational grid  as  a new bottom sediment segment.  During
periods of deposition, the surface layer is allowed to grow in thickness (the bed solids density is
kept constant) until renumbering is  triggered, based on a doubling  of the  surface sediment
volume.  The surface  segment is then split  into two layers  and the sediment segments  are
renumbered accordingly.   Additionally,  the bottom sediment layer is removed  from  the
computational grid and placed on the top of the  archive  sediment  stack.  The archive stack is
allowed to grow or shrink as needed in response to burial or erosion of the  surface sediment
segments.  A significant advantage of using the  sediment archive  stack relates  to  its minimal
effect on the computational requirements  and execution  speed of the model.  This allows  for
improved  vertical resolution of the sediment bed without excessively increasing memory and
runtime requirements.

The HUDTOX approach for sediment scour and burial requires that the upper portion of  the
sediment bed be composed of vertical layers of equal thickness at the beginning of the model
computations.  This insures that long periods of scour and deposition will not  cause changes to
the basic  physical characteristics (e.g. original  volume  and  thickness) of the surface  layer
sediments when deposition or scour triggers the bed handling mechanism.

5.5 PCB DYNAMICS

In the environment, organic chemicals  may transfer across  the different environmental media
(air, water, and sediment)  and may be degraded  and/or transformed by a number of physical-
chemical and biological processes.  Cross-media  PCB transfer processes  within the HUDTOX
model  framework include equilibrium sorption and volatilization (air-water exchange).  PCBs
may also be transformed within HUDTOX through degradation as expressed by a first-order rate
equation to represent the effect of dechlorination and/or destruction as a net mass loss over time.
PCB dechlorination or degradation processes are not represented in the HUDTOX model for this
application to the Upper Hudson River.  Other chemical  transformation processes  (hydrolysis,
photolysis, and chemical oxidation) are included within the overall WASP5/TOXI5 framework.
Detailed descriptions of these processes are contained in Ambrose et al. (1993).

5.5.1  Equilibrium Sorption

Sediment particle dynamics are important in controlling the transport, transformation and fate of
PCBs in aquatic systems due to the tendency of PCBs to sorb, or bind, to both suspended and
bedded solids (Eadie and Robbins, 1987). Karickhoff et al. (1979) and Karickhoff  (1984) have
shown  that organic  carbon is  the  principal sorbent compartment for hydrophobic  organic
chemicals,  such as PCBs, in aquatic systems. In addition to organic carbon in paniculate  form,
dissolved organic carbon (DOC) can also be an important soiption  compartment in  determining
                                           50                       Limno-Tech, Inc.

-------
PCB fate (Eadie et al., 1990; Bierman et al., 1992). Partition coefficients are used to characterize
the distribution of chemical among three  apparent phases: dissolved, particulate-bound, and
DOC-bound.

The assumption of equilibrium partitioning in a natural system is reasonable when PCB sorption
kinetics are rapid  relative to other processes affecting water and sediment concentrations. There
is some evidence of non-equilibrium conditions in the mainstem of the Upper Hudson River;
however, a detailed investigation by USEPA (1997) found that the assumption of equilibrium
partitioning was  a valid  approach for  the spatial-temporal  scales in the HUDTOX model
applications.   Section 6.9 provides further information on the  available site-specific data for
estimating the partition coefficients used for the PCB forms in the HUDTOX model.

The partition coefficients depend upon characteristics of the chemical and the sediments or DOC
on which sorption occurs.  PCBs are non-polar, hydrophobic, organic compounds.  The sorption
of these compounds correlates well with the organic carbon fraction (fOc) of the sediment.  Rao
and Davidson (1980) and  Karickhoff et al. (1979) developed  empirical expressions  relating
equilibrium coefficients to laboratory measurements,  leading to reliable means of estimating
appropriate values. Dissolved organic materials are typically assumed to be composed entirely of
organic carbon (fOc = !)• The partitioning expressions implemented in the HUDTOX model are:

                    Kp = foc x KPOC                                                (5-10)
                    KB = 1.0xKDOc                                               (5-11)
where:
              Kp =  Solids partition coefficient, Lw/kgsolid [L /M].
                   = particulate organic carbon partition coefficient, Lw/kgOc [L /M]
              foc =   organic carbon fraction of sediment, kgOc/kgsolid [M/M].
              KB = DOC-bound partition coefficient, Lw/kgDOc-sorbent material (or DOM) [L3/M]
              KDOC = dissolved organic carbon partition coefficient, Lw/kgooc [L /M].
The dissolved organic carbon (DOC) partition coefficient, KDQC, is commonly estimated as KPQC
times a binding efficiency factor based on analysis of field data measurements of each chemical
phase.

HUDTOX differs from WASP5/TOXI5 in that it includes temperature-dependent partitioning, as
well as segment-specific parameters which allow for both spatial and compartmental (i.e., water
column vis-a-vis  sediment bed) variations in partitioning.  The dependence of partitioning on
temperature was developed and presented in the DEIR (USEPA, 1997). The general form of the
resulting empirical relationship, applicable to both the particulate and DOC partition coefficients,
is represented by:

                     log KP,T = log Kp,25 + tsfx  I _L--__ |                     (5-12)
                                             II  1    Zj  I
                                                   Q        Q
                                           51                        Limno-Tech, Inc.

-------
where:
              Kp 25 = partition coefficient at 25°C, L/kg
              T =   water temperature, °C
              TO =  Absolute zero temperature (0°K) = -273. 15 °C
              tsf =  temperature slope factor, °K.
The HUDTOX model  can include particle interaction effects  on solids  partition coefficients
using the approach proposed by DiToro (1985).  This approach is described in Ambrose et al.
(1993).   Analysis of site-specific data for  the Upper Hudson  River indicated that particle
interaction  effects on PCB partitioning were minimal (USEPA, 1997).  Consequently, none of
the present HUDTOX applications included particle interaction effects on PCB partitioning.

The total chemical concentration is the sum of the three phase concentrations:

                    C=  Cwn+CsMs+CBB                                       (5-13)

where:

              C'w = concentration of dissolved chemical in water, mg/L water
              n =   porosity (Volumewater / Volumewater + solids),  Lwate/L
              Cs =  concentration of solids-sorbed chemical on a mass basis, mg/kgSO|id
              Ms =  concentration of solids, kgsolids/L
              CB =  concentration of DOC-bound chemical on  a mass basis, mg/kgooc
              B =   concentration of DOC, kgDOC/L.

The dissolved fraction fd is given by:


         fd  = —  =  - - - r                                        (5-14)
                 C       1 + KBB +KpMs

The particulate (solids-sorbed) and DOC-bound fractions, respectively fp and fb, are given by:


                cX  =  - KM;                                             (5.15)
          F      C      1+KBB +KpMs
         fb =        =  -     - r                                        (5-16)
                 C       l + KBB+KpM5
where:
              M^ =  Ms/n = solids concentration on a water volume basis, kgsolid/Lw
              B =   B/n = DOC concentration on a water column basis, kgDOC/Lw

These  fractions are determined in time and space throughout a simulation from the partition
coefficients, internally calculated porosities,  simulated  solids concentrations, and externally-


                                           52                       Limno-Tech, Inc.

-------
specified DOC concentrations. Bulk volumetric concentrations for each phase (Cw for dissolved,
Cp for particulate chemical, and CB for DOC-bound chemical) are simply determined from the
product of each relative fraction and the total chemical concentration.

5.5.2 Air-Water Exchange

Air-water exchange, or volatilization, is the mass  transfer of a chemical across  the air-water
interface as dissolved chemical attempts to equilibrate with the gas phase concentration of that
chemical in the  atmosphere. In HUDTOX the air-water mass transfer exchange rate, Sv, is a
function of dissolved chemical gradient between the liquid and vapor phase by the following
equation:
ac,
^J  \vvlat.
                       D
fjC—
       H^
      RT,
                                                                                (5-17)
where:
              Kv
              R
              TK
              HT
              D
              = air-water chemical mass transfer rate, g/m3/day
              = air-water chemical transfer rate, m/day
              = universal gas constant, 8.206xlO~5 atm nrVmole °K
              = water temperature, °K
              = Henry's Law constant at temperature T (°C), atm nrVmole.
              = depth (m)   Ca    = atmospheric chemical concentration (g/m3)
 Equilibrium  occurs  when the ratio of the atmospheric partial pressure of a chemical to its
dissolved  concentration in the water column equals its temperature-corrected Henry's Law
constant.  Atmospheric partial pressure is expressed as a boundary condition in HUDTOX and
the determination of its value is described in Chapter 6.

HUDTOX employs  the  same  two-layer resistance model  (Whitman,  1923)  utilized  by
WASP5/TOXI5 to calculate the air-water exchange rate.  This model assumes that two "stagnant
films" exist at the air-water interface, bounded by well-mixed compartments on either side.  The
air-water  mass  transfer rate  is controlled by the  combined effect of liquid  and gas phase
resistance described by the following equation:
where:
       K
                               KL+
                    = Air-water chemical transfer rate, m/day
                    = liquid phase resistance, day/m
                                                                          (5-18)
                                           53
                                                             Limno-Tech, Inc.

-------
              RQ   = gas phase resistance, day/m
              KL   = liquid phase transfer coefficient, m/day
              KQ   = gas phase transfer coefficient, m/day

Diffusion of chemical  through the liquid (water) layer is driven by concentration differences,
whereas  the gas (air) layer diffusion is controlled by partial pressure differences. The Henry's
Law constant generally increases with increasing vapor pressure and decreases with  increasing
solubility of a compound.  Therefore, highly volatile compounds that have low solubility are
likely to  exhibit mass transfer limitations in water (i.e., high liquid phase resistance).  Similarly,
mass transfer in air is  limited (i.e., high gas phase resistance) when  chemical compounds are
relatively nonvolatile and have high solubility.

Air-water exchange is usually smaller in lakes and reservoirs  than in relatively turbulent rivers
and  streams. Gas  exchanges in  rivers and river-reservoir systems can also be significantly
enhanced by the highly turbulent conditions created as water  flows through and/or over dams.
The present HUDTOX  model does not account for the possible gas exchange losses of PCBs to
the atmosphere as water flows through the various run-of-the-river dams along the Upper Hudson
River between Fort Edward and Federal Dam at Troy. The significance of gas exchange at dams
on PCB  dynamics in  the  Upper Hudson River is evaluated  in the data analysis discussions
presented in Chapter 6 of this report.

Air-water exchange in HUDTOX is the same as in WASP5/TOXI5 (Ambrose, et al., 1993) with
two exceptions that are described in the following paragraphs.

The chemical-specific Henry's Law constant (H) is assumed to describe the equilibrium between
the gas phase and dissolved liquid phase at the boundary between the two layers. In HUDTOX,
the Henry's Law constants are temperature corrected  according to the empirical relationship
presented by Achman et al. (1993) in the following equation:


                       U-^414
      logHT=  logHzjJ	i	°-^(                                         (5-19)
                         .__
                             (25-T0)

where:
             HT =   Henry's Law constant at temperature T, atm m3/mole
             Hos =  Henry's Law constant at 25 °C, atm mVmole
             TO =   Absolute zero temperature = -273.15 °C
             T =    Temperature, °C.
As in WASP5/TOXI5, HUDTOX uses a constant gas film transfer coefficient of  100 m/day,
typically applied to flowing waterbodies such as the Upper  Hudson River.  HUDTOX differs
from WASP5/TOXI5 in that it directly adapts the O'Connor-Dobbins oxygen reaeration formula,
as opposed to the Covar method which selects rates from a range of formulation (including
                                          54                       Limno-Tech, Inc.

-------
 O'Connor-Dobbins) depending on predicted water depth and  current velocity within a river
 cross-section to predict a chemical-specific liquid film air-water transfer rate:
                         2
                  --   x8.64x!04                                            (5-20)
                    D J
where:
              KL   = liquid film air-water transfer rate, m/day
              D     = water depth, m
u      = water velocity, m/sec
  w
                                                     r\
              Dw   = diffusivity of chemical in water, m /sec
The O'Connor-Dobbins formula internally adjusts the air-water transfer rate to determine a
chemical-specific liquid film rate based on the chemical-specific diffusivity:

          Dw  = 22.0E-09 / (MW)273 , as per Ambrose et al. (1993).                  (5-21)

where:
             MW  = molecular weight of the chemical, g/mole
A detailed description of the two-layer resistance model used in HUDTOX and WASP5/TOXI5
is contained in Ambrose et al. (1993).

5.5.3 Dechlorination

Although  the HUDTOX model framework allows  for dechlorination  and other  degradation
processes, these loss processes were assumed to be zero in the HUDTOX model calibration
presented herein.  Rationale for this approach is presented in Chapter 6.  Dechlorination may be
accommodated  in the HUDTOX model  in a variety of ways  (e.g., first-order decay) through the
use of standard WASP5/TOXI5 degradation mechanisms. However, accurate representation of
dechlorination pathways from  degradable higher chlorinated PCB congeners to specific lesser
chlorinated PCB congeners would be an extremely difficult task to undertake in a modeling effort
of this scale. Dechlorination is also not expected to be a significant loss mechanism  for PCB
mass in the Upper Hudson River for future conditions.

5.5.4 Sediment- Water Mass Transfer of PCBs

In river systems, non-flow dependent sediment-water exchange of contaminants, including PCBs,
can result from  many different physical and biological processes, which are discussed in Chapter
6.  These  processes include molecular diffusion in porewater  as well  as  biologically- and
hydrodynamically-enhanced transfer of both porewater and paniculate phase PCBs to the water
column.  The net effect  is observed as changes in PCB loading to the  water column.   The
individual processes have not been directly measured or quantified for the Upper Hudson River,
and not all of them are well understood.  The combined effect of these processes is evident,
however, in observed concentration changes of PCBs in the water column.
                                           55                       Limno-Tech, Inc.

-------
Non-flow dependent sediment-water mass transfer processes are represented in  HUDTOX by
effective participate and/or porewater diffusion mass transfer rates.  These rates move porewater
and particulate chemical across the sediment-water interface based on concentration gradients of
these  phases  between these compartments.  HUDTOX represents  diffusive exchanges  of
dissolved and DOC-bound PCBs between sediment porewater and the overlying  water column
with a diffusion equation similar to Equation 5-3, but with the concentration gradient expressed
in terms of the dissolved and DOC-bound PCB concentrations in the porewater:
                           ffdjcj_fdici]
                              "j     "i
                                                                               (5-22)
        9t        Lij/nij

where:
             M;    = mass of chemical constituent (state variable) in segment i, g [M]
             Q, Cj  = total chemical concentration in segments i and j,
             mg/L (or g/m3) [M/L3]
             Ej.(t)   = diffusion coefficient time function for exchange "ij", m2/day [L2/T]
             AJJ    = interfacial area shared by segments i and j, m2 [L2]
             Ley    = characteristic mixing length between segments i and j, m [L].
             fdj> fdj  = dissolved or DOC-bound fraction of chemical in i and j [dimensionless]
             n-     = average porosity at interface "ij", Lw/L
             (volume of water/volume total solution) [dimensionless]
Depending on the PCB concentration gradients, porewater diffusion may be a source or sink for
the water column.

HUDTOX can represent mass transfer of PCBs from the particulate phase in the sediment to the
overlying water column, without net mass transfer of associated solids, via application  of a mass
transfer coefficient applied directly to the particulate phase PCBs in  the upper sediment layer.

Specific alternative approaches (i.e., porewater only  versus combined porewater and particulate
phase transfer) for specifying PCB sediment-water mass transfer exchanges  within the HUDTOX
model were investigated using data-based mass balances (see Chapter 6), and through  the use of
model simulations presented in Chapter 7 for a range of PCB  forms.

5.6 MODEL SPATIAL SEGMENTATION

5.6.1  Water Column Segments

The HUDTOX water column spatial segmentation was developed to capture the effects of the
principal  factors that influence  spatial  patterns  of  water  column  and sediment  PCB
concentrations within the Upper Hudson  River.  A total of 47 water column  segments were
represented from Rogers  Island (RM 194.6) to Federal Dam (RM 153.9) at  Troy (Figure  5-4,
Parts A through D).
                                          56                       Limno-Tech, Inc.

-------
The criteria for developing the water column segmentation grid were driven by locations of:

       •      Major tributaries to the Upper Hudson River;

       •      Lock and dam structures along the river;

       •      Phase 2 and historical water quality sampling stations;

       •      USGS gaging stations; and,

       •      Sediment PCB "hotspots" along the river.
Hydrographic survey data collected by GE during 1991 (O'Brien & Gere, 1993b) were used to
estimate HUDTOX model segment cross-sections.  The TAMS/Gradient Team also conducted
hydrographic measurements within a portion of the Upper Hudson River; however, the GE data
provides more complete  coverage.  No significant differences were found between the two
datasets in reaches of the river covered by both  surveys, including Thompson Island  Pool.
Consequently the GE data were used exclusively in  determining river cross-section geometry for
HUDTOX.
A two-dimensional segmentation for the  water column was developed within Thompson Island
Pool to better resolve potential differences in impacts from cohesive and non-cohesive sediment
areas. The 28 water column segments within the Pool are configured as three lateral segments
across the river, except at Rogers Island, with longitudinal resolution on the order of Vz to % of a
mile (Figure 5-5). At Rogers Island the east and west river channels are each represented by one
lateral segment. Figure 5-6 presents a schematic representation of the HUDTOX model grid that
includes references to  geographical locations.  Output from the RMA-2V hydrodynamic model
for a flow of 8,000 cfs  at Fort Edward was used to provide flow-routing information for this two-
dimensional segmentation grid within the Pool. An evaluation of the variation in flow  through
the HUDTOX segments at a given TIP cross-section for different upstream flows showed only
minor variations, so the flow routing pattern was held constant over the entire range of  flows
simulated (also see Section 5.2.3.1).

The  19 one-dimensional water column segments between Thompson Island Pool and  Federal
Dam were developed to capture the impacts of hydrologic features of the river, including dams
and locations of tributary inputs.  The  water column segmentation was also specified based on
locations of sediment  PCB  "hotspots".   Consequently, the  longitudinal  resolution of  these
segments is variable, ranging from less than one mile to greater than four miles. The geometry of
the HUDTOX water column  segmentation is presented in Tables 5-la and 5-lb. Figure 5-7
illustrates how the HUDTOX water column segment depths vary from upstream to downstream,
indicating the important impacts of the lock and dam systems on river  geometry.

5.6.2  Sediment Segments

Tables 5-2a  and 5-2b  present the spatial configuration and geometry of the HUDTOX surface
sediment segmentation  (layer 1),  including the  assignment  of cohesive and  non-cohesive
sediment areas. Historically delineated sediment PCB "hotspots" are not explicitly represented by
                                          57                       Limno-Tech, Inc.

-------
individual model segments. The finer model grid in Thompson Island Pool does, however, better
represent these areas than it does in segments downstream of the Thompson Island Dam. As
such, care must be taken in the use of the model for simulating future responses to remedial
scenarios that focus on  bedded sediment  areas which may be much  smaller than the model
segmentation spatial scale. The longitudinal  variation in cohesive sediment abundance in the
HUDTOX model is depicted  in Figure 5-8 and  was developed according  to the procedure
described in the following paragraphs.

Surface sediment segment areas for  the  HUDTOX model  were  computed using  two GIS
coverages. First, a GIS coverage developed from side scan sonar studies conducted as part of the
USEPA Phase 2 investigation (USEPA, 1997) was used to define sediment segments within TIP
and downstream to the Northumberland Dam (RM 183.4).  The side scan sonar measurements
were used to distinguish river  bottom  areas of finer (representing cohesive  solids)  and coarser
(representing non-cohesive solids) sediments.  Rocky and mounded bed areas identified by the
river bottom coverage were excluded from the sediment segmentation grid, as were all islands.

Two additional criteria were used in developing the sediment segmentation from the side scan
sonar data:

       •     Water  column  segments  underlain by 15  percent or  more cohesive
             sediment area were  assigned both cohesive  and non-cohesive sediment
             segments,  unless they contained more than 85 percent cohesive sediment
             area, in which case only a cohesive sediment segment was assigned; and,

       •     Water  column  segments  underlain by  less than 15  percent cohesive
             sediment area were assigned  only non-cohesive sediment segments.

The  second  GIS coverage was based on GE's 1997 sediment bed  type sampling between
Northumberland Dam and Federal  Dam (QEA,  1998). This  coverage was used to define the
HUDTOX sediment segmentation in reaches of the Upper Hudson River that were  not covered
by the side scan sonar surveys.

These two GIS coverages of sediment type were intersected with the HUDTOX water column
segments to  develop a two-dimensional picture of  the surface  sediments, and to define 27
cohesive and 43  non-cohesive  sediment segments for the Upper Hudson River between Fort
Edward and  Federal Dam.  Figure 5-4  (Parts A  through D) depicts  the two sediment types
underlying each water column  segment for the entire upper river.  Figure 5-5 provides a large-
scale view of the same information  within just TIP, which was represented with 15 cohesive and
27 non-cohesive surface sediment segments.

A vertical discretization of two centimeters  was used for the HUDTOX sediment segmentation to
provide adequate resolution of vertical PCB profiles  for simulating sediment-water interactions
and long-term system responses. This resolution also provides flexibility in the use of HUDTOX
model output for PCB sediment exposures  in terms of an "active" surface sediment  layer for the
bioaccumulation  models.   A  summary  of  the HUDTOX  surficial sediment segmentation
geometry is provided in Tables 5-2a and 5-2b.  The model grid includes sediments down to 26
                                          58                       Limno-Tech, Inc.

-------
cm (13 layers), resulting in a total of 1035 water column and sediment segments in the entire
model grid.

5.7 MODEL IMPLEMENTATION

The HUDTOX model was developed from the USEPA WASP toxic chemical model framework.
The model was  originally constructed from the WASP4/TOXI4 version of the code and
subsequently modified to include relevant code corrections and changes that were implemented
by USEPA in the WASP5/TOXI5 version. The WASPS model is documented in Ambrose et al.
(1993)  and is distributed by the Center for  Exposure  Assessment Modeling  (CEAM) at the
USEPA Environmental Research Laboratory, Athens, Georgia.

The HUDTOX model FORTRAN source code was compiled and run using Lahey FORTRAN 90
(Version 4.50b, Lahey Computer Systems, Inc.) for personal computers running Microsoft DOS
or Windows (95, 98 or NT) operating systems.   Development,  testing and application of the
HUDTOX model was conducted on IBM-PC  compatible computers.  The computer hardware
system requirements vary, depending on the type of HUDTOX model  simulations being
conducted.  A Pentium n microprocessor (266 Mhz or higher), 64 Megabytes of RAM, and
available  disk storage space of 1.0  Gigabyte are  minimum  requirements  for the  simulations
presented in this report. As a general indication of model execution speed, a 21-year simulation
from 1977 to 1997 required on the order of 10  hours of real time on a 450 Mhz Pentium n
computer.  This  simulation included a model grid consisting of  1035 spatial segments and
computational time steps ranging from 0.0027 to 0.019 days over the 21-year simulation period.
Most model calibration  and forecast simulations were conducted on 600 Mhz Pentium HI
computers.
                                        59                      Limno-Tech, Inc.

-------
THIS PAGE INTENTIONALLY LEFT BLANK
                  60                Limno-Tech. Inc.

-------
Chapter 6

-------
6. DATA DEVELOPMENT FOR MODEL APPLICATIONS

6.1 INTRODUCTION

The development and application of the HUDTOX PCB mass balance model relied on the
extensive  site data obtained from the database created to  support this Reassessment  RI/FS
(USEPA,  1995) and other sources.  This chapter presents the organization and analysis  of the
available data to specify required model forcing functions, initial conditions, rate coefficients,
and state variable parameters. Additionally, observed spatial and temporal PCB concentration
trends are  presented to support model calibration, which is the topic of Chapter 7.

The following major sections are included in Chapter 6:

             6.2    Available Data
             6.3    Model Application Data Sets

             6.4    Flow Balance
             6.5    Mainstem and Tributary Solids Loads

             6.6    Mainstem and Tributary PCB Loads

             6.7    Sediment Initial Conditions
             6.8    Water Temperature

             6.9    Partitioning
             6.10   Volatilization

             6.11   Sediment Particle Mixing
             6.12   Dechlorination

             6.13   Sediment-Water Mass Transfer


Sections 6.4, 6.5 and 6.6 present development of the 21-year daily flow, solids and PCB inputs to
the model.  Section 6.7 presents development of beginning sediment PCB concentrations for the
entire river from 1977 data for the historical calibration and 1991  data for the short-term hindcast
applications.  Sections 6.8 through 6.13 discuss the  specification of various parameter values and
other inputs to the model based on available data.
6.2 AVAILABLE DATA

The Hudson  River is one of the most extensively monitored PCB contamination sites. The
system has been studied extensively and monitored almost continuously over a period of more
than 20 years.  The  various monitoring studies have provided numerous water column and
sediment datasets  useful to modeling PCB fate  and transport  in the system. Most of these data
have been  compiled by TAMS in the Hudson River Database, which was created to  support this
Reassessment.
                                          61                       Limno-Tcch, Inc.

-------
The development and application of the HUDTOX model relied extensively on the Hudson River
Database, in addition to data obtained from other sources.  The Hudson River Database Report
(USEPA, 1995) and accompanying CD-ROM database provides the validated data for the Phase
2 investigation. This Revised Baseline Modeling Report (RBMR) utilized Release 4.1b of the
CD-ROM database, which was  updated in fall  1998  (USEPA,  1998b).  The Hudson River
Database contains information from a large variety of different sources, including:

       •     New York State Department of Environmental Conservation (NYSDEC)

       •     New York State Department of Health (NYSDOH)

       •     New York State Department of Transportation (NYSDOT)

       •     General Electric Company (GE)

       •     Lamont-Doherty Earth Observatory (LDEO)

       •     Rensselaer Polytechnic Institute (RPI)

       •     U.S.  Geological Survey (USGS)

       •     National Oceanic and Atmospheric Administration (NOAA)

       •     U.S.  Environmental Protection Agency (USEPA).

In addition  to the Hudson River Database, site specific information  was also obtained  from a
number of other sources, which are presented in the bulleted list below.

       •     An update to the GE database, dated 12 October 1998,  was supplied by
             Kerry A. Thurston of O'Brien & Gere.

       •     To supplement the records available in  Release 4.1b  of the database, a
             portion of the  1997 USGS  flow, suspended solids and PCB data  were
             obtained directly from the USGS in Albany, New York (email to Penelope
             Moskus from Brian Wolorby on 8/12/98).

       •     Additional water column dissolved organic  carbon data reported by J.
             Vaughn  (1996), were used in  addition to the measurements by GE and
             USEPA contained in the database.

       •     The 1998 GE Sediment sampling program data are not included in Release
             4.1b  although these data were also used. The poolwide average surface
             sediment PCB  concentrations  reported by QEA (1999) from these  data
             were used as additional  model calibration points  for surficial sediment
             PCB concentrations in Thompson Island Pool (TIP) (Chapter 7).
                                         62                       Limno-Tech. Inc.

-------
       •     Atmospheric PCB concentration data from the  Integrated Atmospheric
             Deposition Network (IADN) station at Point Petre, Ontario (Hoff, et al.
             1996) were obtained to specify atmospheric PCB concentrations in the
             HUDTOX model.

Where necessary and appropriate, information from the scientific literature and various technical
reports was also used to specify values for model process coefficients. These sources are cited in
the report text.

The Data Evaluation and Interpretation Report (DEIR) (USEPA, 1997) and the Low Resolution
Sediment Coring Report (LRC) (USEPA, 1998a) are companion  reports to this Revised Baseline
Modeling Report (RBMR).  The DEIR contains a literature review of current and historical PCB
water column data, and an evaluation of geochemical fate of PCBs in the sediments of the Upper
Hudson River. The LRC contains an assessment of current and historical inventories of sediment
PCBs in the  Upper Hudson  River.  The reader is referred to these  companion reports for
additional details on the available datasets for this Reassessment.

6.3 MODEL APPLICATION DATASETS

The development and application of the HUDTOX model is based on the extensive sediment and
water column monitoring datasets collected by primarily by the USEPA, USGS, NYSDEC, and
the General Electric Company. A summary of the various data collection activities through 1994
is provided in the Hudson River Database Report (USEPA, 1995). In addition to the long-term
record of PCB concentrations in water and  sediment available  from the combined datasets, a
number of specific, focused studies were conducted by USEPA and GE.  The data from these
studies provide additional insight into processes affecting PCB fate and  transport in the system,
which supported parameterization  of key processes in the HUDTOX model.   This section
provides an overview of the primary model application datasets and their use in the HUDTOX
modeling effort.

6.3.1  Sediment Datasets

The primary sediment datasets used in this modeling effort are  the sediment sampling surveys
conducted by:

       •     NYSDEC in 1976-78 (Tofflemire  and Quinn, 1979) and  1984 (Brown et
             al., 1988);

       •     GE in 1991 (O'Brien and Gere, 1993) and 1998 (O'Brien and Gere, 1999);
             and

       •     USEPA in 1992 and  1994 (DEIR and USEPA, 1998b).

The NYSDEC 1976-78, NYSDEC  1984, GE 1991 and the GE 1998  surveys are comprehensive
assessments of PCB levels in the  sediments.  The NYSDEC  1976-78 and  GE  1991 studies
sampled the complete extent of the river from  Fort Edward  to Federal Dam,  whereas, the
NYSDEC 1984 survey was limited to Thompson  Island Pool.  The GE 1998 study extensively
                                         63                      Limno-Tech, Inc.

-------
sampled Thompson Island Pool and included focused coring at a limited number of locations
downstream as well. The USEPA 1994 studies were aimed at assessing PCB concentrations in a
relatively small number of discrete areas in Thompson Island Pool and a few hotspot locations
downstream.

Model  initial  conditions  for  sediment  PCB  concentrations  for  the  1977-1997  historical
calibration were established from 1976-78 NYSDEC sediment PCB data, referred to as the 1977
NYSDEC data throughout this report (Section 6.7). The average surface sediment concentrations
from the 1984, 1991, and 1998 datasets served as calibration targets for the historical calibration.
The GE  1991 and 1998 data were considered primary calibration targets for surface sediments.
The 1984 NYSDEC data and the 1994 USEPA low-resolution sediment  core data were not
primary calibration points for surface sediment PCBs because these data contain measurements
of surface concentrations averaged over large depth intervals.

The sediment survey in Thompson Island Pool by GE in 1998 attempted to 'repeat' portions of
the 1991 O'Brien and Gere, and  1994  USEPA sediment surveys (QEA  1999).  Average
concentrations for cohesive and non-cohesive sediments were computed for the 0-5 cm sediment
layer and reported by QEA (1999). As the raw 1998 sediment data were obtained in a later phase
of this project, the reported concentrations were used as additional model calibration targets for
the Thompson Island Pool.

The "Low Resolution" sediment dataset collected by USEPA in  1994 provides assessments of
sediment concentrations at approximately 20 locations in Thompson Island Pool (15 small zones
and 4 near shore locations) and in 7 hotspots downstream of Thompson Island Pool.  In addition,
the USEPA data provide high-resolution core analyses at 28 selected  locations in the Upper and
Lower Hudson collected in 1992.  The USEPA data are not extensive  enough to serve as primary
calibration information for the model. The main use of the 1994 USEPA sediment data was to
assess changes  in PCB  levels relative to the  1984 measurements by  NYSDEC  at specific
locations, which is a principal topic in the LRC (USEPA, 1998a).  The high-resolution core data
analyses included radionuclide dating, providing an estimate of sediment accumulation rates  at
specific  locations (USEPA, 1997).

The sediment  sampling  effort conducted  by USEPA included  mapping  of fine and coarse
sediment grain size using side scan sonar images from Fort Edward to the Northumberland Dam
(Flood, 1993).  A qualitative sediment bed mapping survey was conducted by GE to characterize
locations of fine and coarse sediment deposits between Northumberland Dam and Federal Dam
(QEA, 1998a).  The combined side scan sonar and qualitative bed mapping data were used  to
develop the model sediment segmentation (Section 5.3) and to classify some sediment samples
for the purpose of determining fine and coarse sediment average PCB concentrations.

The GE 1991 and USEPA datasets included congener analysis in the water and sediments, which
are available in  the Hudson River Database, whereas  the  NYSDEC PCB analysis  reported
concentrations as Aroclors. The GE 1991 data were used to specify  sediment initial conditions
for modeling individual congeners and total PCBs over the period 1991 to 1997 (Section 6.7).
                                          64                       Limno-Tech, Inc.

-------
The GE 1991 data included measurement of porewater PCB and dissolved organic carbon data.
These data were  used to estimate in-situ sediment-water  partition coefficients for individual
congeners and the historical calibration state variable, Tri+.  The development of the partition
coefficients for Tri-f and congeners is presented in the DEIR (USEPA, 1997). The application of
the partition coefficients estimated from these data is discussed in detail in Section 6.9.

A summary of the sediment datasets and their use in this modeling effort is summarized in Table
6-1.

6.3.2  Water Column Data

The principal water  column datasets  used for solids and PCBs in this modeling effort were the
following:

       •      Long-term  monitoring data collected  at  Fort  Edward,  Schuylerville,
              Stillwater and Waterford from 1977 to 1997 (collected by USGS, USEPA
              and GE)

       •      Thompson Island Dam data from 1991  to 1997 (collected by USEPA and
              GE)

       •      Mainstem and tributary solids data from the spring 1994 high flow survey
              (collected by USEPA)

       •      Mainstem data from the USEPA Phase 2 monitoring program in 1993

       •      High  flow sampling data in 1997 (collected by GE)

       •      Thompson Island Pool float study data in 1996 and 1997 (collected by GE)

       •      Thompson Island Dam bias study data (collected by GE).

The long-term water column monitoring by USGS at Fort Edward, Schuylerville, Stillwater and
Waterford, combined with  the more recent sampling by GE and the USEPA at these and other
locations,  provides an extensive history of water column PCB and TSS concentrations in the
Upper Hudson River. Monitoring of PCB and TSS concentrations by the USGS commenced in
1977 and continues to the present.  The USGS data, combined with the more frequent data from
the GE monitoring beginning in 1991, and the  USEPA Phase 2 data collected  in 1993 to 1994,
allow specification of PCB and TSS loading at Fort Edward and the tributaries.  In addition,
these  combined datasets permit development  of in-river load  estimates of PCB and TSS at
Stillwater  and Waterford over the entire historical calibration period.  The long-term record of
solids  and PCB  concentration measurements at Thompson Island Dam  (1991   -  1997),
Schuylerville  (1977  - 1993), Stillwater (1977  - 1997) and Waterford (1977  - 1997) serve as
principal calibration  datasets for the HUDTOX modeling effort.

The GE water column monitoring data, which  spans the period 1991 to the present, provides a
high frequency monitoring dataset at Fort Edward and Thompson Island Dam, in addition to
                                          65                       Limno-Tech, Inc.

-------
periodic data collected at downstream locations.  GE also has conducted a number of specialized
monitoring studies  which provide  insight into  processes  affecting  PCB transport,  localized
sources of PCB loading and seasonal patterns in PCB fluxes.

The high  flow event sampling in March and April of 1994 by USEPA represents  the most
extensive high flow solids monitoring dataset. Samples were collected during the only period
over which tributary and mainstem TSS concentrations  were measured simultaneously during a
high flow event. While this dataset provides the most constrained assessment of solids dynamics
over the course of a high flow event, PCBs were not simultaneously  measured during this
sampling event and a significant fraction of the total tributary flow was  not measured for this
period.

The USEPA also conducted a  number of water column sampling  surveys  to assess PCB
concentrations,  sources,  and transport  in the system.   Six of these surveys were down-river
transects in which composite samples were collected over approximately two-week periods over
six  months spanning a range of  seasonal conditions in the river.  A  series  of seven sampling
events  occurring approximately monthly at 13 stations  was also conducted  in which sampling
was timed to monitor the same parcel of water through the system. These  USEPA Phase 2 data
provide additional information about the spatial and seasonal patterns of  PCB transport in  the
river and provide  a determination of sediment-water partitioning behavior. Interpretation of the
Phase 2 data and  development of partitioning relationships from  these data  is presented in  the
DEIR (USEPA, 1997).

6.3.3 Conversion of PCB Data in Historical Calibration Datasets

Different sediment and water datasets used different analytical methods, which required various
data adjustments  in  order to make the datasets  comparable for  use  in  the HUDTOX model
calibration. The historical modeling state variable was the sum  of tri and  higher chlorinated
congeners (denoted  Tri+ in the remainder of this report).  Individual  congeners and total PCB
could not be used for the historical calibration because neither congener analyses nor equivalent
total PCB quantitations are available in the historical datasets. Individual Aroclors were also not
consistently quantified between datasets. Additionally, Aroclors and total PCBs were considered
unsuitable state variables for historical  calibration because shifts in  congener patterns due to
weathering and/or dechlorination may  result in  variations  in partitioning behavior. The Tri+
quantity could be  determined in all datasets and was selected for the historical calibration. Tri+
was an attractive choice for a historical modeling state variable not only because it is consistently
identified among datasets, but also because its composition is relatively less variable throughout
the  system than PCB forms  including mono- and dichloro-biphenyls.  Methods were previously
developed (USEPA, 1998a,  Butcher, 2000b) to  convert the various  PCB quantifications into
estimates of the Tri+ concentration for each dataset. These methods are  summarized  for each
dataset below.

6.3.3.1 USGS Water Column Data

The USGS water column data  represent whole water analyses,  with PCBs quantified using
Aroclor standards. Packed column analysis was used until  1987, when  data began to be analyzed
                                           66                       Limno-Tech, Inc.

-------
 with capillary columns. Approximately coincident with the USGS switch from packed column
 analysis to capillary column analysis beginning in 1987, a limited number of Aroclor standards
 were used relative to the earlier years.

 Split sample analysis between USGS and Phase 2 data supported use of the USGS-reported total
 PCB concentration from the packed column analysis as  a direct measure of the Tri+ sum. A
 regression relating USGS total PCB to the Tri+ sum gives a good linear fit with an intercept not
 significantly different from zero (USEPA, 1997). Thus, the USGS packed-column total PCB
 results were used directly as Tri+ through 1987.

 Re-analysis of 60 USGS sample chromatograms by QEA (Rhea and Werth, 1999) supported use
 of the USGS reported Aroclor  1242 results or,  when 1242 results are not available, use of
 Aroclor 1248 as the best representation of Tri+ concentration in the USGS data after 1987.

 6.3.3.2 1976-1978 NYSDEC Sediment Data
 Total PCBs were reported by O'Brien and Gere for the 1976-1978 sediment dataset. These were
 based on Aroclor analysis using a limited number of packed column peaks, which tended to miss
 the mono- and di-homologues. Based on reconstruction of the 1976-1978 total PCB results from
 USEPA Phase 2 sediment congener data, a regression between the Tri+ concentration and the
 1977-1978 total PCB concentrations produced a zero-intercept model with  which to estimate
 Tri+ concentrations  from these data (Equation 6-1).  Details of this analysis  are presented in
 USEPA, 1998a and Butcher, 2000b.

                    Tri+ (1977)  = 1.131  x[ Aroclor 1016 + 1254]                   (6-1)

 6.3.3.3 1984 NYSDEC Sediment Data
 Total PCB concentrations reported for the 1984 sediment data were reported by NYSDEC as the
 sum of estimated concentrations of Aroclors 1242, 1254, and 1260. A constant conversion factor
 was determined to correct these data to a basis consistent with the Tri+ quantitation in the Phase
 2 data (Equation 6-2). The analysis conducted to develop this conversion is presented in detail in
 USEPA, 1998a.

                    Tri + (1984) = 0.944x (1984 NYSDEC total PCBs)                 (6-2)

 6.3.3.4 GE Water Column and  Sediment Data

The majority  of GE water column results and all of the GE sediment data collected in 1991
 include congener-specific analyses and homologue fractions. Tri+ concentrations were computed
as the sum of tri-through deca-homologue concentrations.

6.3.3.5 USEPA Water Column  and Sediment Data

All of the USEPA  water column and sediment data  include congener concentrations  and
calculated  homologue concentrations. Tri+ concentrations were computed as the sum  of tri-
through deca-homologue concentrations.
                                          67                       Limno-Tech, Inc.

-------
6.3.4 Data conversion for Total PCB and Congeners

While the primary calibration state variable in the long-term historical calibration is Tri+, short
term hindcast applications over the period Jan. 1, 1991 through Sept. 30, 1997 were additionally
conducted with individual congeners and total PCBs to test the Tri+ historical calibration.

Five congeners were  selected for modeling based on physical  and  chemical  properties and
frequency of detection in all media types (sediment, water and biota).  These  five congeners are
BZ#4  (a  di-chlorobiphenyl),  BZ#28  (tri-chlorobiphenyl),  BZ#52  (tetra-chlorobiphenyl),
BZ#101+90 (co-eluting penta-chlorobiphenyls) and BZ#138 (hexa-chlorobiphenyl).
In the GE congener quantitations, all five  of the congener state  variables co-elute with  other
congeners.  BZ#28, 52 and 138 co-elute, respectively, with BZ#50, 73 and 163, all  of which are
minor congeners not quantitated by USEPA. The BZ#28,  52 and  138 concentrations in the GE
data were used directly as measures of these congeners, ignoring the co-eluting, minor congeners.
BZ#101  and BZ#90 co-elute  in both the GE and the USEPA congener results and  were not
separated for modeling purposes. For the co-eluting BZ#4 and BZ#10 congeners, the average
BZ#4/BZ#10  ratio determined  in TIP water  column and  sediment samples presented by
Hydroqual (1997) was used to compute BZ#4 concentrations in the GE water  column and
sediment data.

The GE database contains total PCB data analyzed by three different methods, which are referred
to as the capillary column method (PCB_cap), the USGS  method (PCB_usgs), and the Webb and
McCall method (PCB_wm). The majority of the  GE samples were  analyzed using capillary
columns, although a relatively small number of samples had only a  USGS or Webb and McCall
result reported. The method to be used was selected as follows: use PCB_cap result  if available,
or else use the PCB_usgs if available, and if neither PCB_cap or PCB_usgs are available, use
PCB_wm.
6.4 FLOW BALANCE

6.4.1  Overview

The HUDTOX model is based on the principle of conservation of mass. Mass balances of flow,
solids and PCBs are represented in the model. HUDTOX requires specification of all  tributary
and upstream flow inputs, in  addition to external solids and PCB loads. The purpose of this
section is to describe the development of daily  flow inputs from  upstream at Fort Edward and
from tributaries and direct drainage flows for the calibration period (January 1, 1977 through
September 30, 1997).  Tributary inflows are specified for eight significant tributaries  and four
direct drainage inputs  between Fort Edward and Federal  Dam at Troy.  Direct  drainage flows
were computed for drainage  areas not included in the eight  tributary watersheds  and they are
treated as additional tributary flows. The Fort Edward daily flow estimates were based on USGS
flow gage data at Fort Edward.  The Hoosic  River and Mohawk  River flow inputs were  taken
from continuous USGS records available for these tributaries.   Ungaged tributary and direct
drainage  flows were estimated based on the Hoosic River  flow records or other available USGS
stream flow data in the Upper Hudson watershed.
                                          68                       Limno-Tech, Inc.

-------
Daily flow estimates for mainstem Hudson River locations downstream of Fort Edward were
based on the sum of Fort Edward, tributary and direct drainage flow inputs. These synthesized
flow time series were used for developing cumulative in-river solids and PCB  load estimates to
supplement the primary, long term sampling stations for use in model calibration (Sections 6.5
and 6.6).

6.4.2  Flow Data

Mainstem and tributary flow gages in operation during the study period are summarized in Table
6-2.  The locations for these flow gages are shown in Figure 6-1. The Fort Edward gaging station
(USGS  # 01327750) was operational for the entire study period, whereas major gaps exist in
daily  flow records for  the other  mainstem stations.  Reported USGS daily flow data for the
Stillwater (USGS # 01331095) and Waterford (USGS # 01335754) stations are flagged by USGS
as estimated values beginning in September 1992 at Stillwater and July 1992 at Waterford.  This
was due to construction activities  that began in 1992 and continued through at least 1995 (USGS
Water Resources Data  1993; and, Charles Fluelling,  NYS Thruway, personal communication,
February 27,  1997). The daily flows at Stillwater continued to be reported as estimates through
the end of 1997 because this gage remained out of operation until  that time. The only direct
tributaries gaged for the entire study period  are on the Hoosic (USGS # 01334500 at Eagle
Bridge) and Mohawk Rivers (USGS # 01357500 at Cohoes, and # 01357499 at Crescent Dam).
Stream  flow  data are available at two locations in the Fish Creek  watershed: Kayaderosseras
Creek (USGS # 01330500), and Glowegee Creek (USGS # 01330000). USGS flow monitoring at
the Kayaderosseras Creek station was discontinued in 1995.

The ungaged tributary  drainage area is a large percentage of the drainage area between Fort
Edward and Waterford. The drainage area of tributaries feeding the Hudson River between Fort
Edward and Waterford equals 1,794 mi2.  Only 33 percent of this area is gaged (Kayaderosseras
Creek near West Milton (90 mi2) and Hoosic River at Eagle Bridge (510 mi2). Approximately 67
percent  of the watershed area between Fort Edward and Waterford is ungaged.  Flows draining
ungaged watersheds were estimated as described in Section 6.4.

The Mohawk River is  a large  gaged tributary to the Hudson River (3,450 mi2) which enters
between Waterford  and Federal  Dam  at Troy.  The drainage area of tributaries  feeding the
Hudson River between Fort Edward and the Federal Dam at Troy equals 5,244 mi2.  Accounting
for the gaged Mohawk River, Hoosic River and Kayaderosseras Creek drainages (3,450 + 510 +
90 = 4,050 mi2), 77 percent of the tributary between Fort Edward and the Federal Dam at Troy
(5,244 mi2) is gaged.

Flood frequency analysis (Log Pearson Type HI) was conducted based on 1930 to 1991 flows at
Fort Edward  by Butcher (2000a). As the period of record at Fort Edward commences in 1977,
this analysis  made use of flows  estimated from the  sum of flows  measured upstream in the
Hudson River at Hadley, NY (USGS gage # 01318500) and  Sacandaga  River (USGS  gage #
01325000) for the period before 1977. The estimated 5, 50,  and 100  year return frequency flows
at  Fort Edward based on this analysis  are 30,126 cfs, 43,671 cfs and 47,330 cfs, respectively
(Figure  6-2).  The peak daily average flow at Fort Edward during the model calibration period
occurred in  1983  (34,100 cfs), which has an estimated return frequency of approximately 11
                                          69                      Limno-Tcch, Inc.

-------
 years.  In 1976, the year prior to the simulation period, a 37-year flow of approximately 42,000
 cfs occurred, only 11 percent lower than the estimated 100 year flow.

 6.4.3 Flow Estimation Methods

 To specify daily upstream and tributary daily flow inputs to the HUDTOX model, daily average
 USGS flow records were used where possible and ungaged inflows estimated by relationship to
 these data. Upstream flow at Fort Edward was specified directly from the USGS data, without
 modification.  Mohawk River flows (sum of  daily flows at  Cohoes and  the  Crescent Dam
 diversion), and the Hoosic  River flows  measured at Eagle Bridge were  also used without
 modification. The USGS Fort Edward flow time series from 1977 to 1997 is  shown in Figure 6-
 3.

 Ungaged tributary flows were  estimated using the drainage area ratio (DAR)  method.  This
 approach relates measured flows to unmeasured flows in similar watersheds  by assuming equal
 flow yield per  unit area of  watershed.  Based on gaged tributary flows,  ungaged flows  are
 computed using the ratio of watershed drainage areas (Equation 6-3).

                                            'DA
                                                  ungaged _ tributary
              ^•ungaged _ tributary   *Ł• gaged _ tributary
                                               DA
                                                  gaged ^tributary
(6-3)
where:
                    = ungaged tributary flow

                   = gaged tributary flow
       DAimgaged_tribu,ary = ungaged tributary drainage area

       DAgaged_tributary = gaged tributary drainage area.

The DAR approach was used to estimate all ungaged tributary and direct drainage flows based on
USGS flow rate data from Kayaderosseras Creek, Glowegee Creek or the Hoosic River at Eagle
Bridge. The ungaged area  includes the Hoosic River watershed downstream of Eagle Bridge.
Reference tributaries for each  watershed in which flows were estimated were selected based on
consideration of similarities in land use, topography,  location and watershed size. Tributary
watershed  areas  were  estimated  by digitizing  the watershed  boundaries  from  USGS
topographical maps in a GIS (Table 6-3).  All estimated tributary and direct drainage flows
between  Fort Edward and  river  mile  180 (just downstream of Schuylerville) were based on
Kayaderosseras Creek or Glowegee Creek flow data, and those downstream of river mile  180
were based on Hoosic River at Eagle Bridge flow data (Table  6-3).  The Kayaderosseras Creek
gaging station is located in the upper portion of the watershed drained by Fish Creek (Figure 6-
1). USGS flow monitoring at  the Kayaderosseras Creek station  was discontinued in 1995. Flow
data collected on Glowegee  Creek were used after 1995. The Glowegee Creek gage is located in
the upper reaches of the same watershed as Kayaderosseras Creek and has a  relatively small
drainage area (26 mi").
                                          70                       Limno-Tech, Inc.

-------
Direct application of the DAR approach does not result in flows from individual tributaries that
are mutually constrained in the sense that these estimates may not sum to observed  flows at
downstream locations in  the  Upper Hudson River. USGS flow estimates at Stillwater and
Waterford were  used to constrain estimated tributary flows  in order to  achieve a long-term
seasonal average flow balance between Fort Edward and Waterford. Comparison of the estimated
flows at Stillwater and Waterford for 1993 to the flow estimates presented in the DEIR (USEPA,
1997) showed the DEIR estimates to be substantially higher during low flow. Correlation of the
DEIR summer average flow estimates with cumulative precipitation data revealed that the DEIR
estimates were biased high (USEPA, 1999a). Consequently, the DEIR flow estimates were not
used in any of the  HUDTOX model applications and the USGS estimates  at Stillwater and
Waterford were used exclusively.

The seasons used in the  seasonal flow balance were defined as follows:

       •      Spring: March, April, May

       •      Summer: June, July, and August

       •      Fall: September, October, and November

       •      Winter: December, January, and February

The seasonal mean flow computed by summing the Fort Edward and estimated tributary flows
was compared to the seasonal mean flow from the USGS gages at Stillwater and Waterford over
the period from March 1, 1977 to June 30, 1992 (Table 6-4).  This period was used because all
three gages (Fort Edward, Stillwater and Waterford) were operational.  After September 1992,
the gages  at Stillwater  Dam and Lock 1 at Waterford were influenced by dam construction
activities and flows reported by USGS after this date are estimated.

The seasonal mean  flows at Fort Edward, Stillwater  and Waterford were  computed from the
USGS data over the period March  1, 1977 to June 30, 1992.  Seasonal mean flow, increases in
seasonal mean flow,  and computed watershed flow yield between these locations are presented in
Table 6-5.   The ungaged  tributary flows estimated by the  DAR method were scaled using an
adjustment factor, a, for season, j, in order to be achieve  a  long-term seasonal average flow
balance  between Fort  Edward  and Stillwater (Equation  6-4)  and  between  Stillwater  and
Waterford (Equation 6-5).  Equations 6-4 and 6-5 were solved for  the adjustment factors (the a
terms) for each season.


                                                                                 (6-4)
                                          71                       Limno-Tech, Inc.

-------
where:
                         iii     = seasonal adjustment factor for season j and for tributaries between
                               Fort Edward and Stillwater

                    a stiii-watfd   = seasonal adjustment factor for season j and for tributaries between
                               Stillwater and Waterford

                    Qreference    = flow of gaged reference tributary

                    fti          = drainage area proration factor for tributary i

      As discussed above, the drainage area proration factor for tributary / is the ratio of the ungaged to
      the gaged (or reference) tributary watershed areas (Equation 6-6).
                           ri  ^™unga ged _ tributary ' *-''* gaged _ tributary                              ^   '

      The mean  seasonal flows, drainage area proration factors and seasonal adjustment factors are
      presented in Table 6-5. The seasonal flow adjustment resulted in average tributary flows that sum
      to the average flows at Stillwater and Waterford for the period considered in the flow balance.

      The required adjustment for the ungaged tributary flow between Stillwater and Waterford was
      much less than 1.0 in the summer and fall (Table 6-5).  This indicates that the extrapolation of
      the Hoosic River flows gaged at Eagle Bridge to ungaged tributaries using the DAR approach
      resulted in a significant overestimate of incremental flows during summer and fall in the reach
      from Stillwater to Waterford.   It is possible that differences  in watershed geology may cause
      different base flow behavior relative to higher flows in the Hoosic River compared to the smaller
      tributaries whose flow was estimated based on the Hoosic flow. Evaporative and other losses
      from the Hoosic River between Eagle Bridge and the Hudson may be significant during the
      summer and fall,  which could result in an overestimate of the ungaged Hoosic  River flows
      between Eagle Bridge and Hudson River for these periods.

      To evaluate resulting tributary flows estimated in the manner described above, resulting flows at
      Stillwater and Waterford computed  by summing the Fort Edward and  tributary flows were
      plotted versus the USGS flow data  (Figure 6-4). Generally, daily flow estimates were within 30
      percent of the USGS estimates, however, during some high flow events, estimated flows differed
      by over 30 percent from the USGS flow. This was is not  surprising, considering the DAR
      approach,  which  assumes  that  unit  hydrograph  responses  seen  at  Eagle  Bridge  and
      Kayaderosserass Creek are instantly translated from the  whole watershed to the Stillwater and
      Waterford gages.  Thus the flow discrepancies are explained in part by relative timing of flood
      pulses.

      To minimize error associated with estimating mainstem Hudson River flows  during high flow
      events, an  adjustment was applied for high flows differing by more than 30  percent from the
      USGS  data.  The USGS gage  readings during the  1977 to 1992  period were assumed to be
      accurate within 30 percent during high flow events.  Estimated tributary flows were adjusted to
      achieve agreement within 30 percent of the USGS flows when flow at Fort Edward was greater
                                                 72                       Limno-Tech, Inc.

-------
than 10,000 cfs. When the difference between estimated and USGS-reported flows was greater
than 30 percent, the tributary flows were reduced according to their percent flow contribution at
mean flow.  This produced flow estimates within 30 percent of the USGS data for flows greater
than 10,000 cfs at Fort Edward.

6.4.4 Results of Flow Balance

Analysis of the flow balance developed for the HUDTOX application period of January 1,  1977
to September 30,  1997 produces mean  flows at  Fort Edward,  Schuylerville, Stillwater and
Waterford of 5,248 cfs, 6,117 cfs, 6,603  cfs and 8,106 cfs, respectively.  Mean Upper Hudson
River flows  increase  54 percent from Fort Edward  to Waterford.  The Fort Edward  flow
represents  79 percent and 65 percent  of the average flow  at  Stillwater  and Waterford,
respectively.   During this period,  the estimated peak flows at Fort Edward, Schuylerville,
Stillwater and Waterford are 34,100, 40,200, 46,800 and 70,500 cfs,  respectively.  Peak flows at
Fort Edward  and Schuylerville occurred in  1983, while at Stillwater and Waterford, peak flow
occurred in  1977.  The 1983  flow  has an  estimated  return frequency at Fort Edward of
approximately 11 years.

Figure 6-5  presents a summary of  average daily flows for the study period, by tributary and
mainstem station.  The three largest tributaries in order of decreasing mean flow are the Mohawk
River, Hoosic River and Batten Kill. Average flows increase by a factor of 1.2 and 1.5 from Fort
Edward to Stillwater and Fort Edward to Waterford, respectively.  Flows over Federal Dam, the
downstream extent of the model are a factor of 2.5 larger than Fort Edward flows.

 A plot of estimated flow contributions from each source along the river allows visualization of
relative  magnitude  of the various tributary inputs (Figure 6-6). Approximately 35 percent of the
flow  volume  at Waterford is  due to tributary inputs entering between  Fort  Edward  and
Waterford.  The Fort Edward flow represents about 65  percent of the flow past Waterford. The
Hoosic River and Batten Kill are the largest sources, providing 16 percent and 7 percent of the
flow at Waterford,  respectively.  At Federal Dam, approximately 62 percent of the total flow is
from tributaries, with the Mohawk being the largest source, providing 41 percent of the total flow
at Federal Dam. Only 38 percent of the flow at Federal Dam is from the  Fort Edward flow
during the 21-year study period.

6.4.4.1 Validation of the Flow Estimation Approach

While the above tributary and mainstem flow balance was determined for the period from March
1, 1977 to June 30, 1992, the adjustment factors in Table 6-5 were applied to the DAR-estimated
tributary flows for  the entire HUDTOX application period from  January 1, 1977 to September
30,  1997. As a measure of accuracy, the estimated daily flows at Stillwater and Waterford were
compared to the reported USGS daily flows and average annual  flows.  After  1992, the USGS
flows are also estimated values at Stillwater and Waterford.

Estimated and USGS-reported flows were compared on a daily and average annual basis. The
estimated and  USGS-reported  average  annual  flow  passing Stillwater  and Waterford  was
compared for each  year of the calibration  period (Table 6-6).  Results indicate that mean annual
                                          73                       Limno-Tech, Inc.

-------
flows are within 7 percent at Stillwater and 9 percent at Waterford. Percent differences for the
1993-1997  period, which is outside of the flow balance period, are consistent with the 1977 to
1992 flow balance period. The difference between estimated and USGS-reported average flow
over the entire calibration period is about  1  percent at each location.   Scatter plots of daily
estimated  versus USGS-reported flows  at Stillwater and  Waterford also  show fairly  good
agreement for the 1993-1997 period, generally within about 30 percent (Figure 6-7).  Inspection
of the estimated and USGS-reported daily flow hydrograph also suggests that the flow estimation
approach produced  good results  for this period (Figure  6-8).  Based on  the good agreement
between the  estimated and USGS-reported flows  for the entire calibration period, the flow
estimation method described above was considered to give good results that were acceptable for
modeling purposes.

6.4.4.2 Application of Estimated Flows in Modeling

The USGS  -reported flow at Fort Edward and the synthesized daily average flow time series for
tributaries were input as discrete daily time functions in the HUDTOX model.

To compute  in-river mass  loads of solids and PCBs for  comparison to model output, the
estimated daily flow time series at TI Dam, Schuylerville, Stillwater and Waterford were used,
instead of the USGS flow estimates,  which are available at Stillwater and Waterford.

6.4.4.3 Summary of Flow Balance

Approximately  20  percent  of the total tributary flow inputs  to  the HUDTOX model  were
estimated. Between Fort Edward and Waterford, approximately 60 percent of the tributary flows
were estimated.  Ungaged drainage  areas between these stations accounts for 67 percent  of the
total  tributary drainage area.  Tributary flow  estimates used the  DAR approach, relating
unmeasured flows to  measured flows on Kayaderosseras Creek (substituting with Glowegee
Creek after 1995), and the Hoosic River.  Flow estimates were  adjusted  to achieve a seasonal
water balance from 1977 to 1992 between Fort Edward and Stillwater and between Stillwater and
Waterford.   Based on comparison  to USGS  data, corrections  were made to estimated flows
during high flow periods to be  within  30 percent  of USGS  flows. Estimated flows  were
compared to USGS-reported flows at Stillwater and Waterford, which are estimated by USGS,
for the period 1992  to 1997 with  good results.  The estimated flows were used  to specify flow
inputs to the HUDTOX model and to develop in-river mass flux estimates for solids and PCBs,
which is explained in the following sections.

Daily average flows  over the study period increase by approximately a factor of 1.2 and 1.5 from
Fort  Edward to Stillwater and Fort Edward to Waterford,  respectively. While the Fort Edward
flow is the largest single flow input upstream of Waterford, the Mohawk River flow is larger than
the Fort Edward flow and is the largest inflow between Fort Edward and Federal Dam.
                                          74                       Limno-Tech, Inc.

-------
6.5 MAINSTEM AND TRIBUTARY SOLIDS LOADS

6.5.1 Overview

The HUDTOX  model requires  specification of  solids and PCB  inputs, analogous  to  the
specification of flow inputs from upstream and tributary sources. Daily average solids loading to
the HUDTOX model from upstream at Fort Edward and from all tributary inputs was estimated
for the entire calibration period: January 1, 1977 through September 30, 1997.

Total suspended solids (TSS) concentrations were not measured at all locations, or continuously
throughout the calibration period, requiring estimation of a significant fraction of the total solids
loading.  Daily sampling  frequency was  approximately  11  percent at  Fort Edward  for  the
calibration period.  Data was especially limited for tributaries, for which daily TSS monitoring
frequency is less than 2 percent for those  tributaries that were monitored. In addition, only 71
percent of the watershed area was monitored for TSS, requiring estimation of TSS  loads from the
other 29 percent of the watershed with no data.

As a consequence of the limited data available,  resulting estimates of tributary  loads are very
uncertain. Initial estimates of tributary solids did  not result in a solids balance for the mainstem
Upper Hudson River.  The sum of upstream and tributary loads were significantly lower than in-
river estimates at Stillwater and Waterford. As a result, assumptions were required regarding in-
river solids dynamics which led to adjustment  of solids loads to achieve a long-term solids
balance.   Significantly  more  solids data are available at  low and high flows for the main
Thompson Island Pool tributaries (Snook Kill and Moses Kill), which allows estimation of solids
loads to  the Thompson Island Pool  with more certainty than to downstream  reaches.  The
significant uncertainty  associated with the estimation of tributary solids loads downstream of
Thompson Island Pool is addressed through model sensitivity analysis (Chapter 7).

Apparent  decreases on solids loads  over  time at  Fort  Edward were observed  during  the
calibration period.  The solids loading was observed to be lower for given flows  at Fort Edward
after 1990, compared before 1990. In  developing estimates of the Fort Edward solids loads,
separate rating curves were developed  for these periods based on this observation.

6.5.2 Solids Data

The available solids data for the mainstem and tributary stations are summarized in Tables  6-7
and 6-8, respectively.  The locations of these solids sampling stations are shown in Figure 6-9.
More frequent  solids  concentration data  were available for mainstem stations  than tributary
stations, with no tributary solids data available prior to 1988.  In addition, as illustrated in Figure
6-10, only 71 percent of the watershed area between Fort Edward and Waterford was monitored
for solids, thus requiring estimation of solids loads from 29 percent of the total watershed area in
the Upper Hudson River.   Furthermore,  for the 71 percent  of the watershed  area that was
monitored, only very limited data are  available for most of the tributaries.  Generally, tributary
samples were collected for only a short period of time  during  the 21-year study period.  Solids
samples were collected for mainstem and tributary  stations  on only 24 percent and 1 percent,
respectively, of the total days in the 21-year simulation period.
                                           75                       Limno-Tech, Inc.

-------
 An  extensive record of suspended  solids concentration data is available  at Fort Edward,
 Stillwater and Waterford over the 21-year study period.  Although numerous measurements are
 available, sampling frequency was sporadic during certain time periods.

 Due to differences in sample collection methods between GE and USGS, there is uncertainty as
 to whether or not  these datasets are comparable, especially at Fort Edward.  The GE samples
 were collected by O'Brien & Gere by a number of methods. For the GE 1991 Thompson Island
 Pool Suspended Solids Study, a Manning automatic sampler was used to collect samples from an
 intake tube positioned at mid-depth (O'Brien & Gere, May 1993a).  In the other GE studies, TSS
 samples were obtained using a 1.2-liter Kemmerer sampler to collect depth-composited samples
 either at 3-foot depth intervals throughout the water column (e.g. O'Brien & Gere, May 1993b),
 or at surface, mid-depth and deep sample depths (e.g. O'Brien & Gere 1998).  The USGS TSS
 sampling method collected a continuous depth-integrated sample throughout the  entire water
 column (personal communication 10/29/99, Ken Pearsall USGS, Albany New, York).

 During periods when a strong vertical gradient in TSS concentrations  existed, it is possible that
 the GE sampling approach may have resulted in a  low bias  in measured TSS concentration
 relative to TSS measurements obtained by the USGS sampling method. This potential is greatest
 during periods of  high  flow, due to  possible occurrence of  bed load.  This may also affect
 estimates of PCB loading because the GE samples were analyzed for both TSS and PCB.

 Relative differences in GE  and  USGS TSS concentrations were assessed several ways.  First,
 daily average GE and USGS TSS concentrations between 5/10/91 and  4/9/97 were paired on the
 basis of date, resulting in 30 daily average data pairs.  Scatter plots of the daily average pairs
 were developed for daily average flows above and below  10,000 cfs (Figure 6-11).  Based on
 observation of the TSS/flow correlation and in-river PCB data,  flow of approximately 10,000 to
 11,000 cfs is considered to be  an approximate threshold  above which resuspension becomes
 significant in the river.  Inspection of these  scatter plots suggests  that  GE  and USGS  TSS
 measurements may be biased relative to  each other, but in different directions at high and low
 flow.  At low flow,  GE TSS  measurements  appear to be  significantly greater than  USGS
 measurements. At high flow, the opposite may occur, at least for concentrations greater than  10
 mg/L.  These observations are  not  well supported  due to the limited  number of  data pairs
 available and uncertainty as to exact times of collection.  A  second  approach to investigating
differences in USGS  and GE data was to group the data by flow range and test for statistically
 significant differences in mean  concentrations in each flow  range.   Two-sample and paired
 sample statistical tests were  done at low flow (less than 10,000 cfs) and at high flow (greater than
 10,000 cfs).  The results of these tests tended to confirm what was observed in the above scatter
plots: the GE measurements were higher at low flow, while the USGS measurements were higher
at high flow.

The  GE  data represents a significant percentage  of  the total  available daily  average TSS
 measurements at Fort Edward  (40 percent) for use in  computing the upstream  TSS loading.
While comparison of USGS and  GE data suggest that these data may have biases relative  to each
other, the limited data pairs  available do not support discrimination between these datasets. The
 data were combined for use  in computing Fort Edward TSS loads (Section 6.5).
                                           76                       Limno-Tech, Inc.

-------
6.5.3 Methods for Estimating Solids Loads

Solids loading estimates were based on sediment rating curves developed for the upstream load
at Fort Edward, all tributary inputs and for long-term mainstem Hudson River TSS sampling
stations.  The mainstem Hudson River solids  load estimates downstream of Fort Edward were
used to develop a long-term solids balance for the river and for comparison to model output. The
solids rating curves relate observed TSS concentrations to flow and thus describe solids loading
as a function of flow. The general form of the rating curve is presented below (Equation 6.7)

                     TSS - aQb                                                     (6-7)

where:
       Q     =flow
       a,b   - fitting parameters

Using measured daily average TSS concentrations where available and concentrations estimated
from rating curves for days on which no measurements were taken, daily average TSS loads were
computed for the HUDTOX calibration period: January 1,  1977 through September 30, 1997.
The daily solids load time series computed for Fort Edward and the tributaries were input directly
into the  HUDTOX  model. The estimated in-river sediment  loads passing Stillwater and
Waterford were used as model calibration targets.

6.5.3.1 Mainstem Solids Loads

To develop suspended solids rating curves for the mainstem Hudson River sampling stations,
daily average TSS concentrations were plotted versus  daily average flow.  Suspended solids
concentrations are generally correlated with flow at Fort Edward, Stillwater and Waterford, with
a stronger dependence of solids concentration on flow observed at higher flows (Figure 6-12).
The relationship between solids concentration and flow is distinctly different at flows above
approximately 1.0 to 1.5  times the  average  flow at each location.   The  flow  at which the
relationship between TSS and  flow changes at each station is referred to as the flow cut-point.
Regression equations were developed for each station that describe the relationship between TSS
and flow above and below the flow cut-point.

A non-linear least squares regression approach was used to fit the data above and below the flow
cut-point.  An alternative approach considered was the Minimum Variance Unbiased Estimator
(MVUE) method of Cohn et al. (1989). Comparison of results from these two methods shows
that differences are small (Figure 6-13).  The non-linear least squares regression approach has
benefit of being applied  using commonly available software  rather  than requiring  special
computer code.

The approach taken consisted of two phases.  The first phase, using log-transformed data,
simultaneously developed linear regression equations and refined the specification of the  flow
cut-point so as to obtain a continuous function relating TSS concentrations to flow over the entire
flow range of interest. The second phase eliminated transformation bias by using non-linear least
squares regression of the un-transformed data; retaining the first  phase cut-point at the cost of a
                                           77                       Limno-Tech, Inc.

-------
discontinuity in the TSS prediction at the cut-point.  This process described above is presented
below in more detail.

    1.  An initial cut-point was selected based on visual inspection of the TSS - flow
       plots.

    2.  The (log-transformed) data above and below the flow cut-point were fit with a
       linear least squares regression equation of the form In(TSS) = A + B  * ln(Q),
       where A and B are the equation parameters.

    3.  The value of Q for which the low flow and high flow equations were equal at the
       cut-point was computed.

    4.  Step 2 was repeated with the value of Q computed in step 3.

    5.  Steps 2 through 4 were repeated until convergence was obtained.

    6.  The (un-transformed) data above and below the  cut-point were fit with a non-
       linear least squares regression equation of the form TSS = aQb, where a and b are
       the equation parameters. However, if the relationship between TSS and flow was
       not significant, the arithmetic average TSS was used.

The resulting rating curve equations for each station are presented below.  The equations fit data
below and above the cut-point, respectively.
Fort Edward    Qcul = 10,829 cfs   Q < Qcut : TSS = 1.767 x Q °08624

                                  Q > Qcut : TSS = 1.431E-8 x Q2101
Schuylerville   Qcu, = 3,866 cfs
Stillwater
Waterford
Qcut = 7,555 cfs
Q Qcul  : TSS = 0.0004238 x Q U3

Q < Qcu,  : TSS = 0.0122 x Q °6937

Q > Qcu,  : TSS = 4.555E-6 x Q 1.595
  cu, = 9,799 cfs     Q < Qcut : TSS = 0.06739 x Q °5237

                   Q > Qcut : TSS = 8.489E-9 x Q 2 213
(6-8a)

(6-8b)

 (6-9a)

 (6-9b)

(6-10a)

(6-10b)

(6-lla)

(6-llb)
Subsequent to development of rating curves based on all available data for the calibration period,
investigation of possible changes over time in the solids load at Fort Edward was pursued.  This
was undertaken primarily in response to comments received from NOAA personal at the June 16
Science and Technical Committee Meeting, Albany, NY.  Investigation of possible decreases in
Fort Edward solids loading  over the calibration period is prudent considering a number of
possible  factors, including:  1) washout and consolidation  of former  Fort  Edward  Dam
impoundment sediments in the river; and 2) stabilization of the remnant deposit areas.
                                           78
                                                     Limno-Tech, Inc.

-------
To initially assess whether the relationship between TSS and flow changed at Fort Edward over
the calibration period, rating curves were developed as described above for the first five years
(January 1,  1977 through December 31,  1982) and the last 5 years (January  1, 1993 through
September 30,  1997) of the calibration period.  Results indicated that there was a statistically
significant reduction in TSS concentrations at both high and low flow between these two periods.
This suggested it was necessary to account for the decrease in TSS loading in order to achieve the
best estimate at Fort Edward over the calibration period.  This required specification of time
intervals over which to represent the change in solids loads.

Reductions in solids erosion over time from bedded sediments and the exposed remnant deposit
areas were considered possible factors contributing to observed decreases over time in solids load
at Fort Edward. A review of site information related to erosion of the former Fort Edward Dam
sediments was used to estimate a reasonable time stratum. A  number of sources document high
erosion rates and solids loading to Thompson Island Pool following removal of the Fort Edward
Dam.  Various stabilization activities were conducted within the calibration  period that were
designed to reduce erosion of the former dam sediments.

Over the period July 1973 to  April 1976, following removal of the Fort Edward Dam in 1973,
approximately 1.0 million cubic yards of PCB laden sediments were  washed downstream  into
Thompson Island Pool (NUS  1984). It is likely that significant erosion continued to occur after
1976.  In 1978, areas of highly  contaminated river-bank sediment that were exposed following
the removal of the Fort Edward Dam were stabilized from a highly erodable state (Brown et al.
1988). The five discrete remnant sediment deposit areas upstream of Fort Edward identified by
NUS  (NUS  1984) were sources of sediment and PCBs until containment of the remnant deposit
sediments in the fall of 1990. Following containment activities in 1990, PCB  loading from the
remnant deposit sediments appears to be small, if any (O'Brien & Gere  1996).

To account for.changes in the TSS to flow relationship in specifying solids loads at Fort Edward,
1990  was considered a reasonable boundary for the two time strata.  While decreases in solids
loads also likely occurred  before  1990,  the stabilization activities  completed by GE  at the
remnant deposit areas in the fall of 1990 provides a logical time stratum for investigating changes
in solids loads.

The daily average TSS data at Fort Edward were grouped before and after Dec. 31,  1990  and
tested for statistical difference  at both high and low flow.   Results indicate that there is a
significant difference in  the solids rating curves that were developed following the procedure
presented above.  Resulting Fort Edward rating curves for the pre  and post  1990 periods are
presented below and in Figure 6-14, for flows below  and above the cut-points, respectively.

1977-1990     Qcut = 10,100 cfs   Q < Qcut : TSS = 0.0674Q0 5024                  (6-12a)

                                  Q > Qcu, : TSS = 9.505E-7Q1'701                  (6-12b)

1991-1997     Qcut = 12,100 cfs   Q < Qcut : TSS = 3.23                           (6-13a)

                                  Q>QCU, :TSS = 8.202E-11Q2592                (6-13b)
                                           79                       Limno-Tech, Inc.

-------
These equations were used to compute cumulative suspended sediment loading at Fort Edward
over the HUDTOX calibration period: January 1, 1977 to September 30, 1997. To illustrate the
estimated change in  loads  before and  after  1991, the average annual solids loading for each
period were  compared, showing 38%  percent  difference  for  the earlier  and later  periods,
respectively.  Part of this difference is attributable to differences in flow. The two largest flow
events in the calibration period (1977 and 1983) occurred before 1991.

Because  the time stratification  approximately coincides  with the time  that TSS data became
available from the GE studies, it is reasonable to consider the influence of data source on the
observed decrease in solids loading.   One way to  assess the influence of data source is to
investigate the justification for time stratification based the USGS data alone.  It is not practical
to attempt this with the GE data because GE data collection began in 1991.

Possible support for time stratification based only on the USGS data was investigated two ways.
First, the USGS data at Fort Edward was time (before and after Dec. 31, 1990) and flow stratified
(above and below  10,000 cfs) and the high and  low flow data groups were compared for each
period. This assesses whether there is a difference in mean concentrations at high and low flow
between  the  two periods,  although  differences  due to flow effects  are  not  accounted for.
Secondly, to account for the influence of flow on observed differences, TSS data from the two
time periods were paired on the  basis of flow. Typically, paired data had differences in flow of
less  than 1 percent.  TSS data  that could not be closely matched  on the basis of flow were
excluded from this paired comparison.  The second approach addresses the question of whether
the  average  difference in TSS  concentrations between the  two time periods  is significantly
different  from zero while removing the obscuring influence of flow.  The following statistical
tests were done using Systat 8.0 (Systat 8.0 Statistics, 1998. SPSS, Inc.):

       •      Two-sample t-tests with logarithmically transformed data;

       •      Mann-Whitney U tests;

       •      Paired samples t-tests with logarithmically transformed data; and,

       •      Wilcoxon Signed Ranks tests.

For both  low and high flows, all tests indicated that USGS TSS measurements for the later time
period decreased relative to the earlier time period. The differences were found to be substantial,
both in absolute terms and  in terms of statistical significance (i.e. p<0.05 in all cases). These
results show  that the use of time  stratification in computing the Fort Edward solids  load is
supported, at both high and low flows, regardless of whether or not the GE data are  included.
Inclusion of  the  GE data  in developing  the time  stratified rating curves tends to  increase
computed TSS loads  at low flow and decrease computed TSS loads at high flow.  Use of the GE
data was considered appropriate considering the large number of GE data available, uncertainty
as to potential differences  between the two  datasets, and the observed  changes in the USGS
TSS/flow relationships independent of the GE data.  In spite of observed differences between GE
                                           80                       Limno-Tech, Inc.

-------
 and USGS data, the use of time stratification in the rating curve is supported based on the finding
 that both the USGS dataset and the combined USGS-GE dataset show statistically significant
 decreases in concentration over time.

 The time stratified rating curves developed based on all the available data (Equations 6-12 and 6-
 13) were used to compute TSS loads at Fort Edward for the model calibration. To develop the
 daily loading time series for the model calibration,  measured daily average TSS concentrations
 were used where available, and TSS concentrations were computed using the rating curves for
 days upon which TSS was not measured.

 In order to develop a long-term solids balance for the Upper Hudson River, cumulative solids
 loads at Stillwater and Waterford were computed  in the same manner as described for Fort
 Edward. While decreases in TSS concentrations over time were also observed at Stillwater and
 Waterford at low flow, no statistically significant differences were observed at high flows.  Based
 on this observation, the use of time-stratified ratings curves was not well supported at Stillwater
 and Waterford, and was not used in the calculation of solids loads at these locations.  Estimates
 of in-river solids  loads  at Stillwater and  Waterford were  used in  model calibration and  in
 developing a long-term solids balance for the river (Section 6.5).

 The solids rating curves  at Fort Edward (Equations 6-8a and 6-8b), Stillwater (Equation 6-10a
 and 6-10b), and Waterford (Equation 6-1 la and 6-1 Ib) were used to estimate TSS concentration
 for days where measurements were not available. This allowed estimates of daily TSS loads at
 these locations for the entire calibration period (1977-1997).

 6.5.3.2  Tributary Solids Loads

 A major obstacle to estimating tributary solids loads was that  available solids data were  very
 limited.  This  factor,  combined with uncertainty  in estimated tributary flows (Section  6.4)
 resulted  in  poorly constrained  estimates  of  tributary  solids  loading.  There  is  significant
 uncertainty in the resulting tributary load estimates,  especially downstream of Thompson Island
 Pool.

The general approach used for estimating tributary TSS loads was similar to that adopted for the
Fort Edward load. TSS  rating curves were developed to relate  TSS  concentrations to  flow.
Similar to  the pattern  observed  at the mainstem stations, tributary  solids concentrations were
positively correlated with flow and the tributary rating curves  generally exhibited a cut-point
above which the slope of the relationship  increases (Figure 6-15).  The average flow of each
tributary was  observed  to reasonably approximate the  flow cut-point  above which  TSS
concentrations  increase  significantly  in   each tributary.    Below  the average  flow,  TSS
concentrations were generally not  flow dependent or only weakly flow dependent,  so average
TSS concentrations were  specified below average flow.

Many  tributaries  were  not  monitored  for  suspended  solids  concentration.   Unmonitored
tributaries represent 29 percent of the drainage area between Fort Edward  and Waterford. For
unmonitored tributaries,  the TSS  rating curves developed for  the  monitored tributaries were
applied.   Each unmonitored  tributary was  matched with  a  monitored  tributary  based on
                                           81                       Limno-Tech, Inc.

-------
consideration of land use distribution, watershed size, topography and location (Table 6-9). The
rating curve for Moses Kill was used to estimate solids loads for all unmonitored drainage areas
between Fort Edward and  Stillwater.   The  Hoosic River  rating curve  was  applied  for
unmonitored drainages between Stillwater and Waterford.  Loads for each of'the unmonitored
tributaries were scaled to account for differences in drainage area from the reference tributary.

In two  cases, unmonitored drainage areas were reduced  in the estimate  of  solids loading to
account for sediment trapping. The unmonitored drainage area between  Fort Edward and TI
Dam includes the Champlain Canal, which is fed by Bond Creek and water diverted from the
Hudson  upstream   of Fort  Edward  (Art Murphy, NYS  Canal  Corporation,  personal
communication).  The canal is highly  regulated and during high flow is allowed to overflow,
delivering water to the Hudson via overland flow.  This likely results in solids retention in the
canal. The effective drainage area of the Champlain canal  was assumed to be  8 mi2, 26 percent
of the canal's watershed area (31 mi2). Fish Creek drains Saratoga Lake, which receives tributary
runoff from the bulk of the Fish Creek watershed. It was assumed that 80 percent of the tributary
solids load from Fish Creek watershed  was retained in Saratoga Lake. Therefore, 90 mi2 of the
245 mi2 watershed was considered in computing tributary solids loads.

The flow-dependent rating curve coefficients developed for each tributary are presented in Table
6-10. As discussed above, constant concentrations were specified below the average flow.

Resulting tributary solids loads from application  of the tributary TSS rating curves in the manner
described  above were  evaluated in the context of  a solids balance for the  mainstem Upper
Hudson River. The  period chosen for the solids balance was Jan. 1,  1977  to Jun. 30, 1992
because all mainstem river flow gages were operational during this time.  Limiting the mass
balance to this period was intended to reduce error associated with flows reported as estimates by
USGS after 1992.

A comparison of annual average mainstem solids yields for the drainage area at Stillwater and
Waterford to tributary solids yield (Table 6-11) shows that  mainstem yields are about a factor of
two  larger than the estimated tributary yields.  Table  6-12 presents the mainstem solids load
increments between  Fort Edward, Stillwater and Waterford.   Estimated  solids loads  passing
Stillwater and Waterford are 1.7 and 2.0 times larger than the sum of Fort Edward and upstream
tributary solids loads.  Assuming the mainstem in-river solids load estimates are accurate, the
observed increases in in-river solids loads must be explained by either:

       •      additional external loads, from tributary inputs or other sources;

       •      internal production of solids from primary production; or,

       •      net erosion of the sediment bed between Fort Edward and Waterford.

External sources other than tributary load inputs are assumed negligible and estimates of possible
contributions of solids  from primary  production were insignificant.   Therefore, if estimated
tributary solids loads  are assumed accurate, this implies that  resuspension  accounts for the
                                           82                        Limno-Tech, Inc.

-------
observed increases and that the Upper Hudson River is on average net erosional between Fort
Edward and Waterford.

Considering that the Upper Hudson is an impounded system with six dams over the 40 miles
between Fort Edward and Waterford, it was considered unlikely that the river is net erosional
over  this  reach.   Typically,  river  impoundments  experience  net  deposition.    Required
navigational dredging over the  extent of the Upper  Hudson also  suggests  that the river is
depositional, however,  erosion-derived  sediment from  other  areas  of the  river could  be
responsible for required navigational dredging in  the main channel.  Based on the  assumption
that the river was net depositional on a reach basis, it was concluded that tributary loads were
likely underestimated and required upward adjustment to achieve a long-term solids balance with
observed loads passing Stillwater and Waterford.

6.5.3.3  Development of Long-term Solids Balance

Development of a long-term solids balance for the Upper Hudson is possible for reaches to which
upstream and downstream in-river solids loads are known. For each reach, the sum of upstream
loads, tributary loads and internal sources (resuspension or primary production) must equal the
estimated  in-river load.   As discussed above, primary production contributions (from algal
growth) and solids  loads from external sources, excluding tributary inputs,  (such  as possible
point source or other direct loads to the water column) were assumed insignificant in developing
the solids balance. Thus, the solids balance required that upstream loads, tributary loads, and net
solids loads from the sediment sum to the estimated in-river loads leaving each reach.  Unless
solids loads to the system being modeled are equal to solids transport out of the system, internal
solids dynamics are unconstrained in the model. For example, if upstream and tributary loads are
estimated without consideration of net solids exchange with the sediment bed, calibration of the
model to observed in-river loads may result in unrealistic predictions of bed behavior.

In adjusting tributary solids loads to achieve a long-term solids balance, all impounded reaches of
the Upper Hudson river from Fort Edward to Waterford were assumed to be net depositional over
decadal time scales, even if there might be localized areas within reaches that are net erosional.
Estimated solids loads  passing  Fort Edward, Stillwater  and Waterford  were assumed to  be
accurate.

Depositional loads to each reach were estimated and tributary  TSS loads were increased between
Thompson  Island Dam and Waterford to equal the sum of observed loads at Waterford and the
estimated depositional load. Tributary solids  loads to Thompson Island Pool from Snook Kill,
Moses Kill and direct drainage inputs were not adjusted. These tributary load estimates are based
on  sufficient TSS and flow data such that their solids loads are reasonably well known. The
Mohawk River suspended solids loads were also not adjusted because insufficient data exist at
Federal Dam to evaluate the solids balance between Waterford and Federal Dam.

A measure of  the  depositional  load  is  the  long-term average  sediment  burial velocity.
Measurements of burial velocity were obtained by USEPA  using radionuclide sediment core
dating at 5 locations between  Federal Dam and Fort Edward (USEPA, 1997).   Two of these
locations are in TIP.  These measured burial velocities represent long-term average deposition at
                                           83                       Limno-Tech, Inc.

-------
the core  sites, however,  these  locations were generally  positioned in low  energy,  highly
depositional areas and  are  not  considered  representative of reach-wide average  conditions.
Therefore, estimation of sediment burial  rates or reach-specific  solids trapping efficiency was
necessary.  Solids trapping efficiency can be used to compute the average burial rate based on
sediment density as shown later below.

Calculations of solids burial rates  were available from a sediment transport model (SEDZL)
developed for the Upper Hudson River by General Electric Company (GE)  contractors  (QEA,
1999).  Flow and solids loading  inputs to SEDZL were developed through discussion with the
USEPA in the development of the HUDTOX model.  As a result, the  SEDZL inputs are nearly
identical to the inputs described in this report.  Initial sediment transport simulations conducted
by GE were used to compute burial rates (and solids trapping efficiency) by reach for  use in
computing the tributary solids loads.  Final results indicated that  initial input assumptions were
reasonable.  The SEDZL  simulation period was  nearly the same as the HUDTOX 21-year
historical calibration period.   This  sediment  transport model was based on the same cohesive
sediment resuspension formulations and site-specific data used to develop the Depth of Scour
Model (DOSM) presented  in Chapter 4. This model also used theoretical formulations for non-
cohesive sediment armoring based in part on the non-cohesive sediment scour calculations in the
DOSM. Earlier versions of SEDZL have been successfully applied on other similar river systems
(e.g. Ziegler and Nisbet, 1994, Pawtuxet River, R.I.; Galiani et al. 1996, Buffalo River,  N.Y.).
Details  of the SEDZL sediment transport model  development  and  application to  the  Upper
Hudson River are provided elsewhere (QEA, 1999).

Available SEDZL calibration results suggest  the  GE sediment transport model achieved
reasonable agreement with estimated  burial rates at the USEPA  high-resolution sediment core
sites (QEA, 1999). Results were within a factor of two with measured solids burial rates from all
but one of the high resolution sediment cores.  Agreement was  within a factor of five for the
remaining sediment core.  The burial rate results from  this sediment transport model contain
uncertainty, however, due  to large uncertainty in model  inputs, especially tributary solids loads
downstream of Thompson  Island Pool. These uncertainties affect  long-term solids burial rates in
both cohesive and non-cohesive sediment  areas. These limitations notwithstanding, burial rates
from the SEDZL  sediment transport model were considered reasonable estimates.  Downstream
of Thompson  Island Pool, these estimates are affected by  high uncertainty as would be any
estimates that could be made, due to the limitations of the tributary flow and solids data.

Based on the above considerations,  the reach-specific estimates of TSS trapping efficiency from
the GE  sediment  transport model presented in Table 6-13 were  used  to develop the long-term
solids balance for Jan. 1, 1977 through 1997. The trapping efficiency estimates by  reach were
area-weighted  to  determine trapping  efficiencies for TI Dam to Stillwater (8.47 percent) and
Stillwater to Waterford (3.66  percent) reaches. Solids depositional loads to the sediments in each
reach were computed from  the trapping  efficiency and used to back-calculate  total tributary
loading to the each reach. The trapping efficiency is generally related to the depositional load as
shown in Equation 6-14.
                                           84                       Limno-Tech, Inc.

-------
                          (upstream load      ^
                        +             (kg/a)  x trapping efficiency (%)                (6-14)
                          tributary load       \


In order to determine  depositional loads to each reach, calculation of tributary loads and
upstream loads was done in succession for Thompson Island Pool, Thompson Island  Dam to
Stillwater,  and Stillwater  to  Waterford.   In Thompson  Island  Pool, the available  data for
tributaries (Snook Kill and Moses Kill) was sufficient to compute loads directly from data-based
rating curves.  A long-term data-based  solids balance could not be conducted explicitly for
Thompson Island Pool because Thompson Island Dam sampling began in 1991.  Based on the
successful calibration of the SEDZL model to available data at the dam (presented by QEA,
1999), the Thompson Island Pool trapping efficiency estimate is assumed reasonably accurate for
use in computing the Thompson Island Dam solids load. This was estimated by multiplying the
TIP  trapping efficiency estimate (8.8 percent) and the sum of tributary and upstream loads
(Equation 6-15).  The incremental  load computed for the Thompson  Island Dam to Stillwater
reach (Equation 6-16) was apportioned to each tributary based on the percent of total tributary
watershed area, excluding watershed area draining to the upstream portion of river (Table 6-2), as
shown below.

Thompson Island Dam to Stillwater solid balance:

LTID  = (1-.088)  x(LFE  +  LSnook + LMmes + LDD )                                   (6-15)
                  I   DA,    ,
      A^fc.r/D.s,,7, XI —	I                                                     (6-17)
where:
       LFE, LTID, Ls,ni       = TSS load at Fort Edward, TI Dam, Stillwater
       LSnook, LMoses, LDD    = tributary TSS load from Snook, Moses, and direct drainage
       ALtrib, TiD-stiti         = incremental load from tributaries between TI Dam and Stillwater
       LI                   = total load apportioned to tributary / in the solids balance
       DA,                 = drainage area of tributary /
       DAriD-stnt            = total incremental drainage area from TI Dam to Stillwater
       %trap              = solids trapping efficiency

In order for rating curve tributary load estimates to equal the apportioned load to each tributary,
an upward adjustment in the data based curves was required.  The largest uncertainty in tributary
TSS load estimates was assumed to be in  the high-flow portion of the rating curve.  Therefore,
the tributary rating curve b coefficient (Equation 6-18) was adjusted iteratively until the resulting
load equaled the specific value  of L, computed  for each tributary.  The  constant low-flow
concentrations were not adjusted.
                                           85                       Limno-Tech, Inc.

-------
 |Q>Q, TSS =  axQ"
    Q< Q, TSS  = c


where:

              c = constant low flow TSS concentration based on data average
Tributary  loads between Stillwater and Waterford were adjusted in the same  fashion.  The
resulting exponent in the rating curve equation, b, for each tributary ranged from  1.1497  for the
Fish Creek to 2.236 for Deep Kill (Table 6-10).

Resulting increases in  tributary solids loads to achieve the long-term  solids  balance were
significant.  Tributary loads between Thompson Island Dam and Stillwater were increased by a
factor of 2.46.  Between Stillwater and Waterford, tributary loads were increased by a factor of
1.91.  The adjusted rating curves required to achieve the solids balance are shown with the data
and data-based rating curves to demonstrate the required adjustments at high flow (Figure 6-15).
While the adjusted Hoosic River rating curve looks reasonable, the adjustment of the Batten Kill
rating curve does not agree well with the limited available  data.  However, considering  the
limited data available and the fact that the Batten Kill flow is estimated based on a much smaller
tributary,  there is considerable uncertainty in the TSS  versus flow relationship observed  for
Batten Kill.  Flow phasing errors (timing of peak flows) based on relating Batten  Kill flow to
Kayaderosseras Creek flow may have significantly affected the Batten Kill rating curve.

6.5.4  Results

A long-term solids balance was developed for the period Jan. 1, 1977 through September  30,
1997. The Fort Edward solids  rating curve and the tributary rating curves adjusted to achieve the
solids balance were applied to develop daily time series inputs to the HUDTOX model  for the
calibration period. Annual average mainstem and tributary solids loads for the calibration period
are presented in Figure 6-16.

Results show that annual average sediment load increases by a  factor of 2.8 between Fort Edward
and Stillwater and by a factor of 5.7 between Fort Edward and  Waterford over the calibration
period. In comparison, average flow increases by a factor of  only 1.2 and 1.5 percent between
these locations, respectively.  Watershed TSS yield increases by a factor of 2.1 and 3.5 moving
downstream from Fort Edward (10.7 MT/yr-mi2) to Stillwater (22.2 MT/yr-mi2) and Waterford
(37.4 MT/yr-mi2), respectively.

To illustrate relative sediment  load contributions, percent of annual average solids loads for  low
and high  flow periods, "non-event" and  "event", respectively,  are plotted in sequence from
upstream to downstream (Figure 6-17).  This plot shows that high and low flow tributary solids
contributions  are about the  same  and  also illustrates  the importance  of tributary loads
downstream of Thompson Island Dam.  The Batten Kill load is about the same magnitude as the
Fort Edward load and the Hoosic River load is approximately twice as  large  as the  Fort Edward
load.  Without accounting for depositional losses, at Stillwater and Waterford, only 36  and 17
percent, respectively,  of the external solids load entering the river is attributed to Ft. Edward.
                                            86                        Limno-Tech, Inc.

-------
Only 5 percent of the suspended solids load at Federal Dam enters the system at Fort Edward,
due to the large contribution from the Mohawk River.

The  distribution of mainstem solids loads over the range of observed  flows was analyzed to
understand the relative importance of high and low flow solids transport (Figure 6-18).  At Fort
Edward,  Stillwater and Waterford approximately 55, 50,  and 70 percent of the TSS  transport
occurs below two times the average flow (Q/Qavg = 2.0). At Fort Edward, two times the average
flow, 10,496 cfs, is approximately equal to the high/low flow strata used to specify the TSS
rating curves,  10,000 cfs.

6.5.5 Summary of Solids Load Estimates

Mainstem solids load estimates were developed for Fort Edward, Stillwater and Waterford using
rating curves developed using non-linear least squares fitting.  At Fort Edward, time stratification
in the load estimates was based on observed changes in the TSS to flow relationship between the
1977-1990 and 1991-1997 periods. Annual average sediment loads for the 1991-1997 period are
40% percent lower than the loads  for the 1977-1990 period at Fort Edward.  The time-stratified
solids rating curves for the 1991-1997 period are considered the best estimates of future TSS
loading at Fort Edward (Chapter 8).

Tributary loads  were  initially computed using rating  curves based on the limited available
tributary  data.  These results required adjustment for tributaries between Stillwater and Waterford
to achieve  a  long-term  solids balance  from  1977  to  1992  consistent  with  the  assumed
depositional character of the Upper Hudson River.  Estimates of solids trapping efficiency by
reach developed by QEA  (1999) using the SEDZL sediment transport model were  used  to
compute  tributary loads between these locations.  The data based tributary rating curves were
scaled up at  high flow  to achieve the necessary TSS  loading increase.  Results produced tributary
solids yields in reasonable agreement with literature ranges (Table 6-14), although the adjusted
rating curves did not agree with the limited  data in  all  cases (Figure 6-15).  Final tributary load
estimates between Thompson Island Pool and Waterford are considered very uncertain.

The solids balance achieved for the 1977-1992 period gave good results for the entire calibration
period when compared to estimated solids fluxes  passing Stillwater and Waterford.   Results
show that mainstem solids loads increase by a factor of 5.7 from Fort Edward to Waterford.  This
illustrates the  significance of tributary loads, which are very uncertain due to limited tributary
solids and flow data.   The large  degree  of uncertainty in these estimates imparts significant
uncertainty to the model calibration below Thompson Island Dam (Chapter 7).

6.6 MAINSTEM AND TRIBUTARY PCB  LOADS

6.6.1  Overview

Application  of the HUDTOX model requires specification of all external flow inputs, solids
loads and PCB loads.  Just as flow (Section 6.4) and solids loading (Section 6.5) time series were
developed for  the calibration period, upstream and tributary loading time series were developed
for the seven  PCB state  variables: total PCB, Tri+, BZ#4,  BZ#28, BZ#52, BZ#90&101, and
                                           87                        Limno-Tech, Inc.

-------
BZ#138.  Tri+ load estimates were developed for the long-term historical calibration period,
January  1, 1977  through  September 30, 1997.  Load  estimates for total PCB and  the five
congener state variables were estimated for the short-term hindcast period: April 1, 1991 through
September 30, 1997.  To aid in model calibration, estimates of in-river fluxes  were also
developed for Tri+ at Schuylerville, from 1977-1992; Stillwater and Waterford from  1977 to
1997; and, for Tri+ and Total PCB at Thompson Island Dam from 1991-1997. In developing the
Thompson Island Dam PCB load estimates, a correction was applied to measurements  taken at
the west shore station to correct for observed biases in these data (QEA, 1998b).  The in-river
load estimates were calculated solely for comparison  to  model output and do not represent
additional loads to the model.

6.6.2 PCB Data

6.6.2.1 Data Availability for Estimating PCB Loads

Mainstem Upper  Hudson River PCB data were available  from the  USGS (1977-present), GE
(1991-present) and  from the USEPA  1993 Phase  2 investigation  (1993).  While the USGS
dataset represents an extensive historical record of PCB concentrations, due to analytical and
sampling limitations these data can only be used to develop approximate estimates of water
column PCB load  (USEPA,  1997).   As discussed in Section 6.3.3.2,  there is uncertainty
associated with  the  USGS PCB  quantitation and the translation  of these  quantitations  to
estimates of the long-term calibration state variable, Tri+.

The most extensively  sampled mainstem  stations in the model  domain are Fort Edward,
Thompson Island Dam, Schuylerville, Stillwater and  Waterford.   Exact sample  collection
locations varied at these stations, especially at Fort Edward.  Samples collected in the vicinity of
Fort Edward by the various USGS, GE and USEPA Phase 2 data collection efforts were grouped
to represent Fort Edward concentrations. The same was done for the other stations.

The data available from each source at the primary mainstem sampling stations are summarized
for  Tri+ by year in Table 6-15 and for total PCB and congeners in Table 6-17.  Significantly
fewer data are available for tributaries, with only Batten Kill, Hoosic River and the Mohawk
River actually being sampled for PCB (Table 6-16).  Additionally,  no tributary PCB data are
available prior to 1991. Figure 6-19 presents the  location of the long-term  PCB monitoring
stations on the mainstem Upper Hudson River and the tributary sampling stations within the
study area.

The long-term combined dataset from USGS,  GE and USEPA represents good coverage of the
high and low  flow regimes.  Figure 6-20 shows the distribution of data over the range of sampled
flows.

The USGS PCB  data are reported as Aroclor quantitations or the sum of Aroclor quantitations
and  neither individual congener nor complete unbiased  total PCB concentrations are available
from these data.  Thus the  total PCB and congener data are limited to the GE and USEPA data
collection periods, which began in 1991. Congener data are available for all five congener state
variables, however, at Fort Edward BZ#4 and 138 were  quantified in only about half of the
                                           88                       Limno-Tech, Inc.

-------
samples in which total PCB was quantified.   BZ#28, 52, and 101+90 were quantified in nearly
all of the samples in which total PCB was quantified.

While continuous sampling was conducted at Fort Edward and Thompson Island Dam from 1991
through  1997, GE  and USEPA conducted little or no sampling at  stations downstream of
Thompson Island Dam  from 1993  to 1997.   As a  result,  a continuous  record of PCB
concentrations over the calibration period (1977-1997) is only available for Tri+ at Fort Edward,
Stillwater, and Waterford.  Sampling at Schuylerville by USGS ended in 1992.

6.6.2.2 Thompson Island Dam West Shore Station Bias Correction

As  summarized in QEA (1998b),  an apparent sampling  bias was discovered in fall of  1997 in
PCB measurements from the routine monitoring station located on the west shore of Thompson
Island Dam. A significant fraction of the GE and USEPA data at Thompson Island Dam were
collected from stations  on the west shore. The samples  collected at this station are not always
representative  of the average PCB concentration leaving the pool, hence  the term "bias", and
must be corrected for use in mass balance analysis.

The bias appears  to be related to contribution of PCB from nearshore contaminated  sediments
under conditions of incomplete lateral mixing. The magnitude of the bias, in terms of percent
difference between the west shore and center channel locations, is related to flow conditions and
upstream PCB concentration.  During high  flow periods  sufficient  lateral mixing  occurs to
prevent any significant lateral gradients at the dam. During periods of high PCB loading at Fort
Edward, the relative contribution of the nearshore hotspots is smaller.

After discovery of the  west  shore station bias, GE modified its monitoring program to better
quantify the magnitude  of the bias (O'Brien & Gere, 1998). The modified program included
collection of samples further upstream and downstream of Thompson Island Dam, in  the center
channel of the river, and on lateral transects. These data allowed assessment of the  relative
degree of bias over different flow and upstream loading conditions.  An analysis conducted by
USEPA,  (USEPA,  1999b) indicated  that the ratio  of west shore  to  center channel Tri+
concentrations approached unity for concentrations and  flows at Fort  Edward greater  than 15
ng/L and 4,000 cfs, respectively.  For flows less than 4,000 cfs and concentrations less  than 15
ng/L at Fort Edward, a significant high bias exists for the west shore concentrations relative to
the center channel. Segregating the observed ratios by these criteria produces the results in Table
6-18, which also presents results for total PCB based on an identical analysis. These values were
used to "bias-correct" the west shore observations to better represent mean concentrations leaving
Thompson  Island Pool,  based on  Fort Edward concentration and flow conditions.  The bias-
corrected west shore concentrations  were used when  center channel observations  were not
available to compute PCB loads at  Thompson Island Dam and for comparison to model output.

6.6.2.3 Data Development for Computing PCB Loads

The combined USGS,  GE, and  USEPA water column PCB  datasets were reduced to daily
average values for  estimating daily average  PCB  loads.  On numerous occasions, multiple
samples were  collected at these  locations on the  same day, especially  at Fort Edward and
                                          89                       Limno-Tech, Inc.

-------
Thompson Island Dam.  Daily average concentrations were computed for days on which multiple
measurements were reported.  In computing daily average concentrations, the Phase 2 flow-
average concentrations (which are from  15-day composite samples) were only used if no discrete
measurements were available from GE or USGS.

Exact sample collection points at the mainstem  sampling stations varied between and within
agencies.  Data from the various sample collection points at the primary sampling stations were
combined to provide a record of concentrations at each location. At Fort Edward, samples were
collected from the east and west channel of Rogers Island.  Where same day measurements were
taken in each channel, these were averaged.  Otherwise, data from east or west channel, included
with data from various other studies  in the direct vicinity of the northern tip of Rogers Island
were included in estimating PCB loads at Fort Edward.

PCB concentrations reported as non-detect were assigned a value of one-half the detection limit
concentration. Detection limits vary  among datasets. Although the USGS laboratory reports a
theoretical quantitation limit of 0.01 |ig/L through  1983, the practical quantitation limit  was often
considered to be 0.1 |ig/L because of the small size of the water samples (Bopp et al., 1985).
With water year 1984, the practical quantitation limit was lowered to 0.01 ^.g/L, however, the
data were often reported as if they adhere to the  previous quantitation limit through  1984 and
1985.  In  1986, the quantitation limit began to be consistently reported as 0.01 |lg/L. For the
purposes of this project, USGS data  in the printed Water Resources Data, New York and the
USGS/Albany NWIS database were cross-checked to recover original quantitations at the 0.01
ug/L theoretical quantitation limit where possible. For the majority of the GE data, the detection
limit was  reported as  11 ng/L.  The Phase 2 results include reported  detection limits for non-
detect values and an adjusted value for non-detects based on a treatment procedure  (USEPA,
1989) for non-detect values put forth by EPA.

6.6.2.4 Overview

Calibration of the HUDTOX model to daily average PCB concentrations  required specification
of daily average  PCB loads at Fort  Edward  and from tributaries.  Thus, it was  necessary to
develop estimates of daily average loads at Fort Edward and for the tributaries for input to
HUDTOX over the 21-year historical  calibration period. While estimates of daily average PCB
load passing  the downstream stations were used for comparison to model output, these  were
developed on a daily basis for consistency with the Fort Edward load estimation and to develop
estimates for the entire 21-year period. Estimates  of annual PCB load at the long-term sampling
stations are presented  in the DEIR for part of the historical calibration period  (USEPA, 1997).
The annual load  results computed  from the sum of estimated daily loads (Section  6.6.3) are
compared to the DEIR estimates.

In order to develop loads for Tri+, it was necessary to estimate concentrations for long  periods of
time between  1977 and  1991  that contained very few measurements.   The estimated loads for
these periods have high uncertainty.  Sampling frequency was sufficiently high from 1991-1997
at Fort Edward and Thompson Island  Dam that load estimates for the period 1991-1997 are
considered more accurate than estimates for 1977-1991, with the  exception of the period
following the collapse of the Allen Mill gate structure at the Hudson Falls plant site.  This event
                                           90                       Limno-Tech, Inc.

-------
led to episodic elevated PCB loading in late 1991 and early 1992 that was probably not fully
captured by routine sampling.

Loads of total PCB and the five congener state variables were also determined at Fort Edward for
the short-term hindcast period (Jan. 1, 1991 to Sept. 30, 1997).  In cases where congeners were
not quantified while total PCB was quantified, congener concentrations were estimated based on
their average observed mass percent in total PCB measurements.

This section presents the development of the mainstem Upper Hudson River load estimates for
Tri+ over the period January, 1 1977 to September 30, 1997 and for total PCB and congeners
BZ#4, BZ#28, BZ#52,  BZ#90&101 and BZ#138 over the period January 1, 1991 to September
30, 1997.

6.6.2.5 Mainstem Tri+ Loads 1977-1997

In-river PCB loads were estimated at the primary long-term USGS monitoring locations, most
importantly, at Fort Edward, the upstream boundary of the HUDTOX model.   Specification of
PCB  loading at Fort Edward  was done on a daily average basis, consistent with the  input
frequency of flows and  solids loading.

Estimates of annual historical PCB loads at  Fort Edward, Schuylerville, Stillwater and Waterford
based on the USGS data are presented in the DEER (USEPA,  1997).  These estimates were based
on  application  of two  methods:  the ratio estimator  by Cochran  (1977), and  the averaging
estimator presented by Dolan et al. (1981).  An overview of these and other methods is presented
by Preston et al. (1991). While these methods are suitable for estimating annual loads, estimates
of daily average PCB load were sought for simulating PCB dynamics in HUDTOX on daily time
scales.  The HUDTOX calibration includes comparison to daily average PCB  concentration
measurements.. To develop  a method for estimating daily average PCB loads, flow-dependent
regression relationships were explored, as  was  linear interpolation of measured concentrations
and use of seasonal average concentrations by year.

Regression methods were eliminated because no significant relationships were observed among
PCB  concentration, flow and  suspended solids concentration.   Relationships between  these
parameters at Fort Edward were explored for 2-year intervals to reduce confounding effects of
long-term reductions in concentrations  (e.g.  Figures 6-21 and 6-22).  Elevated PCB loads do
appear to be  partially  correlated  with flow and  TSS concentrations, however, significant
variability in the correlations is observed at high and low-flow conditions. These observations
did not support use of flow-based regression methods.

Based on exclusion of regression  methods, a combination of linear interpolation of measured
concentrations between sampling dates  and use of seasonal average concentrations by year was
selected. The appropriateness of either of these individual methods depends on data availability.
During periods of low data frequency, linear interpolation has a significant potential for bias due
to the presence of high or low measurements that have biases relative to mean concentrations.
During high sampling frequency, linear interpolation may more accurately describe daily  loads
when concentrations are changing as a result of seasonal effects or upstream source activity.
                                           91                       Limno-Tech, Inc.

-------
 Prior to 1991, a combination of seasonal average concentrations by year and linear interpolation
 was used to estimate daily PCB concentrations at Fort Edward.  Beginning in  1991, GE began
 regular monitoring of PCBs at Fort Edward and data availability over the period 1991 to 1997
 was considered sufficient to support linear interpolation of measured concentrations over time for
 all PCB forms modeled.

 During the earlier  historical period, large  data gaps exist and data collection was  sparse,
 especially from 1977 to 1984.  Due to the limited amount of data available during various times
 in the calibration period,  seasonal  average concentrations  were used in lieu  of the  linear
 interpolation approach.  Seasonal average concentrations were  computed during each year and
 applied in the respective individual years. Periods of application of each approach were specified
 based on inspection of the PCB concentration time series at Fort Edward.  Figure 6-23 presents
 the daily average measured and estimated PCB concentrations at Fort Edward. Inspection of this
 figure reveals the time periods selected for application of each  method. The black  line on this
 figure shows estimated concentrations based on the data, shown as symbols.

 One complication in the use of linear interpolation is the apparent occurrence of random "pulse"
 loading events of PCBs at Fort Edward, identified through inspection of the Ft.  Edward daily
 average concentrations.  These pulses occasionally occur in conjunction with high flow events.
 Two such examples are presented in Figure 6-24.  On May 9, 1994 a concentration of 130 ng/L
 was  observed at Fort Edward and concentrations measured 3 days before and 3 days after this
 measurement both showed concentrations less than 15 ng/L. On April 25,  1983 a concentration
 of 900 ng/L was observed three days following and two days preceding measurements less than
 20 ng/L.

 The  occurrence of these pulse load events appears only partially correlated with flow. Due to
 very high concentrations, however, these events can contribute large mass  loading  of PCBs to
Thompson Island Pool. The use of linear  interpolation during periods  of infrequent sampling
 sometimes exaggerated the apparent contribution of pulse loads that were characterized by only a
 single or very few data points.  Interpolation in these situations caused estimated  loads to be
 strongly affected by individual high concentration measurements for long periods of time prior to
 and following the measurements. This was considered unreasonable based on inspection of the
 high frequency data collected by GE beginning in 1991. These data suggest that these pulse load
events are of short duration, on the order of days rather than weeks. Based on these observations,
the pulse loading  events were assigned a duration of 6 days,  assuming that the measured value
captured  the  peak concentration.   Seasonal  average values were then applied for  the periods
before and after these events to the preceding and following measurements.

Considering the low sampling  frequency,  it is very likely that many pulse load events  were
 missed, which introduces  large uncertainty into  the PCB load  estimates, considering the  large
 magnitude of observed pulse loads. It is noteworthy that a single measured pulse load in  1992
 was  responsible for 19 percent of  the estimated total  PCB  load  in that  year  alone.   The
 uncertainty due to pulse loads is further exacerbated in the early historical period by  the fact that
 sampling frequency was lowest during the period when PCB loads were at their highest levels.
                                           92                       Limno-Tech, Inc.

-------
Estimates of daily average in-river Tri+ loads at Thompson Island Dam, Schuylerville, Stillwater
and Waterford were computed using the same approach as  that implemented for Fort Edward.
These loads were estimated for comparison to model output.  Periods over which interpolation
was used varied among stations due to variations in sampling frequency among stations.

6.6.2.6 Tributary Tri+ Loads 1977-1997

Due to extremely limited data, tributary PCB loads were estimated in a different manner from
mainstem loads. For the monitored tributaries, Batten Kill, Hoosic River and Mohawk River, the
average PCB concentration was calculated and the assumption was made that this concentration
remained  constant  for the entire study period  (Table  6-19).   Measured concentrations were
substituted when available.  Because the three monitored tributaries were also the only tributaries
with known PCB  dischargers, it was assumed that these tributaries would  have higher PCB
concentrations  than the other tributaries in the study  area.  The Tri+ concentrations in  the
unmonitored tributaries were assumed to equal the lowest recorded Tri+ concentration from the
three monitored tributaries, 0.17 ng/1.

These  values  were assumed  to represent  background  concentrations for the unmonitored
tributaries. It is possible that historical tributary PCB concentrations were higher, however, the
relative contribution of tributary PCB loads compared to the upstream PCB load at Fort Edward
is small (less than 5 percent) and has negligible impact on the HUDTOX model predictions.

6.6.2.7 Tri+ Load  Results 1977-1997

To evaluate results, total annual Tri+ loads estimated for each mainstem  station are compared to
annual loads presented in the DEIR by USEPA (1997) (Table  6-20, Figure 6-25). The  DEER
annual load estimates are based on application of a flow-stratified version  of the ratio estimator
developed  by Cochran  (1977).  This comparison  recognizes  that comparison  to  the  DEIR
estimates  is affected  for some periods by use  of different  Tri+ concentration and flow data.
Nonetheless,  the  DEIR  load  estimates  provide a reasonableness check against  the estimates
developed as describe above because the DEIR estimates were based on a different method.

The DEIR estimates used Tri+ concentrations estimated from the USGS data that do not reflect
the  more recent approach developed by Rhea and Werth (1998) to account for analytical bias in
these data.  Therefore, based on the average effect of the bias correction, the DEIR estimates are
estimated  to contain  a high bias of approximately  14  to 16  percent relative to the estimates
presented here. Comparison of the DEIR estimates at the mainstem stations to those developed
herein  for the period 1977-1990 shows that on a cumulative loading basis, the DEIR values are
on average 18 to 26 percent larger on an annual basis, consistent with the  approximate magnitude
of the  analytical bias correction.  This comparison  suggests that the loads estimated here are
consistent with the DEIR estimates, after accounting for the analytical bias.

Several important observations can be made from inspection of the estimated annual Tri+ loads
over the simulation period.  It is clear that a significant declining trend in  Tri+ loads past all of
the  mainstem stations occurred over the period  1977 to 1997 (Figure 6-26). While, the overall
trend is clearly declining, estimated  loads show  large year to year variability (some years' loads
                                           93                       Limno-Tech, Inc.

-------
are greater than previous years). Of particular note is the large increase in Tri+ loads in 1983-84
and 1991-92. The large temporary increase in PCB load in 1991-92 is associated with the failure
of the Allen Mill gate structure in September 1991 (USEPA,  1997). The  1983 load increase
reflects high spring floods occurring in that year.

Average  daily loads at Fort Edward over the period 1980-1984 were approximately  1.21 kg/d
compared to an average of 0.45 kg/d for 1985-1990.  Following an increase in loads in 1991-
1992 due to  the Allen Mill event, loads continued to decline. Average loading at Fort Edward for
the period 1994-1997  was 0.24 kg/d.  The 1997 Fort Edward daily average  load was about 0.8
kg/d.

A conspicuous result is that the estimated Tri+ load passing Fort Edward is much lower than
estimated PCB loads passing Schuylerville, Stillwater and Waterford in  1977, 1978, and 1979,
relative to the remainder of the simulation period (Figure 6-26). This suggests that either large
unmeasured  external loads were entering the river upstream of Schuylerville, or that the sediment
contribution of Tri+ between Fort Edward and Schuylerville was  very large during this period. It
is possible that additional sources were  active during this period, perhaps from land-disposed
PCB laden sediments near the river. The low sampling frequency at Fort  Edward may also have
missed significant high concentration events, resulting in an underestimate of the Fort Edward
load. The most likely explanation, however, is that large amounts of  unstable contaminated
sediment deposits, released  by the 1973  dam  removal at Fort  Edward,  were available for
mobilization within the Thompson Island Pool in this period.  It should be noted, however, that
the period of highest upstream  loading, from  1977 to 1983, contains only 22 percent of the 801
daily average measurements at Fort Edward.  In 1977, when loads were the second highest of any
year in the  calibration, only 3 samples  were collected.  Due to  the high  uncertainty  in Fort
Edward loads in the late 1970's and early  1980's, comparison of model  results to downstream
data was  discounted during this period (Chapter 7).

An important understanding gained from interpreting the estimated loads is that the majority of
PCB transport occurs during non-flood flow periods.  The term "low flow" is used throughout
this  report  to refer  to non-flood flows  less  than  10,000 cfs  at Fort  Edward,  which is
approximately twice the average flow.   Low flow periods are  characterized by relatively  low
sediment scour, although approximately 50% of the total solids transport occurs at low flow. By
comparison,  between 65 and 70% of PCB transport in the Upper Hudson River occurs at  low
flow (Figure 6-27).

Analysis  of  Tri+  PCB load gain across  Thompson Island Pool also shows the significance of
sediment-water exchanges at low flow (Figure 6-28).  For the period  1993-1997, between 60 and
70 percent of the Tri+ load gain across Thompson Island Pool occurs during the summer months,
June through August, when flows are  typically very low. This observation does not diminish the
significance  of high flow events in mobilizing  PCBs due to flow-dependent resuspension,
however,  it   does suggest  that  flow-dependent resuspension  is  not   the dominant  process
controlling sediment-water PCB mass fluxes in the Upper Hudson River.  This focuses attention
on  the importance of low flow sediment  to water  PCB mass  transfer processes,  which is
discussed in  Section 6.12.
                                           94                       Limno-Tech, Inc.

-------
A plot of relative contributions of tributaries to the cumulative Tri+ load over the calibration
period is presented in Figure 6-29. Tributary loads are insignificant relative to upstream loads.

6.6.2.8  Mainstem and Tributary Total PCB and Congener Loads 1991-1997

Specification of daily upstream and tributary PCB loads was also required for the hindcast
application period, Jan.l,  1991 to  Sept.  30,  1997.  Sampling frequency  was approximately
biweekly at Fort Edward and Thompson Island Dam for ice-free conditions during this period. In
winter months, sampling frequency was approximately  monthly.  At Stillwater and Waterford,
sampling by GE and USEPA provided total  PCB and  congener measurements through 1993.
Daily average  loads of total PCB and the  five congener state variables (BZ#4, BZ#28, BZ#52,
BZ#90+101, and BZ#138)  were estimated  by linear interpolation of daily average values at Fort
Edward over this period. For total PCB daily average loads were also estimated by interpolation
at Thompson Island Dam for comparison to model output (Chapter 7).  As discussed in Section
6.3, a correction was developed for the Thompson Island Dam data to account for the observed
bias in the west shore sampling station measurements.  Because sampling frequency was lower
and the  sampling period much shorter at Stillwater and Waterford, daily average loads were not
estimated for total PCB and congeners at these locations.

The estimated loads of congeners BZ#28, BZ#52, and BZ#90+101 are considered  known to
within the certainty of the estimated total PCB  load because the number of data points used was
nearly the same as for total PCB.   BZ#4 and BZ#138,  however, have reported quantitations
greater than zero for only 42 and 34 percent of the total PCB results, respectively in Release 4.1b
of the Hudson  River Database.  Zero, or non-detects were frequent for BZ#4 and BZ#138 when
other  congeners were measured in fairly high concentrations. Concentrations  of BZ#4  and
BZ#138 were estimated based on observed ratios to total PCB. Evaluation of these ratios at Fort
Edward  showed a seasonal pattern, with BZ#4 mass fraction highest in the summer months and
BZ#138 showing the opposite behavior.  This probably reflects enhanced sediment-water release
of BZ#4 during summer months from contaminated sediment deposits upstream of Fort Edward.
This may also reflect enhanced mobilization  of heavier congeners during resuspension events
upstream of Fort Edward, based on  an observed negative correlation  of the BZ#4 mass fraction
with flow.

An additional  observation  was that in  1991  and 1992, average BZ#4 fractions  (0.035)  were
approximate to that in Aroclor  1242 (0.0313),  which is the primary source material of the Fort
Edward  PCB load. In the other years where congener data are available (1993, 1996, and 1997)
BZ#4 mass  fractions were significantly higher, on  average  about 0.09 (Figure  6-30).  Some
measurements  showed BZ#4 fractions greater than 0.25.  These observations probably reflect the
release of fresh Aroclor 1242 material during the Allen Mill event. In  light of these observations,
the monthly average ratios  to total PCB for the period 1991-1992 and  1993-1998 were used to
estimate BZ#4 and  BZ#138  concentrations when total PCB was quantified but  the congener
results reported as zero.

The same approach as used in developing tributary loads for Tri+ was used for specifying
tributary loads for Total PCB.  The average measured  concentrations  was  used  for monitored
tributaries, while the minimum measured concentration  (0.51 ng/1 ) in the monitored tributaries
                                          95                       Limno-Tech, Inc.

-------
was assigned  to  the  unmonitored tributaries  (Table  6-19).  Tributary loads for individual
congeners were assumed insignificant and specified as zero.

6.6.3 Total PCB and Congener Load Results  1991-1997

Annual average total PCB and congener loads computed at Fort Edward are presented in Table 6-
21.  Consistent with  the  long-term declining  trend in Tri+ loads observed at all mainstem
stations, the total PCB and congener loads at Fort Edward are also observed to decline over the
period 1992-1997 following the increase in load associated with the Allen Mill gate structure
event in the fall of 1991 (Figure 6-31).

Comparison of total PCB loads at Fort Edward and Thompson Island Dam shows that annual
average total PCB load gain across Thompson Island Pool is  approximately a factor of two larger
than the  load  gain  of Tri-k  The difference  is primarily  due to the release of BZ#4 from
Thompson Island Pool sediments, which is not reflected in the Tri+ load gain.

6.6.4 Summary of PCB Load Estimates

Daily average estimates of PCB loads at Fort Edward, TI Dam, Stillwater and Waterford were
developed for  use in the long term historical calibration and short-term hindcast applications.
Tri-f loads were developed for the long-term calibration, from January 1, 1977 to September 30,
1997,  using  a combination of  linear interpolation   and seasonal averages of measured
concentrations by year.   Seasonal average  values were specified during  periods  where data
sampling frequency  was too low to support linear interpolation. Total PCB and congener load
estimates were developed for the short-term hindcast period, January 1, 1991 to September 30,
1997, using linear interpolation.  Total PCB loads for this period were estimated at Fort Edward
and  at Thompson Island Dam. Congener loads were only estimated at  Fort  Edward.  The
Thompson Island Dam total PCB loads were developed for comparison to  model  output.
Monitored tributaries  were assigned the average measured concentrations.  For unmonitored
tributaries, loads were specified for total PCB  and Tri+  based on  the minimum measured
concentrations from monitored tributaries.  During the historical calibration  period, tributary
PCB loads represent less than 3 percent  of the  total loading of PCB between Fort Edward and
Waterford.

The resulting daily average PCB loads estimated as described above were used to develop input
time series of  PCB loads  for the HUDTOX  model at Fort Edward and all  12 tributaries.
Estimated loads at Thompson Island Dam, Stillwater and Waterford were used for comparison to
model output.

PCB load estimates at Fort  Edward are very uncertain in  the first few years of the historical
calibration period and again in the fall of 1991  due to low sampling frequency during periods of
high, fluctuating loads. Based on observation of high concentrations at Schuylerville, the Fort
Edward load may have been under-estimated or other sources may have been active from  1977 to
about 1984. An additional  source of uncertainty in the Fort Edward load arises from the apparent
occurrence of random pulse loading events, which are suspected to be only partially captured by
available data.
                                           96                       Limno-Tech, Inc.

-------
Load results show overall, a strong declining trend in upstream PCB loads from the late 70s to
the late 90s, with a noticeable temporary interruption in 1991 due to the Allen Mill event.
Currently, Fort Edward loads and load gain through the system are comparable to loads observed
approximately 10 years ago in the late 1980s.

Analysis of PCB  loads  estimated at Fort Edward and Thompson Island  Dam shows that
approximately 60 percent to 70 percent PCB loading at Fort Edward, and PCB load gain across
Thompson Island Pool, occurs at flows less than 10,000 cfs at Fort Edward.

6.7 SEDIMENT INITIAL CONDITIONS

6.7.1 Overview

The HUDTOX  model requires  specification of  initial  PCB concentrations, in addition  to
sediment specific weight (mass of dry solids per unit volume), sediment paniculate  organic
carbon content (foe) and sediment dissolved organic carbon concentrations (DOC). This section
presents specification of initial PCB concentrations in 1977 and specific weight.  Specification of
sediment foe  and DOC  concentrations  is  presented in Section  6.9,  which also includes
specification of partition coefficients.

Sediment initial  conditions for Tri+ are developed from the 1977 NYSDEC sediment  dataset.
Average  concentrations  were determined on a dry weight  concentration  basis for  specific
sediment layer intervals over discrete areas of the river corresponding to individual segments, or
groups of segments in areas of limited data. Concentrations were averaged over 2 cm intervals in
the sediment bed to develop concentrations for each HUDTOX sediment layer, down to 26 cm.
Sediment specific weight was established based on the USEPA Phase 2 low resolution coring
data.  Average values were determined for cohesive and non-cohesive sediment for the entire
river.

6.7.2 Sediment Specific Weight

Specific weight is defined as the mass of dry solids per unit volume of wet sediment, or the
sediment solids concentration. The HUDTOX model requires  specification of sediment  specific
weight as an initial condition, which is held constant through the simulation (See Chapter 5).

Sediment specific weight values for the HUDTOX sediment layers have been determined using a
subset of the USEPA Phase 2 low resolution sediment core data for which specific weight was
estimated as wet bulk density times percent solids. Some of the reported data were excluded due
to anomalous values for bulk density, percent solids or both.  A total of 535 specific weight
measurements from 169 sediment cores were used (there are  values for multiple core slices at
most locations).

An attempt was made to incorporate the following three criteria used by Zeigler and Nisbet
(1994) to identify cohesive sediments:
                                          97                       Limno-Tech, Inc.

-------
    1.  d50<250um;

    2.  percent clay and silt > 15%; and,

    3.  percent moisture > 75% .

The percent moisture values were deemed unsuitable for this purpose, therefore, samples were
classified as cohesive or non-cohesive based on criteria #1 and #2, using the ASTM sediment
classifications provided in the Hudson River Database.  The sediment classifications (up to 3
classes are identified based on visual description or grain size) are associated with a descriptor
(either abundant, some, trace, or few) indicating relative abundance of the associated material.
The descriptors provided for each sample were used to infer grain size and the percent clay and
silt.

 Samples classified as fine sand, silt, clay, or organics were assumed to meet the criteria of dso <
250 jam, except  those containing "some" coarse sand and/or gravel.  Fine sand samples having
"some" or "abundant"  silt,  clay, or organics were assumed to meet criterion  #2 and were
considered cohesive.  A total of 30 fine sand samples having "few", "trace" or no silt, clay, or
organics were classified as non-cohesive. All other samples were classified as non-cohesive.
This resulted in a total of 366 samples classified as cohesive and 169 samples classified as non-
cohesive.

Specific weight values were grouped by sediment type (cohesive and non-cohesive) for top core
sections, which have an average depth of 10 inches. Mean specific weights for cohesive (0.84
g/cc) and non-cohesive  (1.38 g/cc) sediments were selected to represent the sediment specific
weight in HUDTOX.

The cohesive sediment average specific weight (0.84 g/cc) is 37 percent lower than the average
specific weight for non-cohesive sediment (1.38 g/cc).  The difference in specific weights of the
cohesive and non-cohesive sediments determined from the TAMS low resolution sediment cores
is mainly due to differences in porosity (average porosity is 0.59 for cohesive and 0.37 for non-
cohesive), which accounts for approximately 93 percent of the difference in specific weights.
The difference in particle density (average particle density is 2.16 for cohesive and 2.22 for non-
cohesive) contributes approximately 5 percent of the difference in specific weight.

6.7.3  1977 Tri+ Initial Conditions

6.7.3.1  1977 NYSDEC Sediment Data

Initial conditions were developed from the 1977 NYSDEC data.  These data were collected from
a sediment coring and grab sampling program spanning the Upper Hudson River from the Bakers
Falls  area to Troy.   The  most extensive sampling  occurred  in  Thompson Island Pool.
Approximately 30 samples were collected north of Fort Edward, and 24 samples downstream of
Federal Dam at Troy, falling outside  of the HUDTOX model domain. Approximately 40 other
samples lacked location data (river mile or northing/easting coordinates) and were not usable for
developing sediment initial conditions.  A total of 961  sample locations, consisting of 623 grab
samples and 338 core samples were available within the HUDTOX domain.


                                           98                       Limno-Tech, Inc.

-------
A number of the  1977 samples were excluded from use in specifying initial conditions due to
anomalies in the data.  A group of grab samples all reported to be located at river mile 189.2 and
having very high PCB concentrations were considered suspect and dropped, as was a group of
grab samples reported to have been collected at river mile 156.5 on December 27, 1977.  No
documentation could be found supporting  this  December sampling event (QEA,  1999).  One
additional sample was excluded, ID number 30291, which had a very high PCB concentration
and was surrounded in close proximity by samples with much lower concentrations.

The  1977 NYSDEC data core sample depths varied widely.  Surficial layer sectioning ranged
from 1.5 cm to 56 cm, with an average of 7.5 cm. Grab samples were taken by a Shipek sampler
and  usually obtained a 0-5  inch depth composite (Tofflemire and Quin, 1979).   The average
surface sample depth of core and grab samples combined is 10.9 cm.

The  1977 NYSDEC  samples were not  analyzed for solids specific weight.  Measurement of
principal fraction-phi,  %gravel, %sand, %silt, %clay, %total solids, % volatile solids and texture
is reported for many samples, however numerous samples do not have results for one or more of
these parameters.

6.7.3.2  Methods

HUDTOX requires specification of initial conditions as bulk concentrations (mass of PCB per
unit bulk volume of sediment). Average dry weight Tri+ concentrations (mass of PCB per mass
of dry solids) were computed from the 1977 NYSDEC sediment data.   In order to  specify
sediment initial conditions on a bulk concentration basis, average dry weight  PCB concentrations
were computed for cohesive and non-cohesive sediment from these data and multiplied by the
sediment specific weight values (Section  6.7.2).

Samples  were classified as cohesive or non-cohesive following hierarchy of methods, dependent
on the parameters reported for each sample (Table 6-22). After classification of the individual
core sections according to the criteria in Table 6-22, all core sections in each core were assigned
the cohesive/non-cohesive classification of the  top-most section.  The samples  classified as
cohesive or non-cohesive were  mapped to the  HUDTOX water column  segments based on
location information.

To specify sediment Tri+ initial conditions, the NYSDEC  1977  data were averaged horizontally
and vertically  for cohesive  and non-cohesive sediment types  in each  HUDTOX layer.  The
vertical  segmentation  scheme employs  2 cm layers throughout the  modeled portion of the
sediment bed (0-26 cm).

The core section and grab sample data were mapped onto the HUDTOX vertical sediment layer
segmentation using a length-weighted-average calculation for each sample:

                               .i.
                               -                                               (6-19)
                                          99                      Limno-Tech, Inc.

-------
where:
       Cj = concentration of layer; (mg/L)
       C, = concentration of sample / in layer j (mg/Kg)
       / = length of section / in layer j (cm)
       n = number of sections extending into layer j
Once each sample was mapped onto the vertical segmentation intervals, using equation 6-19,
average Tri+ concentrations were computed for specific intervals of the river based on data
availability.  For some portions  of  the river, data falling in multiple adjacent water column
segments  were grouped and averaged together, by sediment layer interval and sediment type.
Average concentrations were computed for  each  group.  The Thompson  Island Pool  water
column segments were divided  into 7 averaging  groups, and the segments downstream  of
Thompson Island Pool were divided into 10 averaging groups (Table 6-23).

6.7.3.3  1977 Initial Condition Results

The  1977 initial condition  surficial sediment Tri+ concentrations for cohesive and non-cohesive
segments are shown in Figure 6-32.  Maximum concentrations of both cohesive sediment, 290
mg/kg, and non-cohesive  sediment, 44 mg/kg, occurs in Thompson Island Pool (Figure 6-33).
Minimum concentrations are 7.2  mg/kg for cohesive sediment and 3.3 mg/kg for non-cohesive
sediment, occurring just below Stillwater Dam and below the Lock 1  Dam, respectively.   The
vertical profiles computed  for each segment are shown in Figure 6-34 with plus and minus two
standard errors. Inspection of the  vertical profiles shows that peak concentrations typically occur
at depths less than 12 cm. The vertical profiles do not show a significant gradient with depth,
which is attributed to the  variable surface core section thickness used in the  sampling, ranging
from  1.4 to 46 cm.  The average surface sediment sample thickness is 7.5 cm with a standard
deviation of 6.3 cm.

6.7.4  1991 Initial conditions and model calibration targets

Sediment initial conditions for total PCB, Tri+ and the congener state variables were computed
from  the  1991 GE  composite sediment  sampling data for use  in  the short  term hindcast
application conducted from Jan. 1, 1991 to Sept. 30, 1997. Tri+ concentrations were also used as
model calibration targets.

6.7.4.1 Data

In the GE  1991  composite  sampling survey,  approximately  520 individual  samples were
collected in Thompson Island Pool and  approximately 480 from Thompson Island  Dam to
Federal Dam.  The Thompson Island Pool was divided into 6  sub-reaches,  in which 5  to  12
composites samples were  collected.  The composites generally contain on the order of 10 to 20
individual samples collected over a range of 1A mile or  more.  Thus, the  1991 data do not
represent  individual  cohesive or  non-cohesive sediment deposits.  Individual  sediment core
samples were grouped as fine or coarse  sediments  in the  laboratory after determination of
sediment type based on texture and bulk density. After sectioning  into 0-5, 5-10, and  10-15 cm
layers, the individual layer sections were composited in each group and analyzed for PCB.   As
                                          100                      Limno-Tech, Inc.

-------
discussed in Section 6.3, sediment BZ#4 concentrations  were estimated from the  co-eluting
BZ#4&BZ#10 sum.

6.7.4.2  Methods

The calculation of 1991 concentrations involved vertically mapping concentrations on the 2 cm
model sediment layer intervals as conducted for the 1977 NYSDEC data using equation 6-19.
The constant core sectioning employed in the 1991  survey resulted in interpolation at only two
depths, 5 cm and 10 cm. Grab samples were assumed to represent 10 cm.

Each sample in the composite was assigned the composite concentration for each sediment layer.
The composite samples were classified as cohesive or non-cohesive based on the fine/coarse
designation  assigned to the composite samples  during collection.  Similar  to  the  approach
employed for the 1977 NYSDEC data, samples in adjacent water column segments were grouped
by sediment type based  on data availability (Table 6-24).

6.7.4.3 1991 Initial Condition Results

The longitudinal (down-river) surface sediment concentration profiles computed for  each PCS
state variable from the  GE 1991 data are presented with the data in Figures 6-35  through 6-38.
Down-river  trends  show  that PCB concentrations in  Thompson Island Pool  and between
Thompson Island Dam and Schuylerville  are very elevated relative to concentrations observed
downstream of Stillwater.   This  trend has  persisted from  1977, where a similar  pattern is
observed in  Tri+ concentrations (Figure 6-32). Examination of congener concentrations shows
that BZ#4 concentrations exhibit a larger decrease relative to the heavier congeners.  Congener
concentrations normalized to BZ#52 concentrations show  that the ratio of BZ#4/BZ#52  drops
markedly at  Schuylerville (Figure 6-39), while ratios of the heavier congeners increase. This is
attributed to in-place dechlorination of sediment PCBs between Fort Edward and  Schuylerville.
Decreased BZ#4  concentrations downstream of  Schuylerville  may be related to weathering
processes, whereby heavier congeners are  preferentially delivered to downstream  sediments via
deposition and loss of the lighter more mobile congeners via volatilization or export over Federal
Dam.

Poolwide average surface sediment concentrations in TIP  are shown for each state variable in
Table 6-25.

6.7.5  Summary

Sediment initial conditions were computed  in 1977 from  the NYSDEC data for the  historical
calibration period for Tri+ and in 1991 from the GE composite sampling data for the short-term
hindcast period.  Sediment conditions were computed for all seven PCB state variables in 1991
(total  PCB,  Tri+, BZ#4, BZ#28, BZ#52, BZ#90+101, and BZ#138).  The Tri+ concentrations
were used as model calibration  targets for the long-term historical calibration, while the other
PCB  forms were  used  as  initial  conditions  for  shorter-term  1991-1997  simulations.
Concentrations were mapped onto sediment segment layers according to Equation 6-19. Average
concentrations  for cohesive and  non-cohesive  sediment concentrations were computed  for
                                          101                       Limno-Tech, Inc.

-------
specific intervals of the river, based on data availability.  Due to the averaging approach taken,
the specified initial conditions do not represent discrete PCB hotspots in many areas.

Based on the specified initial conditions for the congeners,  BZ#4 represents the largest fraction
of the total PCB mass  in Thompson Island Pool and between Thompson  Island Dam  and
Schuylerville, approximately 25 percent.  Below Lock 5 at Schuylerville the concentration of
BZ#4 declines  significantly and at Waterford, the BZ#4 mass  fraction is  on the order of 5
percent.

6.8 WATER AND Am TEMPERATURES

A number of processes represented in the HUDTOX model are temperature dependent.  These
include: partitioning, volatilization and porewater diffusion rates.  A large number of in-situ
water temperature data are available from the USEPA and GE datasets.  No  in-situ temperature
data were available for sediments. The sediment bed temperatures were assumed to follow the
water column temperature.  Sediment temperatures are likely to  be damped relative to surface
water temperatures by heat exchange with groundwater and underlying bedrock, however,  no
data exist with which to evaluate this.

Monthly-average water column temperatures were computed for the primary Upper Hudson
River sampling  locations for application in HUDTOX. Some smoothing of the monthly average
curves at each station was required (Figure  6-40).  The annual time  series  represented  by the
monthly average was used to describe the HUDTOX calibration and forecast application periods.
The  monthly time series specified at  each  station was applied to segments between  station
midpoints.   For  example,  the  Thompson  Island  Dam  temperature series  applies  to  the
downstream half of Thompson Island Pool and half the distance to the Schuylerville sampling
station.

Year to year variations in mean monthly water temperatures are fairly small.  The largest  year to
year variability  appears  to  occur  during April and May for which the standard  deviation of
observations is approximately 30 to 50 percent (in degrees Celsius), depending on location. Peak
monthly average temperature occurs in  July.  During non-winter months, water temperature
generally exhibits a continual increase from Ft. Edward downstream  to Waterford. In July, at
peak temperature, the mean water column temperature increases 3.6 °C, from Ft. Edward (24.2
°C) to Waterford (27.8 °C) as shown in Figure 6-41. During the winter months, the entire  river is
about the same temperature. Minimum mean monthly temperature is 1.1 °C in January.

Temperature gradients between near shore and center channel  may exist due to a number of
factors, which may result in positive or negative gradients. Shallow, near shore areas of the river
can experience  more solar heating due  to  slower velocities and depth, which may serve to
increase temperatures relative to the center channel. Groundwater inflows may serve to decrease
near shore temperatures relative to the center channel and likely cause sediment temperatures to
lag water column temperatures.
                                          102                      Limno-Tech, Inc.

-------
Very little data exist with which to  evaluate possible  temperature gradients between center
channel  and shallow near shore areas.  In HUDTOX, no lateral  temperature gradients  are
represented.

Monthly average air temperature data were obtained from the NOAA-NCDC daily summaries for
Glens Falls/Warren County, New York.

6.9 PARTITIONING

6.9.1 Overview

In natural systems the fate and transport of PCBs and other hydrophobic chemicals is largely
controlled by their degree of partitioning to  sediment particles and dissolved organic  matter
(DOC-bound).  While sorption and desorption are complex processes, often the dissolved phase
and  particle phase concentrations  are  assumed  to be  in  equilibrium.  This  assumption is
reasonable when sorption kinetics are  rapid relative to other processes affecting concentrations.
Two-phase  and  three-phase  models  of  equilibrium  partitioning assume  that  measured
concentrations in one phase can be used to predict concentration in the other phase(s).

The equilibrium partitioning assumption was evaluated by USEPA (1997) and found to be valid
for the Upper Hudson  River.   TAMS computed  particulate concentrations predicted from
measured dissolved concentrations using two-phase partition coefficients and compared results to
observed values throughout the Upper Hudson River over a range of environmental conditions.
The predictions were unbiased for the majority of samples and average difference  between the
predictions and observations  was 45 percent, and only 33 percent for stations downstream of
Thompson Island Dam.   Results for data collected near Fort Edward suggest non-equilibrium
conditions at this  station.  These  predictions  represent  a high degree of accuracy relative to
similar reported studies and  suggest  it is possible to predict the phase distribution of PCB
congeners to within about 33 percent for the freshwater Hudson below Thompson  Island Dam.
Equilibrium assumptions were proposed to be adequate to represent fate and transport of PCBs in
the Hudson (USEPA, 1997).

Development of two-phase and three-phase equilibrium partition coefficients for 64 congeners
are presented in the DEIR.  Three-phase partition coefficients were  estimated using numerical
optimization from  USEPA Phase 2 water column data and  1991 GE sediment composite data,
both of which report particulate and apparent dissolved  (truly dissolved  plus DOC-sorbed)
concentrations.  The reader is referred to the DEIR  for details of those  analyses.  Partition
coefficients computed from water column data  are considered to be more accurate than partition
coefficients  computed between sediment and  porewater due to differences in  analytical and
sample collection methods. There  is, however, considerable uncertainty in the determination of
three-phase partition coefficients in both media. Results show that for the lightest congeners, the
DOC-bound fraction may comprise up to 50 percent of the total PCB concentration in the water
column, but is generally less than 10 percent for congeners constituting Tri+. In the sediments,
results suggest that a significant fraction of the porewater concentration is associated with DOC
for all congeners.   The  mono-  and di-chlorinated  congeners exhibit different  partitioning
behavior than the other congeners in that their Kpm: and Ktt,)c (see Equations 5-10 and  5-11) values
                                           103                       Limno-Tech, Inc.

-------
 are of approximately the same order of magnitude.  Further, because of their lower partition
 coefficients,  porewater  concentrations  of mono-  and di-chlorinated congeners are enhanced
 relative to the heavier congeners, which may facilitate greater sediment-water transfer of these
 congeners via porewater.  Enhanced flux of these congeners relative to the heavier congeners
 may also occur due to the presence of elevated DOC, considering the differences in ratios of Kpoc
 to Kdoc for the heavy and light congeners (USEPA, 1997).

 Three-phase equilibrium partitioning was adopted for HUDTOX based on two considerations.
 First, because of the importance of the DOC phase in affecting the phase distributions  of the
 lighter congeners, it is necessary to use three-phase partitioning to properly account for the ratios
 between sediment-water transfer of congeners in porewater.  A sensitivity analysis presented in
 the DEIR  (USEPA, 1997) suggests that dissolved and DOC-bound concentrations are likely to be
 of the same order of magnitude.  Second, truly dissolved chemicals are thought to be more
 readily bioavailable than those sorbed to DOC and accounting for the DOC fraction may provide
 a better estimate of exposure to biota.

 The HUDTOX model applies three-phase equilibrium partitioning equations presented in Section
 5.2 (Equations 5-10 and 5-11).  In these equations, the Kpoc and Kdoc values for congeners and
 Tri+ are temperature  corrected according to the temperature correction slope factor (Equation 5-
 12) recommended for all congeners in the DEIR (USEPA, 1997). The temperature correction is
 log-linear  and a 10-degree C change in temperature reduces partition coefficients by 28 percent.

 In addition to the three-phase partition coefficients presented for congeners in the DEIR, three-
 phase coefficients were also developed by USEPA, 2000 for Tri+, following the same approach.
 To compute total PCB partition coefficients, mass-weighted values from Tri+ and the  mono- and
 di-chlorinated congeners were used.  The mass weighting used average congener and Tri+ mass
 fractions in the USEPA and GE water column data.  Details of this analysis are presented below.

 Application of the three-phase partitioning model requires specification of the fraction of organic
carbon (foe) m suspended and bedded sediment particles, as well as concentrations of DOC in the
water column and sediment porewater. For bedded sediments, average foe values were computed
from the GE 1991  composite sediment data based on sediment type and location.  For the water
column foe was found to be correlated with flow. A function relating water column foe to fl°w
was applied in HUDTOX.  Average sediment DOC concentrations were also computed from the
GE data based on sediment  type and location.   Water column DOC  concentrations were
observed to be relatively invariant in the Upper  Hudson  throughout the year and exhibit small
differences between locations. Average DOC concentrations were computed from the GE, Phase
2, and J. Vaughn (1996) data.  The development of the foe and DOC concentrations for sediment
and water  are presented in detail below.

Based on the specified parameters influencing partitioning of PCB, typical phase distributions of
PCB  in the Upper Hudson River are presented for  winter and  summer low and  high flow
conditions.
                                          104                      Limno-Tech, Inc.

-------
6.9.2 Partition Coefficients

The three-phase partition coefficients developed based on the UPEPA Phase 2 water column for
Tri+ and  congeners BZ#4, BZ#28,  BZ#52, BZ#90&101, and BZ#138 are used in  HUDTOX
(Table 6-26).  These  partition coefficients are temperature corrected in the HUDTOX model
according to the temperature correction slope factor developed for all congeners in the DEIR
(Equation 5-12).

Consistent with the equilibrium assumption employed in the HUDTOX model, a single partition
coefficient was applied for the whole Upper Hudson River. Non-equilibrium conditions appear
to occur in some of the Phase 2 data, particularly at Fort Edward, resulting in higher apparent
partition coefficient estimates than downstream.  Partition coefficient estimates were corrected
for the possible  presence of  non-equilibrium samples  by use of the median of  individual
estimates to describe the central tendency of observations.  Partitioning at Thompson Island Dam
and  downstream locations appears to be generally at equilibrium conditions for tri- and  higher-
chlorinated congeners, however, for mono-, di-, and tri-chlorinated congeners, there  appears to
be some local non-equilibrium at Thompson  Island Dam (possibly associated with sediment-
water transfer of predominately the dissolved phase for these congeners).

Generally, the Phase  2 data do not indicate a clear distinction between partitioning behavior
among stations and differences among stations are likely attributable to variations  in organic
carbon concentration and water temperature (USEPA, 1997).

In addition to Tri+ and the five congeners, total PCB is an additional  HUDTOX state variable.
Partition  coefficients  were  not developed  for  total PCB  as part of  the  DEIR or  LRC
investigations.  To estimate KPQC and KDOc  for total PCB, a mass  weighting  approach  was
adopted using values  determined for Tri+ and the mono- and di-chlorinated congeners.   The
average mass fraction  of total PCB represented by Tri+, BZ#1, BZ#4, and BZ#8 were computed
at Fort Edward, Thompson Island Dam, Schuylerville, Stillwater and Waterford from the USEPA
Phase 2 data and the GE data (Table 6-27). Because these mass fractions do not sum to unity due
to exclusion  of other mono- and dichlorobiphenyls for  which partition coefficients were not
calculated, the mass fractions were normalized by the total mass represented by these congeners
and  multiplied by their respective partition coefficients to compute a mass-weighted value for
total PCB  as described by Equation 6-20.
 TS-
  "oc =
                            BZ#4 ^ A BZn T A Tri+
where:
       X = mass fraction for each PCB form
       Kp<)C = particulate organic carbon partition coefficient (I/kg)

This procedure was repeated for KDOC, resulting in KPQC and KDOC estimates for each of the five
stations listed above (Table 6-28, Figure 6-42).  The shift in the congener distribution toward the
mono- and di- fraction due to gain of these constituents across Thompson Island Pool and loss


                                          105                       Limno-Tech, Inc.

-------
downstream of Thompson Island Dam is evident in the  pattern of results.  While partition
coefficients for individual congeners and Tri+ exhibit some  spatially variability (probably related
to differences in dissolved organic carbon and temperature, as discussed above), the changing
composition of total PCB is an additional factor contributing to spatial variability and uncertainty
in the total PCB partition coefficient.  This was a consideration in deciding not to calibrate to
total PCB concentrations, but rather to focus the calibration on Tri+ and use additional congener
calibrations to strengthen the Tri+ calibration. This is discussed in Chapter 7.

To determine a single value of KPQC and KDQC for total PCB  for application in HUDTOX, the
station values were distance-weighted using the  distance between midpoints of each location
divided by the total distance between upstream and downstream locations. The Waterford value
was assumed to represent the reach  from Waterford to Federal Dam in  the distance  weighting
(Table 6-29). While the Fort Edward value may be affected by non-equilibrium conditions (see
above), it receives a fairly small  distance weighting factor  and does not significantly affect the
result. The log Kpnc and log Kdoc values determined by this method for total PCB are 5.64 and
4.22, respectively.

Sediment three-phase partition coefficients were also estimated by USEPA (1997) from the 1991
GE composite sediment sampling data.  These estimates were subsequently updated to account
for corrections to analytical biases in  the GE data (Table  6-26).  While the  GE data allowed
estimates of sediment partition coefficients, a number of important factors in  the sampling and
analysis procedures affect the quality of these estimates. Samples were frozen prior to analysis,
which may alter all phases of the  matrix (USEPA, 1997), and field blank contamination affected
87 percent of the PCB analyses. These limitations suggest that the accuracy of the GE estimates
are low  compared  to  the  values estimated from the USEPA Phase  2  water column data.
Therefore, the Phase 2 water column estimates were chosen to describe both sediment and water
column partitioning behavior in HUDTOX. Application of the water column estimates may have
contributed to difficulty in calibrating the model to individual congeners, which is discussed in
Chapter 7.

6.9.2.1 Water Column Organic Carbon Concentrations

The HUDTOX model employs three-phase equilibrium partitioning formulations (Equations 5-
13 through 5-16), which compute PCB distribution among  paniculate organic matter, dissolved
organic carbon (DOC) and  water  (Section  5.2).   In  these  equations, the  concentration  of
paniculate organic material is computed as the sediment solids concentration times the fraction
of organic carbon in the sediment particles, foe- Values for/0c and DOC  are specified separately
for the water column and sediments.  Values for both of these parameters were determined from
site-specific data and specified as  model inputs.  The determination of water column f0c and
DOC  values considered spatial  and  temporal  patterns in data.  This section presents  the
development of these parameters  for  the water column. The  next section presents development of
the sediment/oc  and DOC values.
                                           106                       Limno-Tech, Inc.

-------
6.9.2.2 Water Column DOC

In addition to the USEPA Phase 2 data and the GE monitoring  data, water column DOC
measurements were also available from investigative studies conducted at Rensselaer Polytechnic
Institute (Vaughn  1996). The GE water column organic carbon data required extensive filtering
due to numerous  inconsistencies among  reported TSS, TOC and DOC concentrations in the
dataset.  Measurements were only used for samples meeting the following criteria:  [TOC]  <
[TSS] and, [TOC] > [DOC]. This resulted in use of 17 percent of the 421 samples for which
TSS, TOC and DOC concentrations are reported in  the Release 4.1b  of the Hudson River
Database.  Table  6-30 summarizes the number of data  used from  each source by mainstem
Hudson River location, along with a statistical summary of the data at each location.

Dependencies of DOC concentration on flow, season, and location were investigated  with the
combined USEPA, GE and J. Vaughn datasets. Consistent with the findings in the DEIR based
on the USEPA Phase 2 data, DOC was observed to be slightly negatively correlated with flow,
and only weakly  correlated with temperature (and season).  The observed decrease  in mean
concentrations in  the spring is explained  by the dependence on  flow.   The lowest DOC
concentrations tend to occur coincident with the highest flows and lowest  temperatures during
the snowmelt runoff period (Figure 6-43).  A plot of the DOC versus river mile suggests some
dependence of DOC concentration on location, which is also evident  in the mean values (Figure
6-44).  Differences between locations are generally small.  Mean values at the primary mainstem
locations (Ft.  Edward, Thompson Island Dam, Schuylerville,  Stillwater and Waterford)  differ by
a maximum of 14  percent.

Even  considering the slight negative correlation  of  DOC on flow and temperature, DOC
concentrations are relatively invariant in the Upper Hudson  River.  Maximum deviations from
mean concentration at each location is less than 30 percent (excluding a possible outlier of 0.94
mg/L at Stillwater from USEPA Transect 4).  Specification of mean values by reach was judged
to give adequate  representation of DOC concentrations  in the model.  The DOC data were
grouped  into  the  four reaches  presented below  for  the purpose of specifying mean DOC
concentrations (Table 6-31).

6.9.2.3 Water Column/oc

Water column foe values were specified based on estimates available from the USEPA Phase 2
data and  the GE data.  As discussed above,  filtering of the GE dataset was required in order to
identify samples  with reported  TSS, total organic carbon (TOC), and  DOC concentrations
consistent with each other.  This  resulted in use of only 17 percent of the 421 samples for which
TSS, TOC and DOC concentrations are reported in  the Release  4.1b  of the Hudson River
Database. The GE data reports concentrations for TSS,  TOC  and filterable TOC (labeled TOC_f
in the database). For each sample, particulate organic carbon (POC) concentration was computed
as TOC - TOC_f. Subsequently, foe was computed as POC /TSS.

In the USEPA Phase 2 studies, POC was not measured directly in the water column. However,
weight-loss-on-ignition (WLOI) data were reported and can be used  to estimate POC (USEPA,
1997).  The Phase 2  data  contain WLOI data at two temperatures, 375 °C and 450 °C (for
                                          107                      Limno-Tech, Inc.

-------
Transect 1  only), however a conversion factor was developed so that all WLOI data could be
converted to a common temperature. Based on zero-intercept regression analysis using sediment
data,  WLOI375 can be converted to organic carbon weight fraction as (USEPA, 1997):

                    foe = 0.611  * WLOI375                                         (6-21)

                    WLOI375 = WLOLtso * 0.864                                    (6-22)

The combination of foe estimates obtained from the USEPA and GE data resulted in  24 to 296
measurements at the primary mainstem sampling locations.

Results in the DEIR (from analysis of the Phase 2 data) show that foe is significantly  correlated
to flow  but not to location at a  95 percent confidence level.  Based on this observation the
dependence of foe on flow was analyzed to develop a functional relationship for the HUDTOX
model. The data were plotted versus flow normalized by mean flow at each location (Figure 6-
45).  While/oc is clearly negatively correlated with flow, there is significant variability in foc
across the range of flows sampled, with the greatest variability observed at low flow.  A power
function regression analysis was  used to fit the  data as a function of normalized flow.  This
produces a  model which generally describes foe well at high flows, but has limited  predictive
ability at low flow due to significant variability in the observations. This function was  applied in
the HUDTOX model to compute foe as a function of flow (Equation 6-23).

             ,   ,-0.3687
foc= 0.175x1^ I                                                              (6-23)


For application in HUDTOX, the average  flow of each model  segment was  computed for
segment below  TIP by  using segment specific  flow, Q, and the  average flow of total flow
estimated upstream flow inputs, Q .

Evaluating the behavior of this equation over the range of flows modeled shows that at the lowest
flow conditions, foe is approximately 0.22 and 0.08 at the low and high end of the flow range,
respectively. At the average flow,/oc is 0.175.

6.9.2.4 Sediment Organic Carbon Concentrations

Average sediment fraction of organic carbon (foc) values and porewater dissolved organic carbon
(DOC) concentrations were developed from GE and USEPA Phase 2 data. The HUDTOX model
requires specification of these input values, which determine PCB phase partitioning in the three-
phase partitioning calculations.  The data were segregated by sediment type  (cohesive  or non-
cohesive) and location  in  the River and average  concentrations were determined for each
sediment type over intervals of the River that were dependent on data availability and apparent
spatial trends in these values. The specification of (foc) and DOC values is described below.

6.9.2.5 Porewater DOC

The sediment DOC measurements available from the GE 1991 Sediment Sampling and Analysis
Program (O'Brien and Gere,  1993)  were  used to specify DOC  concentrations by reach in the
                                          108                      Limno-Tech, Inc.

-------
HUDTOX model. The Phase 2 sediment studies did not measure porewater DOC concentrations.
The GE DOC data are measurements of filterable TOC obtained from sediment core composites.
A total of 86 sediment DOC measurements are available from Fort Edward to Federal Dam.

Spatial differences along the river and between fine and coarse sediment areas were investigated.
When  plotted  versus  river mile, the data show  a  trend  of increasing  porewater DOC
concentrations with distance downstream  from  Fort Edward (Figure 6-46).  The GE samples
were composited by sediment type and composites are identified as being from coarse or fine
sediments.  The available DOC data are biased toward fine sediment composites, with only a
small percentage  of the DOC measurements  being from coarse  sediment..  Based  on  the
distinction of coarse and fine sediments in  the GE composites, and the limited number of coarse
sediment DOC  data  available, no  distinction  between fine and   coarse  sediment DOC
concentrations is supported.

Considering that fine  and coarse sediments  were observed to  have different organic carbon
content, correlation between sediment foe and DOC was investigated as  an alternate approach to
investigating differences in porewater DOC concentration between fine  and coarse sediment.  A
scatter plot of sediment  DOC versus foe  shows no correlation between these  two parameters.
Thus, DOC was specified on river mile intervals, with no distinction between fine and coarse
sediment, as shown in Figure 6-46.

6.9.2.6 Sediment/oc

Sediment foc values were specified using data from the GE  1991 Sediment Sampling and
Analysis Program  (O'Brien   and Gere,   1993).    While measurements  of sediment fOc
concentrations are also available from the USEPA  Phase 2 data, the GE 1991 data are extensive
enough to provide a good estimate of mean foe values throughout the Upper Hudson River. The
GE composites consisted of fine and coarse sediment collected over intervals of about 2 miles
downstream of Thompson  Island  Pool and about 1 mile in Thompson  Island Pool.  The
composite data were assigned river mile location corresponding to the approximate midpoint of
the  sampling interval and plotted versus river mile to investigate changes in foe along the river
(Figure  6-47).   Sediment organic carbon  content  was observed to decline with  distance
downstream from Ft. Edward.  Measured values  ranged from 6.9  to 0.3 percent for fine
(cohesive) sediment and  4.6 percent to  0.2  percent for coarse (non-cohesive) sediments with the
highest values being measured in Thompson Island Pool.

The data were  grouped  by  river mile  interval  to compute average fine and coarse sediment
concentrations for specification  in HUDTOX  (Table 6-32).   The foe values  specified  in
HUDTOX range from 3.7 percent to 1.6 percent for fine sediment and from 1.3 percent to 0.7
percent for coarse sediment.

6.9.2.7 Distribution of PCBs in sediment and water

Based on the specified  parameters influencing  the  partitioning calculations in the  HUDTOX
model, typical phase distributions of PCBs in sediment and water are  presented below for  all
PCB state variables, along with the approximate  range of distributions that may result in the
                                          109                      Limno-Tech, Inc.

-------
 model. Parameters controlling the three-phase partitioning include: KPQCC, KDOC,/OC, DOC, and
 temperature.

 Waterford was chosen to illustrate the typical summer and winter, high and low-flow ranges of
 water column partitioning behavior because it experiences the largest changes in temperature (1.1
 °C to 27.8 °C), and the largest range of observed suspended solids concentration.  Typical low
 and high-flow TSS concentrations of 5 and 100 mg/L were chosen for this illustration. Similarly,
 typical high and low flow/oc values specified are 22 and 8 percent, respectively.  Water column
 DOC was specified as 4.01 mg/L.

 The range of partitioning behavior due to  the range  of  parameter values  specified for this
 illustration is presented for  each state variable,  using water column and  sediment 3-phase
 partitioning coefficients (Table 6-33).   Note that results  are independent of the actual  PCB
 concentration.  Results are displayed as percent of PCB in each phase: truly dissolved, DOC-
 bound and sorbed to particulate organic carbon.  The apparent dissolved phase includes  truly
 dissolved and DOC-bound PCBs.

 6.9.2.8 Partitioning Summary

 Partitioning behavior  of PCB to  particulate matter and colloids  is represented in HUDTOX
 through  the  application  of equilibrium three-phase   partitioning  equations  that  compute
 distribution of PCB among water,  dissolved organic carbon and particulate organic carbon.  The
 equilibrium assumption was evaluated by USEPA (1997)  and found to be reasonable for the
 Upper Hudson River,  although evidence of non-equilibrium conditions was observed.   This
 primarily affected Fort Edward and TI Dam concentrations. Three-phase partition coefficients for
 Tri+  and  congeners estimated from USEPA Phase 2  water column  data were specified for
 HUDTOX.  Partition coefficients  were not varied spatially.  Results suggest  that with accurate
 representation of temperature, foc and DOC it is possible to  predict phase distributions  of
 individual congeners to within 45 percent for the  Upper Hudson River upstream of Thompson
 Island Dam and to within 33 percent below Thompson Island Dam.

 Because  estimates of partition  coefficients  for total PCB were not available  from previous
investigations, these were estimated based on mass weighting of values determined for Tri+ and
mono and di-chlorinated congeners.  Estimates were computed for the primary sampling stations
between Fort Edward and Waterford. A spatial pattern  in results was observed, consistent with
the relative changes in congener distributions through the system.  These results were distance
weighted to obtain an estimate for  total PCB for the entire system. Considering that total PCB is
used only for estimating total PCB transport  and is was not used for primary calibration of the
HUDTOX model, uncertainty in total PCB partitioning behavior does not affect the calibration.

6.10 VOLATILIZATION

6.10.1 Overview

Air-water exchange by volatilization is a transport pathway for water-borne PCBs in the Upper
Hudson that is explicitly represented in the HUDTOX model.  Whereas Chapter 5 presents the
                                            10                       Limno-Tech, Inc.

-------
empirical model formulations  used in the computation of  air-water  exchange, this  section
presents specification of chemical-specific and hydrodynamic parameters affecting the rate of
volatilization. An assessment of volatilization losses at dam cascades is  also presented, with the
conclusion that this  process is not large enough to warrant explicit representation in the
HUDTOX model.

6.10.2 Volatilization Mass Transfer

Volatilization affects PCB  transport in  the Upper Hudson by  serving as a net loss pathway for
water column borne PCB in the truly dissolved phase.  Air-water exchange of truly  dissolved
PCB occurs across the air-water interface of the entire river and is enhanced by induction of air in
cascades such as falls over dams.

The rate of volatilization tends  to be chemical specific and is determined by Henry's  Constant.
Volatilization is enhanced by hydrodynamically-induced and wind-driven shear stresses at  the
water surface.  Due to temperature dependencies, volatilization is also seasonally dependent,
exhibiting higher rates during warm temperature periods. Liquid phase and air-phase resistances
control  the rate of volatilization, which are dependent on the concentration and diffusivity of
PCB in  each phase.

Volatilization rates are computed in HUDTOX according to the O'Connor Dobbins formulation
presented in Chapter 5. This equation computes volatilization mass  transfer coefficients across
the air-water interface based on water column depth  and  velocity, temperature,  and chemical
specific  properties,  including  atmospheric concentrations,  molecular weight  and  Henry's
constant.  Enhanced volatilization due to cascades at dams is not represented in the model based
on a determination that this processes is of small importance  in affecting PCB transport in the
system.  This determination is summarized below.

Also presented in this section are estimates of Henry's constant and  molecular  weight obtained
from literature sources and site-specific data.

6.10.2.1 Henry's Constant and Molecular Weight

Chemical-specific properties, Henry's Constant (H) and molecular weight (MW) were  estimated
for each PCB state variable.  Values for H and MW are  presented for a wide range of PCB
congeners. For Tri+ and total PCB, estimates of these parameters were developed for specific
locations by  mass weighting congener results based on  the average  mass fraction of each
congener.

Henry's coefficients were obtained  for individual congeners from Brunner et al.  (1990).  The H
values in units of (atm-m3/mol) are  presented in Table 6-34 and Figure 6-48.   Average congener
mass fractions for the primary Upper Hudson sampling stations were computed from the GE and
USEPA data (Tables 6-35 and 6-36).  Based on these results, individual congener H values were
mass-weighted to arrive at a value for each location  specific to the GE and USEPA datasets
(Table 6-37 and Figure 6-49).  A weighted  average of these values  for  each location  was
computed based on the  number of  samples in each dataset used to determine average congener
                                                                    Limno-Tech. Inc.

-------
mass fractions.   Results reveal  a down-river pattern in H that reflects the shift in congener
distributions through the system.  H values are highest at Thompson Island Dam, reflecting the
gain in mono- and di-homologues across Thompson Island Pool.  The final  values for each
location were then distance weighted by the distance between sampling midpoints to arrive at a
final value of H for total PCB (1.85e-4) and Tri+  (1.69e-4) for application to the entire Upper
Hudson (Table 6-38).

MW is constant for each congener in a given homologue group and  is a fixed quantity.  MW
values are presented in the DEIR (Table 4-8) for each homologue group. MW was computed for
total PCB  and Tri+ by  mass-weighting congener values in  an identical  manner  as done  to
estimate H. Results of this calculation are presented  in Tables 6-39 and 6-40, and illustrated in
Figure 6-50.

A summary of the H and MW values specified for each state variable is presented in Table 6-41.

6.10.2.2 Film Transfer Coefficients

As described in Chapter 5 (Section 5.2.3), air-water chemical exchange (or volatilization) rates in
HUDTOX  are determined  through application of the  stagnant layer "two-film" theory.  As a
result,  overall volatilization rates (Kv) are controlled by liquid-phase (KL) and gas-phase (KG)
exchange coefficients acting in series (see Equation 5-18). Since these coefficients function in a
series fashion (with KG being adjusted by a chemical- and temperature-dependent Henry's Law
Constant), the smallest of these two factors may be considered to be "controlling" (or limiting)
the overall  volatilization rate.  However, even the non-limiting factor may still have a substantial
effect on the volatilization rate under conditions when both are of similar magnitudes.

In the  Upper Hudson  River, flow and environmental conditions largely determine whether the
liquid-phase or gas-phase coefficient has a more limiting effect on volatilization. The  gas-phase
tends to be limiting during cooler conditions (because KG decreases with temperature)  as well as
during higher flow conditions (because KL generally increases with flow). Conversely, the liquid-
phase tends to more limiting on volatilization during lower flow (average and below) periods and
especially as water temperatures warm up  (e.g., summer low flow conditions). Differences in
chemical-specific diffusivity (Dw) across the range of PCB congeners evaluated in this modeling
study can change the limiting phase between liquid and gas.

Determination of the liquid-phase transfer coefficient  (KL) for a specific river cross-section using
the O'Connor-Dobbins reaeration formulation (Equation 5-20) requires both depth and velocity.
Table 6-42 provides the  Leopold and Maddox (1953) coefficients that were specified for each
HUDTOX  river cross-section to estimate velocity and depth as a function of flow.  Note that
depths were  estimated for average flow  conditions and assumed to  be  constant due to  the
mitigating effects of dams on water level variations as flow changes.

6.10.2.3 Atmospheric PCB Concentrations

Given  the  air-water mass transfer rates, air-water flux depends on  the gradient between the
dissolved  water phases  and  the  atmospheric gas phase; therefore,  computation of this  flux
                                           112                       Limno-Tech, Inc.

-------
requires  specification of the  atmospheric gas phase boundary condition.  For this boundary
condition an annual average value was estimated for Tri+ from 1977-1997 and for total PCBs
and  the two congeners  from 1991-1997.  The procedure for setting this boundary condition
involved establishing a recent reference concentration based on measurement of total PCBs in the
atmosphere and back projecting from that reference value to obtain estimates of historical levels.
The  nearest and most recent reference value was the 1992 annual average atmospheric gas phase
total  PCB  value of  170  ± 86  pg/m3 determined by  Hoff et al.  (1996)  at the  Integrated
Atmospheric  Deposition  Network  (IADN)  station at  Point  Petre,  Ontario.    Historical
concentrations were determined by scaling this value to a curve developed using PCB profiles
collected in dated (1940-1981), ombrotrophic peat bogs (Rapaport and  Eisenreich,  1988) and
observed water column PCB load decay rates for rivers draining Lake Michigan watersheds from
1981-present (Marti and  Armstrong, 1990).  This scaling process produced a curve which reflects
the synthesized time series of atmospheric total PCB concentration from 1977-1997  (Figure 6-
51).   Also included in Figure 6-51 as  a  check on this approach, are  seasonal data reported  by
NYSDEC (undated) and data from Buckley and Tofflemire (1983), both of which represent  air
sampled in the vicinity of the Upper Hudson River. Additionally, the line representing historical
atmospheric PCB concentrations estimated by Mackay (1989) in conducting a modeling analysis
for Lake Ontario is included.

Ideally, the estimate of historical atmospheric concentrations for congeners or the Tri+ mixture
would be made by applying measured ratios of these constituents to the hindcast total PCBs.
This was possible for estimating BZ#4 and BZ#52 levels by using ratios reported by Hornbuckle
(personal communication, 11/18/98) for samples collected over Lake Michigan.  For Tri+, a ratio
was  determined by assuming the atmospheric gas phase concentrations for both Tri+ and total
PCBs in 1992 were in equilibrium with the dissolved phase in the water column and computing a
gas phase Tri+/total PCB ratio for 1992  on that basis. Then Tri+ was hindcast using the same
scaling curve as was  used for total PCBs in Figure 6-51.  The resulting HUDTOX boundary
condition values used for these PCB state variables are presented in tabular form on Figure 6-51.

6.10.3 Gas Exchange at Dams

A method of estimating gas exchange at river cascades presented by Cirpka et al. (1993) was
overviewed in the DEIR (USEPA, 1997) and air-water transfer of Tri+  based on this equation
was  assessed by QEA (1999).  For chemicals with small Henry's constants this model  can  be
expressed as (QEA 1999):
         GH
          Q  )
GH
 Q
  GH
+  Q  )
Cair
~H
                                                                                (6-24)
where:
       G/Q = ratio of entrained air flow rate to river flow rate
       Cd - concentration downstream of cascade
       cu = concentration upstream of cascade
                                          113                      Limno-Tech, Inc.

-------
The air flow to river flow ratio is can be estimated from the cross-section dimensions of the fall
and the river flow rate (c.f. McLachlan et al. 1990). For the two river cascades (shown as a series
of small drops) studied in Cirpka et al. (1990) these ratios were about 0.03 to 0.07 for cascades of
approximately 1 to 2 meters. The falls over dam weirs on the Upper Hudson are approximate to
these heights, varying from about 2 to 6  m,  although the nature of the falls  are  somewhat
different from  the cascades in  Cirpka et  al. (1990), equation 6-24 is assumed to  provide a
reasonable estimate of air-water mass transfer for the dams on the Upper Hudson River.  Based
on this equation, QEA (1999) estimated maximum concentration reductions due to loss at dams
to be less than 3% for Tri+.

Because volatilization at  dams  is  estimated  to  have  a small  impact  on  water column
concentrations, it was not included in the HUDTOX model.

6.11 SEDIMENT PARTICLE MIXING

Vertical mixing of sediment particles and associated porewater in the sediment bed arises from
bioturbation and other physical  processes.   The activities of infaunal organisms  inhabiting the
surface  sediments, called bioturbation, include: burrow and tube excavation and their ultimate
collapse or infilling, ingestion and excretion of sediment, plowing through the surface sediment,
and building of mounds and digging of craters (Boudreau, 1997).  As  discussed in Chapter 5,
particle mixing is  represented as a diffusional process  in  HUDTOX.  The model requires as
input, specification of a depth  over which mixing occurs, and an associated mixing rate, or
particle diffusion rate.

No direct  evidence is  available  for particle  mixing rates in the Upper Hudson  River, however,
Olsen et al. (1981) determined surface particle diffusion  rates  of approximately  1  cm2/yr in
Foundry Cove  and Lents  Cove in  the Lower Hudson River.   This is  a relatively low  rate
compared  to the ranges typically observed,  which is about from 1 to 100 cm2/yr (e.g. Boudreau
1997, Matisoff 1982).  More  specifically, Aller (1982) estimated bioturbation-induced particle
mixing rates in Narragansett Bay to range from 5 to 32 cm2/yr,  Brownawell (1986) estimated a
biodiffusion coefficient of 9.4 cm2/yr in Buzzards Bay, and Thibodeaux et al. (1990) estimated
biodiffusion coefficients of 9-13 cm2/yr. These authors suggest that bioturbation-induced particle
mixing can occur to a depth of 6-10 cm and that benthic organism density and associated mixing
generally decreases with depth from the sediment surface.

Particle mixing depths are  often estimated by inspection of vertical concentration profiles of
tracer material, often radionuclides such as 210Pb, l37Cs, or 7Be.  Observation of contaminant
profiles can also provide an indication of mixed depth.  Finely section sediment cores collected
by USEPA in 1992 and by GE in 1998 (QEA 1999) provide  a means to qualitatively assess
mixed depths.  Inspection of 137Cs  and PCB profiles at five high-resolution core sites in the
Upper Hudson River, shown in Figure 3-53  in the DEIR (USEPA, 1997), suggests mixed depths
may be greater than 20 cm in some locations.

Figures 6-52a-c, presented by QEA (1999), show PCB concentration profiles for 27 sediment
cores  collected in 1998.  Mixed depths appear to vary widely, with a number of cores showing
little or no gradient to 10 cm or more.  Non-cohesive  sediments  are likely  less mixed  due to
                                           114                       Limno-Tech, Inc.

-------
lower  bulk density,  larger grain sizes, and  reduced  sediment deposition relative to  cohesive
sediments.  Due to  the variability in mixed depths  and particle mixing rates, there is large
uncertainty associated with the parameterization of particle mixing in the model.

Considering  the  uncertainty in sediment  mixing depth, this parameter was considered  a
calibration parameter and  was varied spatially to achieve reasonable fits to long-term  sediment
trajectories (Chapter 7).

6.12 DECHLORINATION

Anaerobic and aerobic dechlorination processes have the potential to  alter  PCB  congener
distribution in the water column and sediments. These processes are of particular concern for the
historical  calibration  as  the state variable, Tri+,  is  subject  to  potential mass  loss due  to
dechlorination in the sediments. The influence of dechlorination on the sediment  inventory of
PCBs has been extensively assessed as presented in the DEIR (USEPA, 1997).  This assessment
compared congener patterns in the sediment to known source material (primarily Aroclor 1242 at
Fort Edward)  and found little evidence for extensive dechlorination.  Results showed minimal
aerobic dechlorination and suggested that anaerobic dechlorination of Hudson River sediments is
limited to meta- and para- chlorines, which limits its ability to reduce sediment PCB mass.  The
DEIR concluded that dechlorination mass losses are theoretically limited to 26 percent in Hudson
River sediment.  Dechlorination losses of more than 10 percent were limited to concentrations
greater than 30 mg/kg and below this  level, dechlorination losses were frequently observed to be
zero, compared to the original Aroclor 1242 source material.   Sediments as old as 35 years were
found with little or no dechlorination. No sediments were found with dechlorination mass loss
greater than 25 percent, based on change in molecular weight, and the median mass loss was  7
percent since the time of PCB deposition. The mean mass loss was 8 percent.

Based  on the .interpretations provided by USEPA  (1997)  in the DEIR,  which are partially
summarized above, the overall impact of dechlorination on  the  historical and future fate  of
sediment PCB reservoirs in the Upper Hudson is small.  Therefore, the HUDTOX model does
not include representation of dechlorination processes.

6.13 SEDIMENT-WATER MASS TRANSFER

6.13.1  Overview

Sediment  to water PCB mass transfer in  the HUDTOX model occurs due to either porewater
diffusion,  paniculate phase mass transfer, or  by sediment resuspension, as  discussed in Chapter
5.   During high  flow periods, sediment resuspension  can be the dominant  sediment-water
transfer mechanism, however, under low flow conditions resuspension contributions can be small
relative to other mechanisms giving  rise to  transfer of PCB  from sediment  to  water.  These
include numerous processes that act on paniculate and dissolved phase PCBs.  Possible transfer
mechanisms for the dissolved phase include:
       •      molecular diffusion of dissolved phase PCB in porewater;

       •      diffusion of colloid-bound PCBs in porewater;
                                           115                       Limno-Tech, Inc.

-------
        •      groundwater advection up through the sediment bed;

        •      hydrodynamically induced advective pumping; and,

        •      biologically enhanced porewater transport.

 Non  flow dependent transfer mechanisms may act on paniculate phase PCBs, resulting in
 subsequent desorption to  the water column at the  sediment-water interface.   These processes
 may include:

        •      bioturbation by benthic organisms;

        •      emergence and uprooting of macrophytes;

        •      physical disturbance from wind waves or fish activity; as well as,

        •      direct desorption from surface sediments to the water column.

The magnitude of these various processes can vary seasonally as a function of temperature and
climatological  conditions.   Biologically  enhanced  sediment-water  transfer  of  PCBs  is
temperature  dependent due to  increased  biological  activity  during  warm  temperatures.
Groundwater  advection  transfer  will  vary  with  the  groundwater   hydraulic   gradient.
Measurements of groundwater seepage in the Upper Hudson River indicated large spatial and
temporal variability, ranging from negative (river losses) to  positive groundwater inflows.  The
highest groundwater inflow  rates were measured in late May and early June (HSI Geotrans
 1997). In the absence of any  physical disturbance of the upper sediment layer (e.g., bioturbation,
advection or dispersion), exchange of PCBs between  the sediments and  water takes place  by
molecular diffusion (for dissolved material) or Brownian diffusion (for colloidal bound material).
Valasaraj  et  al.  (1997), using a water diffusivity of 5.6 x  10~6 cm2/sec, estimated  that  mass
transfer rates due to molecular diffusion applied to the dissolved phase of a chemical in sediment
porewater would be on the  order of 0.02 cm/day.  Application of this  mass  transfer rate to
porewater concentrations of PCBs results in a relatively small mass flux from sediments to water.

Direct desorption of paniculate phase PCBs and subsequent  transfer to the water column can be
enhanced  by bioturbation  of surface sediments via the following sequence of processes:  first,
particles can be transported  by mixing processes from depth to the sediment-water interface;
second, while residing briefly at this interface, particles can desorb a fraction of the sorbed  PCB
before being mixed back into deeper sediments; and finally, desorbed PCB can move through the
benthic boundary layer into  the overlying water column (Portielje  and Lijklema,  1999;
Thibodeaux,  1996). Several  authors have shown these processes can increase effective chemical
mass fluxes across the sediment-water interface by a factors of 10-1000 (e.g., Thibodeaux, 1996;
Nadal, 1998; Thorns et al., 1995; Reible et al., 1991). Horn et al. (1979) suggested that this non-
flow-dependent sediment-water exchange process is important  for PCBs in the Hudson River.
They further  suggested that approximately half of PCB transport in the Hudson River occurs at
low to moderate flows and is not the result of solids scour from the sediment bed.  In comparison
to their calculation of molecular diffusion mass transfer of 0.02 cm/day, Valasaraj et al. (1997)
                                           116                       Limno-Tech, Inc.

-------
estimated that a biodiffusion (bioturbation-induced mass transfer of porewater chemical) mass
transfer rate would be approximately 12 cm/day.

Analysis  of low flow PCB  load  gain  of across  TIP reveals that sediment-water transfer
mechanisms are occurring at rates much  greater than those typically associated with molecular
diffusion. This indicates that transfer mechanisms other than molecular diffusion are operative at
high rates under low flow conditions.  While individual sediment-water transfer processes (such
as those listed above) have been  extensively studied  and measured in  other systems (e.g.,
Thibodeaux, 1996), direct measurement of these processes has not been conducted for the Upper
Hudson River.  Due to a lack of site-specific information, development of a process-level model
to describe low-flow sediment to water mass transfer was not supported. Therefore, an empirical
modeling approach was adopted to describe effective sediment-water mass transfer of PCB under
low flow conditions.

A seasonally-variable mass transfer rate coefficient  operating on porewater PCBs was derived
from  observations of  PCB  load gain across  Thompson Island Pool under  low flow.   This
effective mass transfer coefficient (k/) represents the combined effect of all the various  processes
contributing  to low-flow sediment-water transfer of PCBs.  The k/  time series derived from
observations describes the average  low-flow mass transfer occurring during  specific intervals
over which average values were  computed.  This approach provides  a  reasonable estimate  of
mean behavior and was used successfully in the historical calibration to Tri+.

In attempting to apply the calibrated  Tri+  model to individual congeners, it was found that a
single porewater  PCB mass transfer  coefficient could  not be used to  simultaneously  model
multiple PCB congeners. Differences in  sediment-water partitioning behavior apparently cause
differences  in  observed effective  sediment  water  mass transfer coefficients  for individual
congeners. Calibration to individual congeners could have been achieved by deriving congener-
specific mass transfer coefficients,  however, this would have essentially  resulted in multiple
calibrations that  are  not mutually consistent.  In  order to maximize use  of the  congener
simulations in evaluating the historical calibration to Tri+, a modeling approach was sought that
could  simultaneously  describe   sediment-water  mass  transfer for  the  range  of  congener
partitioning behavior represented by the five congeners chosen for modeling.

Analysis of congener patterns in sediment porewater, on sediment particles and in the Thompson
Island Pool load  gain suggested  that the low flow load gain is dominated by  particle-based
processes.  This analysis also suggested that separation  of the porewater and particulate phase
mass  transfer processes may provide a model capable of describing a range of PCB congeners
simultaneously,  with varying only  congener-specific  chemical properties  in  model  inputs.
Separate mass  transfer coefficients for  the particulate and  dissolved  phases were  therefore
derived such that the combined contribution to overall sediment-water mass transfer resulted in
the same amount.  This was done  by picking a ratio between these  processes that optimized
agreement  with the observed  congener distribution in the  water column  at Thompson  Island
Dam.

This  approach,  while subject to a  number  of large  uncertainties, permitted a reasonable
simulation of all five congener state variables, in addition to  the principal state variable, Tri+.
                                           117                       Limno-Tech, Inc.

-------
The historical Tri+ calibration was run with the separate paniculate and porewater mass transfer
coefficients and compared to the calibration achieved with the single kf function.  Results are
presented in Chapter 7.

Because results for simulations with the computed porewater and paniculate mass transfer
coefficients showed good performance for BZ#28 and BZ#52, the two congeners most like Tri+,
the historical calibration to Tri+ based on the k/ series was accepted as the model  calibration
(Chapter 7) and used for model forecasting (Chapter 8).

The analysis of sediment-water mass transfer rates is summarized below.

       •     To describe low-flow  sediment-water transfer of Tri+ for  the historical
             calibration, an empirical modeling approach was used due to a lack of site-
             specific information on individual processes.

       •     A seasonally-variable  mass transfer rate coefficient  was computed  from
             observations of low flow load gain across Thompson Island Pool, which
             was used in the historical calibration to Tri+.

       •     The application of the  model to individual congeners provided insights as
             to the relative importance of dissolved phase versus paniculate phase mass
             transfer processes.

       •     While representation  of these processes provided  better  agreement  to
             individual congener data, results for Tri+ tended to confirm the historical
             calibration based on the effective porewater mass transfer coefficient.

This section presents the development of the effective mass transfer function, kf, and subsequent
investigation of sediment-water  transfer  of congeners.   Sensitivity  analysis are  presented in
Chapter 7 that explore the significance of implementing separate paniculate and porewater mass
transfer processes to describe congener load gain.

6.13.2  Calculation of kf for Tri+

6.13.2.1  Data

The seasonally-variable  low  flow effective mass transfer coefficient, kf, was derived  from
observations of Tri+ load gain across Thompson Island Pool under non-resuspending conditions.
Observations of low flow load gain, determined from paired (same day)  daily average PCB
concentrations at Fort Edward and Thompson Island Dam, were segregated by flow and TSS
concentrations.  Based on the observed knee in the TSS-flow correlation at approximately 10,000
(Figure 6-12), sediment resuspension  is considered significant at flows above 10,000 cfs. Below
10,000 cfs, PCB load gain observations coincident with TSS less than or equal to 10  mg/L were
assumed to minimally affected by sediment resuspension.  To evaluate this  assumption,  the
relationship between same-day TSS concentrations at Thompson Island Dam and  Fort Edward
was examined (Figure 6-53).  The correlation exhibits high variability, with approximately equal
distribution about the  1:1 line, suggesting that on average, TSS  transport may be  considered
                                           118                      Limno-Tech, Inc.

-------
conservative in Thompson Island Pool at flows less than 10,000 cfs and TSS less than 10 mg/L.
A regression of these data suggest that at very low concentrations, TSS is slightly higher at
Thompson Island Dam,  however,  at concentrations above about 3 mg/L, concentrations at
Thompson Island Dam  are lower than  at Ft. Edward.   The  apparent  lack  of significant
resuspension  contributions in these data suggests that  use  of data under these  conditions for
computing low-flow sediment-water mass transfer coefficients is reasonable. Due to the elevated
loading of PCBs observed at Fort Edward beginning in September, 1991 from the Allen Mill gate
failure, none of the 1991-92 data were used in any of the evaluations of mass transfer rates.  The
large pulse loading of PCBs influenced PCB loads at Fort Edward for the later part of 1991 and
early 1992. The effect of this load on surface sediment concentrations in Thompson Island Pool
is unknown, imparting additional uncertainty to calculations of load gain across Thompson Island
Pool for  this  period, therefore the mass transfer analysis was limited to observations collected
from 1993 through  1997.  Observations of load gain across the Thompson Island Pool for this
period  were based on daily average PCB  concentrations at  Fort Edward and Thompson Island
Dam.  At Thompson Island Dam, the bias-corrected concentrations were used, as described in
Section 6.3.

The effective sediment-water mass transfer coefficient relates observations of low flow load gain
to surficial sediment concentrations.  In order to make use of the 1993 -1997 observations of
load gain, estimation of corresponding sediment  concentrations  was required.  The surficial
sediment concentration for Tri+ was estimated for each year by applying a first order rate of
decline computed from observed poolwide average surficial sediment concentrations from 1 99 1
to 1998 (k - "0.076 yr"1). The 1991 average concentrations were computed from the 0-5 cm layer
concentrations in GE 1991 composite sediment data, which were collected before the Allen Mill
Event occurred in the fall of 1991. This event increased surface sediment concentrations by an
unknown  amount and  produced  a noticeable increase on observed PCB  load gain  across
Thompson Island Pool  . The average Poolwide 1998 sediment surface sediment  concentrations
reported  by QEA for cohesive  and non-cohesive sediment (1999) were used.   The unknown
perturbation of sediment concentrations from the Allen Mill Event in 1991 imparts uncertainty to
the estimated rate of sediment concentration declines from  1991 to 1998. Estimated poolwide
sediment and  porewater concentrations for all modeled PCB groups using this approach are
shown  in Table 6-43 and 6-44.

6.13.2.2  Approach

To  compute  the effective mass  transfer coefficient for Tri+,  Thompson Island Pool  was
represented as a single control volume  and the following mass balance equation for the water
column was employed to relate observed load gain to sediment concentrations (Equation 6-25).
                    k             -, e|,244659                           (6_25)
                         A,\       Cf
                                    'PW
where:
             kf      - effective mass transfer rate (cm/day)
                  o = product of flow and concentration at TI Dam, (cfs • mg/L)
                                          119                      Limno-Tech, Inc.

-------
                    = product of flow and concentration at Fort Edward, (cfs • mg/L)
              Cpw   = Apparent porewater Tri+ concentration (mg/L)
              As     = Surficial sediment area (m2)
              244.659 = Conversion factor to cm/day
Application of this simplistic mass balance calculation implies the following assumptions.

    1.  The time of travel between upstream and downstream locations is less than one
       day and therefore samples collected at Fort Edward and TI Dam on the same day
       can be reasonably assumed to represent the same parcel of water.

    2.  Volatilization losses across TIP do not significantly affect the observations of low
       flow load gain.

    3.  The gradient of porewater to water column concentrations can be approximated
       with the porewater concentration. Because porewater concentrations are typically
       at least 1 to 2 orders of magnitude greater than water  column concentrations, this
       assumption is valid.

For consistency between the calculation of kf values and implementation in HUDTOX, sediment
surface area was calculated based on the model segmentation. The percentage of cohesive and
non-cohesive sediment area were used to determine area weighted average values for sediment
properties,  such as:  bulk  density,  porosity, foc, and DOC  concentration  (Table  6-45).  To
compute the porewater PCB concentration,  the 3-phase partitioning equations presented in
Chapter 5 were employed  with the input values in Table  6-43 and the Phase 2 water column
partition coefficients.  Partition coefficients were temperature-adjusted according to the water
column temperature time series in the model (Section 6.8).

Equation 6-25  Avas solved for each individual observation of paired (same day) concentrations at
Fort Edward and  Thompson Island Dam, censored  as described in the above section.  The kf
values for Tri+ ranged from -1.0 to 65 cm/d.  Negative results occurred for days where lower
concentrations were observed at Thompson Island Dam than at Fort Edward.  This affected 7
percent of the observations and these results were excluded in developing the effective  mass
transfer function. (Figure 6-54)

6.13.2.3 kf Results

Individual values  of k/ were plotted versus  Julian day to discern the average seasonal pattern in
low flow load gain for the  1993-1997 period.  The kf values were distinctly higher in summer
months relative to most of the year (Figure 6-54).  Observed high values in March and  April
(Julian days -60-120) maybe a result  of resuspension activity during the spring runoff period,
either preceding these data, or not represented by the associated TSS measurements. The average
mass transfer rate in specific time intervals was used to develop a variable kf annual time series,
which was  incorporated into the HUDTOX model. The approximate mean value (10.2  cm/d) of
the low  temperature  period, September through  April,  was applied for  these months. The
resulting kf series shows that from early May to mid June, kf  increases from about 10 to 25 cm/d
and declines to about 10 cm/d at the end of August (Table 6-46, Figure 6-55). The  seasonal
                                           120                       Limno-Tech, Inc.

-------
dependence on the low-flow mass transfer rate is clearly evident, with the peak mass transfer
occurring in mid June.   The causal factors leading to the peak rate occurring in mid June are
poorly understood.  Peak water column temperature is observed in July (Figure 6-41).  It is
notable that the timing of the peak mass transfer rates are generally coincident with the timing of
the highest measured groundwater influx rates (HSI Geotrans, 1997).

6.13.2.4  Implementation in HUDTOX

The sediment-water transfer of porewater PCBs is computed in HUDTOX by Equation 5-22. To
correctly implement the fc/time series in HUDTOX, Equation 5-22 was rearranged to achieve the
same form of expression of the mass transfer coefficient as in Equation 6-26. This shows that kf
is equal to the following terms.
                          E • nu
                     kf=~ - -                                                  (6-26)
The HUDTOX  model  input describing the transfer rate  is  the  dispersion coefficient,  E.
Therefore, Equation 6-26 was solved for E for each value of k/ in the annual time series and the
resulting series for E was input to HUDTOX.  The mixing length, Ly, was specified as 0.02 m.
The average porosity between sediments and water (ny) computed based on the average sediment
porosity of 0.527 and water column porosity of -1.0 is 0.7635.

6.13.3 Analysis of congener and total PCB mass transfer coefficients

The kf series  developed as  presented above  for Tri+ was  used in  the historical 1977-1997
calibration.  Following the historical calibration to Tri+, the HUDTOX model was tested through
short-term  hindcast  applications to total PCB  and five congeners (BZ#4,  BZ#28, BZ#52,
BZ#90+101, and BZ#138) for 1991 to!997. BZ#4 exhibits the largest deviation in environmental
behavior relative to Tri+ and BZ#4 is not a component of Tri+.  All of the other congeners are
included in  Tri+.   BZ#4 is  the least hydrophobic of these congeners and also has the highest
volatility.

Initial investigations revealed that the porewater mass transfer coefficient, kf, developed for Tri+
was not applicable to all congeners. This is apparent through comparison of observed effective
sediment-water mass transfer rates for total PCB and the five  congeners.   These rates  were
estimated following the same approach as for Tri+ explained in the previous section  (Figure 6-
56) and plotted versus the Rvalues for total PCB. Sediment concentrations used in calculation of
kf for congeners were computed  as described above (Table 6-43 and 6-44). Results for BZ#4
show significantly lower kf  values relative  to the other results.  BZ#28 results were in best
agreement with total PCB results, although still  noticeably higher.   Tri+, BZ#52, 101+90 and
138 show higher values relative to total PCB.  Thus, the fy-for total PCB over-predicts BZ#4 load
gain, while under-predicting load gain for Tri+, BZ#28, 52, 90+101, and 138.  The differences in
apparent /^values among congeners is also shown through comparison of results for 14 selected
days on which quantitations were available for all five congeners  at Fort Edward and Thompson
Island Dam (Figure 6-57).
                                          121                       Limno-Tech, Inc.

-------
 An objective of modeling congeners was to evaluate the Tri+ calibration for PCBs exhibiting
 different environmental behavior. While congener-specific mass transfer coefficients could have
 been developed for the short-term hindcast applications, this would have  somewhat diminished
 the  use  of the model for  this purpose  because  in effect individual calibrations would be
 developed for  each congener.  Therefore, the sediment-water  mass transfer processes were
 investigated through use of the congener data with the goal of representing sediment-water mass
 transfer processes in a consistent manner across all PCB groups  modeled (i.e. Tri+, Total PCB
 and individual congeners). This would allow simultaneous application of the Tri+ calibration to
 all congeners, varying only congener-specific chemical properties.

 Differences in  partitioning  behavior  among congeners was considered in order to explain
 differences in effective mass transfer rates.  The water column  partition coefficients estimated
 from the USEPA Phase 2 water column data are compared to the estimates from the GE 1991
 sediment data  in Figure 6-58.  The estimates of effective Rvalues for  individual congeners used
 pore  water congener concentrations estimated through application of the sediment partition
 coefficients from the GE  1991 sediment data.  Large differences in  estimated sediment-water
 partitioning coefficients exist for the  lighter  congeners,  while the  heavier  congeners show
 approximately the same values in the  water and  sediment.  While initially congener-specific
 estimates of kf used sediment partition coefficients, use of the water column values to compute
 effective mass transfer values did not result in convergence of these  values among congeners.
 This suggests that there are factors other than influences of sediment-water partitioning on pore
 water PCB concentrations controlling the relative flux of congeners out of the sediments.

 Observation of the relative distributions of PCBs  in  pore water, surface sediments, and  in the
 water column at Thompson Island Dam suggest that a pore water source alone cannot account for
 the observed congener patterns in the water column.  This is illustrated through comparison of
 the expected pore water  distribution in sediment pore water,  the measured distribution  on
 paniculate sediments, and  measured distribution in the Thompson Island Pool PCB load gain for
 15 congeners (Figure 6-60).  These congeners are those for which 3-phase partition  coefficients
were estimated  from the Phase 2 water  column data and the GE 1991  sediment data (USEPA
 1997).  Inspection of the congener distribution in these three  compartments suggests  that a
combination of dissolved phase and particulate phase pathways is required to match the observed
congener pattern at Thompson Island Dam.

Using pore water transfer only (represented by kf) means that relative sediment-water flux of the
congeners under  non-resuspending conditions is  fixed by their relative concentrations  and
sediment-water  partitioning, which  does not appear  to be the case.   By implementation of a
particulate transfer mechanism in the description of sediment-water mass transfer,  the relative
flux  of congeners from the  sediments  is determined not only by concentrations and partition
coefficients, but also by the relative ratio of the particulate and pore water transfer mechanisms.

The mechanisms contributing to enhanced sediment-water transfer of PCBs are due to physical
perturbations of the surficial sediments  (see list of possible mechanisms above),  and are largely
independent of chemical properties (assuming dynamic desorption effects are small). The effect
of these processes,  however,  varies by congener due to differences in partitioning  behavior.
Therefore, modeling the relative sediment-water flux ratios of the congeners may be possible by
                                           122                       Limno-Tech, Inc.

-------
 representing the relative contribution of dissolved and  paniculate phase PCBs and congener-
 specific partition behavior. It was postulated that the mechanisms affecting sediment particles at
 the sediment-water interface was resulting in desorption of PCBs from the sediment to the water
 column.   The relative degrees of desorption among congeners was assumed to occur in ratios
 determined by equilibrium phase partitioning on suspended solids in the water column.

 6.13.4 Estimation of Particulate and Pore water Mass Transfer Rates

 As discussed above,  in estimating separate  mass transfer rates for particulate and pore water
 pathways, the sediment partition coefficients derived from the GE 1991 sediment data were used.
 Differences in mass transfer among PCB congeners were assumed to be due only to chemical
 specific properties.  That is, the resulting rates reflect  differences among congeners resulting
 directly from differences in their partitioning behavior.

 Similar to the development of kf for pore water mass  transfer, separation of pore water and
 particulate transfer processes was also represented by simple mass transfer coefficients, which
 combine to produce the total sediment-water flux for Tri+ computed by kf.

 The load gain represented by the effective mass transfer (kf) can be assumed to represent the  sum
 of the load gain of particulate  pathway processes  and the load  gain  of pore water pathway
 processes as in the equation below:
                     ALp+ALd=ALkf                                              (6-27)

 where:

              ALp = load gain from particulate pathway
              ALd = load gain from pore water pathway
              ALkf = total load gain produced by the effective mass transfer rate
 The individual load  terms in this equation can  be expressed  in terms of their respective mass
 transfer rates (Equation 6-28).
                     (kp-As-Cp-p-df)+(kd-As-Cd)=(kf-As.Cd)                    (6-28)

 where:
              kp  = particulate mass transfer rate (cm/day)
              kd  = pore water mass transfer rate (cm/day)
              kf  = effective mass transfer rate  (cm/day)
             A  = surficial area (m2)
              Cp = particulate PCB concentration in the  sediment (mgpcB/KgSOHd)
              p  = sediment dry bulk density (Kgsoiid /Lbuik)
              Cd = apparent dissolved PCB concentration (mgpcB/Lporewater)
             df  = fraction dissolved in the water column
This equation assumes the water component of the concentration gradients are negligible.  The df
 term  reflects the  assumption that  desorption  occurs  from  sediment  particles  according  to
equilibrium partitioning in the water column (based on partition coefficients estimated from the
                                           123                      Limno-Tech, Inc.

-------
Phase 2 water column data). The kj and kp terms can be solved for through specification of R.
This produces two equations  and  two unknowns, from which values of kp and kd can  be
determined (Equation 6-29, 6-30).
                         _      —
                                  .Cd

                                                                                 (6-30)
The value of R was determined through congener pattern matching.  An initial value of R was
specified and kd and kp were solved for using equation 6-29 and 6-30. Then, the relative percent
load gain (RPf) for each of the 15 congeners) for which water column and sediment partition
coefficients were estimated (Table 6-47) was computed according to Equation 6-3 1 .

  RP =
The RP values were plotted for each congener, representing the computed congener distribution
in the TIP load gain,  which was matched to  the observed distribution.  R was optimized to
minimize  cumulative  squared error between computed and observed RP  for each of the  15
congeners as shown in Figure 6-60 for summer and non-summer periods.

The value of R was 710 for summer conditions (June through August) and 725 for non-summer
conditions (September through May). The resulting mass-transfer coefficients for each modeled
congener are shown in Table 6-47.  These rates were used in  short-term hindcast applications
presented in Chapter 7. Results showed that this approach gave reasonable results, however, did
not completely explain differences in sediment-water mass transfer between congeners.
                                          124                      Limno-Tech, Inc.

-------
Chapter 7

-------
7. MASS BALANCE MODEL CALIBRATION

7.1 OVERVIEW

Chapter 5  presented development of the Hudson River Toxic Chemical Model  (HUDTOX)
which  included the conceptual framework, governing equations  and spatial-temporal  scales.
Chapter 6 presented the organization and analysis of available data to specify the required model
forcing functions, initial conditions, rate coefficients and state variables.  This chapter presents
results from calibration of the HUDTOX model to site-specific data for solids and PCB state
variables.   It also includes results from sensitivity analyses for important model inputs and
process mechanisms. The calibration results in this chapter provide the foundation for use of the
HUDTOX  model in  conducting forecast  simulations to  estimate long-term responses  to
continued No Action and impacts due to a 100-year peak flow in Chapter 8.

The principal model application was a long-term historical  calibration for Tri+ for a 21-year
period from 1977 to 1997.  The historical calibration was tested through short-term hindcast
applications for total PCBs and five individual congeners from 1991  to 1997. Consistent with
the Reassessment questions, emphasis was placed  on calibration to long-term trends in sediment
and water column PCB concentrations. Additional, independent model validation is described in
Chapter 9.

The following major sections are included in Chapter 7:

       7.2    Calibration Strategy

       7.3    Solids Dynamics

       7.4    Historical Tri+ Calibration

       7.5    Sensitivity Analyses

       7.6    1991-1997 Hindcast Applications

       7.7    Calibration Findings and Conclusions

The model was successful in its primary objective, representation of long-term trends in PCB
behavior in the Upper Hudson  River.   This was  best demonstrated by comparison to 21-year
trends in surface sediment Tri+ concentrations and in-river solids and Tri-i- mass transport. Tests
of model performance conducted for the 1991-1997 data-intensive period were also  successful in
demonstrating model reliability. Localized and transient discrepancies between the model and
data  were  viewed as having  minimal significance  to the  model's reliability for  long-term
forecasting as required for the Reassessment.  Many different metrics were used to demonstrate
model reliability and they should be used collectively in a "weight of evidence" approach.
                                          125                      Limno-Tech, Inc.

-------
 7.2 CALIBRATION STRATEGY

 The calibration strategy can  be described as minimal and conservative.  It was minimal  in the
 sense that external inputs and internal model parameters were determined independently  to the
 fullest extent possible from site-specific data (as presented in Chapter 6), and only a minimal
 number of parameters were determined through model calibration.  It was conservative  in the
 sense that parameters determined through model calibration were held spatially and temporally
 constant unless there was supporting information to the contrary.

 A 21-year historical  calibration was the principal development vehicle for the  model.   The
 calibration focused on representing long-term Tri+ trends in water and sediment. The Tri+ form
 was the principal focus of the calibration because comparable measurements were available in all
 calibration datasets. Tri+ is the sum of the tri and higher chlorinated PCB congeners. Details of
 selection of Tri+ as the principal state variable are presented  in  Chapter 6.  Also presented in
 Chapter 6 is the development  of all inputs  for model flows and loadings for the calibration
 period.

 The following factors were the most important in controlling long-term trends in sediment and
 water column Tri+ concentrations in the Upper Hudson River:

       •      Hydrology;

       •      External solids loads;

       •      External Tri+ loads;

       •      Tri+ partitioning;

       •      Sediment-water mass transfer under non-scouring flow conditions;

       •      Solids burial rates; and,

       •      Particle mixing depth in the sediments.

The first three of these factors are external inputs largely defined by data,  and the last four are
internal processes within the river defined by data, scientific  literature and calibration.  Long-
term solids burial rates  were  the principal factor controlling long-term Tri+ trends in the  river.
Partitioning controls the distribution of Tri+ mass between sorbed and  dissolved phases, thus
influencing sediment-water and water-air  mass transfer, and bioavailability to fish.  Sediment-
water mass transfer under non-scouring flow  conditions was found to be the principal source of
Tri+ inputs to the water column. Particle mixing depth strongly influenced long-term responses
and the vertical distribution of Tri+ in the sediments. With the exception of solids burial rates
and particle mixed depth, all model inputs and parameter values were  determined using site-
specific data and were not adjusted during the model calibration.
                                            126                       Limno-Tech, Inc.

-------
The principal datasets used in calibration of HUDTOX were the following:

       •      Tri+ surface sediment concentration trends;

       •      Measured solids burial rates from dated sediment cores;

       •      Computed solids burial rates from a sediment transport model;

       •      In-river solids and Tri+ mass transport at high and low flows; and,

       •      Solids and Tri+ water column concentrations.

The historical calibration was conducted simultaneously for solids and Tri+. Operationally, the
approach consisted of adjusting  four model  parameters: gross settling velocities into cohesive
and non-cohesive sediment areas; resuspension rates from non-cohesive sediment areas; depth of
particle mixing in the sediment bed; and, magnitude of sediment particle mixing.

Based on the flow balance and solids loads developed in Chapter 6, solids and Tri+ dynamics in
HUDTOX were calibrated to achieve long-term results consistent with the calibration datasets
listed above.  In the simultaneous solids and Tri+ calibration, primary emphasis was placed on
representing long-term historical rates of decline for Tri+  in the  water column  and surface
sediments from 1977 to 1997. The calibration sought to describe mean high and low flow solids
and Tri+ dynamics in the river.  Calibration  to short-term event dynamics was not emphasized
because detailed  representation of short-term event impacts was not necessary to answer the
principal Reassessment questions.

The model calibration was tested with a short-term 1991-1997  hindcast application for total
PCBs and five congeners (BZ#4, BZ#28, BZ#52, BZ#[90+101]  and BZ#138).  The physical-
chemical properties of the five congeners span a wide range of  partitioning and volatilization
behavior.  These important differences  in environmental behaviors provided opportunity to test
the rigor of the Tri+  calibration, especially  sediment-water and  air-water exchange processes.
For example,  model  results for a highly  volatile congener may be  more sensitive to errors in
sediment-water exchange than a  less volatile congener. Likewise, model results for a strongly
partitioning congener  may be more sensitive to errors in particle-based PCB processes such as
settling than a weaker partitioning congener.

7.3 SOLIDS DYNAMICS

7.3.1  Calibration Approach

Solids dynamics in the Upper Hudson River are strongly driven by hydrology and external solids
loads.  Hydrology  and external  solids  loads were developed in Chapter  6 using data-based
balances for tributary and  mainstem flows  and  solids  mass transport. Internal processes of
settling and resuspension largely determine the long-term PCB fate in the sediment bed.   Long-
term sediment burial or erosion rates are  determined by  the  net effect of deposition  and
resuspension processes.  The calibration approach for solids dynamics in HUDTOX consisted of
adjusting constant gross settling velocities for cohesive and non-cohesive  sediment areas, and


                                           127                       Limno-Tech, Inc.

-------
resuspension rates from non-cohesive sediment areas. Flow-driven resuspension from cohesive
sediment areas was computed internally in the model using algorithms based on the Depth of
Scour Model.

Solids burial  rates  were determined  by model calibration  using  the  following principal
constraints:

       •      Measured burial rates from dated sediment cores;

       •      Computed burial rates from a sediment transport model;

       •      Tri+ surface sediment concentration trends; and,

       •      In-river solids and Tri+ mass transport at high and low flows.

The first two constraints are described below.

Information on solids burial rates was available from two sources: first, measurements from eight
high resolution sediment cores (USEPA, 1997);  and second, results from SEDZL, a coupled
hydrodynamic-sediment  transport  model   for  the  Upper  Hudson  River  (Quantitative
Environmental Analysis, 1999). There are limitations to the high-resolution sediment cores that
preclude direct use of these data as calibration inputs. The cores are few in number and are not
considered representative of average  solids  burial rates on the spatial scale of the HUDTOX
model.  Furthermore, measurements  from these  cores represent burial rates only  in cohesive
sediment areas. Therefore these data were used as upper bounds on burial rates and as only one
of four sources of calibration guidance.

The calibration of solids dynamics in the model was also guided by computed solids burial rates
from  SEDZL,  a coupled hydrodynamic-sediment  transport model  for the Upper Hudson River
developed by General Electric Company contractors (QEA, 1999).  Flow and solids load inputs
to  the SEDZL  model were  developed using  essentially  the same methods and data as
development of flow and load inputs to HUDTOX  and  hence, results are transferable.  The
SEDZL results were in general agreement with estimated burial rates from the USEPA high-
resolution sediment cores (QEA, 1999).  SEDZL results were within a factor of two of measured
burial rates from  all but one  of the high-resolution sediment cores.  Agreement was within a
factor of five for  the remaining sediment core. The  SEDZL  model results contain  uncertainty,
however, due  to  limited data and large uncertainty in model inputs (especially solids loads
downstream of Thompson Island Pool).  These uncertainties affect long-term solids burial rates
in both cohesive  and non-cohesive sediment areas.  These limitations notwithstanding,  results
from  the SEDZL model were considered reasonable and the best available estimates of solids
burial rates on a reach-average basis.

Ultimately, solids burial rates were determined through model calibration using available site
specific information for the four principal constraints  listed above. The model calibration led to
additional upward adjustment of low flow solids loadings between Schuylerville and Waterford,
beyond the estimates presented in Chapter 6.  This adjustment was considered to be within the
                                           128                       Limno-Tech, Inc.

-------
large range of uncertainty in tributary loadings estimated from the sparse available data for the
major tributaries downstream of the Thompson Island Dam.

7.3.2 Solids Calibration Results

Values  for all  solids calibration input parameters are  presented in Tables 7-1 and 7.2. The
calibration of the solids dynamics is demonstrated for:

       •      Long-term solids burial rates;

       •      In-river solids mass transport at low and high flow;

       •      Water column solids concentration time series from 1977 to 1997;

       •      Solids mass balances for the Spring 1994 high-flow event;

       •      Water column solids concentrations during several high flow periods; and,

       •      Scatter  plots   and  cumulative  probability   distributions  of   solids
              concentrations at low and high flow.

Each is discussed below.

7.3.2.1 Burial Rates

Model calibration results for long-term, reach-average burial rates in cohesive and non-cohesive
sediment areas are presented in Figure 7-1.  Over the calibration period, the HUDTOX  model
represents the Upper Hudson River as a whole to be net depositional, based on  the assumption
underlying development of tributary solids loads in Chapter 6.  Computed solids burial rates are
generally an order of magnitude larger in cohesive sediments (0.24 to 1.50 cm/yr) than in non-
cohesive sediments (0.04  to 0.10 cm/yr).  No results for cohesive sediments are reported for the
Federal Dam reach because this reach consists almost exclusively of non-cohesive sediment areas
and it was represented as completely non-cohesive in the HUDTOX model.

Reach average results  for cohesive  and  non-cohesive  sediment areas were compared  to  the
SEDZL model results, one of four primary calibration constraints.   Model burial rates were
generally consistent  with SEDZL results except where differences were necessary to achieve
simultaneous agreement with Tri+ surface sediment concentrations and solids  dynamics. This
resulted in somewhat lower burial rates for Thompson Island Pool than those computed  by the
SEDZL model.  In Thompson Island Pool,  solids burial  rates for cohesive and non-cohesive
sediment areas in the HUDTOX calibration were 0.65 and  0.07 cm/yr,  respectively.   In  the
SEDZL calibration, the corresponding solids burial rates were 0.81 and 0.03 cm/yr (QEA, 1999).

7.3.2.2 High and Low-flow Solids Loads

In addition to achieving agreement with solids burial rates and long-term surface sediment Tri+
trends, another important  calibration test was comparison to estimated high and  low flow solids
                                           129                       Limno-Tech, Inc.

-------
mass transport in the river. These mass transport values can be viewed as in-river solids loads.
Results were stratified using a  river flow of  10,000 cfs  at  Fort  Edward to represent  the
approximate cutpoint above which flow-dependent resuspension is observed  (See Figure 6-12).
Flows below and above 10,000 cfs are referred to as "low flow" and "high flow", respectively,
throughout this report.  Use of a single flow cutpoint is a simple and convenient way to evaluate
model behavior under resuspension and non-resuspension conditions,  however, recognize that no
single flow cutpoint completely separates these conditions at all locations in the River.

There is good agreement between model and data-based estimates of the solids loads at Stillwater
and Waterford for both high and low flows (Figure 7-2). This suggests the model is representing
average high and low flow behavior for the historical calibration period.  Estimated solids loads
were based on the rating curves presented in Chapter 6, which were not developed for Thompson
Island Dam.  There was not a strong relationship between solids concentration  and flow in the
available data at this location.  At Stillwater and Waterford, differences are less than four percent
for both low and high flow. In-river solids loads are split almost equally between high and low
flow conditions, consistent with observations presented in Section 6.5.

It  must be noted that  initial  calibration efforts  were not successful in reconciling estimated
tributary solids loads with solids and Tri+ water column concentrations and in-river solids loads
below Thompson Island Pool  at low flow.  Calibration analyses indicated that decreasing gross
solids settling velocities or increasing solids resuspension velocities produced results that were
not  in good  agreement  with  Tri+ surface  sediment concentrations or   water  column
concentrations. Upward adjustment of low flow tributary solids loads downstream of Thompson
Island Pool provided better  agreement with both  Tri+  and solids  concentrations,  and also
improved  model agreement with  long term sediment Tri+ concentrations.  Solids loads  were
adjusted by adding a total constant additional load of 40 MT/day to the Schuylerville-Stillwater
and Stillwater-Waterford reaches.  These adjustments represent increases of 26  and 17 percent,
respectively to the total tributary loads for these two reaches.  The magnitude  of this adjustment
was considered to be within the large range of uncertainty in estimation of tributary solids loads
below TIP.

7.3.2.3 Water Column Solids Concentrations

The  model calibration  was evaluated by comparing computed water column suspended solids
concentrations to long-term data over the 21-year calibration period and short-term intensive data
during four high-flow events.  Each is described below.

The  calibrated model results for water column solids over the 21-year calibration period show
reasonable fit at both high and low flow observations across the entire period at Thompson Island
Dam, Schuylerville, Stillwater and Waterford (Figure 7-3 a and b).  Results shown for the first
model segment  downstream of Fort Edward represent solids loading inputs at the  upstream
boundary  and are shown only for reference.  Note that data are only available for Thompson
Island Dam  from 1991  to 1997.  The solids concentrations throughout the year, and especially
during high flow events were found to be strongly driven by hydrology and external solids loads.
Solids concentrations are much higher in reaches below Thompson Island Pool and reflect the
much higher external solids loads  to this portion  of the river (Section 6.5). Both computed and
                                           130                      Limno-Tech, Inc.

-------
observed peak concentrations generally range between 50 and 100 mg/1 at TED and Schuylerville,
and between 100 and 400 mg/1 at Stillwater and Waterford.

Although the calibration strategy focused on accurate representation of long-term Tri+ trends and
mean high and low flow solids dynamics, it is of interest to assess model performance for high
flow events when  flow-dependent resuspension  is important.  Suspended solids results for the
spring high flow periods during  1983, 1993, 1994 and 1997 are shown in Figures 7-4 through 7-
7. These  four events are among the most extensively sampled events in the calibration period.
Results are of particular interest in  1993 and 1994  because sampling frequencies were higher
than in 1983 or 1997.

In general, timing  of computed  and observed concentration peaks is in fair agreement.  This is
largely because water column solids concentrations are strongly driven by hydrology and external
solids loads, especially during high flows. The computed peak concentrations at times tend to be
lower than observed peak concentrations. This is especially evident  at Stillwater.  This is  not
unexpected considering the model  calibration  strategy of capturing  mean  high  flow solids
dynamics.   There is good agreement between computed  and  observed concentrations  at
Thompson Island  Dam  during spring 1994, a period during which  daily measurements were
available.

The model calibration to peak concentrations at Stillwater appears weakest, but may be partially
explained by  errors in estimated tributary flow,  especially  for Batten  Kill.  While solids
concentrations were measured for Batten Kill over much of the 1994 event, Batten Kill  flows
were estimated based on Kayaderosseras Creek flows, which drain a much smaller watershed and
thus are expected to exhibit a more "flashy" response to precipitation or snowmelt (See Section
6.4). Closer agreement occurs at Thompson Island Dam, shown for the 1994 and 1997 events.

7.3.2.4 Spring 1994 High Flow Event Solids Mass Balance

The  model calibration  was  also evaluated  by comparing model-estimated and data-estimated
solids mass balances for Thompson Island Pool during the spring 1994 high flow event.  This is
the only reach in the Upper  Hudson River for which there exists a well-constrained solid mass
balance for mainstem and tributary solids loads.  For the 33-day period (March 29 to April  30)
encompassing  this  event,  measurements were  available  for  flows  and  water  column
concentrations for the two major tributaries and upstream inputs. This permitted development of
an input-output solids mass  balance for this event.  The model-based estimate of 400  MT net
erosion during this  event agrees within three percent of the data-based estimate of 411 MT.

7.3.2.5 Further Model-Data Comparisons

To provide insights into model behavior and the limits of model capability, calibration results are
also shown by comparison of computed and observed water column solids concentrations using
scatter plots and cumulative probability functions for model results and data stratified by flow.
Presentation of results in this manner shows the model performance in describing individual data.
However,  it must be recognized that the model calibration approach was not aimed at describing
the full range of observed event-scale behavior.  The solids calibration sought to describe mean
                                           131                       Limno-Tcch, Inc.

-------
 low and  high flow behavior.   Considering this,  model  agreement with mean  or  median
 concentration results is of more interest than a good fit across the range of observed behavior.  In
 fact, the model may be expected to show offsetting errors at the high and low end of each flow
 range.

 Scatter plot results for Thompson Island Dam, Stillwater and Waterford are shown in Figures 7-8
 and 7-9 for low and high flows, respectively. The model and data mean values are shown on
 these figures by the horizontal and vertical crossed lines, as is the 1:1 correspondence line. Even
 on log-log scale, the high variability in agreement between model and data is evident.  Inspection
 of these plots shows that the model tends to over compute low concentrations and under compute
 high concentrations in each flow range, an expected result of the model because it is calibrated to
 mean low and high flow behavior. Note that the model and data means intersect at the 1:1 line
 (meaning  they are nearly identical) for both high and low flows. Agreement is best for Stillwater
 and Waterford. At Thompson Island Dam, the model appears to be biased slightly low under low
 flow conditions and slightly high at high flow conditions.

 Probability distributions provide  similar insights as the scatter plots.  Computed and observed
 cumulative  probability  distributions  for  solids  concentrations  at  Thompson Island  Dam,
 Stillwater  and Waterford  are presented in Figures  7-10 and  7-11  for low and high flows,
 respectively.  The same observations regarding model behavior can be made in these figures as
 the scatter plots. At Stillwater and Waterford, the computed and observed median values tend to
 agree, which was the intent of the calibration, while the model shows offsetting biases at low and
 high concentrations of each flow range.  At Thompson Island Dam, the model results are good at
 low flow,  however, show considerably higher concentrations than were observed at high flow.

 While the model agreement with data at high flow for Thompson Island Dam  is not ideal, the
 overall significance of this  to use of the model for the Reassessment is small.  Fish PCB levels do
 not respond at  any  significant  level to short-term  event concentrations.   The  model was
 successfully calibrated  to estimates  of long-term  solids burial  rates  and  sediment  Tri+
 concentrations.  The water column  Tri+ concentrations  that affect fish levels  are determined
 largely by sediment-water  transfer mechanisms that are not flow driven, and by upstream Tri+
 loadings at Fort Edward.  Therefore, additional model calibration to high flow dynamics was
deemed unnecessary for the Reassessment.

7.3.3  Components Analysis for Solids

Over the 21-year calibration period, the HUDTOX model represents the Upper Hudson River as
 a whole to be net  depositional, based on the assumption underlying development of tributary
 solids loads. The computed average  bed elevation change in Thompson Island Pool over the 21-
year calibration period is approximately 4.5 cm (Figure 7-12).  Computed annual average burial
rates in cohesive and non-cohesive sediment areas from Fort Edward to Federal Dam are shown
over specific river  mile intervals  in Figure  7-13.  Burial rates in cohesive sediments  range from
0.24 to 1.49 cm/yr while burial rates in non-cohesive sediments do not exceed 0.10 cm/yr.
Computed burial rates in  cohesive  sediment areas are approximately an order of magnitude
 greater than those computed in non-cohesive sediment areas.
                                          132                       Limno-Tech, Inc.

-------
A 21-year solids mass balance components analysis from the calibrated model is shown in Figure
7-14 for the four major reaches in the Upper Hudson River.  These  four reaches represent the
river  from Fort Edward to  Thompson Island Dam, Thompson Island Dam to Schuylerville,
Schuylerville to Stillwater, and Stillwater to Waterford.  All four reaches are computed to be
depositional over the 21-year calibration period. There is a computed net load gain to the water
column of 3,043 x 103 MT (497 percent) between Fort Edward and Waterford.  Contributions to
solids load gain are dominated by tributary loads. Gross sediment resuspension accounts for only
21 percent of the total solids inputs to the water column.   Tributary loads (including  Fort
Edward) and sediment resuspension contribute 4,183 x 103 MT (79 percent) and 1,128 x  103 MT
(21 percent), respectively, between these locations.

It is noteworthy that approximately 80 percent of computed solids inputs to the water column are
due to external sources and only approximately 20 percent are due  to sediment resuspension.
Furthermore, these proportions change only slightly when  solids mass balance components
analyses are conducted separately for high and low flows.  It can be concluded that although
sediment resuspension is important, water column solids concentrations and in-river solids loads
are driven primarily by hydraulics and  solids loads from upstream and tributary sources, even
under high flow conditions.

7.3.4  Solids Calibration Summary

The calibration approach for solids dynamics in the historical calibration consisted of adjusting
constant gross  settling  velocities into cohesive  and  non-cohesive  sediment  areas,  and
resuspension rates from non-cohesive sediment areas. These  parameters were adjusted  to meet
the simultaneous constraints of long-term Tri+ concentrations in the surface sediments, solids
burial  rates, in-river  solids  loads, water column solids concentrations,  and  long-term water
column Tri+ concentrations.  The HUDTOX model represents the Upper  Hudson  River  as a
whole  to be  net depositional from  1977  to 1997,  based on  the  assumption  underlying
development of tributary solids loads. Computed solids burial rates in cohesive sediment areas
are approximately an order of magnitude greater than those computed in non-cohesive sediment
areas.   Computed in-river solids loads are  split almost equally between high and low flow
conditions.

There is a computed  net  solids  load gain to the water column of 497 percent between  Fort
Edward and Waterford over the 21-year historical calibration. Contributions to solids load  gain
are dominated by tributary loadings.  Computed tributary loadings  (including Fort Edward) and
gross  sediment resuspension contribute 79  and 21 percent, respectively  to  total solids inputs
between these locations.  Although sediment resuspension is important, water column solids
concentrations and in-river solids loadings are driven primarily by hydraulics and solids loadings
from upstream and tributary sources, even under high flow conditions.

The model calibration was demonstrated as successful for purposes of simulating general solids
behavior in the Upper  Hudson River.  Model performance was deemed satisfactory based  on:

       •     Model computed solids burial rates;

       •     Model computed high and low flow solids loads;
                                           133                       Limno-Tech, Inc.

-------
       •      Representation of the intensely-sampled spring 1994 high flow event; and,

       •      Statistical comparisons of model mean performance.

Additionally, it must be recognized that the HUDTOX calibration was conducted simultaneously
for solids and Tri+ and hence, further support of the calibration is evidenced in the next section
describing Tri+ results.

7.4 HISTORICAL TRI+ CALIBRATION

7.4.1 Calibration Approach

Although the solids  and Tri+ calibrations were  conducted simultaneously,  the Tri+ model
calibration results are presented separately and involved some considerations that did not affect
the solids model. For example, sediment mixed layer depths and mixing rates do not affect water
column suspended solids concentrations, however, these parameters have important impacts on
long-term surface  sediment Tri+ concentrations.  Mixing rates, however, are somewhat  less
influential than long-term solids burial rates.

In addition to simultaneous use of the calibration datasets for  solids dynamics, the additional
principal constraints for Tri+ in the historical calibration were:

       •      Tri+ surface sediment concentrations;

       •      Solids burial rates;

       •      In-river Tri+ mass transport at high and low flows; and,

       •      Tri+ water column concentrations.

As discussed in  Section 7.3,  the application of simultaneous, mutual constraints  on solids and
Tri+ ensured consistency between the solids and Tri+ mass balances in the model.  However,
greater emphasis was placed on trends in sediment and water column  Tri+ concentrations,
because these were the primary objectives of the model calibration.

The historical calibration was conducted on a reach-average spatial scale.  Operationally, the
calibration approach consisted of adjusting  only four model parameters: gross settling velocities
into cohesive and non-cohesive sediment areas; resuspension rates from non-cohesive sediment
areas; depth of particle mixing in the sediment bed; and, magnitude of sediment particle mixing.
No chemical-specific parameters were adjusted during the Tri+ calibration.   External loads,
partitioning, sediment-water  mass transfer and air-water  mass transfer rates  were determined
solely by Tri+ physical-chemical properties  and site-specific data, as described in Chapter 6.

Specification of Tri+ partitioning behavior has significant influence on the model calibration.
The model  uses three-phase  equilibrium  partitioning equations that  require  specification of
organic carbon concentrations. Values for site-specific organic  carbon input parameters to the
model  (determined in Chapter 6) are summarized in Table 7-3.  Input values for Tri+  process
                                            134                       Limno-Tech, Inc.

-------
coefficients and state variable properties  are presented in Table 7-4.   Values for all solids
calibration input parameters were presented previously in Tables 7-1 and 7.2.

7.4.2 Tri+ Calibration Results

Results for the Tri+ calibration are presented in a series of comparisons between computed and
observed values which include:

       •      Long-term surface sediment Tri+ concentrations;

       •      Longitudinal and vertical profiles for Tri+ sediment concentrations;

       •      Water column Tri+ concentration time series from 1977 to 1997;

       •      In-river Tri+ loads at low and high flow;

       •      Scatter plots of water column Tri+ concentrations at low and high flow;

       •      Cumulative probability distributions of Tri+ concentrations at low  and
              high flow; and,

       •      Water column Tri+ concentrations during several high flow periods.

7.4.2.1 Long-Term Sediment Tri+ Concentrations

The principal calibration  metric was comparison of computed and observed  long-term Tri+
concentration  trajectories in surface sediments (as shown in Figures 7-15 a-e).  These figures
show reach-wide average concentrations for five river reaches.  Model results  are shown for
surface sediments, which  correspond to  the first  two  sediment  layers (0-2  and 2-4 cm).
Computed results  for deeper sediments are also shown as average concentrations  over depth
intervals corresponding to the respective sediment datasets.

In addition to the 1977 initial condition data, sediment data were collected in 1991 (for the entire
Upper Hudson) and 1998 (mainly for Thompson Island Pool).  The vertical resolution of these
data (0-5  cm surface layers) permits  direct comparison with HUDTOX results  for the top two
sediment layers in the model. Other sediment data were also collected by USEPA and NYSDEC,
but these  data did not resolve the 0-5 cm surface layer.  As a result, model  comparisons to
sediment data collected by NYSDEC  in 1984 (average depth of approximately 25 cm, TIP only)
and by USEPA in  1994 (average depth of approximately 23 cm)  are displayed as concentrations
averaged over deeper layers.

Computed and observed concentrations in each  reach are expressed in terms of area-weighted
averages for cohesive and non-cohesive sediments.  Data are presented as mean  values plus and
minus two standard errors (2 SE) which corresponds to the 95 percent confidence interval about
the mean.   The model trajectories in Figures 7-15a through 7-15e represent Tri-i- concentrations
for the surface layer.  Symbols denote model  output averaged over layers corresponding to the
average depths of the 1984 and 1994 data.


                                           135                       Limno-Tech, Inc.

-------
Model results show very good agreement between computed and observed surface sediment Tri+
concentrations  in  TIP for both cohesive and  non-cohesive  sediment areas  (Figure  7-15a).
Computed concentrations  in surface sediments decline  by 89 and 80 percent, respectively, in
these sediment areas over the historical period from 1977 to 1997.  These declines correspond to
annual first-order loss rates of approximately  11 and 8 percent, respectively.  Declines in surface
sediment Tri+ concentrations occur due to burial to deeper sediment layers and sediment-water
transfer processes.  Agreement with depth-averaged concentrations in 1984 and  1994 is also
good, although the vertical scales are coarse and data variability is high. This suggests that the
model is  accurately  accounting for changes in the Tri+  mass reservoir in the sediments of
Thompson Island Pool.

There is also generally good agreement between computed and observed surface sediment Tri+
concentrations and deeper concentrations below Thompson Island Dam (Figures 7-15b through
7-15e).  A notable exception is for non-cohesive sediments in the Schuylerville reach in 1991
(Figure 7-15b) where  data are higher than the model. It is speculated that either the  sampling
and/or compositing processes used for  these  data  may have incorporated  high measured
concentrations that were not representative of reach-average conditions.  The 1991 non-cohesive
sediment data for this reach appear unrepresentative when compared to the 1994  data and the
overall declining trend observed at other locations in the river.  Hence this  discrepancy was
viewed as a likely data anomaly and not a model deficiency.

Computed concentrations in surface sediments in reaches downstream of Thompson Island Pool
decline by 91 to 97 percent in cohesive sediment areas and by 82 to 93  percent in non-cohesive
sediment areas.  These declines correspond to annual first-order loss rates of 11 to 14 percent in
cohesive sediment areas and 8 to 12 percent in non-cohesive sediment areas.

7.4.2.2 Longitudinal and Vertical Sediment Profiles

Another way to compare computed and observed sediment concentrations is to assess results at
smaller spatial scales  along the longitudinal  axis of the river and in  the vertical.  Figure 7-16
contains  comparisons of computed  and   observed  depth-averaged  (0  to  25 cm)  Tri+
concentrations for  1984 in Thompson Island Pool in the  longitudinal  direction.  Because the
HUDTOX  model represents three lateral spatial  segments in Thompson Island Pool, multiple
results for computed and observed values are shown at some locations.   The model captures
changes in Tri+ concentrations at depth  along  the length of TIP in both cohesive  and non-
cohesive sediments. This demonstrates that model is representing the approximate magnitude of
changes in the sediment Tri+ mass reservoir in TIP.

Figures 7-17 through 7-19 contain comparisons of computed and observed  depth-averaged Tri+
concentrations in 1991 (0  to 5 cm, 5 to 10 cm and 10 to  23 cm)  from Fort Edward to  Federal
Dam. The two upper reaches (Thompson Island Pool and Schuylerville) between river miles 193
and 183 are more heavily contaminated than  the three lower reaches (Stillwater, Waterford and
Federal Dam) between river miles 183 and 153.  Computed values cluster  well around most of
the data values for the 0-5 cm layer in both the upper and lower reaches.  Model results are
generally good for the 5-10 cm layer with the exception of  a high bias (approximately a factor of
two) in the very lower reaches between river  miles 163 and 153. Model results are not as good
                                           136                       Limno-Tech, Inc.

-------
for the 10-26 cm layer, especially in the lower reaches.  Results for this layer reflect the fact that
grab samples do not represent this depth interval and hence there are fewer data with which to
estimate reach-average concentrations.

7.4.2.3 Water Column Tri+ Concentrations

Figure 7-20 (a and b) shows comparisons between computed and observed Tri+ concentrations in
the water column at Thompson Island Dam, Schuylerville, Stillwater and Waterford for the full
historical period from 1977 to 1997.  Figure 7-20c  shows an expanded view of results at
Thompson  Island  Dam where  flows,  solids  and  Tri+  can all be balanced with the  most
confidence.  Again, results at  Fort  Edward  represent Tri-i- loading inputs at  the  upstream
boundary and are shown only as a reference.  Note that Tri+ detect limits changed in the mid
1980s, showing an  apparent sudden drop in minimum concentrations in the river.  Comparisons
between computed and  observed results are  confounded  by inconsistent temporal  coverage
among stations, changes in detection limits, datasets acquired by different organizations, a bias-
correction applied to data from the west shore of Thompson Island Dam, and revised estimates of
Tri+ concentrations applied to post-1986 USGS data.  These factors are discussed in  detail in
Chapter 6.

Long-term  declining  trends in  observed  Tri+ concentrations  are  captured  by  the model.
Magnitudes and seasonal trends at Thompson Island Dam are well represented by the model,
however, data are available only between 1991 and 1997.  In general, model results tend  to be
higher than observations during the mid-1980s at Schuylerville,  Stillwater and Waterford, and
again during the mid-1990s at Stillwater and Waterford.  There appears to be better agreement
between computed and observed values for the GE data,  especially at Thompson Island  Dam
from 1991 to 1997 and at Schuylerville for 1991 and 1992.

Achieving consistency between  computed and observed values at  Thompson Island Dam and
downstream locations  was a particular concern of the calibration, as the model was often higher
than the USGS data at Stillwater and Waterford. Sensitivity analyses on volatilization, sediment-
water mass transfer rates and gross settling velocities  did not achieve sufficient reductions in
water column Tri+ concentrations.  Increasing settling velocity and decreasing sediment-water
mass transfer rates resulted in over-estimating surface sediment Tri+ concentrations.

This prompted consideration of potential differences between USGS and GE datasets for  Tri+.
Potential  bias between the USGS and GE data could  perhaps explain why the model is well
calibrated  to GE data at Thompson  Island Dam, but over-estimating relative to USGS data
downstream. Figure 7-21 shows same-day comparisons between USGS and GE data for Tri+ at
Fort Edward, Stillwater and Waterford.  These  results suggest that the GE measurements may be
biased high relative to  the USGS measurements; however, these results are not conclusive.

7.4.2.4 High and Low-flow Tri+ Loads

The ability of the model to distinguish between low and high flow  contributions to overall Tri+
mass  transport  was assessed in similar manner as  done in the solids calibration.  Results of
comparisons for model-estimated and data-estimated in-river Tri+ loads  at Thompson Island
                                          137                      Limno-Tech, Inc.

-------
 Dam, Schuylerville, Stillwater and Waterford are contained in Figure 7-22.  The cumulative Tri+
 load estimates are provided,  although these estimates  are  uncertain due to data limitations.
 Perhaps of more interest is the relative contributions of low  and high flow to overall transport,
 shown in the bottom panel of the figure. Note that data for Thompson Island Dam are limited to
 the period from 1991 to 1997  and that there are no data for Schuylerville after 1993 (see Figure
 7-20a).  Consequently, the Tri+ mass loads in the top panel of Figure 7-30 correspond to 1991-
 1997 for Thompson Island Dam, 1977-1992 for Schuylerville and  1977-1997 for Stillwater and
 Waterford.

 Results show that the model accurately computes the relative  high and low flow Tri+ loads at all
 locations, and in addition shows good  agreement with the estimated cumulative loadings. The
 maximum difference in fractional distribution of Tri+ load between high and low flow strata is
 less than seven percent in all cases.  It is of interest to note that 70  to 80 percent of in-river Tri+
 loads at TID and 60 to 70 percent of Tri+ loads below TIP  occur during low flow conditions.
 This is  in contrast to in-river solids loads which  are split almost equally between high and low
 flow conditions (Figure 7-2).

 Model performance is also illustrated through comparisons  of computed and observed values
 over the course of  several high flow events.  Although the model was not developed specifically
 as an event scale  model, it does  include  cohesive  resuspension  formulations  based on site-
 specific  measurements of resuspension behavior (See Chapters 4 and 5).   Resuspension of
 cohesive sediment has higher potential to impact water  column concentrations  because of the
 much higher Tri+ concentrations relative to non-cohesive sediments. The ability of the model to
 represent water column Tri+ concentrations over high-flow events is observed by inspection of
 results for spring high flow  periods during 1983,  1993, 1994  and 1997 (Figures 7-23 through 7-
 26). These results for Tri+ correspond to the results for solids concentrations in Figures 7-4
 through 7-7.

 It is more difficult to compare computed and observed Tri+ concentrations during high flow
 periods than solids concentrations because no high-frequency sampling  was conducted for Tri+
concentrations during high flows.  Nonetheless, results  show  that the model  does  generally
represent the temporal event-scale variability shown in the Tri+ concentration data. Both model
and data exhibit nearly order-of-magnitude increases in water  column concentrations in response
to flow impacts.

7.4.2.5  Further Model-Data Comparisons

As was done for the solids calibration, scatter plots and cumulative probability distributions are
presented to provide insights into model behavior and the limits of model capability. Again it
should be recognized that the  model calibration  approach was not aimed at describing  the full
range of observed event-scale behavior.  The Tri+ calibration sought to  describe long-term Tri+
concentrations in surface sediments and mean low and high flow behavior in the water column.
Considering this, model agreement with mean or median concentration results is of more  interest
than a  good fit across the range of observed behavior.  Given the high variability in  measured
Tri+ concentrations, even within a given year, comparison of computed and observed values on a
point-by-point basis is of marginal value in assessing the calibration. A number of other factors
                                           138                       Limno-Tech, Inc.

-------
also preclude use of such comparisons in assessing model accuracy,  including changes  in
analytical methods and detection limits, a bias-correction applied to data from the west shore of
Thompson Island Dam, and revised estimates of Tri+ concentrations applied to post-1986 USGS
data.  These factors are discussed in detail in Chapter 6.

Comparisons between computed and observed Tri+ concentrations using scatter plots stratified
by river flow are shown in Figures 7-27  and 7-28.  The model and data mean values are shown
on these figures by the horizontal and vertical crossed lines, as is the 1:1 correspondence line.  A
river flow of 10,000 cfs at Fort Edward  is used to represent the cutpoint between low and high
flow, with the rationale explained in the previous section.  Results are good at Thompson Island
Dam,  showing  reasonable  agreement  between  model  and data  mean values,  and  close
correspondence  across the  range of observed values.  A similar behavior as for  solids was
observed for the Tri+ concentration scatter plots at Schuylerville, Stillwater and Waterford. This
is due in small part to the solids calibration approach, however, these  results are also affected by
the  above-mentioned  high bias  in the  calibration  results  compared  to  measured Tri+
concentrations downstream of Thompson Island Dam.

Comparisons between computed and observed probability distributions for Tri+ concentrations
over the entire calibration period stratified by river flow (Figures 7-29 and 7-30)  show similar
results as the scatter plots, but are also affected by the complicating factors mentioned above. At
Thompson Island Dam,  the model computes observed PCB concentrations with good accuracy
over the full range of observed concentrations.  Similar to observations  from the  time  series
results and  scatter plots, results for Stillwater and Waterford also  indicate that  the model is
weaker at Stillwater and Waterford.

7.4.3  Components Analysis for Tri+

A Tri+ mass balance components analysis from the calibrated model is shown in Figure 7-31 for
the four major reaches in the Upper Hudson River. These are the same four reaches for which
the solids mass  balance components analysis was conducted (Figure 7-14).  Over the  21-year
historical period there is a  computed Tri+ net load gain to the water column of 9,141 kg  (39
percent) between Fort Edward and Waterford. Most of this load gain (74 percent) occurs in the
first two reaches, TIP and Schuylerville.

Contributions to Tri+ load gain  are dominated  by non-flow-dependent sediment-water mass
transfer.  Computed  total inputs of  Tri+ to the water column from  1977 to 1997 were 26,597
kilograms.  External loads  (99  percent from Fort Edward) contributed  6,657  kg  (25 percent),
flow-dependent  sediment resuspension  contributed  6,722  kg  (25 percent)  and  non-flow-
dependent sediment-water mass transfer contributed 13,218 kg (50 percent) of the Tri+ inputs
between Fort Edward and Waterford.  Total losses of Tri+ from the  water column were  10,759
kg. These losses consisted of 9,496 kg (88 percent) from gross settling and 1,263 kg (12 percent)
from volatilization.

It is noteworthy that  75 percent of computed Tri+ inputs to the water column are due to internal
sources and  not external loads.  This  is in sharp  contrast to  results  for  solids in which
approximately 80 percent of computed sources to the water column  was due to external  loads.
                                           139                       Limno-Tech, Inc.

-------
Results from Tri+ mass balances conducted separately for  high and low flows indicate that
internal sources are responsible for at least 70 percent of computed inputs to the water column.
The principal difference between low  and high  flow  Tri+ mass  balances  is  the  relative
importance  of   non-flow-dependent  sediment-water  mass  transfer  versus  flow-driven
resuspension. At low flow, non-flow-dependent sediment-water mass  transfer is responsible for
61  percent of computed Tri+ inputs to the water column and flow-dependent resuspension is
responsible for  14 percent.  At high flow, this relationship is reversed and  the  computed
contributions are 12 percent and 60 percent, respectively.

Computed cumulative Tri+  mass load gains between mainstem stations from 1991 to 1997 are
shown in Figure 7-32.  This period represents recent historical conditions and conditions in the
river for the early portion of the forecast simulations for No Action.  Gains in  Tri+ mass are
computed between all four mainstem stations. During  1992 and 1993, load gains are reduced in
the two upper reaches due to large increases in upstream Tri+ loads from failure of the Allen Mill
gate structure in September 1991. In the lower two reaches Tri+ mass is lost from the water
column during 1992  and 1993.  After 1993,  upstream Tri+ loads decline and the influence  of
sediment-water mass  transfer begins to control Tri+ mass  load gains  between stations.  These
load gains appear to increase with time as upstream Tri+ loads continue to decline through the
mid-1990s.

7.4.4  Comparison to Low Resolution Sediment Coring Report (LRC) Results

As  part of the  Reassessment, USEPA (1998a) conducted an investigation of  the change  in
sediment PCB inventories in Thompson Island Pool between 1984 and 1994.  This investigation
involved a comparison of results from the extensive 1984 NYSDEC survey with results from a
series of matched sediment cores collected by USEPA in  1994.  Inventories from a set of 60
sampling locations in Thompson Island Pool were compared on a point-to-point basis to provide
a quantitative indication of the direction and magnitude of change in the  sediment PCB
inventory.   This  analysis was subsequently revised  to include comparisons based on localized
sediment areas as opposed to point-to-point comparisons (USEPA, 1999a and b).  Results from
the revised analysis indicated that the best unbiased mean estimate of mass loss of Tri+ from the
sediments  within historic hotspot areas was 45 percent, with  an uncertainty range from 4 to 59
percent. It was estimated that dechlorination was responsible for approximately 5 percent of the
mean  mass loss.  The  remaining loss was interpreted as a loss of the Tri+ hotspot inventory either
to the overlying water column or through redistribution of contaminated sediments within TIP.
Another conclusion  from  this LRC  analysis was  that there was no evidence of extensive
widespread burial of historically contaminated sediments in the Pool.

Although HUDTOX and LRC findings are in general agreement, a direct comparison of results is
not possible due to the  different assumptions and spatial scales between these two approaches.
The LRC analysis included only cohesive  sediment areas that were historically known to be more
contaminated than average Thompson Island Pool  sediments, whereas the HUDTOX model
includes both cohesive and non-cohesive sediment  areas  over the full range of sediment
inventories estimated to reside in the Pool. The LRC analysis does not account for  Tri+ mass
loss that would  be transported downstream  of Thompson Island Pool or redeposited  in  non-
cohesive sediment areas, or in less contaminated cohesive sediment areas. The HUDTOX model
                                          140                      Limno-Tech, Inc.

-------
accounts for the full mass balance cycle including transport and fate downstream of Thompson
Island Pool, and redeposition in the Pool.

An approximate comparison of results suggests consistency among the HUDTOX, DEIR and
LRC analyses.  A components analysis of the Tri+ historical calibration indicated that 1,288 kg
of Tri+ was lost from the Thompson Island Pool sediment inventory between  1984 and 1994.
Most of this  loss was due to Tri+ mass  transport across Thompson Island  Dam arid a small
portion was due to volatilization. If the  Tri+ inventory in 1984 is taken to be approximately
14,500 kg (USEPA, 1997), then this mass loss out of the pool corresponds to approximately 9
percent.  This  value is  within the range of the  4 to 59 percent estimate of  mass  loss  from
historical hotspots  in the LRC analysis. As an independent check on both of these approaches,
the annual rate of net export of Tri+ from the Pool was estimated to range between 0.36 and 0.82
kg/day over the period April 1991 to October 1995 (USEPA, 1997). Assuming a value of 0.59
kg/day, the net export of Tri+ from the Pool sediments between 1984 and 1994 would be 2,153
kg which corresponds to a mass loss of 15 percent of the 1984 inventory. Because of its focus on
hotspots,  the  LRC does not distinguish  between  loss over  Thompson  Island Dam  and
redistribution to less contaminated areas within the pool.  When coupled with the LRC findings,
HUDTOX  and the DEIR also  suggest  that there  has also been  a significant amount of
redistribution of Tri-i- mass within Thompson Island Pool from historical hotspots.

With  respect  to lack of extensive widespread burial of historically contaminated sediments in
TIP, the HUDTOX model results are again consistent  with  results from  the LRC analysis.
Results in Figure 7-12 indicate that the increase in sediment bed elevation in Thompson Island
Pool between 1984 and  1994 computed by the HUDTOX model is approximately 2.0 cm.  This
is a poolwide result and  it should be understood that there are differences between cohesive and
non-cohesive sediment areas within the Pool (Figure 7-13).  From results  in Figure 7-1 the
computed increase in bed elevation for cohesive  sediments in Thompson Island Pool over 10
years  is approximately 6.5 centimeters. Furthermore, it should be understood that in the actual
river there is variability  within the individual model spatial segments and that certain areas can
be erosional and not depositional. Nonetheless, a net sedimentation rate of 6.5 cm over 10 years
is small compared  to the surface layer depth of 23 cm (9 in) in the LRC  sediment cores.
Considering the differences in spatial and temporal scales  of the two approaches, it can be
concluded that the HUDTOX model and the LRC are in qualitative agreement with respect to the
question of widespread burial of historically contaminated sediments in Thompson Island Pool.

7.4.5  Tri+ Calibration Summary

Summarizing to this point, the calibration approach for Tri+ in the historical calibration consisted
of adjusting only four model parameters: gross settling velocities into cohesive and non-cohesive
sediment areas; resuspension rates from non-cohesive sediment areas; depth of particle mixing in
the sediment bed; and, magnitude of sediment particle mixing.

Computed Tri+ concentrations in surface sediments declined by 89 and 80 percent, respectively,
in the cohesive and non-cohesive sediment areas of Thompson Island Pool between 1977 and
1997.  These declines correspond to  annual first-order loss rates  of approximately 11 and 8
percent,  respectively.  Computed surface  sediment concentrations  in  reaches downstream of
                                          141                       Limno-Tech, Inc.

-------
 Thompson Island Pool decline by 91 to 97 percent in cohesive sediment areas and by 82 to 93
 percent in non-cohesive sediment areas over this period. These declines correspond to  annual
 first-order loss rates of 11 to 14 percent in cohesive sediment areas and 8 to 12 percent in non-
 cohesive sediment areas.  Declines in surface sediment Tri+ concentrations occur due to burial to
 deeper sediment layers and sediment-water transfer processes.

 Computed results indicate that 70 to 80 percent of in-river Tri+ loads at Thompson Island Dam
 and 60 to 70 percent of in-river Tri+ loads below TIP occur during low flow conditions. This is
 in contrast to  in-river solids loads which are split almost equally between low  and high flow
 conditions.  This  finding supports  the  calibration strategy of focusing  on long-term average
 behavior and not on short-term dynamics associated with high flow events.

 There is a computed Tri+ net load gain to the water column of 139 percent between Fort Edward
 and Waterford between 1977 and 1997. Most  of this load gain occurs in the Thompson Island
 Pool and  Schuylerville reaches.  Contributions to Tri+ load gain are dominated by non-flow-
 dependent sediment-water mass transfer.  Computed  external  loads (99 percent from Fort
 Edward), flow-dependent sediment resuspension, and non-flow-dependent sediment-water mass
 transfer contribute 25, 25 and 50 percent, respectively, to total Tri+ inputs to the water column.
 Gross settling and volatilization accounted  for  88  and 12 percent, respectively, of the total
 computed losses from the water column.  It is noteworthy that  75 percent of computed Tri+
 inputs to the water column are due to internal  sources and not external loads.  This is in sharp
 contrast to results for solids in which 80 percent of computed sources to  the water column was
 due to external loads.

 7.5 SENSITIVITY ANALYSES

 Sensitivity analyses were conducted with the  calibrated HUDTOX model to evaluate model
 responses due to uncertainties in important model inputs and calibration parameters. The analysis
elucidates model behavior and identifies parameters which are  important in determining Tri+
exposure concentrations.   The approach was to change a particular model input  or calibration
parameter, and then re-run the model for the 21-year historical calibration period.  Results were
evaluated in terms of changes in long-term Tri+ concentrations in surface sediments and the
water column, relative to base calibration values, and changes in Tri+ mass loadings at mainstem
 stations and  Federal Dam. The sensitivity analyses were designed to assess perturbations to the
base calibration and they do not represent attempts to re-calibrate the model with different model
inputs or calibration parameters.

Sensitivity analyses were conducted for the following model inputs and calibration parameters:

       •      Solids loads at Fort Edward and tributary solids loads;

       •      Tri + partition coefficients;

       •      Tri+ sediment-water mass transfer coefficients;
                                           142                       Limno-Tech, Inc.

-------
       •      Solids burial rates  via variation of gross settling velocity in cohesive
              sediment areas;

       •      Particle mixing in sediments;

       •      Sediment initial conditions for Tri+; and,

       •      Henry's Law Constant affecting volatilization of Tri+.

Table 7-5 contains an inventory of all sensitivity analyses conducted, and results for Tri+ mass
loads at mainstem stations and Federal Dam.

Model calibration results are sensitive to uncertainties in sediment particle mixing depth in non-
cohesive sediments, Tri+ partitioning, solids  burial rates, non-flow-dependent sediment-water
mass transfer rates, tributary solids loads and sediment initial conditions.  The calibration was not
especially sensitive to differences in solids loadings at Fort Edward computed by time-stratified
versus non-time-stratified loading methods, or to changes in Henry's Law Constant.

7.5.1 Solids loadings

During calibration  of the HUDTOX model a  key forcing  function  driving Tri+ exposure
concentrations was external solids loads to the system.  Hence,  upstream solids loads at  Fort
Edward and the tributary solids loads were varied separately and results discussed below.

7.5.1.1 Solids loads at Fort Edward

Solids loadings at the upstream boundary at Fort Edward were developed in Section 6.5. A time-
stratified regression approach was used to develop  these loadings  for the model  calibration
because a significant difference in solids-flow relationships was observed pre- and post-1990.  A
sensitivity analysis was conducted for solids loadings at Fort Edward that was determined using a
non-time-stratified regression  approach.   Both of these  approaches  were conducted using the
same solids concentration and flow data. Although the two approaches used the same data,  they
produced solids loadings that  were distributed differently in time. The non-stratified approach
produced solids loadings that were  lower than the stratified approach early in the historical  period
(1977 to 1990) but higher in the latter part of the period (1990 to 1997).

Results in Figures 7-33 and 7-34 indicate that sediments are  responsive only in TIP and not  at
Waterford, and that only the trajectory in the cohesive sediment area  of TIP is responsive. Two
reasons for  these results are that solids burial rates in cohesive sediment areas are  higher  than
those in non-cohesive areas, and that most of the solids loadings downstream of TIP are from
tributaries and not Fort Edward. The response of cohesive sediments in TIP is consistent with
the lower  solids loadings produced  by  the non-stratified  approach  early in the historical
calibration   period.   With increasing time,  the two  trajectories  converge  and  arrive  at
approximately the same Tri+ concentrations at the end of the calibration period.  Water column
concentrations were not sensitive  to changes in solids loadings due to application  of the non-
stratified regression approach and graphical results are not shown.  Results in Table 7-5 indicate
                                           143                       Limno-Tech, Inc.

-------
 that changes in-river Tri+ loadings and loadings over Federal Dam are small (approximately 3
 percent).

 7.5.1.2 External Tributary Solids Loads

 Approximately 80  percent  of the  solids loadings to the  Upper  Hudson River between Fort
 Edward and Federal Dam  are delivered by  tributaries.   To assess the sensitivity  of model
 calibration these  external solids loads were incremented and decremented by 50 percent and
 compared to the base calibration.  Figures 7-35 and  7-36  show the results of this analysis for
 surface sediment Tri+ concentrations in Thompson Island Pool and the reach from Schuylerville
 to Waterford. Results in TIP are not particularly sensitive to changes in tributary solids loads
 because the loadings at Fort Edward are the principal loads for that reach (see section 7.5.1.1).

 The results downstream of TIP however, show a significant sensitivity to these load changes.
 Concentrations in cohesive areas show  widely different  sediment trajectories  throughout  the
 simulation period. Due to lower net settling velocities in the non-cohesive areas the trajectories
 are not far apart in the earlier part of the historical simulation.  However, for the case where
 tributary loads were 50 percent smaller several segments in the non-cohesive areas become  net
 erosional and consequently expose buried Tri+ concentrations in the latter part of the simulation.
 This behavior also occurs in the forecast simulations (run with  base  tributary loads)  and is
 discussed  at length in Chapter 8.  Figure 7-37 shows the water column Tri+  concentrations at
 Thompson Island Dam and at Waterford  for these sensitivity runs.  As expected, concentrations
 at TED are insensitive to changes in  tributary solids.  Concentrations at Waterford  are much
 higher when tributary loads are reduced by 50 percent and slightly lower  when the  loads  are
 increased by 50 percent.

 7.5.1.3 Tributary Solids Loads Based on the Original Rating Curves

 The details of the methods and data used to compute  the external solids loads  to the system are
 discussed  in detail in Chapter 6 of this report.  Since  the original rating curves for the tributary
 solids loads were adjusted to determine the final calibration tributary loads a sensitivity analysis
 was conducted with the original unadjusted rating curves. Figure 7-38 and 7-39 show the results
for this sensitivity run  compared  with base  calibration  results.   Again  results  in TIP  are
 insensitive to changes in tributary solids loads. Downstream of TID in general higher sediment
Tri+ concentrations would be observed if the original rating curves  were employed. In particular
 note  that several  non-cohesive segments tend  to  be  erosional and  expose buried Tri+
concentrations for this sensitivity run.  The resultant Tri+ concentrations in  the non-cohesive
 areas are  inconsistent with the observed concentrations.  Figure 7-40 shows that  the  water
column concentrations are also significantly higher downstream of TIP.

7.5.2 Partition Coefficients

 A value of log KPOc =  5.845 was used  as the partition  coefficient for  Tri+ in the historical
calibration (Table 7-4).   An  analysis by Butcher et al.  (1998b)  indicated that the range of
 observed  partition  coefficients in  the Upper Hudson River was  approximately  5.4  to 6.6.
 Sensitivity analyses were conducted using these two values of log KPOC for Tri+.
                                            144                       Limno-Tech, Inc.

-------
Results in Figures 7-41 and 7-42 indicate that sediment response trajectories are very sensitive to
these variations in partitioning. The response trajectories for both of these variations violate the
principal  sediment  calibration  constraints  (reach average cohesive  and  non-cohesive Tri+
concentrations in  1991  and 1998) in TIP for cohesive and non-cohesive sediments. Response
trajectories at Waterford for higher partitioning are in better agreement with observations than
base calibration results. Water column concentrations (Figure 7-43) are also sensitive, however,
the sensitivity of these responses declines with time.

Sediment water exchange processes in  the model  were parameterized as a mass transfer rate
from the  dissolved  and DOC phases in the  sediment porewater.  Thus, any changes  to  the
partition coefficient directly changes the porewater concentrations and thus affects the flux out of
the sediments. In addition,  the larger the concentration, the greater the proportional  flux out of
the sediments.  Hence, for example the behavior observed in Figure 7-43.  In the early part of the
historical  simulation for  the  case  of the lowered partition coefficients,  the flux  out of  the
sediments is very  large and results  in water column exposure much greater than the base case.
However, this large  flux depletes  surficial sediment reservoirs  and in the latter part of  the
historical  simulation the concentrations  are closer to the base case (and in some locations even
smaller).  For the higher partition coefficient the reverse is true.  Hence, Figure 7-43 shows the
sensitivity results crossing each other in the water column.

It is noted here that the empirically determined mass transfer coefficient was dependent on the
choice of the partition coefficient.  A change in the partition coefficient  would necessitate re-
computation of this parameter to achieve the same net flux out of the sediments.  Hence, though
the numerical value of the mass transfer parameter is dependent on the choice of the partition
coefficient, the base calibration result in TIP would not be  any different were a different value of
the partition coefficient utilized.

7.5.3 Sediment-Water Mass  Transfer  Rates

To  assess the sensitivity of  historical calibration results to this important mechanism  two
approaches were undertaken: first, variation of the seasonally dependent rate between upper and
lower bounds; and second, specification of different rates between the cohesive and non-cohesive
sediment areas.

7.5.3.1  Variation of Sediment-water Transfer Rate

Figure 7-44  contains the time  series used in the model calibration for non-flow-dependent
sediment-water mass transfer rates.  This time series was determined using data-based, site-
specific mass balances (Section  6.13) and was not adjusted during the model calibration.  The
estimated range of uncertainty in this time  series is shown in Figure  7-44 and corresponds to
approximately plus and minus 50 percent. Sensitivity analyses were conducted for this range of
values about the base time series.

Results  in Figures 7-45 and 7-46 indicate that sediment responses are sensitive to these variations
in  sediment-water mass transfer.   The Tri+ surface sediment responses violate  the principal
calibration constraints in non-cohesive sediments in TIP,  and remain within approximately two
                                            145                       Limno-Tech, Inc.

-------
standard errors of the constraints in the cohesive sediments.  The response trajectory for non-
cohesive sediments at Waterford for the lower  sediment-water mass transfer rate is in  better
agreement with observations than base calibration results. Water column concentrations (Figure
7-47) are also sensitive, however, the sensitivity  of these responses declines with time.  Results
in Table 7-5  indicate that in-river Tri+ mass loadings and loadings over Federal Dam are
sensitive to  changes in  sediment-water mass transfer.  Loadings over Federal Dam change by
approximately 10 percent in response to variations in sediment-water mass transfer rates.

The data to determine the mass transfer rates is confined primarily to TIP.  It is uncertain if the
same  rates are operative throughout the Upper Hudson River.  Thus, water column exposure
concentrations downstream of TID and the export  of PCBs  over Federal  Dam to the Lower
Hudson River contain uncertainties due to this assumption in the historical calibration.

7.5.3.2  Differences in Sediment Water Transfer between Cohesive and Non-Cohesive
    Areas

The base HUDTOX model calibration assumes that  the sediment-water exchange processes are
identical for cohesive and non-cohesive sediment areas. Even though the exact mechanisms
which govern these  processes  are unclear,  a hypothesis with some merit is to consider the
possibility of  higher exchange in cohesive  sediment areas as compared to the non-cohesive
sediment areas. The reasons are discussed in Chapter 6 and include, for example the observation
of greater benthic activity and mixing in cohesive  areas as compared to non-cohesive areas.

A sensitivity run was conducted by assuming that the mass transfer rate is twice as large in the
cohesive sediment areas as compared to the non-cohesive sediment areas.  However, the overall
net flux from the entire sediment bed was constrained as previously by the data-based Tri+ mass
balance. The results are shown in Figures 7-48  and 7-49 and are compared to the base model
calibration. As expected the surface sediment Tri+ trajectories are shifted below the base case in
the cohesive areas and shifted higher in the non-cohesive areas. Thus, if this hypothesis were to
be incorporated into the model calibration it would require revisions to the base model parameter
choices.  Water column results as expected show no effect since the same net flux out of the
sediments was maintained.

7.5.4  Burial Rates in Cohesive Sediments

The gross settling velocity into cohesive sediments in the model calibration  was 4.15 m/day
(Table 7-1).   This parameter was adjusted during the model calibration and  its value was
determined by the principal calibration constraints. Sensitivity analyses were conducted in which
gross  settling velocity was varied so as to produce plus and minus 50 percent changes in solids
burial rates in TIP cohesive sediment areas.  Although these sensitivity analyses were conducted
by varying gross  settling velocity, the effect of this variation is  the  same as if external solids
loadings were varied. The reason is that there is a direct relationship between external solids
loadings and solids burial rates in the river.

Results in Figures 7-50  and 7-51 show responses of solids burial rates to the sensitivity analyses
conducted.  Although only the gross settling rate  into cohesive sediment areas was varied, burial
                                           146                       Limno-Tech, Inc.

-------
rates respond in both cohesive and non-cohesive areas because solids are redistributed due to the
dynamics of settling and  resuspension.   Solids burial rates  respond in opposite directions,
increasing in cohesive sediment areas and decreasing in non-cohesive sediment areas because
external solids loadings were not changed.

Results in Figures 7-52 and  7-53 indicate  that sediment Tri+ trajectories are sensitive to these
variations in solids burial rates. The principal model calibration constraints are violated in the
cohesive sediment areas of TIP, however, responses for lower settling velocities are in better
agreement with observations in cohesive sediments at Waterford than in the base calibration.
Water column concentrations  are more responsive at Waterford  than  in TIP (Figure 7-54).
Results in Table 7-5 indicate that in-river Tri+ mass loadings and loadings over Federal Dam are
sensitive to changes  in solids burial rates.  Loadings over Federal Dam change by approximately
10 to 15 percent in response to these variations in solids burial rates.

7.5.5 Particle Mixing in Sediments

Sediment particle mixing in the model was determined on the basis of observed sediment core
depth profiles, judgments on distributions  of biological activity and model calibration to long-
term Tri+ concentration trajectories  in the sediments.  Table  7-1 presents  model calibration
values for particle  mixing depths and  mixing  rates.  Particle  mixing depths were  10 cm in
cohesive sediments in all reaches, 6 cm in the non-cohesive  sediments in TIP and 4 cm in the
non-cohesive sediments downstream of TID. The most uncertain of these values is mixing depth
in non-cohesive sediment areas in  reaches  downstream of TED.   Some  reaches achieved
calibration constraints better with 4  cm depth of mixing while other areas were better at 6 cm
depth of mixing. To avoid use of differing parameters in the  calibration the choice of 4 cm was
selected for all reaches downstream of TID.  This sensitivity analysis presents results for the case
of non-cohesive sediments mixed to a depth of 6 cm for all reaches downstream of TID.

Results in Figures 7-55 through 7-58 indicate that sediment  trajectories  are sensitive to these
changes.  The deeper mixing depths and higher mixing rates  contribute additional Tri+ mass to
the   surficial  sediment layer by upward  mixing from deeper  contaminated layers. Modeled
historical  trajectories are more consistent  with this choice of mixing depth in the reaches at
Stillwater and Waterford.  The calibration choice of 4 cm yields better  results than a mixing
depth of 6 cm in the non-cohesive sediments in the Federal Dam reach. Based on this sensitivity
analysis a choice of 6 cm mixing in the non-cohesive sediments downstream of TID  appears
reasonable and may represent an alternate choice for the historical calibration.

7.5.6 Sediment Initial Conditions

There is large uncertainty in the 1977 data used to specify sediment initial concentrations for
Tri+ in the historical calibration.   Figure 6-32, 6-33, and  6-34  show  the variability in the
available historical data for 1977.  In many locations concentrations measured in 1977 can vary
by an order of magnitude or more.  Hence, sensitivity analyses were conducted for variations of
plus and minus  one standard error about mean values for Tri+ concentrations in cohesive  and
non-cohesive surface sediments in the 1977 data.
                                            147                       Limno-Tech, Inc.

-------
 Results  in Figures 7-59  and 7-60  indicate that  sediment trajectories are sensitive  to  these
 variations  in  initial conditions,  especially early  in  the  historical  calibration  period.   With
 increasing time  the  sensitivity  response trajectories  converge  closely to  the  base model
 calibration trajectories.  At the end of the calibration period, differences are small between the
 sensitivity results and the base calibration.  Both  sensitivity trajectories agree reasonably well
 with the model calibration targets in  1998 in TIP. Water column responses (Figure 7-61) follow
 the same trends as sediment responses.  Results in Table 7-5 indicate that in-river Tri+  mass
 loadings and loadings over Federal Dam are sensitive  to changes in initial conditions. Loadings
 over Federal Dam change by approximately 20 percent in response to variations in sediment
 initial conditions.

 7.5.7 Henry's Law Constant

 Among the parameters which affect volatilization  of Tri+ there  is some uncertainty associated
 with the choice if the Henry's Law constant. The historical calibration used a value of 1.69E-04
 atm m3/mole.   A sensitivity  analysis was conducted  to assess the sensitivity of the model to
 choices of 1.93E-04 and 0.68E-04 atm nvVmole.  This range was  established based on published
 literature. Water column results at Thompson Island Dam and at Waterford are shown in Figure
 7-62. Results  show that the modeled exposure concentrations are insensitive to the choice of this
 parameter.

 7.6 1991-1997 HINDCAST APPLICATIONS

 7.6.1  Overview

 Following  successful  calibration  of  the  HUDTOX model to Tri+ over the 21-year historical
calibration period, the model  calibration was tested with short-term hindcast applications for the
 period 1991 to 1997 to five  congeners and total PCB.    The five congeners  chosen for  these
 applications (BZ#4, BZ#28, BZ#52, BZ#[90+101]  and BZ#138) have different physical-
chemical properties, spanning  a  wide range of partitioning and  volatilization  behavior. The
 primary objective of the individual congener applications was to strengthen and support the  long-
term Tri+ historical calibration, and thereby, the use of the  model in  addressing the principal
questions of the Reassessment.

The differences in environmental  behavior among the  five congeners  provides an opportunity to
test the rigor of the Tri+ calibration, especially sediment-water and air-water exchange processes.
For example,  model results for a highly volatile congener may be more  sensitive  to errors in
sediment-water exchange than a less  volatile, strongly  partitioning congener.  Conversely, model
results for a strongly partitioning congener may be more sensitive to errors in particle-based PCB
processes such as settling and resuspension than a weaker partitioning  congener.

Overall,  the short-term hindcast applications to congeners demonstrate that the Tri+ historical
calibration is technically sound and appropriate for use in the Reassessment.  The Tri+ historical
calibration was confirmed in that the two congeners  whose  physical-chemical properties  most
closely resemble Tri+ (BZ#28  and BZ#52) were accurately represented using the same model
parameters as  the Tri+ historical calibration. While changes to the model (discussed below) may
                                            148                       Limno-Tech, Inc.

-------
permit better  simultaneous representation of all five PCB congeners, these  changes do not
enhance  model performance for Tri+.   Therefore, the Tri+ historical calibration was judged
technically sound  based on the calibration results presented  above and on the  confirmatory
results presented below for BZ#28 and BZ#52.

7.6.2 Approach

The approach  to the congener hindcast applications was to use the same flow and solids mass
balances as in the Tri+ historical calibration, along with input loads  for total PCBs and the
individual congeners.  Total PCB and congener loads were developed in Chapter 6, in addition to
sediment initial  conditions based on the 1991  GE composite data.  All of the  same  model
parameters used for the Tri+ historical calibration were used initially for the total  PCB and
congener applications, with the exception of those that were chemical-specific. These  chemical-
specific  parameters  include  only  Henry's Law  constant,  molecular weight and  partition
coefficients (Table 7-4).  Partition  coefficients for total  PCBs and the five congeners were
developed in Section 6.9.

Initial  1991 to 1997 results showed  that the empirical sediment-water mass transfer coefficient
developed  for Tri+ (Chapter 6) was not applicable to BZ#4. The Tri+ sediment-water mass
transfer coefficient produced unreasonably high sediment-water transfer of BZ#4. An analysis of
apparent sediment-water mass transfer rates  for  all congeners  revealed  large  differences,
seemingly  related  to  differences  in sediment-water partitioning.   This led to estimation of
separate particle-based and porewater sediment-water mass transfer coefficients, which allowed a
reasonable  simultaneous representation of sediment-water mass  transfer for all congeners.
Details of the development of these coefficients is presented in Chapter 6.

7.6.3 Results

Model testing  through the short-term hindcast applications was accomplished by comparison of
computed congener and total PCB concentrations to water column observations.  As  discussed
above, initial  simulations used the  same  empirical sediment-water mass transfer coefficient
developed for  Tri+ (Chapter 6).  Results of these initial simulations, shown  for BZ#4, BZ#28,
and  BZ#52 from  1991 to 1992  at Thompson Island Dam,  revealed that while the  model
performed  well for BZ#28 and BZ#52, the BZ#4 sediment-water mass transfer was too high
(Figure 7-63).  Water column BZ#4 concentrations were significantly higher than the data for the
first two years of the forecast.  This  is due to rapid loss of BZ#4 from the sediments, evidenced
by a comparison of the BZ#4, BZ#28, and BZ#52 sediment concentration trends for these three
congeners (Figure 7-64).

To  determine  whether  potential differences between measured water column and sediment
partition coefficients was contributing to the over-prediction of BZ#4, simulations using partition
coefficients from the 1991 sediment  data were conducted.  Sediment partition coefficients were
used as computed  from the GE 1991 sediment core composite data (USEPA,  1997), however,
they were considered less reliable than the estimates from the Phase 2 water column  data. This is
because the GE 1991 samples were frozen and composited prior to analysis, likely altering PCBs
in porewater measurements.  Results showed that use of the sediment partition coefficients may
                                          149                      Limno-Tech, Inc.

-------
 improve model fit to BZ#4. This indicated sediment-water transfer for this congener is over-
 computed using the Tri+ mass transfer coefficient.

 This led to incorporation of the  separate particle and  porewater-based sediment-water mass
 transfer coefficients,  as described in Chapter 6, which generally improved model for BZ#4 and
 total PCB while  also  showing  reasonable results  for the other congeners.   Results are
 demonstrated several ways:

       •     Model versus data concentration time series at several locations for 1992
             to 1997;

       •     Comparison to congener concentrations and congener ratios to BZ#52 in
             the GE 1996 and  1997 float study surveys which measured down-river
             profiles in Thompson Island Pool; and,

       •     Comparison of down-river congener ratios to BZ#52 from Fort Edward to
             Waterford.

These are each discussed below.

Computed  water column concentrations for each congener  are compared to observations at
Thompson  Island Dam for the period  1991 through 1997 (Figure 7-66).  Results are shown in
Figure 7-67 (a to f) for Schuylerville, Stillwater and Waterford for the period 1991 through 1993.
Very little data are available below Thompson Island Dam after 1993.  The model shows very
good comparison  to  observed concentrations of BZ#28 and BZ#52, the two congeners with
environmental behavior most similar to  Tri+.  Results are also good for total PCB and BZ#4,
although summer time concentrations are slightly below  observed values.   Performance for
BZ#[101+90] and BZ#138 is very good at Schuylerville,  however, the model under-computes at
Stillwater and Waterford. This may be due to differential settling losses of these congeners as a
result of their stronger partitioning behavior.

Overall, based on the specification of separate paniculate and porewater mass transfer pathways,
the model does  reasonably  well for   simulating  BZ#28,  BZ#52,  BZ#4  and  total  PCB.
Performance is weaker for BZ[101+90] and BZ#138,  however, the best agreement with BZ#28
and BZ#52 is a confirmation of the historical calibration because these congeners have the most
similar environmental behavior to Tri+.

A unique series of datasets was collected by GE in the summer of 1996 and 1997 that provide a
useful evaluation of the model performance on a small spatial scale.  These data were collected
from boats at a large number of sequential locations,  floating down through Thompson Island
Pool.  The  down-river profile of PCB concentrations in  these data show the highest PCB load
gains over  high sediment concentration  areas and consequent changes  in the  congener pattern.
Model   results  were  compared   to  observed concentrations  and  the ratio  of  congener
concentrations to  BZ#52 concentrations.  BZ#52 was chosen as  a normalizing congener for
evaluating the congener pattern because it is consistently present in higher concentrations and is a
stable congener. Results are shown BZ#4, BZ#28 and BZ#[90+101] in Figure 7-68 (a to d). The
                                          150                       Limno-Tech, Inc.

-------
model describes the shift in congener ratios across the Pool reasonably well, representing both
the down-river concentration profiles and the observed higher release of BZ#4 relative to the
other congeners.

In addition to evaluation of the down-river profile comparisons in Thompson Island Pool using
the above float study data, comparisons were also made to the down-river profile of the BZ#28 to
BZ#52 ratio using data for Fort Edward, Thompson Island Dam, Schuylerville, Stillwater and
Waterford. The ratios were evaluated for summer and non-summer conditions (Figure 7-69), as
well as high and low flow conditions (Figure 7-70). These results show that the model captures
the BZ#28/BZ#52 ratio reasonably well  for the entire Upper Hudson River over these seasonal
and flow conditions.

To assess the significance  of the alternative parameterization of sediment-water mass transfer
rates on the historical calibration, the calibration simulation was run with these rates.  That is, the
historical calibration  was  re-run  with  non-flow-dependent sediment-water  mass  transfer
decomposed into separate particle-based and porewater-based pathways.  Results showed that
this did  not alter the performance of Tri+ in  the historical  calibration.  Cohesive  sediment
concentrations  were insensitive  to  use of the  separate porewater  and particulate  transfer
pathways, however, the rate of decline  in non-cohesive sediment  concentrations was slowed,
resulting in concentrations somewhat higher than the observed values in 1991.

7.6.4  Hindcast Applications Summary

The short-term hindcast applications confirmed the Tri+ historical calibration because use of the
exact same parameter set used for the historical calibration resulted in  very good predictions of
observed BZ#28  and BZ#52 concentrations.  The hindcast applications revealed that alternate
specification of the sediment-water mass transfer coefficient was required to allow simultaneous
simulation of all congeners; however, these changes resulted in somewhat poorer calibration to
non-cohesive  sediment concentration  trends  in the  historical  Tri+  calibration.    While
investigations conducted in the hindcast applications to individual congeners provided insights
into sediment-water mass transfer behavior, no  changes to the Tri+ historical calibration  were
supported.  Model performance in the hindcast applications was strongest for BZ#28 and BZ#52,
the congeners whose environmental behavior most closely resembles that of Tri+. Testing of the
historical calibration through short-term hindcast applications to individual  congeners strongly
supported the technical soundness of the Tri+ historical calibration and  use of the calibrated
HUDTOX model in the Reassessment.

7.7 CALIBRATION FINDINGS AND CONCLUSIONS

The HUDTOX 21-year historical calibration to Tri+ served as  the main development vehicle for
the PCB  fate and transport model to be used this Reassessment.  This calibration was successful
in reproducing observed long-term trends in water and sediment PCB concentrations over the 21-
year period. This was primarily demonstrated through comparisons between model results and
available data for the following parameters:
                                           151                       Limno-Tech, Inc.

-------
       •      Long-term solids burial rates;

       •      Long-term Tri+ surface sediment concentrations;

       •      In-river solids and Tri+ mass transport at low and high flows;

       •      Water column solids and Tri+ concentrations;

       •      Solids mass balance for the Spring 1994 high flow event; and,

       •      Testing  of  the  historical  calibration  through  short-term  hindcast
              applications to individual congeners.

Many  different metrics were  used  to demonstrate  model reliability and they were used
collectively in a "weight of evidence" approach.

The following factors were found to be the most important in controlling long-term trends in
Tri+ responses in the Upper Hudson River:

       •      Hydrology;

       •      External solids loads;

       •      External Tri+ loads;

       •      Tri+ partitioning;

       •      Sediment-water mass transfer under non-scouring flow conditions;

       •      Solids burial rates; and,

       •      Particle mixing depth in the  sediments.

The first three of these factors are external inputs largely determined by site-specific data, and the
last four are internal processes within the river.

The principal findings and conclusions from the calibration  analyses are the following:

       •      The HUDTOX model represents the Upper Hudson River as a whole to be
              net depositional from 1977  to 1997, based on the assumptions underlying
              development of tributary solids loads.   Computed solids burial rates in
              cohesive  sediment areas are approximately an order of magnitude greater
              than those computed in non-cohesive sediment areas;

       •      Computed in-river solids mass loads are split almost equally between high
              and low flow conditions;
                                           152                       Limno-Tech, Inc.

-------
There  is a computed net solids load gain to the  water column of 497
percent between Fort Edward and Waterford over the 21-year historical
calibration;  computed tributary loadings  (including  Fort Edward) and
gross sediment resuspension contribute 79 and 21 percent, respectively, to
total solids inputs between these locations;

Although sediment  resuspension is important,  water  column  solids
concentrations and in-river solids loads are driven primarily by hydraulics
and  solids loads  from upstream and tributary sources, even under high
flow conditions;

Computed Tri+ concentrations in surface sediments declined by 89 and 80
percent, respectively, in the cohesive and non-cohesive sediment areas of
Thompson  Island Pool  between  1977  and  1997.   These  declines
correspond to annual first-order  loss rates  of approximately  11  and 8
percent, respectively;

Computed  Tri+   concentrations  in  surface  sediments  in  reaches
downstream of Thompson Island Pool  decline by 91 to  97 percent  in
cohesive sediment areas and by 82 to 93 percent in non-cohesive sediment
areas.  These declines correspond to annual first-order loss rates of 11  to
14 percent in cohesive sediment areas and 8 to 12 percent in non-cohesive
sediment areas;

Computed results indicate that 70 to 80 percent of in-river Tri+ loads at
Thompson Island Dam and 60 to 70 percent of in-river Tri+ loads below
Thompson Island Pool occur during low flow conditions;

There is a computed Tri+ net load gain to the water  column of 139 percent
between   Fort  Edward  and  Waterford  over  the  21-year  historical
calibration; most  of this Tri+ load gain occurs in the Thompson Island
Pool and Schuylerville reaches;

Computed external loads (99 percent from Fort Edward), flow-dependent
sediment  resuspension,  and non-flow-dependent  sediment-water  mass
transfer contributed 25,  25  and 50  percent, respectively, to total Tri+
inputs to the water column during the 21-year historical calibration;

Gross  settling  and  volatilization accounted for  88  and  12  percent,
respectively, of the total computed losses of Tri+ from the water column;

Testing of the 21-year historical calibration for Tri+  through short-term
hindcast  applications to  individual  congeners  strongly  supported  the
technical soundness of the historical calibration, and use of the calibrated
HUDTOX model in the Reassessment;
                             153                      Limno-Tech, Inc.

-------
       •      Model  calibration  results  are  sensitive  to  uncertainties in solids  burial
              rates, Tri+ partitioning, non-flow-dependent sediment-water mass transfer
              rates, tributary solids loads, sediment particle mixing depth, and sediment
              initial conditions; and,

       •      Model calibration results were not especially sensitive to uncertainties in
              solids loadings at Fort Edward  or  to  volatilization,  as  influenced  by
              uncertainties in Henry's Law Constant.

Based  on the  above tests  of model  performance  and reliability,  the  HUDTOX  model is
considered adequately calibrated for predicting long-term PCB responses in the Upper Hudson
River, which is the primary use of the model in the Reassessment.
                                           154                       Limno-Tech, Inc.

-------
Chapter 8

-------
8. FORECAST SIMULATIONS FOR NO ACTION

8.1 OVERVIEW

In 1984 the U.S. Environmental Protection Agency (USEPA) issued an interim decision of No
Action concerning PCB contaminated sediments in the Upper Hudson River. In December 1990,
USEPA issued a Scope of Work for a Reassessment of the  1984 No Action decision.  The
modeling  work presented  in  this report  is part  of Phase 2  of the  three-phase Remedial
Investigation and Feasibility Study being conducted for the Reassessment. The HUDTOX model
was developed to answer two of the three principal study questions posed in the Reassessment:

       •     When will PCB levels in fish populations recover to levels meeting human
             health and ecological risk criteria under continued No Action?

       •     Can remedies other than No Action significantly shorten the time required
             to achieve acceptable risk levels?

Question one is addressed in this chapter through  prediction  of the future course of PCB
concentrations in the Upper Hudson River using the HUDTOX model developed and calibrated
as described in Chapters 5 through 7.  Question two will be addressed through application of the
HUDTOX model in Phase 3, the Feasibility Study. The third principal study question:

       •     Are  buried  contaminated  sediments  likely  to  become  "reactivated"
             following a major flood, possibly resulting in an increase in contamination
             of the fish population?

was answered through use of the Depth of Scour Model (Chapter 4) and HUDTOX.

The Depth of Scour Model (DOSM) presented in Chapter 4 was developed  specifically to
address the third principal question of the Reassessment.  The DOSM computes the depth of
scour expected under peak flow conditions and was used to evaluate the likelihood that buried
PCB sediments would become reactivated.  However, DOSM does not account for subsequent
longer-term transport and redistribution of resuspended sediment  and PCBs. Hence, the DOSM
formulations for cohesive sediment resuspension were incorporated into the HUDTOX model,
and HUDTOX  was  used  to  calculate the  long-term  system  response  to a  flood  induced
resuspension event.  HUDTOX was not designed to simulate short-term transient events, hence
the focus of the 100-year peak flow simulation with HUDTOX is on long-term system response.

The No Action  simulation  was conducted for a 70-year forecast period beginning January 1,
1998.  The 70-year water and sediment concentrations computed by the model were provided as
input to the fish bioaccumulation modeling effort (presented in Books 3 and 4 of the  Revised
Baseline Modeling Report) and the  Human Health  Risk Assessment, which are also  being
conducted under this Reassessment.  The results of the  No Action forecast and 100-year peak
flow simulations are presented in this chapter, along with the design and implementation of these
scenarios.   An important aspect of the forecast simulations is  that the load at the upstream
boundary at Fort Edward was specified as a constant  PCB (Tri+) concentration, ranging from
                                         155                      Limno-Tech, Inc.

-------
zero to 30 ng/L, the average concentration in 1998.  Also presented are results of sensitivity
analyses conducted for upstream solids loading conditions.

The following major sections are included in Chapter 8:

              8.2   No Action Forecast Simulation Design

              8.3   No Action Forecast Results

              8.4    100-Year Peak Flow Simulation Design

              8.5    100-Year Peak Flow Simulation Results

              8.6   Sensitivity Analysis

              8.7   Exposure Concentrations  for the Bioaccumulation Model  and Human
                    Health Risk Assessment

              8.8   Principal Findings and Conclusions

8.2 No ACTION FORECAST SIMULATION DESIGN

The forecast state variable for the No Action simulation and the 100-year flow simulation is Tri+,
the sum of the tri through decachlorinated PCB homologues.  This PCB group was the principal
model calibration parameter. This  PCB  group was selected for the historical calibration and
forecasts because it provided a consistent basis for comparison among the site-specific datasets
over the historical period,  and because it is a good represention of the distribution of PCB
congeners that bioaccumulate in fish.

In order to conduct forecast simulations with  the HUDTOX model, it was necessary to specify
future conditions in the Upper Hudson River  for flows, solids loads,  and Tri+ loads. Estimates
were made based on  historical observations  and current  information regarding PCB loading
trends.

It is important to recognize  that forecast results have inherent uncertainty due to uncertainties in
estimating future flow and loading conditions. This uncertainty can be assessed and accounted
for in  management decision making  by evaluating predictions across a  range of  alternate
scenarios for these inputs. To the extent that one or another scenario may be considered the most
likely, it is possible to compute a best estimate of future PCB concentration trends  with  the
model.

Presented in  the next few  subsections are discussions of forecast model inputs  for  the flow
hydrograph, solids loads, PCB loadings, model initial (start-up) conditions, and specifications of
the other inputs.
                                           156                      Limno-Tech, Inc.

-------
8.2.1 Hydrograph

Specification of daily flow inputs at Fort Edward and all tributaries was required for the 70-year
No Action forecast simulation. Flow inputs were based on the historical 1977-1997 Fort Edward
and tributary hydrographs presented in Chapter 6. To construct the 70-year forecast flow inputs,
randomly selected annual hydrographs from the 1977-1997 period were linked sequentially until
a continuous 70-year daily flow series was produced. If the average flow over the synthesized 70
years was within  10 percent of the  average flow of the actual 1977-1997 period, the hydrograph
was accepted for  possible use in the No Action forecast. This process was repeated to produce
four randomly generated 70-year Fort Edward hydrographs.  For each of the four hydrographs
developed for Fort Edward, corresponding tributary hydrographs were used (See Chapter  6),
tributary flows were assembled sequentially in the same order as the Fort  Edward hydrograph to
produce 70-year flow inputs for all tributaries.

The four sets of alternative  Fort  Edward and  tributary  hydrographs developed above  were
assessed by  conducting No  Action  forecast simulations with each and selecting  the one
producing the approximate median result.  The hydrograph selected for forecast modeling is
presented in  Table 8-1 and Figure 8-1.  The  other three alternative hydrographs were used to
show sensitivity of the forecast predictions to the choice of hydrograph in Section 8.6.1.

8.2.2 Solids Loads

Solids loads play  an important role in determining Tri+ concentrations in  the river, due to burial
of PCB  in the sediment bed by solids deposition. As with flow, future solids loads are uncertain
and forecast inputs are best  based on historical data.  As discussed in Chapter 6, upstream solids
loading  relationships at Fort Edward changed over time.  Comparison of solids rating curves  for
1991 through 1997 data and for 1977 through 1990 data indicate that solids loading decreased
over time. This suggests that future solids load estimates at Fort Edward  may be best estimated
by the solids rating curves developed  for the  period 1991-1997, Equation 6-13. This equation
was used with the forecast hydrograph selected as described in the preceding section, to compute
solids loads at Fort Edward for the baseline No Action forecast.

Sensitivity of the  forecast simulations to the method of developing the solids rating curve at Fort
Edward was assessed using Equation 6-8, which is based on all available data from 1977 to 1997.
These results are presented in Section 8.6.2.

8.2.3 PCB Loads

Specification of Tri+ loads at Fort Edward has a large influence on forecast results of Tri+
concentrations in  the Upper Hudson River.  However, due to the variable nature of Tri+ loading
at Fort Edward, specification of future loading conditions is very uncertain.  In spite of extensive
remediation  of upstream sources conducted by GE over the last decade, PCB loading continues
from contaminated sediment deposits  and direct discharge  of PCB oil through bedrock fissures
(USEPA, 1997).   While  average  loads  generally continue to decline,  high concentrations
continue to  be periodically observed at  Fort  Edward.  For example on  January 10,  1998 a
concentration of 190 ng/L was observed at Fort Edward during  the spring high-flow event and a
                                           157                       Limno-Tech, Inc.

-------
concentration of 85 ng/L was observed on September 9, 1998 during a period of low flow (Figure
8-2).  Pulse loads due to these periodic high concentrations have been observed in the past and
the total annual load delivered to Thompson Island Pool has been estimated to be as much as 660
kg/year based on analysis of data from 1977 to 1998 (QEA 1999).

Year to year variability in PCB loads at Fort Edward are  seen in Figure 8-3 for 1991 through
1997 (and previously shown in Figure 6-26 for 1977 through 1997).  Available data collected by
GE in 1997 and 1998 show that annual average Tri+ concentrations were about 9.9 ng/L and 30.4
ng/L, respectively (based on using one-half detect limit concentrations for non-detect values).

As a result of the uncertainty associated  with estimates of future Tri+ loads  at Fort Edward,
baseline No Action forecasts were conducted with a range of assumptions that may reasonably
bound future behavior. Simulations were run with constant Tri+ concentrations of zero, 10 and
30 ng/L specified for the incoming flows at Fort Edward. The 30 ng/L simulation represents an
upper bound  for the 70-year forecast based on the  assumption that  future annual average
concentrations will not increase substantially above 1998 observed average levels.  The 10 ng/L
simulation roughly approximates the 1997 Tri+ loads and approximates the analytical detection
limit.   The zero boundary sensitivity  provides a  lower  bound  for simulation and  assumes
complete upstream remediation. Based on the average flow, a constant concentration of 10 ng/L
Tri-t- produces an annual load of about 47 kg/yr at Fort Edward, and 30 ng/1 produce a load of 141
kg/yr.

All downstream tributary Tri+ loads were assumed to be negligible and  set to zero for the
forecast simulations because historical concentrations were very low. This assumption may be
significant in calculating loads to the Lower Hudson River (over Federal Dam).  Conditions in
the Mohawk River are particularly important because of its large flow; however, the Mohawk
River enters just above the downstream boundary of the model and it has little impact on model
results for the  Upper River.  For a Lower Hudson River assessment, different assumptions for
tributaries (especially the Mohawk River) might be  considered, but these were not the focus of
this report. Any independent  estimates of Tri+ loading to the Lower  Hudson River from the
Mohawk River could simply be added to the HUDTOX estimates of load over Federal Dam.

8.2.4 Initial Conditions for the Forecast

Sediment initial conditions for the 1998  through 2067 forecast simulations were based on the end
results of the  1991 through 1997 model hindcast application.   In effect,  this corresponds to
initializing the forecast simulations with measured conditions in 1991. This approach utilizes the
most recent, reliable  and comprehensive dataset to begin the model forecasts.  In particular, in
terms of accuracy, vertical resolution and spatial coverage for the Upper Hudson River, the 1991
sediment measurements represent the best match between available sediment data  and model
requirements.  The initial conditions for the forecast are presented on a reach-average basis for
cohesive and non-cohesive sediments in Table 8-2.  These concentrations are  model results on
the last day of the simulation from January 1, 1991 to September 30, 1997.
                                           158                       Limno-Tech, Inc.

-------
8.2.5 Specification Of Other Model Inputs

In addition to specification of flows, solids loads and Tri+ loads, the forecast simulation required
specification of air and water temperatures and atmospheric Tri+ concentrations.  The  same
annual air and water temperature series applied  in the model calibration (Section 6.8)  were
repeated for the forecast simulations. Atmospheric Tri+ concentrations were assumed to be zero
for the entire forecast period.  All other model inputs remained the same as those in the historical
Tri+ calibration.

8.3 No ACTION FORECAST RESULTS

As discussed  in Section  8.2.3, there  is  large  uncertainty in  estimated upstream  Tri+
concentrations  entering Thompson Island  Pool  at Fort Edward  over the forecast period.
Consequently, a single most likely estimate of future concentrations is not presented.  Rather, a
likely range of estimates is presented, bounded by results based on specification of constant 30
ng/L and zero Tri+ concentrations at Fort Edward.  An additional simulation was done with a
concentration of 10 ng/L,  the approximate analytical detection limit.  Results of the 70-year No
Action forecast simulation are presented for:

       •      Surface sediment (0-4 cm) Tri+ concentration time series for  the whole
              Upper Hudson River (Figure 8-4,a-e);

       •      Annual  average and summer average water column Tri+ concentration
              time series for the long-term monitoring stations (Figures 8-6, a and b; and
              8-7, a and b); and,

       •      Annual Tri+ mass loading to the Lower Hudson River over Federal Dam
              (Figure 8-9, a and b).

These results are discussed in the subsections below.

8.3.1 Forecast Results: Surface Sediment PCB Concentrations

Several  important observations  are made from  the  No Action forecast  results for surface
sediment PCB concentrations (Figure 8-4, a-e). First, the influence of the upstream PCB load on
surface sediment  PCB concentrations is immediately apparent.  Following the first few years of
the forecast period, large separation occurs between the three upstream Tri+ load scenarios (zero,
10, and 30 ng/L Tri+ at Fort Edward). For the zero Tri+ loading simulation, concentrations in
the sediment and water column exhibit a continual decline, at a rate of between 7 and 8 percent
per year. This rate  is consistent  with observed  historical trends.  For  the simulations  using
constant  upstream concentrations  of 10 ng/L and  30  ng/L, similar recovery  rates are  initially
observed, however, upstream loads control sediment concentrations by about 2030.

All three loading simulations show concentrations  approaching  various asymptote values
throughout the river. For the zero Tri+  loading simulation, surface sediment concentrations
asymptotically approach zero. For the  simulations using constant upstream concentrations  of 10
ng/L and 30 ng/L, results asymptotically approach a quasi steady-state concentration determined
                                           159                      Limno-Tech, Inc.

-------
 by the upstream load and processes acting on Tri+ in the river. This is not a true equilibrium
 because the declining sediment releases still have minor effects on concentrations, and natural
 variability causes year to year fluctuations. The quasi steady-state concentrations approached in
 Thompson Island Pool  cohesive sediment for the 30 ng/L and  10 ng/L are approximately  1.8
 mg/kg  and 0.7 mg/kg, respectively.   Respective  non-cohesive  sediment concentrations  in
 Thompson Island Pool level off at about 0.75 mg/kg and 0.27 mg/kg for these simulations.   In
 downstream reaches, asymptote concentrations are slightly lower due to "clean" tributary flow
 and solids load inputs as well as volatilization losses.

 The 70-year forecast results  showed increases in surface  sediment concentrations in localized
 areas in Thompson Island  Pool (Figure 8-4a) and Stillwater Pool  (Figure 8-4c) as a result of
 small long-term average sediment erosion in some model segments  that eventually exposes Tri+
 contamination in deeper layers.  These results do  not occur in response to singular high flow
 events.   This response was not observed during the  calibration period, although  erosional
 behavior did occur.  This behavior is predicted to occur 40 to 50 years in the future because of
 the long-term impact of small sediment erosion rates  in certain locations.  These  computed
 increases in surface  sediment Tri+  concentrations  serve to slow or interrupt apparent rates of
 recovery. However,  due to factors described later below, the exact timing and magnitude of the
 events are dependent on forecast assumptions.

 In year  2044, two of the 15 cohesive sediment segments in the  Thompson Island Pool (below
 water column Segments 10 and 25,  Figure 5-5) increase approximately 0.4 mg/kg and again in
 2051  by about  0.3 mg/kg.   The minimum poolwide average cohesive sediment concentration
 before the increase in 2044 ranges from 0.27 mg/kg for the zero upstream loading simulation to
 1.92 mg/kg for the simulation with a constant Tri+  concentration  of 30 ng/L at Fort Edward.
 Non-cohesive sediments in the Stillwater reach (below water column Segments 37, 38 and 39,
 Figure 5-4c) also experience a series of increases, most notably in 2039 when concentrations rise
 by about 0.4  mg/kg.  Prior to this increase, average non-cohesive concentrations ranged from
 0.03 to 0.47 mg/kg for the zero and 30 ng/L Fort Edward  Tri+ simulations, respectively.

 These results are manifested as relatively sharp rises in  surface concentrations that occur when
 buried sediments of  higher concentration are incorporated  into the mixed layer due  to erosive
 loss of the surface sediment layer. Gradual increases are also observed in non-eroding surface
 sediments downstream  due to transport and  subsequent  deposition from upstream  sediment
erosion.  The relative magnitude of these increases is very small when compared in the context of
 historical observations (Figure 8-5, a-e). These forecasted increases will be considered during the
 development of the Feasibility Study.

The actual occurrence,  magnitude and timing of  the predicted increases in surface sediment
concentrations is uncertain due to the uncertainty in future conditions, especially tributary solids
loads. Also, the occurrence of these increases may be  in part due to solids dynamics in the model
calibration and/or the specification of sediment initial conditions that would not be apparent in
21-year  historical calibration period.   Solids burial  rates, which  were determined  by model
calibration, determine the net erosional or depositional nature of each sediment area in the model.
 Solids burial  rates are strongly driven by external solids  loadings.  Variations in these solids
 inputs, which may be within calibration uncertainty, can  result in slight net deposition instead of
                                           160                       Limno-Tech, Inc.

-------
slight net erosion for these localized areas.  However, it is equally possible that such variations
could  enhance the occurrences of predicted  increases. Also,  alternate  specification of  the
sediment initial condition profile and/or the mixed layer depth may reduce the magnitude of these
events. These possibilities are assessed through sensitivity analysis, presented below.

Model results indicating that  not  all areas  of the river are net depositional is a  reasonable
expectation and is corroborated by independent modeling and  data analysis.  The computed
erosional behavior of some areas of the river is not unique to HUDTOX.  The HUDTOX solids
results are based on a calibration  guided, in part, by calculations from  the SEDZL sediment
transport model (QEA, 1999).  The SEDZL model also computed some areas of both cohesive
and non-cohesive sediments to be net erosional over the historical calibration period from 1977
to 1997.  This is discussed in more detail in Chapter 7.

General findings of the Low Resolution Sediment  Coring Report (USEPA,  1998a) seem to also
offer corroborative support for the findings of  the forecast simulations that localized sediment
areas in the river are experiencing long-term erosional behavior. The general conclusion from the
Low Resolution Sediment Coring Report that burial  is not sequestering PCBs on a widespread
basis is borne out to some extent by the results  observed in the model.  It  is likely that there are
localized areas of continuing scour and PCB erosion on scales smaller than the HUDTOX model
segmentation. The HUDTOX results may not show these areas to be net erosional because they
fall within larger model segments  that are  on average  (across  the  whole  segment),  net
depositional. While the findings presented in the  Low Resolution Sediment Coring Report are
based  on  analyses at much finer  spatial scales than the HUDTOX calculations,  results  are
comparable in a general  sense and seem to  indicate that long-term burial  may not effectively
sequester buried PCBs at all locations in the river.

8.3.2 Forecast Results: Water Column PCB Concentrations

Forecasted annual average and summer average water column concentrations are also presented
for all three  loading  simulations at  Thompson Island  Dam,  Schuylerville,  Stillwater and
Waterford  in Figures 8-6 (a and b) and 8-7  (a and b).  These results show large separation in
water  column  Tri+  concentrations throughout the system under the different assumptions
regarding PCB loads at Fort Edward.  For the  zero Tri+ concentration assumption at upstream
boundary,  water  column concentrations  decline  to  less than  10 ng/L within  10 years  at all
locations.   With  upstream concentrations at Fort Edward set to 30 ng/L  and  10 ng/L,  water
column concentrations at Thompson Island Dam begin to level off around 2015 to about 27 ng/L
and 9  ng/L Tri+  respectively.   Downstream, annual average concentrations level off at  lower
values  due  to dilution from tributary inputs and losses from the water column.

Water  column concentration  results show significant year  to  year variability,  caused by
variability  in flows and  solids loading.  The  noticeable difference  in average  concentrations
observed in the third year of the forecast period  versus those in the first  two  years is  due to
differences  in  summer  flow  conditions (Figure 8-7).   Computed summer average Tri+
concentrations in  2000 are much lower than in  1998 and 1999 because summer time flows were
higher, diluting inputs of Tri+ from the sediment.  This  point is  illustrated below  in  the
simulations showing model sensitivity to choice of forecast hydrograph.
                                           161                       Limno-Tech, Inc.

-------
The predicted increases in surface sediment Tri+ concentrations in Thompson Island Pool and in
non-cohesive sediments in  the Stillwater reach do not  significantly impact the water column
concentrations.  This is because water column concentrations are being driven by upstream Tri+
loads at the time the sediment increases occur.  Additional Tri+ inputs due to the increased
surficial sediment concentrations is small relative to the upstream load contribution.

Figure 8-8 (a and b) presents the forecasted water column PCB  concentrations along with the
historical  calibration concentrations.   At this  temporal scale, forecasted  increases  in  water
column concentrations in the Stillwater and Waterford reaches are very small. From a transport
and fate standpoint,  these increases may be considered insignificant; however, they may be
important in a potential remediation decision.

8.3.3 Forecast Results: PCB Loads to the Lower Hudson River

Annual and cumulative forecasted Tri+ loads over Federal Dam to the Lower Hudson River are
also presented for all three upstream PCB loading simulations (Figure 8-9 a and b). Tri+ loads to
the Lower Hudson River  mirror water column  concentration  declines  and  decrease  from
approximately 150 kg/yr to less than 10 kg/yr by about 2015 under assumptions where Tri+
loading at Fort Edward ceases in 1998 (0 ng/L boundary concentration). For the upstream load
simulations based on 30 ng/L and 10 ng/L Tri+ concentrations at Fort Edward, loads continue to
decline until about 2015 and then essentially level off at  about  125 kg/yr and  45 kg/yr,
respectively.  These loads assume zero downstream tributary inputs which may be an important
consideration especially for the Mohawk River at the downstream boundary of the model.

8.4 100-YEAR PEAK FLOW SIMULATION DESIGN

As discussed in Section 8.1, one of the principal questions of this Reassessment concerns the
possibility that  deeply buried  contaminated sediments  might  become "reactivated"  during a
major flood, possibly resulting in an increase in PCB contamination of the fish population. The
available historical data for the Upper Hudson River can not  provide a direct answer to this
question because there were no very large floods during the period 1977 to 1997.  The peak flow
during this period was 35,200 cfs in May of 1983, a 15-year peak flow.  The 100-year peak flow
in the Upper Hudson River at Fort Edward is estimated to be 47,330 cfs (Butcher, 2000a).  The
Depth  of  Scour Model presented  in Chapter  4 was developed specifically to  address this
question, through application  to the Thompson Island Pool. The 100-year flood impacts were
also assessed for the whole Upper Hudson River through application of the HUDTOX model.

8.4.1 Specification of the 100-year Flood Hydrograph and Loadings

In the  100-year peak flow  simulation,  the HUDTOX model  was  run for a 70-year forecast
simulation identical to the No Action forecast with the exception of a peak flow corresponding to
the 100-year flood  peak imposed in the  spring of the first year.  The 100-year flood simulation
was conducted with zero PCB loading at Fort Edward and compared to  the corresponding No
Action simulation  to maximize the observation of an effect.  External  solids  loads  were not
increased during simulation of the 100-year peak flow, but remained the same as in the continued
No Action simulation. This design was a single factor, worst-case experiment in which the only
                                          162                      Limno-Tech, Inc.

-------
differences between continued No Action and the 100-year peak flow would be due to changes in
flow-dependent sediment resuspension.  This design was a worst-case  scenario because the peak
flow was placed in the first year of the simulation when sediment contamination was greatest and
because there was no increase in  external loads of  "clean" solids to sorb PCBs in the water
column or enhance PCB burial rates.

Figure 8-10 illustrates the base flow at Fort Edward during the continued No Action simulation
and the scaling of the first spring peak to match the 100-year peak flow. The value of 12,000 cfs
for the maximum spring peak in  1988 (the hydrograph for  1988 was randomly  selected to
represent the hydrology of the first year of the 70-year forecast) was nearly half of the historical
average spring peak flow of 21,339 cfs at Fort Edward. Due to the relatively quiescent nature of
this spring flood during  the first year of the forecast, daily flows at Fort Edward were scaled by
applying exponential trends to the rise and fall periods  between March 26 and April 18.  The
1988 peak spring flow in the time  series was scaled up from 12,000 cfs to a value of 47,330 cfs
on April 5, approximately a factor or four. The duration, rise and fall of the 100-year flow in this
modified hydrograph were generally consistent with other observed peak flows in  the 21-year
historical period. The rise period from 50 percent of the peak flow was approximately five days,
and the fall period to 50 percent  of the peak flow was just under seven  days. This modified
hydrograph represents a 100-year peak flow but does not necessarily represent the duration, rise
and recession characteristics of an actual 100-year flood event.

8.5 100-YEAR PEAK FLOW SIMULATION RESULTS

Table 8-3 presents the impact of the 100-year peak flow on sediment bed solids in Thompson
Island  Pool  (TIP) and downstream reaches of the Upper Hudson River. Cohesive sediments in
the Pool were computed to be scoured with a net bed elevation decline of approximately 0.28 cm
(on a poolwide average basis).  The poolwide average non-cohesive sediment bed elevation
declined  by approximately 0.05 cm.  Some individual  HUDTOX cohesive and non-cohesive
sediment segments experienced higher erosion depths. As noted above and in the discussion of
the HUDTOX calibration, resuspension of non-cohesive sediment is treated in a relatively simple
fashion so these predictions are more uncertain for simulation of event  impacts.

The small erosion depths predicted  throughout the  Upper Hudson in response to the 100-year
peak flow produce only minor fluctuations in surface sediment concentrations. This finding is in
agreement  with the findings reported in Chapter 4 from  the separate DOSM application to
cohesive sediments in Thompson Island Pool.

The 100-year event does not produce increased surface sediment concentrations of the  nature
observed later in the 70-year forecast. This is because the increases observed in the forecast are a
result of long-term net erosional behavior that over time works to expose buried concentrations.
The incremental  impacts of the  100-year  event on  the long-term  erosion depth speeds the
occurrence of these increases by about one year relative to the No Action forecast.

Differences in water column Tri+ concentrations between the No Action scenario and imposition
of a  100-year flood event are of relatively short  duration. A comparison of predicted water
column concentrations between the 100-year peak flow and No Action is  shown for Thompson
                                          163                      Limno-Tech, Inc.

-------
Island Dam and Federal Dam in Figures 8-11  and 8-12, respectively. Although  a significant
increase in water column Tri+ levels occurs during the event due to resuspension, the differences
are short-lived and  decline to a fairly  insignificant  levels fairly rapidly once  the event  is
completed.  The impact of the 100-year peak flow on surface  sediment levels  (i.e., the top 2
HUDTOX sediment layers) was minimal. Cohesive sediment Tri+ concentrations in the TIP
increased  marginally (on the order of 0.1 ppm), and then decline nearly back to No Action
forecast levels within less than 4 years. Impacts on non-cohesive  surface sediment contamination
levels were insignificant.

Figure 8-13 shows the impact on Tri+ mass loading at various  locations in the Upper Hudson
River caused by imposing the 100-year peak flow on  the base No Action scenario. The event
causes an  increase of slightly less than 26 kg (57 Ibs.) in cumulative Tri+ mass  loading across
Thompson Island Dam by  the end of the first year of the forecast. This increase represents
approximately 13 percent of the average annual Tri+ mass loading across Thompson Island Dam
during the 1990s.  The Tri+ mass loading increase at Federal  Dam during the first year was
approximately 73 kg (161 Ibs.)

Also note  that almost all of the increase in Tri+ mass loading over No Action levels depicted by
Figure 8-13 occurs during the course of the flood event (March 26 through April 18). In general,
the in-river mass loading effect of the 100-year peak flow is very short-lived. Subsequent (post-
first year)  increases over the No Action predictions generally amount to less than 2 kg per year at
Thompson Island  Dam and downstream locations, and  this  impact  eventually  declines to
negligible levels.

A similar  analysis of 100-year peak flow impacts was  conducted by  General Electric (QEA,
1999). Both analyses used the same peak daily average flow for the 100-year event.  On the basis
of predicted sediment scour depths, results for these two modeling analyses were comparable
when corrected for the flood plain effects, which were not included in the GE model.

8.6 SENSITIVITY ANALYSIS

Results of sensitivity analyses are provided for the No Action forecast simulation to illustrate the
impact of the assumptions made regarding future flow and solids  loading conditions. In addition,
sensitivity  analyses  for non-cohesive sediment  particle  mixed depth and  sediment  initial
conditions are presented to  show the influence  of these parameters on the magnitude of the
increases in surficial sediment concentrations observed in the No Action forecast results.  All
sensitivity results presented here  are compared to the No Action forecast based on  specification
of a constant Tri+ concentration of 10 ng/L at Fort Edward.

8.6.1 Sensitivity to Specification of Forecast Hydrograph

To demonstrate  the forecast sensitivity  to  the choice  of  simulation  hydrograph, No Action
forecasts were conducted with three alternative synthetic 70-year hydrographs in addition to the
baseline forecast hydrograph. These hydrographs were developed as described in Section 8.2.1.
Sensitivity results are presented for surface sediment Tri+ concentrations (8-14 a-e) and annual
average water column concentrations (Figure 8-15 a-b).
                                           164                      Limno-Tech, Inc.

-------
In general, specification of the forecast hydrograph does not change either the long-term rates of
decline in concentrations or the quasi  steady-state concentrations based on the 10 ng/L Tri+
concentration at Fort Edward. Hydrograph choice does, however, significantly affect the timing
of the computed increases in surface sediment Tri+ concentrations.  In Thompson Island Pool,
surface sediment concentration increases due to sediment erosion occur as early as 2013 versus
2044 with the baseline forecast hydrograph.  In non-cohesive sediments in the Stillwater reach,
timing of the largest increase in sediment concentrations is affected by about 13 years among the
four hydrographs.

The water column results from the  alternative  hydrographs show significant year-to-year
variability relative to each other, although the overall trends are the same. The impact of the
hydrograph on the  forecast results does show  notable differences in the  first few years of the
forecast period.  However, these differences are based on the choice of forecast hydrograph.  The
apparent rate of recovery is different over the first few years, but then overall trends normalize to
similar conditions for all 4 simulations.

8.6.2 Sensitivity to Solids Loads at  Fort Edward

The magnitude  of solids loading plays  an important  role  in determining future Tri+
concentrations in the Upper Hudson River. As discussed in Section 8.1.1.2, the specification of
future  solids loads  at Fort Edward  was estimated by the solids rating curve developed  for the
1991-1997 period.  The possibility that solids loads may occur at a level observed during earlier
periods of the historical calibration  cannot be  excluded.  To assess the significance of  this on
forecast results, a sensitivity analysis was conducted with a solids rating curve based on all
available historical data from 1977 to 1997 at Fort Edward (Equation 6-8).

Results indicate that higher solids loads at Fort Edward produce a faster rate of decline of Tri+
surface sediment concentrations  and concentrations  asymptotically approach a  slightly lower
quasi steady-state concentration with the upstream load (Figure 8-16). The higher solids loads
also result in more  solids deposition and delay the occurrence of the predicted concentration
increases in cohesive surface sediments.  Overall the impact of using the solids rating curve
estimated based on the 1991 to 1997  period relative to use of the 1977 to 1997 period is small.

8.6.3 Sensitivity to Tributary Solids Loads

The solids mass balance described in Chapter 6 estimated tributary solids loads downstream of
Thompson Island Pool  from limited available data. This solids  balance  was premised on the
assumptions  that the Upper Hudson River below  TIP was net depositional  over the  21-year
historical calibration period, and that solids  trapping efficiencies estimated using data  for TIP
could also be applied to reaches downstream of TIP.  To show the sensitivity of the  forecast
results to tributary solids loads, sensitivity analyses  were conducted with 50 percent upward and
downward adjustments to all tributary solids loads to the river.

The result of 50 percent increases and decreases in tributary solids loads is  generally to speed and
slow rates of Tri+ concentration declines, respectively.   Responses in Thompson Island Pool
(Figure 8-17a) are  smaller than  responses  in  downstream reaches (Figure 8-17b, e)  because
                                            165                       Limno-Tech, Inc.

-------
 tributaries account for most of the solids inputs to downstream reaches and only a small portion
 of solids loadings to  TIP.  The occurrence, magnitude and timing of computed increases in
 surface sediment Tri+ concentrations are  strongly influenced  by changes in solids loadings.
 Increases of  50 percent in tributary solids  loadings  cause the complete  disappearance of
 computed increases in Tri+ concentrations. Decreases of 50 percent in tributary solids loadings
 cause  increases in   both  the  frequency  and  magnitude of  computed increases in  Tri+
 concentrations.  Furthermore, new increases in Tri+ concentrations are now computed to occur in
 the surface sediments of the Waterford  and  Federal Dam reaches (Figure 8-17d, e).  The 50
 percent adjustments  to tributary solids loads in this sensitivity analysis are considered  to be
 within the uncertainty of the load estimates, therefore, these alternative results are all considered
 plausible.  Use  of the forecast results in decision making should consider the possibility  that
 long-term erosional  behavior downstream of Thompson Island Pool may occur on  a  more
 frequent basis than indicated by the baseline forecast results, or may not occur at all depending
 on future solids loadings.

 8.6.4 Sensitivity to Particle Mixing

 The sediment mixed  layer depth and particle mixing rate  are parameters  for  which  direct
 measurements are not  available.  In the historical calibration and the  base forecast simulation,
 sediment  mixed layer depths were 10 cm and 6 cm,  respectively, in the cohesive and non-
 cohesive  sediment areas of Thompson Island Pool.  In reaches downstream of TIP, sediment
 mixed layer depths were 10 cm and 4 cm, respectively, in cohesive and non-cohesive sediment
 areas.  Initial estimates of sediment particle mixed depths were guided by available information
 from vertical profiles of sediment Tri+ concentrations, and then adjusted  as part  of the model
 calibration process.  A sensitivity analysis to non-cohesive sediment mixed layer depth was
 conducted for the forecast by changing the mixed layer depth from  4  to  6 cm in  reaches
 downstream of TIP.  This was done primarily to assess sensitivity of computed sediment Tri+
 concentration  increases  in  the  Stillwater  reach.   These simulations used  sediment  initial
 conditions for 1991-1997 simulations also computed with mixed layer depth set to 6 cm.

 Results indicated  slower  declines in surface sediment  Tri+   concentrations  in all  reaches
 downstream of TIP (Figure  8-18 a-e), compared to the baseline No Action case.  These responses
 were due to greater  upward mixing of Tri+ mass from  the depth interval between 4 and 6
centimeters. The computed increase in surface sediment Tri+ concentration in the  non-cohesive
 sediments of the Stillwater reach was approximately 20 percent less than the increase in the base
 forecast simulation.  Water column Tri+ concentrations are shown in Figure 8-19 at  Stillwater.

8.6.5 Sensitivity to Sediment Initial Conditions

 Sediment initial conditions for the 1998 through 2067 forecast simulations were based on the  end
results of the 1991 through  1997 model hindcast application.   In  effect, this corresponds to
initializing the forecast simulations with measured conditions in 1991. This approach utilizes the
 most recent, reliable  and comprehensive  dataset to begin the model forecasts.  In  particular, in
 terms of accuracy, vertical resolution and spatial coverage for the Upper Hudson River, the 1991
 sediment  measurements represent the best  match between available sediment data and model
 requirements.  An alternate  approach would have been to use the end results of the  1977 through
                                           166                       Limno-Tech, Inc.

-------
1997 historical calibration.  This would correspond to initializing the forecast simulations with
measured conditions in 1977.

A forecast sensitivity analysis was conducted to investigate the impacts of changes in sediment
initial conditions.  This analysis involved initializing the forecast simulation  with measured
conditions in 1977.   The principal result from  this sensitivity analysis was  that computed
increases in surface sediment Tri+ concentrations in TIP and the Stillwater reach were magnified
relative to computed increases in the base forecast simulation (Figure 8-20 a-e).  Water column
Tri+ concentration increases were also magnified (Figure 8-21 a, b).  This indicates that the
magnitudes of computed increases in surface sediment Tri+ concentrations during the forecast
simulations  depend on  the  temporal  history  of Tri+ vertical concentration profiles in the
sediments.

8.7 EXPOSURE CONCENTRATIONS FOR AUGUST 1999 AND DECEMBER 1999 RISK
             ASSESSMENTS

The HUDTOX model was developed and refined over a period of years, and EPA conducted the
risk assessments for the Reassessment concurrently with this modeling effort.  Accordingly, EPA
used the most updated version of HUDTOX and the latest model results that were available at the
time the risk assessments were conducted. The processing of HUDTOX results for linkage with
the FISHRAND bioaccumulation model is discussed in Book 3 of this report.

The computed total PCB concentrations in water and surface sediment in the No  Action forecast
from the May 1999 BMR  were used in the August 1999 Ecological Risk Assessment and the
Human Health Risk Assessment for the Upper Hudson River.  These results are based on initial
conditions in sediment specified from the  1991 GE  composite data set and 10 ng/L  PCBs in
water at the upstream boundary (see. Appendix A).

The computed Tri+ concentrations in water and surface sediment in the No Action forecast from
this RBMR report were used in the December 1999 Ecological Risk Assessment for Future Risks
in the Lower Hudson River and the Human Health Risk Assessment for the Mid-Hudson River.
These results are based on initial conditions in sediment specified from the 1977  data set and 10
ng/L PCBs in water at the upstream boundary (see, Appendix A).

8.8 PRINCIPAL FINDINGS AND CONCLUSIONS

Several  important conclusions were  drawn from  the  No Action and  100-year peak  flow
simulations provided in this report.  The conclusions drawn from these simulations are based on
the No  Action  forecast and  100-year event applications of the HUDTOX model that was
successfully calibrated to  long-term  trends  of  water column and  surface  sediment  Tri+
concentrations.   Findings  and conclusions from  the No Action forecast, the  100-year event
simulation and the selected sensitivity analyses are addressed in the sections below.
                                          167                       Limno-Tech, Inc.

-------
8.8.1 No Action Forecast

The  principal findings  and conclusions from the No Action  forecast simulations are the
following:

       •     Forecasted surface sediment  Tri+ concentrations continue to decline  at
             approximately  7  to 9  percent  per year over  the next two decades,
             consistent with long-term historical trends.

       •     Forecasted surface sediment Tri+ concentrations eventually  reach  levels
             determined by upstream boundary Tri+ loadings at Fort Edward.  Under
             the assumptions in the forecast simulations, this occurs after  the first two
             decades of the forecast period.  For the first two decades, the in-place Tri+
             reservoir in the  sediments and sediment-water transfer processes control
             long-term responses of surface sediment concentrations.

       •     Forecasted water column Tri+ concentrations continue to decline for the
             first one to two decades and are very sensitive to  Tri+ loading at Fort
             Edward.   Based on specification of constant Tri+ concentrations  of  10
             ng/L and 30 ng/L at Fort Edward, the Fort Edward load begins to control
             average annual water column responses after 12 to 15 years.

       •     Declines in Tri+ loads to the Lower Hudson River  mirror water column
             Tri+ declines. They reach a quasi steady-state asymptote of 45kg/yr and
             125 kg/yr for the  10 and 30 ng/L Tri-i- concentration assumptions  at Ft.
             Edward.

       •     Surface sediment  Tri+  concentrations  in localized areas in Thompson
             Island  Pool  and the Stillwater reach are  forecasted  to increase 40  to  50
             years in the future. These computed increases occur due to the long-term
             consequences of small sediment erosion rates that eventually expose Tri+
             contamination originally present in deeper sediment layers.

       •     The relative magnitudes of computed increases in surface sediment Tri+
             concentrations  are small within  the  context of  long-term trends  in
             historical concentrations; however, they may be important in a potential
             remediation decision.   The occurrence, magnitude  and timing of these
             computed increases are dependent on forecast assumptions.

       •     Forecasted  responses  of water column  and  surface  sediment  Tri+
             concentrations in the  Upper Hudson River were sensitive to changes  in
             hydrology, solids loadings, sediment particle mixing depth and sediment
             initial conditions.  Long-term responses were most sensitive to changes in
             tributary solids loadings and sediment mixing depth.  Computed increases
             in surface sediment Tri+ concentrations were most sensitive to changes in
             tributary solids loadings and sediment initial conditions.
                                           168                       Limno-Tech, Inc.

-------
The No Action forecast findings are affected by uncertainty in upstream Tri+ loads. In general, if
Tri+ loads stay at or below levels observed in the past few years (1997 through 1999), surface
sediment  Tri+ concentrations  are expected to show declines consistent  with  historical rates.
Model forecasts show that concentration declines are likely to exhibit half-lives of about 7 to 10
years. In other words, every 7 to 10 years, concentrations will decrease by  50 percent. However,
beyond two  decades,  forecasted  surface  sediment  Tri+  concentrations  will reach  levels
determined by the assumed constant upstream boundary concentrations at Fort Edward.

8.8.2 100-Year Peak Flow Simulation

The principal findings and conclusions from the 100-year peak flow simulation are the following:

       •      Results of the 100-year peak flow simulation show that a flood of this
              magnitude would result in only a small additional increase  in sediment
              erosion beyond what might be  expected for a reasonable range of annual
              peak flows. A 100-year peak flow is 39 percent larger than the peak flows
              included in the base No Action forecast simulation.

       •      The small sediment scour depths produced by the  100-year peak flow
              result in only very small increases in surface sediment Tri+ concentrations.
              These increases are short-lived and decline to values in the base forecast
              simulation (without the 100-year peak flow) in approximately four years.

       •      Increases in water column Tri+ concentrations in response to a 100-year
              peak flow are very short-lived and decline rapidly after occurrence of the
              event.  The event causes an increase of 26 kg (57 Ibs) in cumulative Tri+
              mass loading across Thompson Island Dam by the end of the  first year of
              the forecast.  This increase  represents approximately  13 percent of the
              average annual Tri+ mass loading across Thompson Island Dam during the
              1990s.

       •      The occurrence of a 100-year peak flow is not likely to have a substantial
              effect on the future course of Tri+ concentrations in the water or sediments
              of  the  Upper Hudson River relative  to the base No  Action  forecast
              simulation.

Results from  simulation of a 100-year peak flow with  HUDTOX are consistent with those
reported for the Depth of Scour Model in Chapter 4.
                                           169                       Limno-Tech, Inc.

-------
THIS PAGE INTENTIONALLY LEFT BLANK
                  170                Limno-Tech, Inc.

-------
Chapter 9

-------
9. HUDTOX VALIDATION

9.1 OVERVIEW

Model validation is the process of confirming the ability of a model to predict observed behavior
using datasets that are independent of the datasets used to calibrate the model.  Validation is a
test of the scientific rigor of a model and of its utility as a predictive  tool.  Should a model fail a
validation exercise, its predictive ability is suspect.  Conversely,  should a model's predictions
compare well with validation datasets, its suitability for forecasting  is considered validated and
conclusions drawn from model predictions are strengthened.   With this in mind, a validation of
the HUDTOX model  was pursued using an independent dataset collected after the calibration
period.

Water column data collected by GE in 1998 were available to conduct an independent validation
of the HUDTOX model. These data are not included in Release 4. Ib of Hudson River Database,
the main data source for the modeling work presented in this report.  The GE 1998 data include
PCB measurements of water column and sediment concentrations.  A portion of these data are
presented in O'Brien and Gere 1999a and 1999b. Validation of the  model to the water column
dataset provided an assessment of the sediment-water mass transfer processes in the model.

Observed sediment concentrations served as the primary calibration targets for the 1977  to 1997
historical calibration.  The calibrated model  predicts the observed  sediment concentrations in
1998  reasonably well.  However, year to year changes  in sediment concentrations are small.
Therefore, focus of the validation testing of HUDTOX was on the water column data for  1998,
available at  Fort Edward, Thompson Island  Dam and Schuylerville.  Model results for 1998
agreed very well with the observed seasonal variations in water column data.

9.2 VALIDATION APPROACH

The validation simulation was conducted for the period of January  1, 1998 through September
30, 1998 because USGS flow records extending beyond this  date were  not available when this
analysis was conducted. Beyond specification of 1998 flows and loads,  all other model  input in
this validation runs were unchanged from the historical calibration.

Tributary solids loads and input hydrographs were calculated based  on USGS flow data, solids
data,  and solids to flow  regressions  used  in the  calibration.   The major tributaries which
contribute significant  solids  and flow (Hoosic River and the Mohawk River) are all located
downstream of the two locations where 1998 data was available for comparison to model output.
Thus, any uncertainty  in assumptions for these sources should not significantly affect the results
of this validation exercise. Improved estimates of solids loads from Batten Kill would make the
comparisons of model output and data at the Schuylerville location slightly more accurate.

Linear interpolation of Fort  Edward Tri+ concentrations  was used to  specify the  upstream
boundary condition for the validation. Computed surficial sediment concentrations at the end of
the model calibration period were specified as sediment initial conditions,  the same as described
                                          171                      Limno-Tech, Inc.

-------
in Chapter 8 for the No Action simulation.  All other model parameters and coefficients were
identical as employed in the calibration period simulations.

9.2.1 Validation Results

Year to year changes in surface and sediment concentrations are small and hence 1998 sediment
conditions were not used as a primary measurement of validation.  In fact, 1998 sediment data
were used in the historical calibration in Chapter 7.  However, water column PCB concentrations
can  vary significantly  year to year,  season to season or even day to day due to changes in
hydrology, loads, and sediment effects.

Water column PCB observations are  available for Tri+ at Fort Edward, Thompson Island Dam
and  Schuylerville. The sampling frequency was approximately weekly with a few exceptions.
Model results were compared with these data by visual inspection of time series concentrations.
Results are shown for the last three years of the calibration period and the validation period, 1998
(Figures 9-1 and 9-2).  Model comparisons to just the 1998 data are presented in Figure 9-3.

Good agreement is observed between model results and observations.  The model  generally
reproduces the observed concentrations through the entire 1998 validation period.  Figures 9-4
and 9-5 show scatter diagrams comparing model output and data at Thompson Island  Dam and
Schuylerville on a monthly  average  basis.  Based on the good agreement of the model with
observed concentrations at Thompson Island  Dam and Schuylerville, the HUDTOX model
validation appears reasonably successful.  While the extent of this validation period does not
necessarily lead to a validation of the model's ability to make accurate long-term projections of
exposure concentrations, it does provide a test of the model  to represent annual fluxuations in
water column PCB concentrations outside of the calibration period.  This tends to strengthen the
model's utility as a predictive tool.

9.2.2 Validation Summary

The  model validation was conducted  by comparing predicted water column Tri+ concentrations
to observed concentrations  in 1998.   Results indicate good  agreement between predicted  and
observed concentrations at both Thompson Island Dam  and Schuylerville over an entire year,
spanning a range  of environmental conditions.  The validation is considered successful  and it
enhances the model's credibility as a  predictive tool for use in assessing the future course of the
river's recovery from historical contamination under continued No Action.
                                           172                       Limno-Tech, Inc.

-------
References

-------
REFERENCES

Achman,  D.R., B.J. Brownawell,  L.  Zhang, 1996.   "Exchange of  polychlorinated
biphenyls between sediment and water in the Hudson River Estuary." Estuaries.  19(4):
950-965.

Achman,  D.R.,  K.C. Hornbuckle,  and S.J.  Eisenreich,  1993.   "Volatilization of
Polychlorinated Biphenyls from Green Bay, Lake Michigan." Environmental Science and
Technology. 24(1), pp. 75-.

Aller, R.C., 1982. "The  effects of macrobenthos on chemical properties of marine
sediment and overlying water." In Animal-sediment Relations, The Biogenic Alteration
of Sediments. P.L. McCall and J.J.S. Tevesz (eds.), Plenum Press, NY.

Ambrose, R.B., Jr.; T.A.  Wool; and J.L. Martin, 1993.  "The Water Quality  Analysis
Simulation Program, WASPS - Part A: Model Documentation." Environmental Research
Laboratory, Office of Research and Development, USEPA, Athens, Georgia.

Baun, K., 1982. "Alternative Methods of Estimating Pollutant Loads in Flowing Water."
Tech. Bulletin 133, Dept. Natural Resources, Madison.

Beale, E.M.L., 1962.  "Some uses of computers  in operational research." Industrielle
Organisation. 31: 51-52.

Berger,  R.C.,  1990.   "Mass   Conservation  In  The  RMA2V Code."   Hydraulic
Engineering:  Proceedings of the 1990 National Conference, San Diego, CA, July 30-31,
1990. American  Society of Civil Engineers, Boston Society of Civil Engineers Section,
Boston, MA, 2, 873-878.

Bierman,  V.J., Jr.; J.V. DePinto; T.C. Young;  P.W. Rodgers; S.C. Martin; and R.
Raghunathan, 1992.  "Development  and Validation of an Integrated Exposure Model for
Toxic Chemicals  in Green Bay,  Lake  Michigan."  Large  Lakes and Rivers Research
Branch Environmental Research Laboratory - Duluth.  U.S.  Environmental Protection
Agency.

Bopp, R.F., 1979.  "The geochemistry  of polychlorinated biphenyls  in the Hudson
River." Ph.D. Dissertation.  Columbia University. Univ. Micro. Intern. 8008704, 191 pp.

Bopp, R.F., H.J. Simpson, and B.L.  Deck, 1985. "Release of Polychlorinated Biphenyls
from Contaminated Hudson River Sediments."  Final Report NYS C00708 to NYSDEC,
Albany, New York.

Borah, D. K.,  1989.  "Scour Depth Prediction Under Armoring Conditions", Journal
Hydraulic Engineering 115(10), p. 1421.
                                      173                        Limno-Tech, Inc.

-------
Boudreau, B.P., 1997.  Diagenetic Models and Their Implementation.  Springer Verlag,
Berlin, Heidelberg. 414 pp.

Brown, J.F Jr., R.E. Wagner and D.L. Bedard,  1984.  "PCB Transformations in Upper
Hudson Sediments." Northeastern Environmental Science 3:184-189.

Brown, M P.; M. B. Werner; C.R. Carusone, M. Klein, 1988.  "Distribution of PCBs in
the Thompson Island Pool of the Hudson River; Final Report of the Hudson River PCB
Reclamation Demonstration Project Sediment Survey." May 1988.

Brownawell,  B.,  1986.  "The Role  of Colloidal  Organic  Matter  in  the Marine
Geochemistry of PCBs". Ph.D. dissertation. Woods Hole Oceanographic Institution,
Woods Hole, MA. WHO 1-8619-19.

Brunner, S.  E. Hornung, H. Santi, E. Wolff, and O.G. Piringer,  1990. " Henry's Law
Constants  for Polychlorinated  Biphenyls: Experimental Determination and Structure -
Property Relationships." Environmental Science Technology 24(11): 1751-1754.

Buckley,  E.H. and T.J.  Tofflemire,  1983.    "Uptake of  Airborne  Polychlorinated
Biphenyls in Foliage." Science 216: 520-522.

Butcher, J.B.,2000a. "Flood Frequency Analysis, Fort Edward at Rogers Island Gage,
Hudson River. Memorandum" from J.B. Butcher of Cadmus to: V. Bierman, S. Verhoff,
Limno-Tech Inc.; E. Garvey, TAMS; and, D. Merrill, Gradient, Inc. June 18, 1993.

Butcher, J.B., 2000b. Memorandum from J.B. Butcher of Tetra Tech to V. Bierman and
S. Hinz of Limno-Tech, Inc. and E.  Garvey and A. DiBernardo of TAMS Consultants,
Inc. regarding Historic Sediment PCB Data Conversions dated June 9, 1998.

Butcher, J. B., E. A. Garvey and V. J. Bierman, Jr., 1998. Equlibrium Partitioning of PBC
Congeners  in the Water Column :   Field  Measurements  From the Hudson River.
Chemosphere, Vol. 36, No. 15, pp. 3149-3166.

Capel,  P.D. and SJ Eisenreich, 1990.  "Relationship between chlorinated hydrocarbons
and organic carbon in sediment porewater." Journal of Great Lakes Research 16(2): 245-
257.

Chapra, S.C., 1997.  Surface Water-Quality Modeling. McGraw Hill.  New York, New
York.

Chow,  V.T., 1959. Open-Channel Hydraulics. McGraw-Hill. New York, New York

Chow,  V.T., 1960. Open-Channel Flow. McGraw-Hill. New York, New York.

Cochran, W.G., 1977. Sampling Techniques. 3rd ed.  John Wiley and Sons,  Inc.  New
York, New York.
                                      174                        Limno-Tech, Inc.

-------
Cohn, T.A.; L.L. DeLong; E.J. Gilroy; R.M. Hirsch; and O.K. Wells, 1989. "Estimating
constituent loads."  Water Resources Research 25(5):  937-942.

Decisioneering, Inc., 1996.  Crystal Ball® Version 4.0 User Manual.  2530 S. Parker
Road, Suite 220, Aurora, CO 80014 U.S.A.

DiToro, D.M., 1985. "A  particle  interaction model of reversible organic chemical
sorption" . Chemosphere 14(10): 1503-1138.

Eadie, B.J. and J.A. Robbins, 1987.  "The role of particulate matter in the movement of
contaminants in the Great Lakes."  Advances in Chemistry Series No.  216, Sources and
Fates of Aquatic Pollutants.  R.A. Kites  and S.J. Eisenrech (eds.), American Chemical
Society. Chapter 11.

Eadie, B.J.; N.R. Moorehead, and P.P.  Landrum, 1990.  "Three-phase partitioning of
hydrophobic organic compounds in Great Lakes waters."  Chemosphere.  20(1-2): 161-
178.

Eadie,  B.J., N.R.  Moorehead,  J.  Val Klump, P.P.  Landrum,  1992.  "Distribution of
hydrophobic organic compounds and particulate organic matter in Green Bay waters."
Journal of Great Lakes Research 18(1): 91-97.

ESRI, 1997. ARC/INFO Version 7.1.2. Environmental Systems Research Institute, Inc.,
Redlands,  California.

Federal Emergency Management Agency (FEMA), 1982. Flood Insurance Study for the
Town ofFL Edward, New York,  Washington County. Washington, D.C.

Fischer, et al.,  1979 Mixing in Inland and Coastal  Waters,  Academic Press, Inc.  San
Diego, CA.

Flood,  R.D.,  1993.   "Analysis  of Side-Scan Sonar, Bathymetric,   Subbottom,  and
Sediment Data from the Upper  Hudson River between Bakers Falls and Lock 5."  State
University  of New York at Stony Brook, Marine Science Research Center.   Report to
TAMS Consultants, Inc. for the Hudson River PCB Reassessment   RI/FS.  (Original
Report September,  1993; Report Errata March, 1994).

Gailani, J., C.K. Ziegler, and W. Lick, 1991.  "Transport of suspended solids in the lower
Fox River." Journal of Great Lakes Research 17(4): 479-494.

Gailani, J., W. Lick, C.K. Ziegler, and D. Endicott, 1996.  "Development and calibration
of a fine-grained sediment transport model for the Buffalo River." Journal of Great Lakes
Research 22(3):765-778.
                                       175                        Limno-Tech, Inc.

-------
Hicks D.M. and P.O. Mason, 1998.  "Roughness Characteristics of New Zealand Rivers"
National Institute of Water and Atmospheric Research Ltd. Water Resources Publication.
September.

Hoff, R.M., W.M.J.  Strachan,  C.W. Sweet, 1996.  "Atmospheric deposition of toxic
chemicals to  the  Great Lakes: A review  of data through  1994."   Atmospheric
Environment 30(20):3505-3527.

Horn, E.G., L.J. Hetling, and T. J. Tofflemire, 1979.  "The problem  of PCBs in the
Hudson River." Ann. N.Y. Acad. Sci. 320: 591-609.

HydroQual, Inc, 1995.  "The Erosion Properties of  Cohesive  Sediments in the Upper
Hudson River." General Electric Company.

HydroQual, Inc, 1997.  "Development of Corrections for Analytical Biases in the 1991-
1997 GE  Hudson  River PCB  Database."   Prepared  for General Electric Company,
Albany, New York. June 1997.

Karickhoff, S.W.,  D.S. Brown,  and T.A. Scott,  1979.  "Sorption  of  hydrophobic
pollutants on natural sediments." Water Resources. 13:241-248.

Karickhoff, W.W. 1984.  "Organic  pollutant sorption in aquatic systems."  Journal of
Hydraulic Engineering.  110: 707-735.

Karim, M. F., and J. F. Kennedy,  1982. "IALLUVIAL: A computer-based flow and
sediment routing model for alluvial streams and its application to the Missouri River",
Report No. 250, Iowa Institute of Hydraulic Research, Iowa City, Iowa.

Leopold, L.B., and T. Maddox,  1953. "The Hydraulic Geometry of Stream Channels and
Some Physiographic Implications." Professional Paper  252, U.S. Geological  Survey,
Washington, D.C.

Lick, W., J. McNeil, Y.  Xu, and C. Taylor, 1995. 'Resuspension Properties of Sediments
from the  Fox, Saginaw, and  Buffalo Rivers."  Journal  of Great Lakes  Research,
International Association for Great Lakes Research. 21(2):257-274.

Limburg,  K.E., 1984.  "Environmental  Impact Assessment of  the  PCB  Problem:  A
Review." Northeastern  Environmental Science. 3(3/4): 122-136.

Lin, H. J., and D.R. Richards, 1993. "Numerical  Model Investigation  of Saugus River
and Tributaries, Massachusetts, Flood Damage Reduction," Technical Report HL-93-5,
U.S. Army Engineer Waterways Experiment Station, Vicksburg, MS.

Little,  W. C.,  and  P. G. Mayer, 1972.  "The role  of sediment gradation on channel
armoring." School of Civil Engineering,  Georgia Inst. of Tech., Atlanta, GA.
                                      176                        Limno-Tech, Inc.

-------
MacKay, D., 1989.  "Modeling the Long-Term Behavior of an Organic Contaminant in a
Large Lake:  Application to PCBs in Lake Ontario." Great Lakes Research. 15(2): 283-
297.

Mackay, D., W.Y. Shiu, and K.C. Ma, 1992. Illustrated Handbook of Physical-Chemical
Properties and Environmental Fate of Organic Chemicals.  Monoaromatic Hydrocarbons,
Chlorobenzenes, and PCBs. Volume 1. Lewis Publishers, Ann Arbor, MI.

Marti,  E.A.  and D.E. Armstrong,  1990.  "Polychlorinated Biphenyl Levels in Lake
Michigan Tributaries."  Journal of Great Lakes Research. 16(3): 396-405.

Matisoff, G., 1982. Mathematical models of bioturbation. In Animal-sediment Relations,
The Biogenic Alteration of Sediments. P.L. McCall and J.J.S.  Tevesz (eds.), Plenum
Press, NY. pp. 289-330.

McAnally,  W H., Jr.;  J. V. Letter, Jr.; J. P. Stewart; N. Brogdon; N. James, Jr.; T.W.
Anthony,  1984.   "Columbia River  Hybrid Modeling System,"  Journal of Hydraulic
Engineering 110(3), 300-311.

McCarl, P. L. and M. J. S. Tevesz, 1982. "The effects of benthos on physical properties
of freshwater sediments." In: Animal-Sediment Relations:   The Biogenic Alterations of
Sediments. Eds. P.L. McCall and M. J. S. Tevesz, Plenum Press, New York.

McNeil, J.  D., 1994. "Measurements of the Resuspension and Erosion of Sediments in
Rivers." Ph.D. dissertation. University of California at Santa Barbara.

Mehta, A.J., E.J. Hayter, W.R. Parker, R.B. Krone, and A.M. Teeter, 1989. "Cohesive
sediment transport. 2. Application."  Journal of Hydraulic Engineering 115 (8), pp. 1094-
1112.

Nadal,  E.O.,  1998. "Evaluation of physical/chemical mechanisms  controlling PCB
release  from river  sediments."   Ph.D.  dissertation.  Department  of  Civil  and
Environmental  Engineering,  Carnegie  Institute  of  Technology,  Carnegie Mellon
University.

Neter, J., W. Wasserman, and M. H. Kutner,  1990. Applied Linear Statistical Models.
Third Edition, Irwin, IL 60430.

New York  State Department of Environmental Conservation. "New York State Toxics
Air Monitoring Report" by Department of Environmental  Conservation Division  of Air
Resources 1982-1984.  Bureau of Air Quality Surveillance,  Bureau of Air Research.

NUS, 1984. "Volume  1, Feasibility Study,  Hudson River  PCBs Site, New York.  April,
1994."  EPA Contract No. 68-01-6699.  Table 4-12, pp4-13.
                                      177                       Limno-Tech, Inc.

-------
 O'Brien & Gere Engineers, Inc., 1993a. "Hudson River Project Sampling and Analysis
 Program;  1990 Sediment  Sampling and  Analysis Program."   Prepared  for  General
 Electric Company, May 1993.

 O'Brien & Gere Engineers, Inc., 1993b. "Data Summary Report, Hudson River Project
 1991  Sediment  Sampling and Analysis Program."  Prepared  for General  Electric
 Company  Corporate Environmental  Programs.  O'Brien  and  Gere  Engineers,  Inc.,
 Syracuse, New York.

 O'Brien & Gere  Engineers,  Inc.,  1993c.   "1991 Hydrographic survey of the Upper
 Hudson River Hudson River project sampling and analysis program." General  Electric
 Company, Corporate Environmental Programs. Albany, New York.

 O'Brien & Gere Engineers, Inc., 1993d. "Hudson River Project 1991-1992 Sampling and
 Analysis  Program Temporal Water Column Monitoring Program." General  Electric
 Company, Corporate Environmental Programs. Albany, New York.

 O'Brien & Gere  Engineers, Inc.,  1995.  "1994  Post  Construction Remnant  Deposit
 Monitoring Report." Syracuse, NY: O'Brien & Gere Engineers, Inc. November 1995.

 O'Brien & Gere  Engineers, Inc.,  1996.  "1995  Post  Construction Remnant  Deposit
 Monitoring Report". Syracuse, NY: O'Brien & Gere Engineers, Inc.

 O'Brien & Gere Engineers, Inc.,  1998. "Hudson  River Project 1996-1997 Thompson
 Island Pool Studies." Syracuse, NY: O'Brien & Gere Engineers, Inc. February 1998

 O'Brien & Gere Engineers, Inc., 1999.  "1998 High Flow Monitoring Program,  Hudson
 River Project." General Electric Company, Corporate Environmental Programs.  Albany,
 New York.

 Olsen,  C.R., H.J.  Simpson, T.H. Peng, R.F. Bopp and  R.M. Trier, 1981. "Sediment
 mixing and accumulation rate effects on radionucleide depth profiles in Hudson estuary
 sediments." /. Geophys. Res., 86:

 Parchure, T.M., and Mehta, A.J., 1985. "Erosion of soft cohesive sediment deposits."
Journal of Hydraulic Engineering., ASCE, 111(10), 1308-1326.

 Partheniades, E., 1965.  "Erosion and deposition of cohesive soils."  Journal Hydraulic
Division, ASCE, 91 (HY1), pp 105-138.

 Phillips, P.J. and D.W. Hanchar,  1996.  "Water Quality Assessment of the Hudson River
 Basin in  New York and Adjacent States- Analysis of  Available Nutrient, Pesticide,
Volatile Organic Compounds, and Suspended-Sediment Data, 1970-90," U.S. Geological
 Survey, Water-Resources Investigations Report 96-4065.
                                      178                        Limno-Tech, Inc.

-------
Portielje, R. and L. Lijklema, 1999. "Estimation of sediment-water exchange of solutes
in Lake Veluwe, The Netherlands." Water Research 33(1): 279-285.

Preston, S.D., V.J. Bierman, Jr., and S.E. Silliman,  1989.  "An evaluation of methods for
the estimation of tributary mass loads." Water Resources Research.  25(6):  1379-1389.

Preston, S.D., V.J. Bierman, Jr., and S.E. Silliman, 1992. "Impact of flow variability on
error in estimation of tributary  mass loads."  Journal of Environmental Engineering
118(3): 402-419.

Quantitative Environmental Analysis, LLC, 1998a. "Sediment Bed Mapping Study of
the Upper Hudson  River from Northumberland Dam to Troy Dam."  Prepared for the
General Electric Company.  Final Report.

Quantitative Environmental Analysis, 1998b. "Thompson Island Pool Sediment Coring
Program - Field Sampling Plan". General Electric,  Albany, New York.

Quantitative Environmental Analysis, 1999. "PCBs in the Upper Hudson River; Volume
2; A Model of PCB  Fate, Transport, and Bioaccumulation."  Prepared for General
Electric, Albany, New York. May 1999.

Rao,  P.S.,  and J.M.  Davidson,  1980.   "Estimation  of Pesticide  Retention  and
Transformation Parameters Required  for Nonpoint  Source  Pollution  Models." In
Environmental Impact of Nonpoint Source Pollution. Ann Arbor Sciences, Ann Arbor,
Michigan, pp. 23-67.

Rapaport,  R.A. and S.J. Eisenreich, 1988.  "Historical Atmospheric  Inputs of High
Molecular Weight Chlorinated Hydrocarbons to Eastern North America. Environmental
Science Technology. 22(8): 931-941.

Raudkivi, A.J., 1990. Loose Boundary Hydraulics. 3rd edition. Pergamon Press.

Reible, D.D., K.T. Valsaraj, and L.J. Thibodeaux,  1991.  Chemodynamic models for
transport  of contaminants  from sediment beds. In  The Handbook of Environmental
Chemistry. Vol. 2, part F. Hutzinger, O. (ed). Springer-Verlag, Berlin, pp. 185-228.

Rhea, J. and M. Werth, 1999. "Phase 2 Evaluation of Analytical Bias in the USGS Water
Column Database."  Technical Memorandum from Quantitative Environmental Analysis
to John Haggard, General Electric. March 22, 1999. 12 pp.

Rhea, J. and M. Werth, 1999.  "Memorandum  from  James  Rhea and Michael Werth to
John Haggard", GE  Corporate  Environmental  Programs,  Albany,  NY.   Phase 2
Evaluation of Analytical Bias in the USGS Water Column Database. March 22,  1999.

Richards, D.R., 1990.  "Flow Separation Around A  Solitary Dike; Eddy  Viscosity and
Mesh Considerations."  Hydraulic Engineering:  Proceedings of the 1990 National
                                      179                       Limno-Tech, Inc.

-------
 Conference, San Diego, CA, July 30-31, 1990.  American Society of Civil Engineers,
 Boston Society of Civil Engineers Section, Boston, MA, 2, 867-872.

 Sanders, J. E.,  1989.  The PCB-Pollution Problem in the Upper Hudson River: From
 Environmental  Disaster  To  "Environmental Gridlock,"  Northeastern Environmental
 Science, Vol. 8, No. 1

 Schlichting, H., 1979, Boundary-Layer Theory. McGraw-Hill, New York, New York.

 Shields, A., 1936. "Anwendung der Ahnlichkeitsmechanik und der Turbulenzforschung
 auf  die  Geschiebebewegung",  Mitteilungen  der  Preussischen Versuchsanstalt fur
 Wasserbau und Schiffbau, No. 26, Berlin, Germany.

 Sofaer, A.D., 1976.  "Interim Opinion and Order in the Matter of Alleged Violations of
 the Environmental Conservation Law of the State of New York by General Electric Co.,
 Respondent." NYSDEC File No. 2833. February 9, 1976, Albany, New York.

 SPSS Inc., 1998. SYSTAf® Version 8.0 for Windows® Statistics.  444 North Michigan
 Avenue, Chicago, IL 60611 U.S.A.

Tennekes, H., and Lumley J.L.,  1972.  A First Course in Turbulence The MIT Press,
Cambridge, MA.

Thibodeaux, L.J., 1996.  Environmental Chemodynamics: Movement of Chemicals in
Air, Water, and Soil. John Wily & Sons, Inc., New York.

Thibodeaux, L.J., D.D. Reible, W.S.  Bosworth,  L.C.  Sarapas,  1990. "A theoretical
evaluation  of the effectiveness  of  capping  PCB  contaminated  New Bedford Harbor
 Sediment."  Final  report. Hazardous Substance Research Center, Louisiana State
 University, Baton Rouge, LA.

Thomann, R.V., and J.A. Mueller, 1987. Principles of Surface Water Quality Modeling
and Control. Harper & Row, New York, New York.

Thomas, W. A. and W.H. McNally,  Jr., 1990. "User's manual for the generalized
computer program system: open-channel flow and sedimentation, TABS-2."  US Army
Engineer Waterways Experiment Station, Vicksburg, Mississippi.

Thorns,  S.R., G. Matisoff, P.L. McCall, and  X. Wang, 1995. "Models for alteration of
sediments by benthic organisms." Final technical report for Project 92-NPS-2, prepared
for Water Environment Research Foundation,  Alexandria, VA.

Thurston,  Kerry  A., 1998.  "Database Transmittal 10/13/98  (0612.246)."  E-mail from
 Kerry A. Thurston, O'Brien & Gere Engineers, Inc.  to John Haggard, General Electric
 regarding GE database update as of October 12, 1998.
                                      180                       Limno-Tech, Inc.

-------
Tofflemire, T.J., and S.O. Quinn, 1979. "PCB in the Upper Hudson River: Mapping and
Sediment Relationships". NYSDEC  Technical Paper No. 56. March 1979.  NYSDEC,
Albany, New York.

Tsivoglou,  E.E.  and J.R.  Wallace,  1972.   "Characterization  of Stream  Reaeration
Capacity"  EPA-R3-72-012. U.S. Environmental Protection Agency, Washington, D.C.

U.S. Environmental Protection Agency, (US  EPA).  1989.  "Statistical Analysis of
Groundwater Monitoring Data at RCRA Facilities - Interim Final Guidance." EPA/530-
SW-89-026.

U.S. Environmental Protection Agency, (USEPA),  1991.  "Phase I Report - Review
Copy.  Interim Characterization and Evaluation" Volume 1.   Hudson  River  PCBs
Reassessment RI/FS.  Prepared for USEPA, Region  II by TAMS Consultants,  Inc. and
Gradient Corporation,  August 1991.

U.S. Environmental Protection Agency, (USEPA), 1992.  "Final  Phase 2 Work Plan and
Sampling Plan."  Hudson River PCBs Reassessment RI/FS.    Prepared for  USEPA,
Region n by TAMS Consultants, Inc. and Gradient Corporation, September 1992.

U.S. Environmental Protection Agency, (USEPA), 1995. "Further Site Characterization
and Analysis, Volume 2A Database Report."  Phase  2 Report -  Review Copy.  Hudson
River PCBs  Reassessment RI/FS.    Prepared  for  USEPA,  Region H  by  TAMS
Consultants, Inc. and Gradient Corporation, October 1995.

U.S. Environmental Protection Agency, (USEPA),  1996. "Preliminary Model Calibration
Report. Phase 2 Report, Further Site Characterization and Analysis Volume 2B". Hudson
River PCBs Reassessment RFFS. USEPA Region H. Prepared  for USEPA  by Limno-
Tech, Inc., Menzie Cura and Associates, Inc., and The CADMUS Group, Inc., October
1996.

U.S. Environmental Protection Agency, (USEPA), 1997. "Further Site Characterization
and Analysis, Volume 2C - Data Evaluation and Interpretation Report." Hudson River
PCBs Reassessment RI/FS. Prepared for the USEPA by TAMS Consultants, Inc., The
Cadmus Group, Inc. and Gradient Corporation, February, 1997.

U.S. Environmental Protection Agency, (USEPA) 1998a. "Further Site Characterization
and Analysis" Volume 2C-A - Low  Resolution Sediment Coring Report, Hudson River
PCBs Reassessment RI/FS.  Prepared for USEPA, Region II and the U.S. Army  Corps of
Engineers, Kansas City District by TAMS Consultants, Inc., Gradient Corporation and
Tetra Tech, Inc., July 1998.

U.S. Environmental Protection Agency, (USEPA), 1998b. "CD-ROM Database for the
Hudson River PCBs Reassessment RI/FS, Phase 2 Report Further Site Characterization
and Analysis Database Report", Volume 2A, Release 4.1b. Prepared for USEPA, Region
                                     181                        Limno-Tech, Inc.

-------
 II and the U.S. Army Corps of Engineers, Kansas City District by TAMS Consultants,
 Inc., August, 1998.

 U.S.  Environmental  Protection  Agency  (USEPA),   1998c.  "Hudson  River PCBs
 Reassessment RI/FS  Responsiveness Summary for  Volume  2A:  Database  Report,
 Volume 2B: Preliminary Model Calibration Report, Volume 2C: Data Evaluation and
 Interpretation Report."  Prepared for USEPA and the U.S. Army Corps of Engineers,
 Kansas City District by TAMS Consultants, Inc.; Limno-Tech, Inc.; Tetra Tech, Inc.;
 Menzie Cura & Associates, Inc., December 1998.

 U.S. Environmental Protection Agency (USEPA), 1999a. "Responsiveness Summary for
 Volume 2C-A Low  Resolution  Sediment  Coring Report,  Addendum  to  the  Data
 Evaluation and Interpretation Report, Book 1 of 2".  Hudson River PCBs Reassessment
 RI/FS. Prepared for USEPA, Region II and  the U.S. Army Corps of Engineers, Kansas
 City District by TAMS Consultants, Inc., and Tetra Tech, Inc.,  February 1999.

 U.S. Environmental Protection Agency (USEPA), 1999b. "Responsiveness Summary for
 Volume 2C-A Low  Resolution  Sediment  Coring Report,  Addendum  to  the  Data
 Evaluation and Interpretation Report, Book 2 of 2."  Hudson River PCBs Reassessment
 RI/FS. Prepared for USEPA, Region II and the U.S. Army Corps of Engineers, Kansas
 City District by TAMS Consultants, Inc. and Tetra Tech, Inc., February 1999.

 U.S. Environmental Protection Agency (USEPA), 1999c. "Further Site Characterization
 and Analysis, Volume 2D - Baseline Modeling Report, Book 2 of 4."  Hudson River
 PCBs Reassessment RI/FS. Prepared for USEPA by Limno-Tech, Inc., Menzie-Cura &
 Associates, Inc., and Tetra Tech, Inc., May 1999.

 U.S. Environmental Protection Agency, (USEPA).  1999d. "Phase 2  Report -  Review
 Copy.  Further Site Characterization and Analysis.  Volume  2F-  Human Health  Risk
 Assessment." Hudson River PCBs Reassessment RI/FS.  Prepared for USEPA, Region II
 and the U.S. Army Corps of Engineers, Kansas City District by Gradient Corporation,
August, 1999.

U.S. Environmental Protection Agency, (USEPA),  1999e.  "Phase 2 Report -  Review
Copy.  Further Site Characterization and Analysis,  Volume 2E - Baseline Ecological
Risk Assessment, Hudson River PCBs Reassessment RI/FS." Prepared for USEPA by
TAMS Consultants, Inc., and Menzie-Cura & Associates, Inc., August 1999.

U.S. Environmental Protection Agency,  (USEPA).  1999f. "Phase 2  Report -  Review
Copy. Further Site Characterization and  Analysis.  Volume 2F-A Human Health Risk
Assessment for the Mid-Hudson River."  Hudson  River PCBs Reassessment RI/FS.
Prepared for USEPA, Region II and the U.S.  Army Corps of Engineers,  Kansas City
District by TAMS Consultants, Inc. and Gradient Corporation, December, 1999.

U.S. Environmental Protection Agency, (USEPA).  1999g. "Phase 2 Report -  Review
Copy. Further Site Characterization and Analysis.  Volume 2E-A Baseline Ecological
                                     182                       Limno-Tech, Inc.

-------
Risk Assessment for Future Risks in the Lower Hudson River." Hudson River PCBs
Reassessment  RI/FS. Prepared for USEPA,  Region II and  the U.S. Army  Corps of
Engineers,  Kansas  City  District  by TAMS  Consultants,  Inc. and Menzie-Cura &
Associates, Inc., December, 1999.

U.S. Environmental Protection Agency, (USEPA), 2000.  "Responsiveness Summary for
Volume 2D - Baseline Modeling Report, Phase 2."   Hudson  River PCBs Reassessment
RI/FS. Prepared for USEPA, Region H and the U.S. Army Corps of Engineers by TAMS
Consultants, Inc., Limno-Tech, Inc., Menzie-Cura & Associates, Inc., Tetra Tech, Inc.,
January, 2000.

USGS, 1969. "Time-of-Travel Study Upper Hudson River -  Fort Edward, NY to Troy
Lock and Dam, Troy NY." RI-10. Prepared by Harold L. Shindel, US Geological Survey,
US Department of the Interior, Washington DC.

USGS, 1993. "Water Resources Data Report" for New York, Volume 1.

Valsaraj, K.T., L.J. Thibodeaux, and D.D. Reible, 1997. "A quasi-steady-state pollutant
flux methodology for determining sediment quality criteria."  Environmental Toxicology
& Chemistry.  16(3):391-396.

van den Berg,  J., and  A. van Gelder,  1993.  "Prediction of Suspended Bed Material
Transport in Flows Over Silt and Very Fine Sand",  Water Res.  Res., Vol. 29 No. 5 pp.
1393-1404.

Vaughn, J.A., 1996. "Dissolved Organic Carbon Dynamics in the Upper Hudson River
Basin,"  thesis presented to Rensselaer Polytechnic  Institute, Troy New York, in partial
fulfillment of the requirements for the degree of Master of Science.

Wagner, L.A., 1982. Drainage Areas Of New York Streams, By River Basins: A Stream
Gazetteer. U. S. Geological Survey Water Resources  Investigations Open File Report 81-
1055. U. S.  Department of the Interior.

Wasserman N. J., and M.  H. Kutner, 1985.  Applied Linear Statistical Models, 2nd Ed.,
Richard D. Irwin, Inc.

Whitman, W.G., 1923.  "The two-film theory of gas adsorption."   Chemical and Met.
Engineering. Vol. 29, 147 pp.

Xu,  Y., 1991.  "Transport  properties of fine-grained sediments."  Ph.D. dissertation.
University of California at Santa Barbara.

Ziegler, C.K. and B. Nisbet, 1994. "Fine-grained sediment transport in Pawtuxet River,
Rhode Island." Journal Hydraulic Engineering. 120(5):561-576.
                                      183                       Limno-Tech, Inc.

-------
Zimmie, 1985.  "Assessment of erodibility of sediments in the Thompson Island Pool of
the Hudson River."  Final Report to the  Office of Special Projects, New York State
Department of Environmental Conservation. US EPA Grant No. C-36-1167-01.

Zreik, D.A., E.G.  Krishnappan, J.T.  Germaine, O.S.  Madsen, and  C.C.  Ladd,  1998.
"Erosional  and mechanical  strength of  deposited cohesive  sediments."  Journal of
Hydraulic Engineering, ASCE, 124(111):1076-1085.
                                      184                       Limno-Tech, Inc.

-------