PHASE 2 REPORT - REVIEW COPY
FURTHER SITE CHARACTERIZATION AND ANALYSIS
VOLUME 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
ADDENDUM TO THE DATA EVALUATION AND INTERPRETATION REPORT
HUDSON RIVER PCBs REASSESSMENT RI/FS
JULY 1998
fjfej
*1 PRO^"
For
U.S. Environmental Protection Agency
Region II
and
U.S. Army Corps of Engineers
Kansas City District
Volume 2C-A
Book 1 of 2
TAMS Consultants, Inc.
Gradient Corporation
Tetra Tech, Inc.
-------
PHASE 2 REPORT - REVIEW COPY
FURTHER SITE CHARACTERIZATION AND ANALYSIS
VOLUME 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
ADDENDUM TO THE DATA EVALUATION AND INTERPRETATION REPORT
HUDSON RIVER PCBs REASSESSMENT RI/FS
JULY 1998
15&J
For
U.S. Environmental Protection Agency
Region II
and
U.S. Army Corps of Engineers
Kansas City District
Volume 2C-A
Book I of 2
TAMS Consultants, Inc.
(iradient Corporation
Tktra Tkch, Inc.
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
REGION 2
290 BROADWAY
NEW YORK, NY 10007-1866
JUL 23 I998
To All Interested Parties:
The U.S. Environmental Protection Agency (EPA) is pleased to release the Low Resolution
Coring Report for the Hudson River PCBs Superfund site. This report presents findings based
on the analysis of the low resolution sediment core data collected by EPA in 1994 for the
Reassessment. As part of this report, the 1994 sediment data is compared to sediment data
previously collected by the New York State Department of Environmental Conservation in 1984
and 1976 - 1978.
The Low Resolution Coring Report, Volume 2C-A, is an addendum to the Data Evaluation and
Interpretation Report, Volume 2C, which was released in February 1997. Combined, these
volumes are considered the third report in the series of six reports that will make up the Phase 2
Report of the Reassessment. The Phase 2 Report was divided into sections at the request of
members of the community interaction program, thereby allowing interested parties to comment
on the reports prior to the incorporation of the work into the risk assessments and the feasibility
study. It also separates the Phase 2 Report into more manageable individual documents.
As with the previous Phase 2 Reports, it is important to recognize that the conclusions in this
report, although significant, do not indicate whether or not remedial action is necessary for the
PCB-contaminated sediments of the upper Hudson. The numerical analysis (computer modeling)
of fate and transport of PCBs, the associated ecological and human health risk assessments, and a
feasibility study must be completed before any such conclusion can be reached.
EPA will accept comments on the Low Resolution Coring Report until Monday, August 31,
1998. Comments should be marked with the name of the report and should include the report
section and page number for each comment. Comments should be sent to:
Douglas Tomchuk
USEPA - Region 2
290 Broadway - 20th Floor
New York, NY 10007-1866
Attn: LRC Comments
EPA will present the findings of this report as well as the Scope of Work for the Human Health
Risk Assessment at a joint liaison group meeting in Albany, New York. Notification of this
meeting was sent to liaison group members several weeks prior to the meeting. In the interim,
between the release of this report and the end of the comment period, EPA will hold two public
availability sessions to further answer public questions regarding the Low Resolution Coring
Report and the Scope of Work for the Human Health Risk Assessment. These sessions will be
Recycl«cyR«cyclabla • Printed with Vegetable Otf Based Inks on Recycled Paper (40% Poslconsumer)
-------
2
held on Wednesday, August 19, 1998 at the Holiday Inn Express in Latham, New York from
2:30 to 4:30 p.m. and from 6:30 to 8:30 p.m., and on Thursday, August 20, 1998 from 6:30 to
8:30 p.m. at Marist College in Poughkeepsie, New York.
If you need additional information regarding this report, the availability sessions or with respect
to the Reassessment in general, please contact Ann Rychlenski, the Community Relation
Coordinator for this site, at (212) 637-3672.
Sincerely yours,
Richard L. Caspe, Director
Emergency and Remedial Response Division
-------
TABLE OF CONTENTS
€
&
8
©
U
<*4
0
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 1 of 2) Page
TABLE OF CONTENTS i
LIST OF TABLES iv
LIST OF FIGURES v
LIST OF PLATES viii
EXECUTIVE SUMMARY ES-1
ACRONYMS Acro-1
GLOSSARY G-l
1. INTRODUCTION 1-1
1.1 Purpose of Report 1-1
1.2 Report Format and Organization 1-2
1.3 Project Background 1-3
1.3.1 Site Description 1-3
1.3.2 Site History 1-3
1.4 Background for the Low Resolution Sediment Coring Program 1-5
1.5 Low Resolution Sediment Coring Program Objectives 1-6
2. SAMPLING DESIGN AND METHODS 2-1
2.1 Technical Approach for the Low Resolution Sediment Coring Program .. 2-1
2.2 Field Sampling 2-4
2.2.1 Sample Locations 2-5
2.2.2 Sample Preparation 2-6
2.3 Sample Analyses 2-8
2.3.1 PCB Congener Analysis 2-10
2.3.2 Radionuclide Analysis 2-12
2.3.3 Total Organic Carbon and Total Kjeldahl Nitrogen 2-13
2.3.4 Physical Properties 2-13
2.4 Summary of Analytical Results 2-15
2.4.1 PCB Congener Analysis 2-15
2.4.2 Radionuclide Analysis 2-19
2.4.3 Total Organic Carbon and Total Kjeldahl Nitrogen 2-22
2.4.4 Physical Properties 2-23
i TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 1 of 2) Page
3. INTERPRETATION OF LOW RESOLUTION SEDIMENT CORING RESULTS 3-1
3.1 Comparison between the PCB Results for the Low Resolution Cores and
the High Resolution Cores 3-1
3-2 Interpretation of the Relationships Among the Low Resolution Core
Parameters 3-13
3-3 Interpretation of the Low Resolution Core and the Side-Scan Sonar Results 3-27
3-4 Summary of Chapter 3 3-35
4. AN EXAMINATION OF HUDSON RIVER SEDIMENT PCB INVENTORIES: PAST
AND PRESENT
4-1 Sediment Inventories of the Thompson Island Pool 4-1
4.1.1 A Comparison of 1984 and 1994 Conditions 4-7
4.1.2 Assessment of Sediment Inventory Change Based on the Original 1984
£Tri+ Sediment Inventory 4-13
4.1.3 Assessment of Other Potentially Important Characteristics 4-18
4.1.4 Implications of the Inventory Assessment 4-18
4-2 Sediment Inventories of the Upper Hudson Below the Thompson Island
Dam 4-19
4.2.1 Calculation of the Length-Weighted Average Concentration (LWA) and
Mass Per Unit Area (MPA) for Sediment Samples Below the TI
Dam 4-20
4.2.2 Comparison of 1976-1978 Sediment Classifications and the
Side-Scan Sonar Interpretation 4-23
4.2.3 Comparison of Sediment PCB Inventories: NYSDEC 1976-1978 Estimates
versus 1994 Low Resolution Core Estimates 4-26
4.2.4 7Be in Surface Sediments 4-38
4.2.5 Hot Spot Boundaries 4-39
4.2.6 Comparison of the 1994 Hot Spot Inventories with Other 1977
Estimates 4-40
4.3 Sediment Contamination in the Near-Shore Environment 4-42
4.4 Summary and Conclusions 4-44
4.4.1 Sediment and PCB Inventories in the TI Pool 4-44
4.4.2 Sediment and PCB Inventories Below the TI Dam 4-45
ii TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 1 of 2) Page
4.4.3 Sediment Contamination in the Near-Shore Environment 4-49
4.4.4 Summary 4-50
References
APPENDICES
APPENDIX
A
APPENDIX
B
APPENDIX
C
APPENDIX
D
APPENDIX
E
APPENDIX
F
R-l
Data Usability Report for PCB Congeners Low Resolution Sediment
Coring Study
Data Usability Report for Non-PCB Chemical and Physical Data Low
Resolution Sediment Coring Study
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the
Thompson Island Pool
1994 Low Resolution Core Profiles Below the Thompson Island Pool
Memoranda from John Butcher of TetraTech Inc. Concerning Historical
PCB Quantitation
Statistical Summary Sheets for Chapter 4
iii
TAMC
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 2 of 2)
LIST OF TABLES
2-1 Summary of Low Resolution Sediment Core Collection Program
2-2 Sediment Core Segment Summary
2-3 Summary of Low Resolution Sediment Core Analytical Results
2-4 Comparison of Sediment Types for Complete and Incomplete Low Resolution Cores
3-1 Parameters Obtained for the Low Resolution Sediment Coring Program
3-2 Summary Statistics for Total PCBs, AMW, and MDPR
3-3 Regression Coefficients (r) for Correlations Among Laser Grain-Size Distribution
Parameters
3-4 Regression Coefficients (r) for Correlations Among Total PCBs, AMW, MDPR, and Laser
Grain-Size Distribution Parameters
3-5 Regression Coefficients (r) for Correlations Among Total PCBs, AMW, MDPR, and
ASTM Grain-Size Distribution Parameters
3-6 Regression Coefficients (r) for Correlations Among Total PCBs, AMW, MDPR,
Chemical, and Radionuclide Parameters
3-7 Regression Coefficients (r) for Correlations Among Total PCBs, AMW, MDPR, and Bulk
Sediment Properties for all Sediments and Shallow Sediments
3-8 Regression Coefficients (r) for Correlations Among Length-weighted average Total PCB,
Total PCB Mass/Unit Area and Several Important Ancillary Parameters
4-1 Assessment of Core Profiles in the TI Pool
4-2 Summary Data for Hot Spots Surveyed by the Low Resolution Coring Program
4-3 Assignment Classifications for 1976-1978 Samples for Solid Specific Weight Based on the
Low Resolution Coring Results
4-4 Assignment of Grain-Size Distribution Bins for Determination of Principal Fraction for
1977 NYSDEC Samples
4-5 Assignment of Principal Sediment Fraction Based on 1977 NYSDEC Visual Sediment
Classifications
4-6 Shapiro-Wilk Statistics for 1976-1978 and 1994 Hudson River Sediment Samples Below
the TI Dam
4-7 Estimates of Mean Values for PCB Mass per Unit Area and Length-Weighted Average for
Sediments Below the TI Dam
iv
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 2 of 2)
4-8 Comparison of MPI (1992) and Low Resolution Inventory Estimates for Dredge Locations
4-9 Characterization of the 1976-1978 and 1994 Sediment Sample Types
4-10 Assessment of the Studied Hot Spot Areas Below the TI Dam
4-11 Comparison of Historical and 1994 PCB Inventories for Hot Spots Below the TI Dam
4-12 Summary of 1994 Hot Spot Inventories Below the TI Dam
4-13 Estimates of PCB Concentration in Shallow, Near-Shore Sediments
LIST OF FIGURES
2-1 Distance Between 1984 and 1994 Sediment Sample Locations
2-2 Distribution of Core Segments Depths
2-3 Low Resolution Sediment Core Preparation
2-4 Distribution of Total PCB Concentrations in Low Resolution Sediment Core Samples
2-5 Example Regressions for Low Resolution Sediment Core Field Split Pairs
2-6 Precision in Total PCB Concentration for Low Resolution Core Field Splits
2-7 High Resolution Sediment Core Profiles in the Upper Hudson: Examples of the
Coincidence of 137Cs and PCBs Over Time
2-8 Classification of Shallow Sediment Samples
Comparison of Visual Inspection and Laser Grain-Size Analytical Technique
2-9 Classification of Sediment Samples
Comparison of Visual Inspection and ASTM Grain-Size Analytical Techniques
2-10 Classification of Sediment Samples
Comparison of Grain-Size Analytical Techniques (ASTM and Laser Methods)
3-1 Molar Dechlorination Product Ratio vs Fractional Difference In Mean Molecular Weight
Relative to Aroclor 1242 for All Low Resolution Sediment Core Results
3-2 Total PCB Concentration vs Molar Dechlorination Product Ratio and Fractional Difference
In Mean Molecular Weight Relative to Aroclor 1242
3-3 Total PCB Concentration vs MDPR and AMW Showing Cores With and Without 137Cs
Present
3-4 Congener Pattern Comparison Between Upper and Lower Segments on Potentially Cross-
Contaminated Cores
3-5 Congener Pattern Comparison Between Upper and Lower Segments on Cores without
Cross-Contamination
v
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 2 of 2)
3-6 Comparison of Low Resolution Core and High Resolution Core Subsampling Processes
3-7 Sample Points Excluded as a Result of the Selection Criteria
3-8 Examination of the Relationship of MDPR and AMW to Total PCBs for Selected Low
Resolution Sediment Core Results
3-9 Comparison of Low Resolution Core and High Resolution Core Regressions for MDPR
and AMW vs Total PCBs
3-10 Comparison of the Low Resolution Core and High Resolution Core Slicing Techniques on
Measured Sample Values for High Resolution Core 19
3-11 Comparison of the Low Resolution Core and High Resolution Core Slicing Techniques on
Measured Sample Values for High Resolution Core 21
3-12 Comparison of Calculated Results for High Resolution Cores with the Low Resolution
Core Regression Lines for AMW and MDPR vs Total PCBs
3-13 Total PCBs Grouped by Bulk Density
3-14 Total PCBs Grouped by Percent Solids
3-15 Total PCBs Grouped by Solid Specific Weight
3-16 Total PCBs Grouped by Particle Density
3-17a AMW and MDPR Grouped by Bulk Density for All Sediment Segments
3-17b AMW and MDPR Grouped by Bulk Density for Shallow Sediment Segments
3-18 Total PCBs Grouped by Geologist's Classification
3-19 Total PCBs Grouped by Silt Fraction in Shallow Sediments
3-20 Total PCBs Grouped by Median <|) (Phi) in Shallow Sediments
3-21 Total PCBs Grouped by Total Organic Carbon
3-22 Total PCB Concentration and Mass per Unit Area Grouped by 7Be
3-23 Total PCBs Grouped by 137Cesium for Shallow Sediments
3-24 AMW and MDPR Grouped by 137Cs in Shallow Sediments
3-25 Comparison of the Mean DN Value for 10-ft and 50-ft Circles
3-26 Three Dimensional Correlation Plot of Digital Number vs Grain-Size Distribution
Parameters: Comparison Between Confirmatory and Low Resolution Core Samples
3-27 Classification of Sediment Samples
Comparison of Visual and Analytical Techniques to the Interpretation of the Side-Scan
Sonar Images
3-28 Acoustic Signal Mean (DN50) Based on 50-ft Circles Grouped by Laser Analysis Principal
Fraction
vi
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 2 of 2)
3-29 Comparison of the Regression Lines for the Confirmatory and Low Resolution Core
Results against the DN50 for the 500 kHz Side-Scan Sonar Images
3-30 Comparison of 500 kHz Acoustic Signal (DN50) and Low Resolution Core PCB Levels
in Shallow Sediments
3-31 Comparison of 500 kHz Acoustic Signal (DN50) and Low Resolution Core PCB
Mass/Area
4-1 Typical Low Resolution Core Profiles for the TI Pool and Their Classification
4-2 High Resolution Core 19 from the TI Pool
4-3 Core Locations Exhibiting Sediment Scour
4-4 Comparison Between 1984 and 1994 MPA for Total PCBs Showing Core Classifications
4-5 Relationship Between 1984 and 1994 Sediment Inventories (MPA) for Total PCBs and
Trichloro and Higher Homologues
4-6 Relationship Between the 1984 £Tri+ Mass Per Unit Area (MPA3+) and the Change in
Sediment PCB Inventory for the TI Pool
4-7 1984 Trichloro and Higher Homologues as MPA vs Mass Difference and Mole Difference
Relative to 1994 - Log Scale
4-8 Determination of the Molecular Weight of the Trichloro and Higher Homologues (]TTri + )
at the Time of Deposition
4-9 Distribution of Mass Difference (g/m2) and Mole Difference (mole/m2) between 1984 and
1994
4-10 Distribution of the Percent Change in PCB Molar Inventory (DeltaM)
4-11 Change in (Moles/m2) by 1984 £Tri+ PCB Inventory
4-12 Change in Mass per Unit Area (MPA) by 1984 £Tri+ PCB Inventory
4-13 Percent Change in PCB Molar Inventory (DeltaM) by 1984 £Tri + pcB Inventory
4-14 Percent Mass Change (Delta p^) by 1984 £Tri+ PCB Inventory
4-15 Statistical Analysis of DeltaM as a Function of 1984 Sediment £Tri+ Inventory and
NYSDEC Sample Type
4-16 Implications of the Inventory Change Analysis for the 1984 TI Pool Inventory
4-17 Relationship Between Total PCB Concentration and Solid Specific Weight for Low
Resolution Core Samples
4-18 Comparison of 1977-1978 Sediment Classifications and Interpretationof the Side-Scan
Sonar Images
vii
T4MS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 2 of 2)
4-19 Distribution of Length-Weighted Core Averages in 1976-1978 NYSDEC Survey and Low
Resolution Sediment Core Samples
4-20 Distribution of MPA in 1976-1978 NYSDEC Survey and Low Resolution Sediment Core
Samples
4-21 Comparison of Geometric Mean PCB MPA and Length-Weighted Core Averages from the
1976-1978 NYSDEC and Low Resolution Core Surveys in Dredge Locations
4-22 Comparison of Geometric Mean PCB MPA and Length-Weighted Core Averages from the
1976-1978 NYSDEC and Low Resolution Core Surveys in Hot Spots
4-23 Relationhips Between 0-4", 0-12" and Entire Core PCB Concentrations
4-24 Core Profiles in Areas of Continuous Deposition
4-25 Typical 1994 Sediment Core Profiles from Hot Spot 28
4-26 Typical 1994 Sediment Core Profiles from Hot Spots 25 and 35
LIST OF PLATES
1-1 Hudson River Drainage Basin and Site Location Map
2-1 Low Resolution Sediment Coring Locations in the Upper Hudson River
3-1 Key to Location of Plates 3-2 Through 3-20
3-2 Determination of the DN50 Values for the Low Resolution Coring Locations in Cluster
14
3-3 Determination of the DN50 Values for the Low Resolution Coring Locations in Cluster
13
3-4 Determination of the DN50 Values for the Low Resolution Coring Locations in Clusters
12, 15, and 17
3-5 Determination of the DN50 Values for the Low Resolution Coring Locations in Clusters
10 and 11
3-6 Determination of the DN50 Values for the Low Resolution Coring Locations in Clusters
8 and 9
3-7 Determination of the DN50 Values for the Low Resolution Coring Locations in Clusters
6 and 7
viii
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 2 of 2)
3-8 Determination of the DN50 Values for the Low Resolution Coring Locations in Cluster
19
3-9 Determination of the DN50 Values for the Low Resolution Coring Locations in Clusters
4, 5, and 18
3-10 Determination of the DN50 Values for the Low Resolution Coring Locations in Cluster
3
3-11 Determination of the DN50 Values for the Low Resolution Coring Locations in Cluster
1 and 2
3-12 Determination of the DN50 Values for the Low Resolution Coring Locations in Hot Spot
25
3-13 Determination of the DN50 Values for the Low Resolution Coring Locations in Cluster
25
3-14 Determination of the DN50 Values for the Low Resolution Coring Locations in Hot Spot
28
3-15 Determination of the DN50 Values for the Low Resolution Coring Locations in Hot Spot
28
3-16 Determination
31
3-17 Determination
31
3-18 Determination
34
3-19 Determination
34 and 35
3-20 Determination
34 and 35
4-1 Key to Locations of Plates 4-2 Through 4-9
4-2 Comparison Between 1984 and 1994 Coring Results in Thompson Island Pool
to
4-9
4-10 Key to Locations of Plates 4-11 Through 4-19
of the DN50
Values
for
the
Low Resolution Coring Locations in Hot Spot
of the DN50
Values
for
the
Low Resolution Coring Locations in Hot Spot
of the DN50
Values
for
the
Low Resolution Coring Locations in Hot Spot
of the DN50
Values
for
the
Low Resolution Coring Locations in Hot Spots
of the DN50
Values
for
the
Low Resolution Coring Locations in Hot Spots
ix
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
CONTENTS
Volume 2C-A (Book 2 of 2)
4-11 Changes in Inventory in the Thompson Island Pool 1984 vj 1994
to
4-19
4-20 Low Resolution Coring and 1984 NYSDEC Sampling Results in the Thompson Island
Pool: Change in Sediment Inventory for Trichloro to Decachlorohomologues
4-21 Low Resolution Coring and 1976-78 NYSDEC Sampling Results Near Hot Spot 25
4-22 Low Resolution Coring and 1976-78 NYSDEC Sampling Results Near Hot Spot 28
4-23 Low Resolution Coring and 1976-78 NYSDEC Sampling Results Near Hot Spot 31
4-24 Low Resolution Coring and 1976-78 NYSDEC Sampling Results Near Hot Spots 34 and
35
4-25 Low Resolution Coring and 1976-78 NYSDEC Sampling Results Near Hot Spot 37
4-26 Low Resolution Coring and 1976-78 NYSDEC Sampling Results Near Hot Spot 39
4-27 Low Resolution Coring and 1976-78 NYSDEC Sampling Results In TAMS' Location 41
and 42
4-28 Low Resolution Coring and 1976-78 NYSDEC Sampling Results In TAMS' Location 43
and 44
4-29 Key to Locations of Plates 4-21 Through 4-28
x
TAMS
-------
TAMS
-------
Low Resolution Sediment Coring Report
Executive Summary
July 1998
This report presents the findings from the analysis of data relating to the low resolution sediment
coring program for the Hudson River PCBs Site Reassessment Study. The low resolution sediment
coring program was designed to evaluate changes in sediment PCB inventory over time and the
degree of burial of PCB-contaminated sediments.
Low resolution coring refers to the relative thickness of the sediment slices analyzed during the
sampling program. In the low resolution coring program, the average thickness of a sediment slice
was 9 inches (22 cm), compared to the 0.8-inch (2 cm) to 1.6-inch (4 cm) slices analyzed in the high
resolution coring program.
Background - The Low Resolution Sediment Coring Report is a companion to the Data Evaluation
and Interpretation Report, which was issued in February 1997. Similar to the Data Evaluation and
Interpretation Report, the Low Resolution Sediment Coring Report is based on geochemical analysis.
However, whereas the Data Evaluation and Interpretation Report primarily evaluated the transport
of PCBs through the analysis of water-column data and dated sediment cores, the Low Resolution
Sediment Coring Report assesses the inventory of PCBs found in the Upper Hudson River
sediments. The geochemical analysis in these reports will be complemented and verified to the
extent possible by additional numerical analysis (i.e., computer modeling). Results of the computer
modeling will be reported in the Baseline Modeling Report, to be released in May 1999. In addition,
the Low Resolution Sediment Coring Report does not explore the biological uptake and human
health impacts, which will be evaluated in future Reassessment documents.
The Reassessment Remedial Investigation and Feasibility Study is being conducted by the U.S.
Environmental Protection Agency in order to determine an appropriate course of action to address
the PCB-contaminated sediments in the Upper Hudson River. This study is a reassessment of the
Agency's interim "no-action" decision made in 1984. During the first phase of the Reassessment,
EPA compiled existing data on the site and conducted preliminary analyses of the data. As part of
the second phase, EPA conducted field investigations to characterize the nature and extent of PCB
contamination in the Upper Hudson. The Phase 2 data, along with data from other sources, has been
used to better understand the fate and transport of PCB contamination in the river.
Two large-scale sediment investigations were previously conducted by the New York State
Department of Environmental Conservation (NYSDEC) to characterize the extent and magnitude
of PCB contamination in the sediments—one from 1976 to 1978, and one in 1984. On the basis of
data gained from these investigations, approximately forty zones of highly contaminated sediments,
designated as hot spots, were identified. These data were used to estimate total PCB inventory in
Hudson River sediments at the time of the completion of both the 1976-1978 and 1984 studies.
Objectives - The low resolution sediment coring program, conducted in July and August 1994, had
two main objectives:
ES-I
TAMS
-------
• Obtain new estimates of sediment PCB inventories at selected locations in the Thompson
Island Pool to compare against the existing PCB sediment database constructed from the
1984 NYSDEC survey; and
• Refine the PCB mass estimates for a limited number of historic hot spot locations defined
by the 1976-1978 NYSDEC survey in the Upper Hudson below the Thompson Island Dam.
Low resolution sediment coring was performed to examine PCB contamination in a limited
number of areas and to augment and improve estimates of the sediment inventory and spatial
distribution of PCBs previously developed for these areas. The comparison of previous surveys to
current conditions is significant to the Hudson River PCBs Reassessment because changes in PCB
inventories can indicate areas of the river where PCBs are being lost and/or gained.
In the Thompson Island Pool, locations for most of the cores were selected by grouping the samples
into zones with minimal local sediment heterogeneity, thereby minimizing uncertainties due to
factors other than PCB inventory changes. It was then possible to compare the 1994 data to the 1984
NYSDEC data and draw general conclusions regarding PCB inventory change in the Thompson
Island Pool. The coring locations below the Thompson Island Dam were selected to generate PCB
mass estimates for areas which, based on previous estimates, represented approximately 75 percent
of the hot spot inventory below the dam. While these areas should be indicative of areas with similar
levels of contamination (i.e., hot spots), the results cannot be extrapolated to inventory changes in
other areas below the Thompson Island Dam. Other locations sampled were selected to characterize
near-shore sediments.
Analytical Tools - Low resolution sediment core samples were analyzed on a congener-specific
basis for PCBs, which allows for the same type of analysis of dechlorination products as was
conducted for the Data Evaluation and Interpretation Report. In addition, radionuclide analyses of
the isotopes cesium-137 and beryllium-7 in sediments were conducted. Beryllium-7 data were used
to determine whether the top of the core showed evidence of recent deposition. The cesium-137 data
were used to establish that the core had penetrated all post-1954 sediment, which correlates to the
PCB-containing sediments. The radionuclide data allowed the low resolution cores to be assessed
for completeness, because the absence of cesium-137 in the bottom of the core was a reliable
indicator that the core did not miss any PCBs at depth (i.e., the core was complete). Because the
1994 inventories are from cores that were determined to be complete, and because the previous
inventories did not utilize radionuclide analyses and therefore may have been underestimated, the
observed losses in PCB inventory are minimum estimates. Likewise, the observed gains in PCB
inventories in certain areas are maximum estimates.
Major Findings - The analyses presented in the Low Resolution Sediment Coring Report lead
to four major findings as follows:
1. There was little evidence found of widespread burial of PCB-contaminated sediment by clean
sediment in the Thompson Island Pool. Burial is seen at some locations, but more core sites showed
loss of PCB inventory than showed PCB gain or burial.
2. From 1984 to 1994, there has been a net loss of approximately 40 percent of the PCB
inventory from highly contaminated sediments in the Thompson Island Pool.
ES-2
TAMS
-------
3. From 1976-1978 to 1994, between the Thompson Island Dam and the Federal Dam at Troy,
there has been a net loss of PCB inventory' in hot spot sediments sampled in the low resolution
coring program.
4. The PCB inventory for Hot Spot 28 calculated from the low resolution coring data is
considerably greater than previous estimates. This apparent "gain" in inventory is attributed to
significant underestimates in previous studies rather than actual deposition of PCBs in Hot Spot 28
These conclusions are briefly described and explained below.
1. There was little evidence found of widespread burial of PCB-contaminated sediment
by clean sediment in the Thompson Island Pool. Burial is seen at some locations, but more
core sites showed loss of PCB inventory than showed PCB gain or burial. Thirty percent of
coring sites do not exhibit burial, or may exhibit erosion, based on the absence of beryllium-7 in core
tops. Comparisons of sediment core profiles between the 1984 and 1994 data indicate that burial is
not occurring at more than half of the locations investigated. Burial does occur at some hot spot
areas, but there is also evidence of sediment PCB loss occurring, often within the same hot spots.
Again, there is more evidence for sediment PCB loss rather than burial.
Beryllium-7 is a naturally-occurring isotope whose presence in sediments indicates recent deposition
or interaction with surface waters within the six months prior to sample collection. The absence of
beryllium-7 was shown to be a statistically significant indicator of inventory loss. Absence of
beryllium-7 is attributed to a core collected in a non-depositional area or an area that has undergone
scour (erosion) of river sediment. Thus, this radionuclide was used to test a core top (0 to 1 -inch) for
the presence of recently deposited sediment. Surficial sediments in which beryllium-7 was not
detected (no burial) had lower PCB inventories than cores in which beryllium-7 was detected,
indicating that burial of PCB mass by less contaminated sediments is not occurring at these
locations. Although this analysis does not offer proof of sediment scour, it does show that burial of
contaminated sediments is not occurring in at least 30 percent of the coring sites.
The core profiles, or core results presented by depth, show an important finding. PCB maxima are
principally found in the top-most core layer in approximately 60 percent of the samples, which
represent shallow sediment (median core segment depth of 9 inches). These results indicate that
burial of PCB-bearing sediments is not occurring on an extensive basis and that high concentrations
of PCBs remain relatively close to the sediment/water interface. In addition, in areas where burial
does occur, the newly deposited sediments commonly contain PCBs.
In addition, the average depth to the maximum total-PCB concentration (taken as the bottom of the
core section in which the PCB maximum was found) varied considerably according to whether the
area showed a gain or loss of PCB inventory. In the hot spots, for cores exhibiting a PCB inventory
increase, the average depth to the maximum total-PCB concentration was 18.7- inches (46.8-cm),
contrasted to 10.6 inches (26.5-cm) in the cores exhibiting a loss of PCB inventory. The difference
in mean depth between areas of PCB loss and gain is statistically significant. This finding confirms
that the PCB-maximum moves downward in areas of PCB (and accordingly, sediment) gain, and
does not exhibit such burial in areas where PCB loss is occurring.
ES-3
TAMS
-------
2. From 1984 to 1994, there has been a net loss of approximately 40 percent of the PCB
inventory from highly contaminated sediments in the Thompson Island Pool. Sediments in the
Thompson Island Pool with total PCB inventories of greater than 10 g/m2 (typical of hot spot
sediments) exhibit a statistically significant loss of PCBs. This inventory loss includes loss to the
overlying water column as well as dechlorination. Specifically, there has been roughly a 30 percent
decline in inventory due to actual loss from the sediments (from erosion, diffusion, groundwater
advection, or other mechanisms) and a 10 percent loss via dechlorination. When the 30 percent loss
is combined with an average dechlorination loss of approximately 10 percent, the result is a total
PCB inventory loss of approximately 40 percent.
The PCBs lost from hot spot areas enter the water column and may be available to the food chain
or deposited in other areas.
3. From 1976-1978 to 1994, between the Thompson Island Dam and the Federal Dam at
Troy, there has been a net loss of PCB inventory in hot spot sediments sampled in the low
resolution coring program. When the 1994 total PCB inventory is compared to the 1976-1978
inventory, a statistically significant loss of 50 to 80 percent of PCB inventory is seen for Hot Spots
31, 34 and 37. This represents a potential loss of approximately 3 metric tons into the water column,
although some loss may be due to dechlorination. The other hot spots evaluated either appear
unchanged or have not experienced significant gains in PCB inventory, with the exception of Hot
Spot 28 , as noted in Major Finding 4, below.
Hot Spot 39 exhibits burial. Total PCB concentrations are at greater depths than previous sediment
surveys. Because of the inability to obtain "complete" cores in this hot spot, there is uncertainty in
our current estimate as well as the previous estimates, making it difficult to determine whether there
is inventory loss or gain. Given this uncertainty, Hot Spot 39 is considered to not have experienced
a significant change in inventory.
The sediment inventories of three other areas appear unchanged (Hot Spots 25, 35 and dredge
location 182), but only one (Hot Spot 25) had a sufficient number of samples to confirm the lack of
change.
It should be noted that the calculations for Hot Spots 28 and 39 show large PCB inventories of 20
and 4 metric tons, respectively. This is greater than the inventory of the entire Thompson Island
Pool, which was estimated to be between 14.5 and 19.6 metric tons in the Data Evaluation and
Interpretation Report, based on the 1984 NYSDEC data.
Overall, hot spot sediments below the Thompson Island Dam exhibit both losses and gains based
on 1976-1978 and 1994 inventory estimates. Losses total a minimum of 3 metric tons. Apparent
gains in certain hot spots were likely due to previously inaccurate estimates.
4. The PCB inventory for Hot Spot 28 calculated from the low resolution coring data is
considerably greater than previous estimates. This apparent "gain" in inventory is attributed
to significant underestimates in previous studies rather than actual deposition of PCBs in Hot
Spot 28. An evaluation of the 1994 data collected for the low resolution coring program found that
the PCB inventory in Hot Spot 28 was substantially greater than had been estimated in previous
studies. The Low Resolution Sediment Coring Report estimates the mass of PCBs in Hot Spot 28
ES-4
TAMS
-------
to be 20 metric tons. Previous estimates varied between two to seven metric tons. Therefore, based
on a comparison of these estimates there would appear to be a large gain in PCB inventory.
However, further examination of the core profiles for Hot Spot 28 shows that less than 50 percent
of the sample locations have evidence of deposition (burial). The remaining sites are either
unchanged or have undergone scour based on the presence of the maximum total-PCB concentration
in the shallow sediment. The deposition history recorded by the nearby high resolution cores
indicates that this type of profile can only be caused by scour. Only between two to five percent of
PCB mass was deposited between 1977 and 1991 for two nearby high resolution cores, thus making
such a large gain in inventory unlikely. Therefore, the apparent "gain" of PCBs in Hot Spot 28 based
on a comparison of historical estimates to the current estimate is not real. There have been losses
of PCBs from several locations within the hot spot but, overall, the evidence suggests no significant
change in inventory in the hot spot. The previous mischaracterization of the inventory probably
results from an initial inaccurate assessment of Hot Spot 28 by the 1976-1978 sediment survey
caused by too many shallow cores and grabs (i.e., "incomplete" cores). EPA's current estimate is
based on cores that have been found to be "complete" based on radionuclide analysis.
Additional Findings
The interpretation of the low resolution coring data is consistent with the findings of the Data
Evaluation and Interpretation Report. The analysis in the Low Resolution Sediment Coring Report
supports the conclusions from the Data Evaluation and Interpretation Report that the extent of
dechlorination is proportional to sediment concentration, and that the water-column PCB load
originates primarily from the sediments of the Upper Hudson River.
There has been a net gain of PCB inventory in areas of the Thompson Island Pool outside of the hot
spots. The additional PCBs may have come from redistribution of PCBs from high concentration
areas or from PCBs that entered the pool from upstream sources, such as the GE Hudson Falls Plant
site. Sediments with total PCB inventories of less than 10 g/m2 exhibit a statistically significant gain
of PCBs. However, this net gain is found at a limited number of locations, insufficient to support
such a finding for all non-cohesive (coarse-grained) sediments in the Thompson Island Pool. The
increase of PCB inventory in the Thompson Island Pool outside of the hot spots has an upper bound
of approximately 100 percent, although the actual gain is probably much less.
Comparison of the river bottom texture type indicated by the side-scan sonar images with the 1976-
1978 NYSDEC sediment survey grain-size data resulted in good agreement with one another. This
indicates that side-scan sonar can be used to classify large areas of the river bottom in terms of
sediment properties and that the river bottom depositional types remained constant. The comparison
between side-scan sonar results and PCB levels in shallow sediments implies that side scan sonar
images can be used to estimate both shallow sediment PCB concentrations and PCB inventories.
Hot spot boundaries appeared accurate, although in some instances hot spot areas needed to be
increased to include all nearby areas of high contamination.
Historical estimates of PCB mass in hot spots below the Thompson Island Dam assumed a solid
specific weight of 1 g/cc. Based on the low resolution core relationship between solid specific weight
and total PCB concentration, more appropriate values of solid specific weight ranged from 0.5 to
0.79 g/cc for the majority of the 1976-1978 hot spot sample locations. Applying a solid specific
weight based on length-weighted average concentrations yielded about a 20 to 30 percent decrease
ES-5
TAMS
-------
in the original PCB inventory estimates. In other words, the previous calculations of PCB
inventories were somewhat overestimated.
Sediments in the near shore environment, which was defined as within 50 feet of the shoreline, had
higher PCB concentrations than estimated in the Phase 1 Report. The Phase 1 Report estimated an
exposure point concentration of 66 mg/kg (parts per million) for the 95 percent confidence interval
of the arithmetic mean of the shallow sediment concentration based on the 1984 data, whereas the
current estimate would be within the range of 135 to 264 mg/kg. Implications of this change will
be addressed in the swimming or wading exposure scenario in the Human Health Risk Assessment.
Summation
The decrease in PCB inventories in the more contaminated sediments of the Thompson Island Pool
and from several of the studied hot spots below the Thompson Island Dam, along with the indication
of an inventory gain in the coarse sediments of the Thompson Island Pool, indicate that PCBs are
being redistributed within the Hudson River system. These results show that the stability of the
sediment deposits cannot be assured.
Burial of contaminated sediment by cleaner material is not occurring universally. Burial of more
PCB-contaminated sediment by less contaminated sediment has occurred at limited locations, while
significant portions of the PCB inventories at other hot spots have been re-released to the
environment. It is likely that PCBs will continue to be released from Upper Hudson River
sediments.
ES-6
TAMS
-------
PHASE 2 REPORT - REVIEW COPY
FURTHER SITE CHARACTERIZATION AND ANALYSIS
VOLUME 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
ADDENDUM TO THE DATA EVALUATION AND INTERPRETATION REPORT
HUDSON RIVER PCBs REASSESSMENT RI/FS
ACRONYMS
ASTM American Society for Testing and
Materials
'Be Beryllium-7
cm Centimeter
,,TCs Cesium-137
DEIR Data Evaluation and Interpretation
Report
DN Digital Number
ECD Electron Capture Detector
GC Gas Chromatograph
GE General Electric
1QD Interquartile Distance
ITD Ion Trap Detector
kg Kilogram
kHz Kilohertz
LQ Lower Quartile
LWA Length-Weighted Average
IMDPR Molar Dechlorination Product Ratio
MPA Mass Per Unit Area
MPI Malcolm Pimie, Inc.
MW Molecular Weight
NPDES National Pollution Discharge Elimination
System
NYSDEC New York State Department of
Environmental Conservation
PCB Polychlorinated Biphenyl
PPB Parts per Billion
PPM Parts per Million
QA Quality Assurance
QAPjP Quality Assurance Project Plan
RPD Relative Percent Difference
RRT Relative Retention Time
RSD Relative Standard Deviation
s Standard Deviation (also as SD)
SAP Sampling and Analysis Plan
SAS Special Analytical Services
SOP Standard Operating Procedure
SSW Solid Specific Weight
TC Total Carbon
TCL Target Compound List (Organics)
TI Thompson Island
TKN Total Kjeldahl Nitrogen
TN Total Nitrogen
TOC Total Organic Carbon
UQ Upper Quartile
USGS United States Geological Survey
Aero-1
T» 1UC
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
Arithmetic Mean
The sum of values divided by the number of values.
ASTM
7Be
Box & Whisker Plot
Bulk density
BZ U
C/N
Chow's F test
The sieve and hydrometer-based grain size distribution analysis. This analyis is a
modified version of ASTM Methods D421-85 and D422-63.
Radioisotope Beryllium 7 (pCi/Kg).
A plot which enables the quick examination of a number of variables and extraction
of the major characteristics including the median, upper and lower quartiles,
interquartile distance and outliers.
The mass of water and solids per unit volume of sediment (g/cc).
The PCB congener nomenclature system developed by Ballschmiter & Zell (1980).
Total Carbon to Total Nitrogen Ratio (molar).
Chow's F test (Fisher, 1970) addresses the hypotheses that the parameters
have or have not changed between regressions developed for two data sets.
It is developed by calculating error sum of squares or sum of squared
residuals (SSEs) for regression models on each of the data sets individually
and an SSE for a regression on the pooled data. The comparison is made by
forming an F statistic with k and ((, + t: - 2k) degrees of freedom, formed as
(Kennedy, 1979)
P _ \SSE (constrained) - SSE (unconstrained)}/k
SSE(unconstrained)!(tx + t2 - 2k)
in which
SSE(unconstrained')
SSE(constrained)
'i
the sum of the SSEs from the two separate
regressions,
the SSE from the regression on the pooled data,
the number of observations in the first sample set,
the number of observations in the second
sample set, and
the number of parameters in the model, including
the intercept term.
G-l
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
Clay %
Coarse Sand%
Congener
137Cs
DEIR
Delta,
84
d(x)
DN
Fine Sand%
Fines
Geometric Mean
Geometric Mean
Diameter
The resulting statistic can then be compared to a tabulation of the F distribution
with k and (/,+ t2 - 2k) degrees of freedom to test the hypothesis that parameters
have changed significantly between data sets I and 2.
Percent Clay - ASTM Classification by Laser Analysis (%).
Percent Coarse Sand - ASTM Classification by Laser or Sieve Analysis (%).
The 209 different configurations of a PCB molecule resulting from multiple
combinations of hydrogen, chlorine and position on the byphenol molecule (two
benzene rings linked at a single point).
Radioisotope Cesium 137 (pCi/Kg).
Data Evaluation and Interpretation Report.
The total PCB mass/area in 1994 minus the total PCB mass/area in 1984 divided by
the total PCB mass/area in 1984.
A grain-size distribution measure. The effective diameter of a theoretical sieve that
would retain X percent of the sample. Equally the effective diameter that is larger
than (100 - X) percent of a sample. Values for X are 10, 20, 30, 40, 50, 60, 70, 80,
90, or 99.
Digital acoustic signal values of the 500 kHz sonar images. Each pixel of the side
scan sonar images has a number assigned according to the shade of grey with values
ranging from 0 to 255. The mean value of the assigned values in a region is the
DN.
Percent fine sand - ASTM Classification by Laser or Sieve Analysis (%).
Percent fines by ASTM Analysis - ASTM Classification (%). Corresponds to the
sum of clay and silt fractions by Laser analysis.
The antilog of the sum of the log-tranformed values divided by the number of
values.
Antilog of the arithmetic mean of the logs of the diameters. See also mean phi.
Mean diameter of mass weighted phi value (mm).
G-2
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
Gravel %
Homologue
HSD
Laser
Percent Gravel - ASTM Classification by Laser or Sieve Analysis (%)
A grouping of PCB congeners based on the number of chlorine atoms on the
molecule. A PCB molecule can have from one to ten chlorine atoms. The
homologue groups are mono, di, tri, tetra, penta, hepta, hexa, octa, nona and deca.
Honestly significant difference. (See Tukey-Kramer HSD.)
Method for obtaining sediment grain size distribution. This method provides greater
resolution of fines.
Mahanalobis Distances Mahalanobis distances (SAS, 1994) are a means of determining outliers in a
correlation analysis. The Mahanalobis distance from the multivariate mean
(centroid) depends on estimates of the mean, standard deviation and correlation
for the data.The points with the largest distances are the multivariate outliers. For
each observation number the distance is denoted dt and computed as:
d, = J(a. - a)' S x (a,-a)
where:
a; is the data for the ith row
a is the row of means
S is the estimated covariance matrix for the data
and ( a, - a )' is the transpose of (a, - a ) matrix.
The distances are then compared to a reference distance determined by
VF * "v
where:
ny is the number of variables
and F is a statistic with parameters including 95% confidence level,
nj {n - nv- 1) degrees of freedom, and is centered at zero.
(n=number of observations)
If any of the distances are further from the center than the refernce distance they
are considered outliers. With the jackknife method, is calculated 'without
including the ith observation in the mean, standard deviation or correlation matrix.
In this case, outliers cannot distort the estimates of means or covariance.
G-3
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
MDPR Molar Dechlorination Product Ratio is the ratio of the sum of five congener molar
concentrations (BZ# 1, 4, 8, 10, and 19) to the total sample molar concentration.
MPA The EPCB mass per unit area. The MPA was calculated from the product of the
concentraion of PCBs (mg/kg), length (m) and density (kg/m3).
m
MPA = £ C, • Lt • p,
/*!
MPA,+ The 2Tri + mass per unit area. The MPAJ+ was calculated from the product of the
concentraion of Tri + (mg/kg), length (m) and density (kg/m3).
m
MFA = £ Cm. *L,* Pi
/ - 1 '
AMW
Mean Phi
Median
Median Diameter
Medium Sand%
Non-target
NYSDEC
Particle density
The fractional difference in the mean molecular weight relative to Aroclor
1242.
Mass weighted mean phi value (phi). The inverse log (base 2) of this value
is the geometric mean diameter.
The data value located halfway between the smallest and largest values.
The diameter in a grain size distribution located halfway between the smallest
and largest values.
Percent medium sand ASTM classification by Laser or ASTM Analysis (%).
One of the additional 36 congeners reported in the Phase 2 data set. Reported
values may or may not be calibrated.
New York State Department of Environmental Conservation..
Sediment particle density (g/cc).
G-4
TAN/15*
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
Percent Similarity A means of comparing grain size distributions. The lower value for each of the
grain size parameters is summed. The closer the sum is to 100% the more similar
the distributions are.
Percent Solids Mass of solids per unit mass of wet sediment. Max value = 100%.
Phi -0.5 Mass fraction with a diameter larger than -0.5 phi and smaller than -1.0 phi (%).
Phi -1.0 Mass fraction with a diameter larger than -1.0 phi and smaller than -1.5 phi (%).
Phi -1.5 Mass fraction with a diameter larger than -1.5 phi and smaller than -2 phi (%).
Phi -2.0 Mass fraction with a diameter larger than -2 phi (%).
Phi 0.0 Mass fraction with a diameter larger than 0.0 phi and smaller than -0.5 phi (%).
Phi 0.5 Mass fraction with a diameter larger than 0.5 phi and smaller than 0.0 phi (%).
Phi 1.0 Mass fraction with a diameter larger than 1.0 phi and smaller than 0.5 phi (%).
Phi 1.5 Mass fraction with a diameter larger than 1.5 phi and smaller than 1.0 phi (%).
Phi 2.0 Mass fraction with a diameter larger than 2.0 phi and smaller than 1.5 phi (%).
Phi 2.5 Mass fraction with a diameter larger than 2.5 phi and smaller than 2.0 phi (%).
Phi 3.0 Mass fraction with a diameter larger than 3.0 phi and smaller than 2.5 phi (%).
Phi 3.5 Mass fraction with a diameter larger than 3.5 phi and smaller than 3.0 phi (%).
Phi 4.0 Mass fraction with a diameter larger than 4.0 phi arid smaller than 3.5 phi (%).
Phi 4.5 Mass fraction with a diameter larger than 4.5 phi and smaller than 4.0 phi (%).
Phi 5.0 Mass fraction with a diameter larger than 5.0 phi and smaller than 4.5 phi (%).
Phi 5.5 Mass fraction with a diameter larger than 5.5 phi and smaller than 5.0 phi (%).
Phi 6.0 Mass fraction with a diameter larger than 6.0 phi and smaller than 5.5 phi (%).
G-5 TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
Phi 6.5 Mass fraction with a diameter larger than 6.5 phi and smaller than 6.0 phi (%).
Phi 7.0 Mass fraction with a diameter larger than 7.0 phi and smaller than 6.5 phi (%).
Phi 7.5 Mass fraction with a diameter larger than 7.5 phi and smaller than 7.0 phi (%).
Phi 8.0 Mass fraction with a diameter larger than 8.0 phi and smaller than 7.5 phi (%).
Phi 8.5 Mass fraction with a diameter larger than 8.5 phi and smaller than 8.0 phi (%).
Phi 9.0 Mass fraction with a diameter larger than 9.0 phi and smaller than 8.5 phi (%).
Phi 9.5 Mass fraction with a diameter larger than 9.5 phi and smaller than 9.0 phi (%).
Phi 10.0 Mass fraction with a diameter larger than 10.0 phi and smaller than 9.5 phi (%).
Phi 10.5 Mass fraction with a diameter larger than 10.5 phi and smaller than 10.0 phi (%).
RPD Relative Percent Difference is the absolute value of the difference between values
divided by the average of the values.
Sand % Percent Sand by Sieve Analysis - ASTM Classification (%).
Shallow Sediment Sediment 0-12 inches below the sediment/water interface.
Shapiro and Wilk The W test (Gilbert, 1987) developed by Shapiro and Wilk (1965) is an effective
W Test method for testing whether a data set has been drawn from an underlying normal
distribution. Furthermore, by conductiong the test on logarithms of the data, it is
an equally effective way of evaluating the hypothesis of a lognormal distribution.
The W test compares the range of values to the number and mean of the samples
collected to assess the probability of an underlying normal distribution. The
hypothesis that the underlying distribution is normal is rejected at the 95%
confidence level.
G-6
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
52 ai i]
w = —
£(*, -* )2
where:
= data drawn at random from some population
k = n/2 if n is even
= (n-1 )/2 if n is odd
a, = coefficient for the Shapiro-Wilk test dependent on n
Silt Percent Percent Silt - ASTM Classification by Laser Analysis (%).
Skewness Statistical characterization of the degree of asymmetry of the phi values about the
phi class mean - Laser Analysis (Unitless).
Solids specific weight Sediment solids specific weight (g/cc) i.e., mass of solids per unit volume of wet
sediment.
Sorting
Surficial Sediment
Target Congener
Standard deviation of the phi values - Laser Analysis (Unitless).
Sediment 0-2 inches below the sediment/water interface.
One of the original 90 target congeners chosen for the Phase 2 investigation.
Reported values are based on congener specific calibration and undergo data
validation.
Theil's U Statistic Theil's U statistic, otherwise known as Theil's inequality coefficient (Theil, 1961),
is often used for the evaluation of model simulation error. The U statistic gives a
measure of the consistency between forecasts (e.g., Low Resolution predictions
using the High Resolution model) and the data used to develop the forecasts. It
ranges from 0 to 1, with 0 indicating perfect prediction. The variance of the U
statistic can be approximated (for U less than 0.3) as U2/T, where T is the number
of samples in the "forecast".
The U statistic is defined as (Pindyck and Rubinfeld, 1981)
G-7
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
(K
* 1 = ]
>',V
* r = l
where:
Y' = simulated value for observation t,
Y* = actual value for observation t, and
T = total number of observations.
The numerator of U is simply the root mean square simulation error, but the scaling
of the denominator is such that U always falls between 0 and 1.
The U statistic may also be decomposed into portions attributed to bias or
systematic error (Um), variance or ability of the model to replicate the degree of
variability in the variable of interest (Us), and covariance or unsystematic error (Uc).
These proportions of inequality, which sum to 1, are defined as:
U
Us
(rJ - y")2
(i/r>£ (r/ - r,mf
fc - a.f
(i/r>£ (y; - r,'f
u
c = 2(1 - p)o,o,
(i/7-)E(r/ - y;)2
where:
Y* = the mean of the series Y',
Ya = the mean of the series Y",
a = the standard deviation of the series Y,',
G-8
TAMS
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION
a
a
P
the standard deviation of the series Y', and
the correlation coefficient of the two series.
TI Dam
TI Pool
TKN
TOC
Total PCBs
Tukey-Kramer
Honestly Significant
Difference
When U is non-zero, a desirable evaluation of a model will show that the non-zero
component is dominantly attributable to the covariance or unsystematic
component, which represents non-controllable random variability. Weight on the
bias component indicates that the linear relationship differs between the two data
sets. Weight on the variance component indicates that the difference is attributable
primarily to differing variances between the two data sets.
Thompson Island Dam.
The Thompson Island Pool is the segment of the Hudson River between Rogers
Island and the Thompson Island Dam.
Total kjeldahl nitrogen (mg/Kg DW), a measue of organic nitrogen.
Total organic carbon (percent DW).
The sum of the 126 congener concentrations used throughout the Phase 2 analysis.
Tukey- Kramer HSD (Box et ai, 1978) is used to calculate the confidence interval
for n, - nj when comparing k averages. Where n ^nd n ai~e the number of
observations made when populations / and j were tested. The confidence limits for
n, - nj are given by:
/- — \ . v.a/2
-------
PHASE 2 REPORT
FURTHER SITE CHARACTERIZATION AND ANALYSIS
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
GLOSSARY
PARAMETER: DEFINITION'
Tukey Kramer
Honestly Significant
Difference (Conu
q , „ = the appropiate upper significance level of the studennzed
range for k and v
Any difference in averages greater than the confidence limit calculated is
considered to be significant. This formula is exact if the numbers of observ ations
n in all the averages are the same, and approximate if the averages are based on
unequal numbers of observations. Since the range statistic q , v is used rather than
the s statistic all possible comparisons of averages may be made and the size of the
confidence interval for any given level of probability is larger. With a larger
confidence interval statistically significant differences can be detected with greater
certainty.
ETri+
The sum of trichloro to decachloro homologues.
G-10
TAMS
-------
Introduction
TAMS
-------
1. Introduction
1.1 Purpose of Report
This volume is part of a series of reports describing the results of the Phase 2 investigation
of the Hudson River sediment polychlorinated biphenyls (PCB) contamination. This report, entitled
Phase 2 Volume 2C-A, The Low Resolution Coring Report - Addendum to the Data Evaluation and
Interpretation Report, discusses the results from the Low Resolution Sediment coring Program and
their interpretation. This investigation is being conducted under the direction of the United States
Environmental Protection Agency (USEPA) as part of a three-phase Remedial Investigation and
Feasibility Study (RI/FS) intended to reassess the 1984 No Action decision of the USEPA
concerning sediments contaminated with PCBs in the Upper Hudson River. 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 the Federal Dam
at Troy. Plate 1-1 presents a map of the general site location and the Hudson River drainage basin.
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 (TAMS/Gradient, 1991) is Volume 1 of the Reassessment documentation and
was issued by the 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 (TAMS/Gradient, 1992a) detailed the
following main data-collection tasks to be completed during Phase 2:
• High and low resolution sediment coring;
1 -1
-------
• Geophysical surveying and confirmatory sampling;
• Water column sampling (including transects and flow-averaged composites);
and
• Ecological field program.
The Database for the Hudson River PCBs Reassessment RI/FS, which is described in the
Database Report (Volume 2A in the Phase 2 series of reports; TAMS/Gradient, 1995) provides the
validated data for the Low Resolution Sediment Coring Report. USEPA issued Release 3.0 of the
database on CD-ROM in March 1996. Subsequently, there have been several updates to the database.
The Low Resolution Sediment Coring Report utilizes Release 3.5, which was updated in June 1997.
Subsequent revisions of the Hudson River database, 3.6, 3.7, and 4.0, contain the same low
resolution coring data as found in Release 3.5, with one exception. Releases 3.7 and 4.0 contain an
additional parameter, sediment principal fraction, which was generated from the sieve and laser
grain-size data as part of the preparation of this report. Release 4.0 of the Database for the Hudson
River PCBs Reassessment RI/FS will be available in July 1998.
This report is Volume 2C-A in the series of Reassessment documents presenting results and
findings of the Phase 2 characterization and analysis activities. It contains the results and findings
of the 1994 low resolution coring program and provides a comparison to the 1976 to 1978 and 1984
New York State Department of Environmental Conservation (NYSDEC) sediment data.
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 reporting results strictly
according to Work Plan tasks. In particular, results are presented in a way that facilitates input to
other aspects of the project. Chapter 2 describes the technical approach for the Low Resolution
Coring Program, field sampling procedures, and sample analyses. Chapter 3 interprets the results
of the program and presents evidence to show how the low resolution coring results build on
previously collected Phase 2 data. The three subchapters of Chapter 3 cover: 1) a comparison of
the low resolution coring program with high resolution core PCB results; 2) examination of
-------
correlations among the various low resolution core analyses; and 3) interpretation of low resolution
core results along with the side-scan sonar data. Chapter 4 examines PCB inventories in the areas
of study. Sediment inventories estimated from the low resolution cores are compared with historical
studies of the sediments conducted by NYSDEC in 1976 to 1978 and 1984.
This report is organized in two books. The first contains the document text and appendices.
The second contains the figures, plates, and tables referenced throughout the main body of the report.
Appendix tables are found at the end of each appendix section.
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 their 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, from Federal Dam to the Battery.
1.3.2 Site History
Over a 30-year period ending in 1977. two 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;
Sanders, 1989). 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,
1 - 3
-------
1979, citing 1976 litigation; Limburg, 1986. citing Sofaer, 1976). In 1977. manufacture and sale of
PCBs within the U.S. was stopped under provisions of the Toxic Substances and 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 lbs/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. However, the Aroclors that were discharged varied over time, with Aroclor
1254 being 75 percent or more of the total until about 1955; Aroclor 1242 being at least 95 percent
of the discharges from about 1955 through 1971; and Aroclor 1016 being close to 100 percent of the
discharge from 1971 through 1977 (Brown et al., 1984).
A significant portion of the PCBs discharged to the river adhered to suspended particulates
and subsequently accumulated downstream as sediment 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. Because of its deteriorating condition, the dam was removed in 1973. During
subsequent spring floods, the highly contaminated sediments trapped behind the dam were scoured
and transported downstream. A substantial portion of these sediments were stored in relatively
quiescent areas of the river. These areas, which were surveyed by NYSDEC in 1976 to 1978 and
1984 have been described as PCB hot spots. 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 site near the Bakers Falls Dam through migration
of PCB oil through bedrock has also occurred, although the extent and magnitude of this release are
not well known. This release through bedrock continued until at least 1996, when remedial activities
by GE brought the leakage under control. Despite some evidence for its existence prior to 1991
based on United States Geological Survey (USGS) data, this leakage was not identified until the
partial failure of an abandoned mill structure near GE's Hudson Falls plant site in 1991. This failure
1-4
-------
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 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; TAMS et al.. 1997).
1.4 Background for the Low Resolution Sediment Coring Program
Two previous large-scale sediment investigations were conducted by NYSDEC; one in 1976
to 1978 (reported in Tofflemire and Quinn, 1979), and one in 1984 (reported by Brown et al., 1988).
The 1976 to 1978 sampling covered the area from Fort Edward to Troy (RM 194.8 to RM 154);
whereas the 1984 sampling was restricted to the Thompson Island (TI) Pool (RM 194.6 to RM
188.5). On the basis of data gained from these investigations, approximately 40 zones of high
contamination, designated as hot spots, were identified. These data have been used to estimate total
PCB inventory in Hudson River sediments (Tofflemire and Quinn, 1979; Brown et al., 1988). As
part of this investigation, the 1976 to 1978 and 1984 data have been subjected to more sophisticated
mathematical evaluation (including a kriging analysis of the 1984 data) to develop detailed maps of
contamination (TAMS et al., 1997).
These surveys served to describe Upper Hudson sediment conditions at the time of their
completion. An issue of great significance to the Hudson River PCBs Reassessment is the
applicability of previous surveys to current conditions. To some degree, this issue was addressed
by the geophysical surveying of the Upper Hudson as part of the Phase 2 investigation. This survey
covered 31 of the 36 hot spots previously defined by NYSDEC. Although 40 hot spots were
originally defined, four hot spots near Rogers Island were removed by dredging prior to 1980. As
discussed in the DEIR (TAMS et al.. 1997), the geophysical survey documented the continued
presence of fine-grained sediments in many of the areas previously defined as hot spots. This
indicates that much of the PCB-contaminated sediment deposited from 1955 to 1978 is probably still
in place. Nonetheless, a more direct assessment of the current PCB inventory still contained within
the sediments was desired to confirm this finding.
1 - 5
-------
The DEIR (TAMS et al.. 1997) discussed the results of several of the earlier Phase 2
investigations, including the high resolution coring program. The high resolution coring involved
slicing each core into sections at two to four centimeter intervals to provide a detailed (i.e., highly
resolved) history of PCB deposition at the coring location. In contrast, the low resolution cores were
sliced into three relatively thick sections at approximately 22 cm (9 in) intervals. These thickly sliced
cores do not provide sufficient resolution to examine the depositional history at the coring site, hence
the term low resolution, but they do provide an excellent basis on which to estimate the sediment
PCB inventory.
1.5 Low Resolution Sediment Coring Program Objectives
The Low Resolution Sediment Coring Program had two main objectives:
1) Obtain new estimates of sediment PCB inventories at a number of locations in the
TI Pool to compare against the existing PCB sediment database constructed from the
1984 NYSDEC survey; and
2) Refine the PCB mass estimates for a limited number of historic hot spot locations
defined by the 1976 to 1978 NYSDEC survey in the Upper Hudson below the TI
Dam.
The implementation of the sampling program was based upon a review of the historic sediment
records from NYSDEC, GE, and others, in conjunction with the results from the geophysical
survey. The low resolution coring effort in the TI Pool was intended to establish the extent to
which the sediment inventory calculated from the 1984 NYSDEC data (TAMS et al., 1997)
accurately reflected current conditions. The low resolution coring in the lower reaches of the Upper
Hudson River (below the TI Dam) provided information to refine the accuracy of previous hot spot
sediment PCB inventory estimates which were based on a relatively small number of samples (from
1976 to 1978), as well as providing a measure of the contribution of areas below the TI Pool to the
Upper Hudson River sediment PCB inventory. It should be noted that the 1988 NYSDEC report
indicated that "no major change in the distribution of PCBs in the bed of the TI Pool between 1977
1 -6
-------
and 1984 is evident" (Brown et al1988), even though the PCB inventory reported by Brown el
al. (1988) was less than half of the PCB mass estimated by Tofflemire and Quinn (1979). Brown
et al (1988) concluded "that most of the difference between 1977 and 1984 PCB mass estimates
was due to differences in calculation methods and assumptions" after taking into account USGS
estimates of PCB transport from Fort Edward to Schuylerville from 1978 to 1984 (Brown et al.,
1988).
Low resolution coring was performed to examine PCB contamination in a limited number
of areas and to augment and improve estimates of the spatial distribution of PCBs initially
developed for these areas. The low resolution coring program was not intended to provide the
extensive spatial coverage obtained as part of the earlier NYSDEC investigations. Rather, the
program focused on obtaining information to be used in conjunction with the other Phase 2
sampling programs as well as to augment and update the data obtained in the prior NYSDEC
surveys.
In the subsequent chapters, the following conventions are used concerning the description
of sediments near the sediment/water interface and differentiate between the samples taken for the
low and high resolution coring programs. Shallow sediments refer to sediments within 12 inches
(30 centimeters [cm]) of the sediment/water interface. The top segment of each low resolution core,
beginning at the sediment/water interface and extending down about 9 inches (22.5 cm), is
considered to represent shallow sediments. Surficial sediments refer to sediments within 2 inches
(5 cm) of the sediment/water interface. The top two slices of a high resolution core are considered
to represent surficial sediments by this definition. Surficial sediment samples were also obtained
for radionuclide analysis exclusively as part of the low resolution coring program. In order to obtain
and refine the estimates of the entire sediment PCB inventory at each sampling location as
described earlier in the objectives, low resolution core collection included both shallow and deeper
sediments for analysis.
1 - 7
-------
SAMPLING DESIGN AND METHODS
-------
2. Sampling Design And Methods
2.1 Technical Approach for the Low Resolution Sediment Coring Program
As discussed in Section 1.5, the Low Resolution Sediment Coring Program was designed
with two basic investigation objectives. First, the program was to assess the current status of the
sediment PCB inventory of the Thompson Island (TI) Pool relative to the NYSDEC survey
completed in 1984. The second objective was to assess, to a limited degree, the sediment PCB
inventory of several previously defined hot spots in the Upper Hudson below the TI Dam. Each
of these objectives was focused on limited study areas so as not to repeat the earlier, more
extensive surveys. In both cases, the intent was to provide a basis on which to assess the current
applicability of the historical surveys.
The TI Pool was sampled intensively in 1984, with over 1,200 samples collected on a
triangular grid with 125-foot centers. The spatial coverage of that effort was adequate for the
purpose of estimating PCB sediment mass inventory, as discussed in Brown et al., 1988 and
TAMS et al., 1997. Rather than resurvey the entire TI Pool, the 1994 low resolution coring
effort focused on replicating a representative subset of the 1984 locations. The intent was to
assess the comparability of the current PCB inventories at these locations relative to the 1984
conditions. Based on these comparisons, an assessment of the current applicability of the entire
1984 data set could be made. The locations sampled in the TI Pool as part of the low resolution
coring program were grouped into 15 relatively small zones or clusters. Within these clusters,
samples were collected as near as possible to the same locations sampled by NYSDEC in 1984.
Each zone generally consisted of about four or five sampling points, corresponding to the original
NYSDEC locations. The new results were then used to compare the current sediment PCB
concentrations in the TI Pool with the previously determined ones. The analysis and comparison
of these results are presented in the subsequent chapters of this report. In addition to the 15
clusters, four additional clusters were located in near-shore areas of fine-grained sediment where
the original 1984 NYSDEC coverage was poor. These samples were intended to characterize near-
shore sediment contamination. These near-shore locations represent potential human exposure
2-1
~\MS
-------
routes for direct sediment contact as well as a potentially important but an undocumented reservoir
of PCBs.
The locations for re-sampling under this program were selected based on a review of the
1984 data. In order to assess the changes in sediment PCB inventor}' and avoid uncertainty due
to variability stemming from causes other than PCB inventory changes, the sampling locations
were grouped into selected zones with minimal local sediment heterogeneity. Sediment
heterogeneity was evaluated based on the following information:
• 1984 PCB data, showing PCB inventories within each zone to be similar;
• Side-scan sonar data, establishing the presence of similar material within each
zone; and
• Field logs or other observations made during the 1984 sampling and, where
available, 1992 confirmatory coring data to verify similar sedimentological
properties.
In general each sampling zone (cluster) consisted of three to six locations with similar sediment
properties and PCB concentrations (expressed as mass per unit area, e.g., g/m2) which typically
varied by a factor of two or less, and no more than a factor of three. One "null set," that is a
zone in which PCBs were previously reported as not detected, was also included in the low
resolution coring program. A total of 19 clusters and 76 cores were collected from the TI Pool
during the Low Resolution Sampling Program.
Below the TI Dam, the 1976 to 1978 NYSDEC sampling program identified 20 hot spots,
numbered 21 to 40 (Tofflemire and Quinn, 1979). However, the spatial coverage of the 1976 to
1978 NYSDEC sampling was not as intensive as the 1984 survey so the areas of these hot spots,
and the resulting PCB inventory estimates, were based on a relatively small number of samples.
As part of the low resolution coring program, seven of these hot spots were re-sampled to
determine the current PCB inventory. These data were designed to assess how well estimates
based on the limited 1976 to 1978 data reflect the current extent and magnitude of PCB
contamination. The hot spots selected for the low resolution coring program are those in which
2-2
TAMS
-------
previous estimates have indicated the highest PCB inventories and which physically span a
substantive length of the river. Based on the NYSDEC estimates, these hot spots represented
about 74 percent of the total mass of PCBs in hot spots below the TI Dam. The spatial intensity
of the low resolution coring effort in these areas (below the TI Pool) was adjusted so that the
entire area previously identified as the zone of contamination was covered by seven to ten low
resolution cores. In addition, four near-shore or exploratory locations were selected in this region
in areas not covered by the 1976 to 1978 NYSDEC survey to examine sediment contamination
in unstudied zones of fine-grained sediments.
The sample collection procedure used for the low resolution coring program was similar
to that used for the high resolution coring program although there are important differences which
are discussed later. The term low resolution coring arises from the subdivision of the cores. In
general, low resolution cores were separated into 9-inch (22.5-cm) layers, as opposed to the 0.8
to 1.5-inch (2 to 4-cm) thick layers used in the high resolution coring program. However, since
a goal of the low resolution coring program was to verify total PCB inventories, the thickness of
the low resolution core slice was modified as necessary so that all of the apparently contaminated
material is included in the samples submitted for PCB analysis from each core. "Apparently
contaminated material" was determined by the field geologist, based on the depth to which wood
chips or cellulose-type material were present in the core. In addition, core intervals were chosen
to coincide with visible sediment horizons, such as a sand/silt boundary or a change in sediment
color. Where applicable, the low resolution core length and sampling interval were made to
correspond to the same lengths as the original NYSDEC sample at that station when no other
sedimentological criteria could be used. The low resolution cores ranged in length from 6 to 54-
inches (15 to 137-cm) of sediment. Cores were advanced through the sediment until the coring
apparatus could not penetrate any further.
One to three core slices were obtained from each core for PCB analysis, typically
representing a total of about 18 to 20-inches (45 to 50-cm) of sediment. Below the deepest layer
analyzed for PCB contamination in each core, an additional 3-inch (7.5-cm) slice was collected
for radionuclide analysis only. The results from this layer were used to verify that the core
2-3
TAMS
-------
thickness was sufficient to recover all contaminated sediment deposition based on the discussion
below.
One of the areas of uncertainty between the assumptions used for PCB mass estimates
made by Tofflemire and Quinn (1979) and Malcolm Pirnie, Inc. (MPI, 1992) is the depth to which
PCB contamination extends. In cores collected from throughout the Hudson, PCBs are first
detectable in strata deposited in the late 1940s to early 1950s, with concentrations increasing from
the mid-1950s to a peak in the early 1970s (Bopp and Simpson, 1989; TAMS et al., 1997).
However, the contribution of PCBs from deposition prior to 1955 is relatively insignificant, as
confirmed by the 1992 high resolution core data. Cesium-137 (137Cs) has a similar release history,
with the radionuclide first appearing in 1954 with the onset of atmospheric nuclear weapons
testing, from which 137Cs is derived. Given the coincidence of the PCB and 137Cs histories in the
1950s, 137Cs was analyzed in the deepest layer from each core. Absence of 137Cs in this layer
indicated that the core had been advanced to pre-1954 sediments and to a sufficient depth to
account for all substantive PCB deposition.
2.2 Field Sampling
The low resolution sediment coring program was designed to further characterize and
analyze site conditions at the Hudson River PCBs Superfund site. The low resolution sampling
program took place from July 13, 1994 to August 12, 1994. The Phase 2B Sampling and Analysis
Plan/Quality Assurance Project Plan Volume 4 (TAMS/Gradient, 1994) detailed the sample
collection and analytical procedures, which are summarized below.
In addition to the interpretation of the low resolution sediment coring samples, this report
also discusses the side-scan sonar results obtained in the earlier portion of the Phase 2
investigation. A detailed discussion of the side-scan sonar program can be found in the DEIR
(TAMS et al., 1997) while a brief description is provided here. Side-scan sonar data covering the
Upper Hudson from above Bakers Falls to Lock 5 were collected in 1991 and 1992. Essentially,
acoustic signals (digitally recorded sound) were used to create false-color grey-scale images of the
river bottom. These images were then combined to form mosaic maps of the river bottom wherein
2-4
TAMS
-------
brighter areas on the images corresponded to coarser-grained sediments and darker areas
corresponded to finer-grained sediments. The sediment classifications as defined by the side-scan
sonar images were compared to low resolution coring grain size data. These data area discussed
in Section 3.3 of this report. The comparison of the side-scan sonar results to previous sediment
collection efforts (confirmatory sampling) is described in the DEIR (TAMS et ai, 1997).
2.2.1 Sample Locations
Low resolution sediment coring was conducted in the Upper Hudson River between Rogers
Island at Fort Edward (RM 194) and Lock #2 (RM 163.5). The low resolution coring sample
locations (zones) in the TI Pool are shown in Plate 2-1 and summarized on Table 2-1. The
locations were selected to represent a range of sediment types and sediment PCB inventories, with
emphasis placed on areas of greatest PCB contamination. Fifteen small zones or clusters were
selected within the TI Pool for a total of 60 low resolution core sites. In addition, four near-shore
locations within the TI Pool were sampled during the low resolution coring program.
Downstream from the TI Dam, seven previously identified hot spots in the Upper Hudson were
sampled. The locations of the NYSDEC hot spots in the project area along with the associated
low resolution coring locations are listed on Table 2-1 and shown on Plate 2-1. The number of
cores taken in each of the hot spots was roughly proportional to its mass and spatial extent. To
establish limits on the spatial extent of contamination at some of the larger hot spots, a few
samples were placed beyond the original boundary defined by NYSDEC to examine the accuracy
of the boundary. In addition to the cores associated with the hot spots, thirteen low resolution
cores, roughly equivalent to the sampling intensity of one large hot spot (approximately 30
samples), were used as near shore/exploratory samples in four areas in which previous sampling
was minimal or non-existent, and where bathymetric and/or sedimentological data suggested
potential contamination.
Sample locations were surveyed at the time of collection using shoreline control points
established prior to core collection. Estimated accuracy of all coring locations is ± 3-feet.
Within the TI Pool, field personnel attempted to re-occupy the NYSDEC locations. The original
specification in the sampling plan stated a goal of ± 20-feet about the NYSDEC location. In fact,
2-5
TAMS
-------
the field team was much more successful, achieving a median distance of 3-feet between the 1984
and 1994 locations and only one 1994 location differed by 20-feet from the targeted 1984 location.
Figure 2-1 shows the distribution of the distances between the 1994 locations and their targeted
1984 sites. Because of the success of the coring location control, 1984 to 1994 changes in total
PCB inventory due to sample location differences were kept to a minimum.
2.2.2 Sample Preparation
The low resolution coring sampling plan (TAMS/Gradient, 1994) detailed a coring
program to collect 39-inches (97.5-cm) of sediment at each coring location, conditions permitting.
A core length of thirty-nine inches was selected based on the earlier NYSDEC studies which
found PCB contamination principally in the top 24 inches of sediment. Twelve inches more were
added to the goal to generously allow for additional deposition since the earlier surveys. Lastly,
three inches more were added for a radionuclide sample to be collected below the main core
segments. Since the main objective was not to simply collect a 39 inch core but rather to obtain
all recent deposition, cores less than the prescribed goal were expected. Low resolution cores
were obtained by a "vibra-coring" technique where a 4-inch (10-cm) diameter clear plastic tube
was advanced through the sediment by applying pressure and vibration to the top of the tube.
This served to partially liquefy the sediments along the walls of the tube and permit greater
penetration of the tube into the river sediments. A sediment catcher or shoe was attached to the
bottom of the coring tube to minimize the loss of sediment from the tube during retrieval. By
comparison, the high resolution cores were collected in 2.5-inch (6.3-cm) diameter plastic tubes
by applying pressure only. A check valve at the top of the core, rather than a shoe at the bottom,
was used to prevent sediment loss during retrieval of the high resolution cores. Once obtained
from the river, low resolution cores were then intended to be sliced into three 12-inch (30-cm)
sections and one 3-inch (7.5-cm) section for subsequent analyses, when a full 39-inch core was
obtained. Cores less than 39 inches were acceptable subject to field review by the field geologist.
In essentially all cases, the coring apparatus was advanced until the equipment could not push it
further. The core was then retrieved and the recovered material was examined by the field team.
The average core length of 23 inches (57.5 cm) was substantially less than the goal of 39 inches
(97.5 cm) and many core recoveries were less than 15 inches (37.5 cm). Generally, four segments
2-6
TAMS
-------
were taken from cores 27 inches or more. As a result, the median segment thickness for the first
three core layers was about 9 inches (22.5 cm), while still leaving a bottom 3-inch layer. For
cores less than 20 inches, the field geologist evaluated the stratigraphy of the core and decided
where to slice the core based on factors such as changes in color or grain size. Nonetheless, as
will be discussed later, the evidence suggests that most cores were advanced through all
potentially PCB-contaminated sediment.
Of the 170 cores collected for the program, 73 cores were sufficiently long so as to yield
four segments per core. Fifty-five cores yielded three segments per core while 41 cores yielded
only two segments per core. One core (LR-02C) yielded only one segment. Besides two to four
main core segments, in each instance the top portion of the top segment (0 to 1-inch) was used for
7Be and I37Cs radionuclide analyses and the bottom segment was used exclusively for 13ts
radionuclide analysis. The overlying one to three main segments were analyzed for PCB and other
analytes as discussed below. Core LR-02C was not analyzed for radionuclides. Table 2-2
provides a summary of the cores collected in terms of layer thickness and depth, the number of
each core type collected (e.g., number of two-layer cores), and the overall core lengths for each
core type. Figure 2-2 is a summary diagram of the core segment thickness and depth. Figure 2-3
provides an illustration of how the cores were subdivided for analysis.
The actual slicing intervals for each low resolution core were selected by the field team
based on visible sediment stratigraphy and analytical sample requirements as well as the prior
NYSDEC core slicing intervals when applicable. Once the slicing intervals were selected, the core
was extruded into pre-weighed stainless steel bowls for bulk density determination and
homogenization. After homogenization, the sample was sub-sampled for various analytes as
discussed below. During processing, samples were also field classified as clay, silt, fine-to-coarse
sand, and gravel. The presence of minor quantities of other soil types, as well as organic
materials and soil color, were also noted by the field geologist.
Sample nomenclature was similar to the high resolution coring program with one important
difference. Core intervals were incorporated as inches and not centimeters, as was done
previously. This was done to avoid lengthening the standard Phase 2 eleven-character
2-7
TAMS
-------
identification (ID) string to account for samples greater than 40-inches (100-cm) in depth. The
information encoded within the low resolution sample ID is as follows:
AA-OOA-OOOO-A
where UA" represents a letter and "0" represents a number.
For the low resolution coring effort, the first letter was assigned an "L". The second letter
was assigned an "R" for samples collected from clusters within the TI Pool. For samples
collected downstream of the TI Dam, the second letter was assigned an "H". The next three
characters represent the core location. The first two digits were assigned the number of the
investigation zone in which the core was collected. The Thompson Island Pool locations were
assigned numbers in the range of 01 to 19 corresponding to the cluster number. The Hudson
River sample locations below the TI Dam were assigned values corresponding to the two-digit hot
spot number (i.e., 25, 28, 31, 34, 35, 37, or 39). The alpha character which is the third character
of this triad designates one of the three to 15 cores obtained from the cluster or hot spot area.
Thus, UA" would represent the first core from the area, "B" the second, and so on. The last four
digits of the ID represent the depth interval of the core segment in inches, with the first two digits
corresponding to the top of the core segment and the last two digits corresponding to the bottom
of the core segment. As was the case for previous efforts, the final letter is reserved for QC
sample designations where applicable - D (Duplicate), and M (Matrix Spike or Matrix Spike
Duplicate).
As an example, the sample designated LH-28K-0918 would be the low resolution core
sample (LH) from Hot Spot 28. The "K" indicates the eleventh core in that Hot Spot, and 0918
indicates that the segment (section) was from the 9 to 18-inch (23 to 45-cm) depth interval.
2.3 Sample Analyses
Analytical parameters for the low resolution coring included:
2-8
TAMS
-------
• PCB congeners by capillary column gas chromatograph/electron capture detector
(GC/ECD; with a limited number of confirmation analyses by GC/ion trap detector
[ITD]);
• Radionuclide analysis;
• Total Organic Carbon;
* Total Kjeldahl Nitrogen;
• Grain-size distribution by sieve and laser particle analyzer (Laser);
* Grain-size distribution by sieve and hydrometer analysis (American Society for
Testing and Materials IASTMJ);
* Percent solids; and
* Bulk density.
The analytical method for PCB congeners is provided in Appendix A. Non-PCB chemical and
physical properties analytical methods are discussed in Appendix B.
As mentioned briefly in the previous sections, not all samples were analyzed for all
parameters. This choice was based on the analysis type and intended use of the data. The typical
sample set generated for each core was as follows:
* Radionuclide (beryllium-7 fBe] and cesium-137 [137Cs]) analysis of sediment from
a portion (half) of the 0 to 1-inch (0 to 2.5-cm) interval;
• PCB congeners from up to three layers of sediment, each about 9-inches thick;
• Laser grain-size distribution analysis of sediment from 0 to 9-inches;
• Bulk density for all sediment layers;
* Percent solids for all sediment layers greater than 1-inch; and
* Radionuclide (l37Cs only) analysis of sediment from a 3-inch layer immediately
below the deepest layer analyzed for PCBs.
Roughly two-thirds of the cores had one to three additional grain-size distribution analysis by a
standard ASTM sieve and hydrometer technique performed on one of the core intervals analyzed
for PCBs. A subset of these samples was obtained from the top core slice permitting a direct
comparison between the laser and ASTM techniques, discussed later in this report. A small subset
2-9
TAMS
-------
of the core samples, roughly one in twenty, was analyzed for total organic carbon and total
kjeldahl nitrogen. Figure 2-3 provides an illustration of how the cores were subdivided for
analysis and Table 2-3 provides a tally of the number of analyses conducted for each analyte,
excluding duplicates. As shown in Figure 2-3, only one half of the top 1-inch slice was used for
radionuclide analysis. The other half of the slice was added to one half of the core portion from
1 to 9- inches and homogenized prior to subsampling for PCBs and other analytes.
Two additional analyses were slated to be collected or performed during the low resolution
sediment coring program, including reduction/oxidation potential and total carbon/total nitrogen
content. The results for reduction/oxidation potential were all rejected during the data quality
review due to field calibration problems. Samples collected for total carbon/total nitrogen content
were inadvertently held beyond acceptable sample holding times while finalizing a laboratory
contract and were therefore never analyzed. The original low resolution sampling plan
(TAMS/Gradient, 1994) also called for grain-size distribution analysis based on sieve and
hydrometer to be performed for all samples. However, prior to beginning the field operation, this
requirement was reduced to 150 samples, including duplicates.
Sample quality assurance followed the guidelines defined the Low Resolution Sediment
Coring Sampling and Analysis/Quality Assurance Project Plan (TAMS/Gradient, 1994).
Duplicate analyses were performed on field-generated sample splits at the rate of 1 in 20 samples.
Data usability was reviewed for all analytes and is discussed in Appendix A. In general, data
quality objectives were met with the exception of the reduction/oxidation measurements as
discussed above.
2.3.1 PCB Congener Analysis
PCB congener analyses were performed on a rigorously homogenized core segment by a
wet sediment soxhlet extraction followed by analysis on a dual capillary column gas
chromatograph with election capture detectors (GC/ECD). Roughly 10 percent of the sample
analysis was confirmed via a second analytical technique, gas chromatography with an ion trap
detector (GC/ITD). This confirmation was limited to samples with high PCB concentrations dufc
2-10
TAMS
-------
to the higher (less sensitive) detection limits characteristic of this technique. These techniques are
the standard congener-specific techniques used throughout the Phase 2 investigation. Use of these
techniques guaranteed comparability of the low resolution coring results for PCBs with those
obtained in the earlier Phase 2 studies. Due to improvements in the techniques achieved during
the earlier Phase 2 studies, a total of 145 congeners was reported for each low resolution core
PCB analysis, including 108 calibrated (target) congeners. However, to maintain comparability
among the Phase 2 program, only the original 126 congeners including 90 targets are discussed
and presented here. In general, the additional 19 congeners did not contribute substantively to the
sediment PCB inventory, the main focus of this report. At the beginning of the Phase 2
investigation in 1992, standards only existed for 90 of the 209 PCB congeners. Retention times
were known for an additional 36 congeners, bringing the total number of identifiable congeners
to 126 at the initiation of the Phase 2 program. Subsequent to the start of the program, many more
individual congener standards have been made available, thus enabling the increase to 145
identified congeners in the low resolution sediment core samples. As discussed in the DEIR
(TAMS et al., 1997), use of the original 126 congeners including the 90 calibrated congeners
captures more than 90 percent of the original Aroclor 1242 and a higher percentage of the other
Aroclors. Results for all 145 congeners analyzed in the low resolution cores are reported in the
Phase 2 database (Release 3.5, June, 1997).
The congener-specific GC/ECD analysis performed for the low resolution coring does not
rely on arbitrarily assigning Aroclor identification, but enables a total PCB concentration to be
calculated, regardless of peak or congener composition. This is especially significant for congener
mixtures which may have been subject to dechlorination, and no longer resembles the original
Aroclor mixture. Since the major goal of this investigation was to examine the PCB inventory
in a variety of sediments, this ability to calculate total PCB concentration is particularly important.
However, in order to compare the congener-specific data to previously reported data, Aroclor data
were calculated by summing the concentrations of the congeners present in each Aroclor mixture
based on Aroclor standards analyzed via the Phase 2 congener-specific technique. In this manner
the Phase 2 low resolution coring results can be compared with earlier studies which were based
on Aroclor quantitation techniques.
2-11
TAMS
-------
2.3.2 Radionuclide Analysis
The radionuclide analytical procedure was restricted to the isotopes cesium-137 (137Cs) and
beryllium-7 CBe). 7Be data were obtained and used to date the top of the core, and the 137Cs data
were used to verify that the core has penetrated all recent (i.e., post-1954) sediments. Analysis
of radionuclides in sediment cores provided a means of establishing the sediment core chronology.
Studies of sediment cores in the Hudson have demonstrated the occurrence of well-documented
radionuclide events which can be used to establish sediment accumulation rates at various
locations throughout the Hudson. By determining the activities of 137Cs and 7Be, it is possible to
verify that the core collected includes the depositional period from the time of core collection back
to 1954.
137Cs is a persistent (half-life of 30 years) anthropogenic radionuclide that has two distinct
events associated with it. The first event corresponds to the onset of atmospheric atomic weapon
tests in 1954, which is indicated in the sediments by the first appearance of 137Cs. Background
levels prior to 1954 are essentially zero for this radionuclide. Therefore, 137Cs analysis of the
bottom core segment verified for most cores that they extended at least as far back as 1954. A
non-zero result for 137Cs indicated that the core collected did not include sediment deposition prior
to 1954. Previous data have indicated that PCB deposition in the Hudson River prior to 1954 is
not significant in comparison to PCB discharges since then (TAMS et al., 1997).
7Be is a short lived (half-life of 53.6 days), naturally-occurring isotope whose presence in
the sediments indicates recent deposition or interaction with surface waters within the six months
prior to sample collection. Approximately 90 percent of the 7Be activity dissipates within 180
days. Thus, this radionuclide was used to initially test a core top (0 to 1-inch) for the presence
of recently deposited sediment. Absence of 7Be is attributed to a core collected in a non-
depositional area or an area experiencing scour (erosion) of river sediment.
2-12
TAMS
-------
2.3.3 Total Organic Carbon and Total Kjeldahl Nitrogen
Total organic carbon and total kjeldahl nitrogen analyses (TOC/TKN) were performed on
about seven percent of the samples analyzed for PCBs, representing about 15 percent of the coring
sites. These analyses were performed on individual core segments randomly selected from the
various cores. Specifically, one sediment segment was analyzed per core in 26 cores with nine
surface segments, seven second layer segments, and 10 third layer segments analyzed. TOC
analyses were accomplished by the USEPA Region II method involving the direct combustion of
the sample. (TAMS/Gradient, 1994). TKN analyses were accomplished by a modified version
of USEPA method 351.2 (TAMS/Gradient, 1994). TOC and TKN analyses were performed on
the same subset of samples so as to permit the calculation of a molar carbon-to-nitrogen ratio.
These analyses were originally intended solely to supplement the planned total carbon/total
nitrogen (TC/TN) analyses which were to be run on every core sample. This approach was the
same as that used for the high resolution cores. The TOC/TKN analyses were intended to confirm •
that the TC/TN results closely matched the organic C/N results. However, due to exceedances of
holding time, none of the TC/TN analyses were run. As a result, only the randomly selected
TOC/TKN samples were available to characterize organic carbon and nitrogen levels in the low
resolution coring program study areas.
2.3.4 Physical Properties
Grain-Size Analysis
Grain-size analysis was performed on a large subset of the low resolution sediment core
slices (except the bottom segment which is analyzed for 137Cs only). In the top core segment only,
grain size distribution was determined by mathematically combining results from an ASTM sieve
method and a laser-particle analyzer method (Phase 2B SAP/QAPjP Appendix B-5). The smaller
particle (less than 2-mm) grain-size distribution was determined by the laser particle analyzer-
based methodology. Larger particle size fractions (greater than 1-mm) were determined by the
ASTM methods. There was an overlap between the methods in the 1 to 2 mm-range, thus
providing a means of correlating and cross-checking data between the two methods. The principal
2-13
TAMS
-------
purpose for these samples was to further support the grain-size distribution and side-scan sonar
analysis presented in the previous Phase 2 report (TAMS et al., 1997). An additional set of
sediment samples was analyzed using ASTM Methods D421-85 and D422-63 including
hydrometer analysis. These samples will be used to characterize the sediments for engineering
analyses.
Bulk Density
Review of the method for deriving bulk density used in the previous studies indicated that
the method was somewhat crude and may not be representative of in situ density. Although the
data review in general indicated that the average bulk density values from the NYSDEC studies
were reasonable, the data for some individual points were not (TAMS et al., 1997). Therefore,
to address concerns regarding the accuracy of previously-derived bulk sediment densities, new
data for this property was obtained during the low resolution coring program.
Bulk density was measured in the field by extruding a known volume from the core tube
into a pre-weighed stainless steel bowl and then re-weighing the bowl. Bulk density was
calculated simply as the ratio of the mass weighed to the known volume. The results were used
to calculate the total PCB inventory. Reported bulk density values were restricted to a range of
1 to 3 g/cc. Values outside this range were rejected.
Percent Solids
Percent solids were determined for all samples analyzed for PCBs as well as the bottom
slice analyzed for 137Cs. This method involved the simple weighing of a sample portion before
and after drying. The original results were reported as percent moisture. These values were
converted to percent solids by simply subtracting the percent moisture value from 100 percent.
The percent solids data were combined with the bulk density data to calculate the solids specific
weight (the weight of solids per unit volume of sediment) and the solids density (the mass of solids
per unit volume of solids). Typically, the solids density should be in the range of 2 to 2.5 gm/cc.
Sample values for solid bulk densities outside the range of 1 to 3 gm/cc were excluded from
2-14
TAMS
-------
further analysis. This exclusion incorporated the bulk density, percent solid values, solids specific
weight and the solids density. That is, unless the bulk density and percent solids yielded
internally consistent results, all four values (two measured, two calculated) were excluded from
further analysis.
2.4 Summary of Analytical Results
This section presents a brief general description of the analytical results obtained for the
low resolution coring program. The interpretation of the results is presented in Chapters 3 and
4 of this report.
2.4.1 PCB Congener Analysis
PCBs were detected in every sample analyzed for PCBs. The range of total PCB
concentrations (<0.05 to 1,352 ppm) was slightly less than the range obtained for the high
resolution cores (<0.05 to 2,500 ppm). Nonetheless, these values represent substantive PCB
inventories, especially when the segment thickness is considered. The low resolution sediment
values represent a 9-inch median thickness, as opposed to the 0.8 to 1.6-inch high resolution core
layers. Thus, the high values obtained in cores from the Upper Hudson do not appear very
unusual, given their occurrence in both the low resolution and high resolution programs.
A summary of the results from the cores is presented on an area basis and a depth basis
in Table 2-3. The arithmetic mean and median concentrations in the TI Pool are comparable to
the areas downstream of the TI Pool. Mean sediment concentrations obtained from the low
resolution core results should not be directly compared between the two regions because the 76
cores analyzed in the TI Pool and 94 cores taken downstream of the TI Pool were intended to
characterize local conditions in several areas and do not comprise a spatial coverage sufficient to
calculate PCB inventories for these areas directly. Histograms of the total PCB concentrations in
all core segments and only in shallow (top) core segments are shown in Figure 2-4. The total
PCB concentrations are statistically skewed, with an approximately normal distribution when data
are log normalized. The log-normal nature of the shallow segment data was confirmed by a W
2-15 TAMS
-------
test for normality, although the entire log-transformed data set {i.e., all core segments) did not
pass the W test for normality. The lack of normality is largely the result of the inclusion of very
low PCB concentrations, typically found in deep core segments. In many of the subsequent
analyses, these samples (total PCBs < 100 Mg/kg) are excluded due to quantitation issues, thus
making the remaining data set more log-normal. The majority of statistical calculations dealing
with PCB concentrations included in this report are based on the log-transformed data.
The core results presented by depth show an important finding. PCB maxima are
principally found in the top-most core layer (61 percent or 104 out of 170 cores), representing
shallow sediments (median depth of 9-inches). These results indicate that burial of PCB-bearing
sediments is not occurring on an extensive basis and that high concentrations of PCBs remain
relatively close to the sediment/water interface. The assignment of the sediment PCB maximum
to a given layer is based on several criteria including:
• Occurrence of the maximum PCB concentration for the core in that layer; and
either
• Occurrence of a deeper layer with lower PCB levels; or
• Absence of 137Cs in the radionuclide layer at the bottom of the core when the
maximum PCB layer is immediately above.
These criteria are based on the known PCB deposition history for the Upper Hudson as
recorded in sediment cores (Bopp and Simpson, 1989; TAMS, et al., 1997). In particular it
assumes there is only one PCB maximum and that 137Cs presence indicates PCB presence. Thus,
when a core has an assigned PCB maximum, it is inferred that the maximum concentration in the
sediments at that location has been captured by the core, and that the core represents the majority
of the PCB inventory. When 137Cs is not detected in the core bottom, the core is assumed to
represent the entire PCB inventory. On 137Cs alone, 119 cores are assumed to represent complete
inventories. By utilizing the criteria given above, an additional 15 cores can be added to this
category, bringing the percentage of complete (and nearly complete) cores to 134 or 79 percent
of the cores collected.
2-16
TAMS
-------
In a portion of the cores collected (36), these criteria were not met and the core PCB
maximum was unassigned. In these instances, the possibility that the PCB maximum
concentration lies below the bottom of the core cannot be ruled out. Thus, these cores are listed
as having an unknown maximum. It should be noted that the majority of these unknown
maximum cores (22 of 36) represent short (< 15 inches) 2-layer cores where additional sediment
could not be obtained. Given the inability to further advance the collection apparatus, it is unlikely
that these areas are underlain by large amounts of high PCB-bearing sediments, typically silts and
fine sands which are easily cored. Also, as noted above, the PCB maxima are generally found
in the shallowest sediments. Thus, even though these cores are incomplete, it is unlikely that the
PCB inventories are much greater than calculated.
For the multiple layer cores with unknown PCB maxima, the depth of penetration is
similar to many of the complete cores. The results from the complete multiple layer cores indicate
that the majority of PCB contamination resides in the two uppermost layers (top 9 to 18-inches).
This would suggest that the incomplete multiple layer cores are likely to capture the majority of
the sediment PCB inventory since they typically extend below 18 inches. However, this is much
less certain than for the incomplete 2 layer cores. In particular, these incomplete multiple layer
cores were more commonly found below the TI Dam and have important implications for the
sediment mass estimates from this area. Further interpretation of these cores is presented in
Chapter 4 during the discussion of the hot spots below the TI Dam.
The fact that the PCB maxima are found largely in the shallow sediments provides useful
information for the cores labeled incomplete, as noted above. This implies that for most of these
cores, the material retrieved probably represents the majority of the PCB inventory at the coring
site. PCB estimates derived from these incomplete cores probably underestimate the actual
sediment inventory in the affected cores by less than 50 percent. Discussions of sediment-PCB
inventories and their comparison to earlier surveys are presented in subsequent chapters of this
report.
The total PCB values discussed above were based on congener-specific analysis, just as
all Phase 2 results have been. Although the 145 congeners were reported for the low resolution
2-17
TAMS
-------
core samples, the total PCB results as well as the congener-specific discussions to follow are based
on the original 126 congeners to maintain consistency in the Phase 2 discussion. Exclusion of the
19 additional congeners has little effect on estimates of sediment PCB mass since these congeners
represent less than 1.5 percent of the total PCB concentration, on average. In all but five of the
371 PCB analyses, these 19 congeners represented less than six percent of total PCB mass. For
the five samples where these 19 congeners were greater than six percent, the total PCB
concentration was less than 32 yug/kg (0.032 ppm). Thus, these congeners do not represent
substantive PCB mass and are ignored in subsequent discussions in this report. Some of these
congeners may be used in PCB pattern comparisons as part of the ecological assessment.
Sample splits were generated in the field and run as blind duplicates by the laboratory.
A total of 23 split pairs was generated. The fields were compared using a relative percent
difference (RPD) calculated as follows:
RPD =
*, - *2
(*, +*2)/ 2
* 100%
where:
R.
r2
result for original sample
result for duplicate sample
An RPD of zero is ideal, meaning the paired measurements are identical. An RPD of 50
percent represents a difference of 40 percent between the smaller and larger measurement based
on the larger measurement. For example, a pair of measurements of 6 and 10 would have an RPD
of 50 percent. Figure 2-4 shows the level of precision attained for field replicates. The average
RPD was 36 percent, and the median RPD was 27 percent. These results suggest that, on
average, measured results would be expected to fall within ±36 percent of the true PCB
concentration for the sample.
Field duplicate sample pairs were also examined for the reproducibility of the congener
patterns by performing a simple regression on the normalized congener values. Congeners were
2-18
TAMS
-------
normalized to BZ #52, as was done in previous Phase 2 work. An ideal match between the
sample pairs would yield a regression slope of unity (1) and a regression coefficient (R2) of unity.
Figure 2-5 shows these example regressions run on field split pairs. The close agreement between
the field split congener ratios is evident by the low level of scatter in each diagram. Figure 2-6
provides a histogram of the regression slopes for the field pairs. The zero intercepts for all of
these regression pairs fell within a few percent of zero (Figure 2-5), largely dictated by the large
number of non-detect results. Thus the slope of these regressions is the indicator of agreement.
Note that the slopes for two-thirds of the replicate pairs fall within 10 percent of unity, indicating
close agreement of the congener patterns. Also notable are the two outliers. The most distant of
the two outliers is the result of detection limit differences between the replicate pairs. One sample
was roughly four times more concentrated. This led to several congeners being detected in one
sample while not being detected (effectively zero) in the other, which yielded the poor slope.
When the non-detect/detect pairs were excluded, the remaining results yielded a slope close to
unity. The second outlier was characterized as having a few congeners whose ratios were very
disparate while the remaining congeners agreed well. This was typical of all of the poorer fits
only to a lesser degree, indicating that in general, most congener ratios were reproducible to
within a few percent. In general, agreement between congener patterns in replicate pairs was
substantially better than the ability to reproduce absolute mass. This suggests that sediment
heterogeneity in concentration as well as the ability to completely homogenize sediment samples
will probably be the main source of analytical uncertainty for PCB results.
2.4.2 Radionuclide Analysis
The sampling program was roughly split in half between the TI Pool and the areas below
the TI Dam. Seventy-six cores were collected from the TI Pool and 94 from the area below the
TI Dam, yielding 170 cores in all. Each of these cores were analyzed for radionuclides in the top-
most and bottom-most layers (Table 2-3). It is important to note here how radionuclides were
used in the interpretation of the low resolution coring data. Specifically, radionuclides were
examined primarily on an absence/presence basis. The radionuclide l37Cs was first introduced to
the Hudson (and most watersheds in the Northern Hemisphere) in 1954 as atmospheric fallout
with onset of atmospheric weapons testing by the US. As a result of the retention of 137Cs in the
2-19
TAMS
-------
soils of the watershed and subsequent erosion of these soils, all post 1954 sediments in the Hudson
have readily measurable levels of 137Cs. The introduction of PCBs into the Hudson potentially
predates ,37Cs by a few years since PCB manufacturing at the GE facilities began prior to the
appearance of 137Cs. However, as was shown in the analysis of the high resolution cores (TAMS
et al., 1997), little if any detectable levels of PCBs are present below the first appearance of 137Cs
in a core. This is illustrated in the core profiles shown in Figure 2-7. On this basis it is evident
that if l37Cs is present in a Hudson River sample collected below Hudson Falls, the sample must
contain PCBs. Similarly, if 137Cs is not detected in a sample, it is likely to contain little or no
PCBs. When l37Cs was not detected in a core bottom, the core was considered to be "complete"
(i.e., it represents all recent [post-1954] deposition and all substantive PCB deposition). If 137Cs
was present in the bottom layer, the core did not capture the entire PCB inventory and was
labeled "incomplete". In these instances, nearby cores may be applied to estimating what was
missed. In any case, the magnitude of the 137Cs level in the sediment was principally interpreted
internally to the core since the concentration could be affected by several factors including the
fine-grain content of the core, the mixing of 137Cs-bearing and non-bearing sediments within the
segment, and the actual sediment age which make core-to-core comparisons less certain.
Similarly, the7Be results were considered on an absence / presence basis. Presence of7 Be
indicated very recent (six months or less) deposition at the coring site. Absence of 7Be indicated
a non-depositional site at a minimum, as well as the potential for sediment scour.
l37Cs was not detected in the bottom layer of the 120 of the 170 cores collected. These 120
cores, representing 70 percent of the coring effort, were considered complete. This indicates that
the PCB contamination contained in these 120 cores is representative of the entire current PCB
inventory at these sites; i.e., no substantive PCB contamination exists below the depth of core
penetration. In looking at the cores above and below the TI Dam, the success rate for complete
cores was greater above the TI Dam than below it. Sixty-one cores from the TI Pool (roughly 81
percent) included all recent deposition as demarcated by 137Cs. Conversely below the TI Dam,
the rate for successful core collection was only 60 percent, with 137Cs detected in 36 of 94 cores.
The I37Cs and 7Be results for one core in the TI Pool were lost and so this core is excluded from
the above tallies.
2-20
TAMS
-------
The reason for the difference in success between the TI Pool and areas downstream is
unknown, but is at least partially due to differences in sampling approaches. Within the TI Pool,
sites were preferentially selected where previous NYSDEC coring work was successful. In
addition, sites were selected in relatively homogeneous areas based on both sediment and PCB
criteria. In the study areas below the TI Dam, sampling locations were selected to demarcate hot
spots. This included placement of some cores in areas beyond hot spot boundaries where
recoveries were less certain.
The length of the 137Cs-bearing cores was consistently less than those considered complete,
about 5 to 7-inches shorter based on the median core length. This suggests that the l37Cs-bearing
cores were collected in areas more difficult to penetrate, perhaps due to sediment type, or due to
underlying harder substrate, relative to the successful core sites. A comparison of complete and
incomplete cores above and below the TI Dam based on core lengths, layer thickness, and the
sediment types associated with each layer is provided in Table 2-4. Median overall core lengths
for complete cores are relatively comparable, but in both instances the incomplete cores are
distinctly shorter than the complete cores. It was originally thought that this distinction might
result from coarser sediment types for the incomplete cores, making core collection more difficult.
However, as can be seen in Table 2-4, the proportions of silt, fine-sand, and coarse-sand samples
in the complete and incomplete core sets are quite similar. This suggests that core sediment type
does not affect the ability to collect a complete core. From this conclusion, it may be inferred that
other causes, such as a change in the underlying substrate, may be responsible for the collection
of incomplete cores. Other potential causes include loss of the core bottom during retrieval. It
is important to remember here, however, that the majority of cores (70 percent) were complete.
Based on the PCB data discussed below, it is likely that an additional 15 cores may be considered
nearly complete, bringing the fraction of successful cores to 79 percent.
The 7Be results for surface sediments for each core were used to establish the current
deposition condition at each site. Of the 169 cores analyzed for 7Be, 119 cores indicated the
presence of 7Be and, therefore, recent deposition. However, this was not proof of a continuous
depositional site but only deposition within the last six months. This issue is discussed in
subsequent chapters of the report when 1984 and 1994 sediment inventories are compared. It
2-21
TAMS
-------
should also be noted that 50 cores, or more than 30 percent of the core sites, had no 7Be,
indicating that the site was currently non-depositional and potentially undergoing scour.
All coring sites contained n7Cs in the surface layer, indicating that some PCB-
contaminated sediment had accumulated between 1954 and 1994.
2.4.3 Total Organic Carbon and Total Kjeldahl Nitrogen
Total organic carbon and total kjeldahl nitrogen (TOC/TKN) analyses were originally
intended to confirm the scheduled total carbon and total nitrogen (TC/TN) analyses, respectively.
The latter two analyses measure all carbon and all nitrogen while the former analyses measure just
the organic forms. Based on the high resolution sediment cores results, it had been shown that
TC/TN and TOC/TKN yield similar values for Hudson sediment samples, indicating that most
sediment carbon and nitrogen were in organic forms. Nonetheless, the results of the TC/TN and
TOC/TKN from the Low resolution Sediment Coring Program were intended to provide further
confirmation of this finding.
Lacking the TC/TN results, the TOC/TKN results are still useful as measures of sediment
properties. The mean and median TOC values were five and six percent of sediment mass on a
dry weight basis with a range of 0.2 to 11 percent (Table 2-3), which is quite comparable to the
high resolution sediment cores. The results are also typical for organic-rich, fine-grained
sediment.
The TKN results for the sediments had similar values for mean and median at 1,640 and
1,370 ppm, respectively (Table 2-3). The range in TKN was similar in scale to that for TOC, as
might be expected since both measures are tied to organic matter.
The molar ratio of carbon to nitrogen is sometimes used as an indicator of the source of
the organic material in the sediment. Specifically high ratios (>80) are indicative of a wood
cellulose source while low values (about 10) are typical of soil carbon and algal production
(Soderlund and Svensson, 1976). Values in between would be expected in blends of wood
2-22
TAMS
-------
cellulose and soil or algal-based organic materials. Wood cellulose would not normally be an
important carbon source in a river system. However, in the Hudson, because of its history as a
wood processing area, wood cellulose is much more common. In addition, it has been reported
in the literature (e.g., Bopp and Simpson, 1989) that wood chip-bearing sediments tend to contain
enhanced levels of PCBs relative to other sediments. Thus the C/N ratio may serve as a flag for
these layers. The C/N ratio in the limited low resolution coring data set varied over a range of
11 to 82, with a median ratio of 40 (Table 2-3). The results indicate the presence of some woody
material in the organic matter contained in many of the samples since a value around 10 would
be expected in the absence of wood cellulose. This confirms the visual classification which notes
wood chips in 19 of the 27 samples run for TOC and TKN. Unfortunately, no correlation was
seen between the visual presence of wood chips and the C/N ratio. The range and mean C/N ratio
for the 19 samples with wood chips noted was not statistically different from the range and mean
ratio for the seven samples where wood chips were not noted. This may be the result of the small
samples size (26 samples), or of the difficulty in homogenizing wood chips in the sediment
sample. Given the median C/N value of 40, which is well above the expected value of 10 for soil
and algal-based organic matter, these results suggest that wood chips or woody material are
present as part of the organic matter throughout much of the Upper Hudson sediments.
2.4.4 Physical Properties
Grain-Size Distribution
Low resolution sediments were classified by three separate techniques, specifically:
• visual field inspection;
• combined sieve and laser particle analysis (Laser); and
• combined sieve and hydrometer analysis (ASTM).
Results from these techniques are summarized in Tables 2-3 and 2-4. Both Laser and ASTM
techniques were applied to a large subset of the samples collected. Visual field inspections were
performed for every sediment sample. Samples are represented in Table 2-3 based on the largest
2-23
TAMS
-------
grain-size fraction. It should not be inferred from this table that any sample is 100 percent silt,
fine- sand, etc. In fact, most sediment samples were considered to be poorly sorted with a
significant fraction (> 10 percent) of other sediment types. In many instances the largest fraction
represented less than 50 percent of the sample.
Evident in all three data sets is the predominance of samples classified as silt (fines in the
case of the ASTM results). The predominance of this fraction reflects the orientation of the
sampling program, i.e., to obtain cores from areas of substantive PCB contamination, generally
areas of fine-grained sediments. In general, the three methods yield similar results for most
samples. The results of these methods are compared by principal fraction in Figures 2-8 to 2-10.
In Figure 2-8 the results of the visual and Laser classifications are compared for the
shallow sediments only, (i.e., just the top slice of each of 169 cores). The uppermost diagram
shows the coincidence between principle fraction by visual inspection versus that obtained by the
Laser technique. The two lower diagrams represent the distribution of matched samples as
classified by each method. In most instances, the two methods agree on the principal fraction for
samples classified as silt and fine-sand, effectively verifying the subjective visual classification.
When the two methods disagree, it is usually by only one class (i.e., fine-sand by visual inspection
is assigned silt by the Laser technique). In most of these instances, the actual fractions are very
close (e.g., 35 percent silt and 32 percent fine-sand). The coarser materials, i.e., medium- or
coarse-sand and gravel, were not as constant as silt and fine-sand for the two methods. In
particular, the medium-sand as classified by visual inspection could be found in every class by the
Laser method. This is indicative of the poor sorting of the coarse sediments, which made visual
classification more difficult.
In Figure 2-9, the visual inspection results are compared with the ASTM method for
samples (n= 143) from a range of depths and locations, as opposed to the shallow sediment
samples presented in Figure 2-8. Again, the two methods generally agree for silt and fine-sand;
however, the coarser fractions are more problematic. As discussed above, this is attributed to the
poorly sorted nature of the sample materials.
2-24
TAMS
-------
Figure 2-10 compares the results for the Laser and ASTM methods directly for the 69
shallow sediment samples run by both methods. The top diagram shows the agreement of the
principal fractions between the two methods. Although the methods agree for most fines, the
Laser method characterizes more samples as silt than does the ASTM method. This trend is
apparent for all sediment classes, with the Laser method tending to characterize more samples into
smaller fractions than the ASTM method. The lower half of Figure 2-10 is a histogram of the
percent similarity calculated for each Laser-ASTM measurement pair. Percent similarity is
calculated by summing the smallest value in each of the sediment classes for a pair of
measurements as shown below:
Sediment and Class Fraction
Silt
Fine
Sand
Medium
Sand
Coarse
Sand
Gravel
Laser Analysis of Sample 1
45
28
12
15
0
= 100%
ASTM Analysis of Sample 1
35
32
18
12
3
= 100%
35
28
12
12
0
= 87%
Similarity
The range of percent similarity for this data set is 34 to 98 percent with a mean value of 76
percent. This is quite similar to the work of Shillabeer, et ai, 1992, where a set of 406 sediment
sample pairs was analyzed by both Laser and sieve techniques. A mean percent similarity of 79
and a range of 55 to 97 percent was obtained, with the Laser technique consistently predicting
larger fractions of the finer sediments. This matches the results obtained for the low resolution
coring program quite well. The authors attributed the difference to the way the techniques
measure particles. Essentially the Laser technique reports the particle-size distribution by volume
while the ASTM (sieve) method is sensitive to particle diameter and shape.
Thus, the two methods report different distributions for the same sample. Since the
primary goal of these analyses was to classify sediments in a qualitative sense for potential PCB
contamination, this difference is unlikely to be important. In particular, the Laser results can be
applied directly to the existing Phase 2 database, to expand and confirm the correlations seen
2-25
TAMS
-------
between the side-scan sonar and the confirmatory samples (TAMS et al., 1997). This application
is presented later in this report.
2-26
TAMS
-------
INTERPRETATION OF LOW RESOLUTION
SEDIMENT CORING RESULTS
-------
3. Interpretation Of Low Resolution Sediment
Coring Results
This chapter presents evidence to show how the low resolution coring results build on the
previously collected Phase 2 data. Specifically, the PCB results of the low resolution coring
program are compared and contrasted with the high resolution core (dated sediment core) results.
Correlations among the various low resolution core analyses are also examined. Finally, the low
resolution core results are interpreted along with the side-scan sonar data. As noted in Section 2.4.1,
PCB concentration data from the low resolution coring program are log-normally distributed. For
this reason, most graphical presentations in this chapter utilize a log-scale total PCB axis when
displaying these results. Median and geometric mean values are good measures for the central
tendency of log-normal data and will be used extensively throughout these discussions. Use of the
median and geometric mean to characterize the log-normally distributed PCB data permits the use
of various statistical tests to examine nature of the PCB data and its correlation with various ancillary
parameters such as total organic carbon and bulk density. A summary of parameters obtained in the
low resolution sampling program is provided in Table 3-1. It should be noted, however, that the
geometric mean is not an appropriate value for the calculation of sediment mass. In this instance, the
arithmetic mean is required. Characterization of sediment mass is examined in detail in Chapter 4
of this report.
3.1 Comparison between the PCB Results for the Low Resolution Cores and
the High Resolution Cores
The High Resolution Sediment Coring Program obtained cores from fine-grained sediments
from throughout the Hudson. As part of the interpretation of these results (TAMS et al., 1997),
several important correlations concerning anaerobic dechlorination were found, including the
following:
1. All sediment PCB dechlorination losses could be explained by loss of the outer
chlorine atoms (meta and para positions) from the PCB molecule. No evidence
3-1
TAMS
-------
for the loss of the inner chlorine atom (ortho position) or destruction of the PCB
molecular structure was found. In support of this, it was shown that the molecular
weight of a sample decreased in direct proportion to increases in the molar ratio of
the sum of five congeners, BZ# 1, 4, 8, 10, and 19, to that of the entire sample. This
relationship would only hold if dechlorination was strictly limited to outer chlorines
and if other possible destructive processes were minor.
2. PCB mass loss by dechlorination was limited to 26 percent of the original
deposited mass and that few samples ever approached this limit. In fact, the
mean mass loss was less than 10 percent assuming Aroclor 1242 as the original
mixture.
3. The degree of dechlorination increased with the logarithm of the total PCB
concentration. Thus, samples with PCB concentrations greater than 30 ppm
exhibited various levels of dechlorination, while samples with concentrations of less
than 30 ppm were relatively unaltered. This conclusion was perhaps the most
important of the three.
One of the concerns with the high resolution sediment coring program was the issue of
representativeness for other Hudson sediments. Since the high resolution sediment cores were
obtained from select, high-deposition rate, fine-grained sediment environments, would they be
representative of conditions throughout the Hudson where deposition conditions were not as
favorable? The low resolution coring program was intended to generate a data set of spatially
representative samples from a number of areas of the Upper Hudson. This data set was used to
reexamine the relationships derived from the high resolution sediment cores.
As the first step in the re-examination, the change in the samples' molecular weight relative
to Aroclor 1242 (AMW) and the molar dechlorination product ratio (MDPR) were calculated for the
entire low resolution coring data set. These terms are calculated as follows:
3-2
TAMS
-------
where: AMW is the fractional difference in the mean molecular weight relative to Aroclor
1242;
MWA1242 is the mass-weighted mean molecular weight of Aroclor 1242; and
MW^ is the mass-weighted mean molecular weight of the sample calculated by:
C,
MW~ . ~ Total
Sample
10
i
1 *Yc
E
As j
/ = i
mwi , = ,
where: I is the homologue group number froml to 10;
mw, is the molecular weight of homologue group I in g/mole;
n, is the number of measured congeners in homologue group /;
Cj is the concentration of congener j in /Ug/kg; and
Ctoiai is ^e total concentration in the sample in /^g/kg.
E wzm
MDPR ~ ' " '• 4> 8> l0, 19
(3-2)
126 (3"3)
E
7=1
where: [BZ#I] and [BZ#j] are the molar concentrations of congeners I and j, respectively in
the sample (mole/kg); and 126 is the number of congeners for which consistent
reliable identification and quantitative data were generated in the Phase 2 analytical
program.
The term AMW equals zero for Aroclor 1242, reported to be the main PCB mixture released
to the Hudson (Brown et al., 1984). The AMW is positive for a decrease in molecular weight in the
sample. If a sample is limited to dechlorination by loss of outer chlorines, the maximum value for
AMW is 0.223, corresponding to a mass loss of 26 percent. AMW can have negative values if
heavier Aroclor mixtures are present or if lighter congeners are lost from the mixture.
3-3
TAMS
-------
The MDPR equals 0.14 for a fresh Aroclor 1242 mixture. The maximum value is unity,
assuming complete conversion of a mixture to BZ# 1,4, 8, 10, 19 via dechlorination. The MDPR
can never be negative by definition, since it is the ratio of two sums.
These terms were calculated for all low resolution sediment core samples with a total PCB
concentration greater than 100 //g/kg. The concentration of 100 /^g/kg was select'ed as a lower
bound to avoid quantitation uncertainties associated with the lighter congeners that can occur at low
concentrations. Table 3-2 provides a summary of the range, median and mean values for AMW,
MDPR, total PCBs, and an estimated mass loss based on the AMW value. Immediately evident from
the table is the consistency of the maximum level of dechlorination with the high resolution core
results (i.e., no sample attains the theoretical limit values for AMW of 0.223 or for MDPR of 1.0,
and few samples even get close). The average level of dechlorination as measured by AMW is 0.1,
corresponding to a mass loss of 12 percent. Considering that the low resolution coring sites were
focused on the hot spots of the Upper Hudson, this result is very consistent with the DEIR
conclusion that the average level of dechlorination throughout the freshwater Hudson is less than 10
percent.
It should be noted that the degree of dechlorination observed (AMW = 0.10) for the median
sediment concentration of 19 mg/kg is noticeably higher than would be expected from the results
reported in the DEIR. Based on the high resolution core analyses, a sediment concentration of 19
mg/kg would have an expected dechlorination level AMW of only 0.05, corresponding to a six
percent mass loss. The reasons for the difference between the observed and expected dechlorination
levels are discussed at length later in this chapter, as part of the analysis of AMW, MDPR, and total
PCB concentration.
Before exploring the relationship between AMW, MDPR, and total PCBs (log transformed
length-averaged) for the low resolution cores, it is first important to examine the relationship
between AMW and MDPR for these cores. This relationship is shown in Figure 3-1, which presents
the regression line determined by the low resolution coring results, as well as the high resolution
core regression line and the theoretical relationship. There is a close, but not exact, reproduction of
3-4
TAMS
-------
the high resolution core regression by low resolution core data for AMW versus MDPR. Although
the difference between the two regression curves is minor, a test for significance using Chow's F test
(Fisher, 1970) shows the coefficients of the two curves to be statistically different. A second
statistical test, Theil's U statistic, indicated that the difference was minor, and likely due to
differences in variance between the high resolution core and low resolution core results. Based on
these statistical tests, both curves are sufficiently close to the theoretical relationship that they serve
to support the original premise, i.e., that dechlorination is limited to outer chlorines (meta- and para-
dechlorination) and that little, if any, in situ destruction of PCB molecules is occurring. Definitions
for Chow's F test and Theil's U statistic can be found in the glossary as well as in the references (
Fisher, 1970 and Theil, 1996).
The plot of AMW versus MDPR confirms the first two high resolution core conclusions
discussed earlier. Confirmation of the last conclusion from the high resolution cores that the degree
of dechlorination increases with the total PCB concentration by the low resolution core results was
more problematic. As part of the analysis, the low resolution core results were plotted as MDPR and
AMW vs total PCB concentration in the same way as the high resolution cores. Figure 3-2 presents
the results for the low resolution cores as well as the regression line and confidence limits for the
high resolution cores. Note that the confidence limits represent the 95 percent confidence limit of
the individual data points and not the regression line itself. There is a distinct left shift in the low
resolution core results relative to the high resolution core results and greatly increased data scatter.
Figure 3-2 also shows the layering information for the low resolution cores, with separate
symbols for the top, second and third slices. With this information displayed, it is apparent that the
majority of the scatter away from the high resolution core regression line stems from the deepest
slices. The separation of these points becomes even clearer when the presence of 137Cs in the
bottom-most layer is considered. Deep layers from cores considered complete (no l37Cs in the
bottom layer) are almost exclusively responsible for the scatter away from the high resolution core
regression line and confidence limits, as shown in Figure 3-3.
Although this was an interesting correlation, the underlying reason for this was less clear.
The basic issue with these samples was their extensive degree of dechlorination given their very low
3-5
TAMS
-------
sediment PCB concentration. In addition, these points were unique to those cores where the
maximum PCB concentration was usually in the surface layer and where no l37Cs was detected at
depth. Cores with deeper PCB maxima and 137Cs present generally fell within the high resolution
core regression and confidence limit domain. When congener patterns of the deeper layers were
compared with those from the top layers, it was found that in many instances, the deeper layer
pattern closely matched that of the surface layer. The results for the selected core samples are very
similar to those of all core segments, as illustrated in Figure 3-4 which shows three examples of the
match between the surface and deeper layers. In each of the diagrams, a perfect match would have
both a slope and R2 of unity (one). The values obtained are somewhat poorer than the values
obtained for field split analyses, as discussed in Section 2.3.2. However, when they are compared
to other top and deeper segment pairs which fall close to the high resolution core regression, the
slope and R2 are distinctly higher for the cores with no l37Cs present at depth. (See Figures 3-4 and
3-5).
Given the strength of the trend observed for high resolution cores and that the low resolution
cores generally followed or paralleled this trend when deeper layers from complete cores were
excluded, it seemed clear that the scattered points must result from a different cause and that the
relationship of increasing dechlorination with increasing PCB concentration was valid for results
from both coring programs. The simplest and most likely explanation of the widely scattered points
given the absence of 137Cs in the bottom-most layer is cross-contamination of deeper sediments by
overlying ones. Cross-contamination is of greatest concern when layers originally differed by
several orders of magnitude. Incorporation of as little as a few tenths of a percent of overlying
material would serve to create samples with low PCB concentrations that exhibited much higher
degrees of dechlorination. The level of dechlorination in the PCBs in the deeper layer would be
expected to match that of the more contaminated upper layers since any cross-contamination process
would largely serve to dilute the mixture and not cause any change in the congener ratios. The terms
AMW and MDPR represent various PCB congener ratios and thus would also be unaffected by
dilution.
The sensitivity of the results to cross-contamination stems in part from the way ,37Cs and
PCBs are handled. 137Cs measurements are performed directly on samples without any concentration
3-6
TAMS
-------
or dilution steps. This technique is sufficient to discern recent deposition from pre-1954 sediment.
Cross-contamination is not an issue since 137Cs levels are generally measured over only a range of
one to two orders of magnitude. Thus, a minor cross-contamination event (one percent of overlying
material mixed into a deeper layer) will not be detected and a pre-1954 sediment will be correctly
identified. On the other hand, PCB measurements are reported over nearly five orders of magnitude
via laboratory dilution of sample concentrations. Thus a deeper sediment layer incorporating 0.1
percent by mass of an overlying layer at 500 mg/kg (ppm) yields a layer at 0.5 mg/kg (ppm), well
within the range of PCB measurements. This layer would exhibit AMW and MDPR values
commensurate with the level of dechlorination expected for a 500 mg/kg (ppm) sample and not a 0.5
mg/kg (ppm) sample.
The potential to cross-contaminate low resolution samples relative to high resolution samples
stems from the way in which cores were collected and processed. As mentioned in Chapter 2, low
resolution cores were collected by a "vibra-coring" method wherein coring tubes were driven while
being vibrated, enabled greater penetration of the sediments. This technique serves to partially
liquefy the sediments around the tube walls, potentially permitting mixing between sediment layers.
Once collected, low resolution core segments were rigorously homogenized so that the section was
well represented in any subsample, such as that collected for PCB analysis. Cross-contamination
of the magnitude of one percent would have little impact on the major goal of the program, i.e., to
establish current sediment inventories, since the amount of PCB represented in the deeper cross-
contaminated layer would be roughly one percent of the total, well below the uncertainty of the PCB
mass determination of the upper layers (26 to 36 percent based on the field split RPD). However,
in the determination of the relationship of PCB mass and dechlorination, it has the potential to add
significant variability to the data which is unrelated to the dechlorination process.
The issue of cross-contamination was avoided in the high resolution cores in several ways.
Push coring, rather than "vibra-coring" was used. Thus less energy was available to mix adjacent
layers. Additionally, instead of homogenizing each layer prior to subsampling, samples for PCB
analysis were obtained by collecting a three-quarter-inch diameter mini-core from the center of each
core slice. Because of the small size of the original 0.8 to 1.6-inch (2 to 4-cm) slice, subsampling
in this manner integrated the slice thickness while avoiding any sediment layer mixing which may
3-7
TAMS
-------
have occurred along the coring tube walls. Figure 3-6 illustrates the difference between the low
resolution and high resolution subsampling processes.
To avoid the cross-contamination problem, it was decided to include only a portion of the
low resolution coring results. Based on the measured presence of PCBs in all top core segments and
the tendency for the PCB maximum to occur in this layer, all top segments were kept, regardless of
the l37Cs result for the bottom segment. This added 170 samples, one for each core. In addition, all
core segments from cores with 137Cs present in the bottom segment (incomplete cores as defined in
Section 2.4.2) were included, which added 43 more samples. Lastly, the segment of maximum
concentration in a core, if it was not the top layer, was added. This added 24 more samples, bringing
the total to 237 out of a total of 371 core segments, or 64 percent of the data set. The remaining core
layers were deemed to have too much potential for cross-contamination to be included here. All of
the excluded segments failed the criteria given above. These samples met all of the following
conditions:
• The segment was not the top-most segment in the core;
• The segment did not contain the maximum concentration for the core;
• 137Cs was not detected in the bottom layer of the core, which indicated that the core
was complete and had greater potential for cross-contamination since PCBs may
not have been present in the bottom layer ; and
• The maximum PCB concentration for the core was found in a shallower segment,
increasing the potential for cross-contamination of the lower layers as the coring
tube was pushed down through the mud.
These criteria were sufficient to remove the vast majority of the widely scattered points.
However, a few apparent outliers to the AMW and MDPR vs total PCB regressions remained in
the data set. Outliers were selected using the Mahalanobis distances (Mahalanobis, 1930) which
are calculated for each point based on the mean, standard deviation and correlation for the data.
All points that were excluded are points marked with an "X" on Figure 3-7. (A definition of
Mahalanobis distances is given in the glossary.) Exclusion of these points yielded the final data
set of 229 points, as shown in Figure 3-8, with the statistics for the final data set provided in the
3-8
TAMS
-------
lower half of Table 3-2 (selected core segments). These results were compared with those in the
upper half of the table (all core segments) to see the impact of removing the potentially cross-
contaminated samples. The removal made almost no impact on the AMW and MDPR, based on
their mean and median values. The AMW of the selected core segments had a geometric mean
AMW of 0.101, identical to that of the entire data set, with a median of 0.098 and a range of -
0.106 to 0.195. This is notable since the mean total PCB concentration of the points doubled from
15.3 mg/kg to 31.8 mg/kg as a result of the selection process. This result would be expected given
the removal from the data set of many relatively low level PCB samples whose congener patterns
(and therefore AMW and MDPR) matched those of the overlying, more contaminated core
segments. That is, this would be expected if cross-contaminated samples are removed from the
data set and not simply low level contaminated sediments which would presumably conform with
the high resolution AMW to total PCB relationship. The mean AMW for the selected points
corresponds to a mass loss of 12 percent, with no sample exceeding the theoretical limit AMW
of 0.223, derived in Table 4-8 of TAMS et al. (1997).
Removal of the potentially cross-contaminated samples from the data set examined here
improves the general trend greatly, but still does not yield the regression line determined from the
high resolution core data. Specifically, although the points tend to fall within the 95 percent
confidence interval for individual measurements as defined by the high resolution cores, the low
resolution core data are still shifted to the left of the regression line. In fact, a regression line fit
to the low resolution core results was shown to be statistically different from the high resolution
core regression (Figure 3-9), indicating a different relationship between dechlorination and total
PCB mass in the two data sets. Chow's F test showed a greater than 99.99 percent probability
that the coefficients of the two trends were different. This difference was primarily due to the
intercept; slopes were similar (Butcher, 1998b). The difference between the two regression lines
depends on whether AMW or MDPR is used. At the mean conditions for the selected low
resolution core segments (AMW = 0.10 and MDPR = 0.54), the high resolution core regressions
yielded a total PCB value of 107 mg/kg, 3.5 times higher than the mean low resolution core
condition. This suggests that the low resolution core sediments have dechlorinated as if their
concentrations were 3.5 times higher than the measured values.
3-9
TAMS
-------
This result was again disconcerting given that the two programs were measuring the same
media in approximately the same manner spanning a period of less than two years. Further
exploration of the PCB data and the sampling techniques was again warranted. As it turned out,
the explanation again lay in the difference in sampling techniques.
For the high resolution cores, the 0.8 to 1.6-inch (2 to 4-cm) slicing intervals appear to
have been small enough to capture one to five years of deposition. If we examine a typical high
resolution core, we can see that changes in sediment PCB concentrations appear to be well
captured by these sampling intervals. Figures 3-10 and 3-11 illustrate this point further by
examining two individual high resolution cores. Based on the measured changes between adjacent
core layers, it is evident that most high resolution core layers ought to be relatively homogeneous
in concentration throughout their thickness. The only exceptions to this are the points around the
PCB maxima where change is rapid and a given layer might span a concentration range of three.
Conversely, the low resolution core slice or segment is nominally nine-inches thick. Given that
the high resolution cores record the range of sediment concentrations deposited in a given area,
it is likely that a low resolution core segment would span a range of two or more orders of
magnitude, based on the PCB ranges measured in the high resolution cores over a nine-inch
interval. If we then apply the total PCB versus AMW and MDPR relationships derived from the
high resolution cores, it is apparent that the PCBs contained in a low resolution core segment
would have seen a broad range of dechlorination conditions. However, the process of collecting
and homogenizing the low resolution core sample would serve to effectively dilute the
concentrated and dechlorinated layers with the less concentrated, relatively unaltered layers, found
above and below the PCB maximum. The dilution would modify the measured concentration but
have little effect on the AMW and MDPR since these parameters represent congener ratios, which
tend to be unaffected by dilution. The congener ratios would be unaffected because of the large
PCB mass represented by the peak concentration layers. The collection and homogenization
process would effectively yield a relatively dilute sample with the level of dechlorination normally
found in a concentrated layer as measured by the AMW or MDPR. In Figure 3-9, this would
serve to shift the AMW and MDPR relationship with total PCB to the left relative of the high
resolution regression, just as it is plotted. The extent of the shift would depend upon the range
of concentrations contained within a given low resolution core segment; the greater the range, the
3-10 TAMS
300079
-------
greater the shift. Presumably, the shift would be greatest for cores where sediment deposition was
undisturbed by biological processes, yielding a range of values akin to those seen in the high
resolution cores. Conversely, vertical mixing by biological or other processes would minimize
this effect. A similar effect would be obtained for a core segment which contained portions of
pre- and post-1954 sediment, wherein the uncontaminated layers would serve to dilute the
overlying contaminated ones without changing the AMW and MDPR values.
To further illustrate this point, we can examine the two high resolution cores presented in
Figures 3-10 and 3-11, that differ by about an order of magnitude. Assuming that the cores in
the figure were each represented by a single low resolution core segment, the length-weighted
average concentration of the full core (that which would be obtained by a thorough
homogenization of the core) of Core 19 (Figure 3-10) would be 512 mg/kg (ppm) with a AMW
of 0.1871 and an MDPR of 0.8828. This is to be contrasted against the conditions of the high
resolution core PCB maximum concentration layer of 2,083 mg/kg (ppm) with a AMW of 0.2074
and an MDPR of 0.9156. Calculating the MDPR and AMW from the original high resolution
core regression line (MDPR = -0.714 +0.248 log [Total PCBs] and AMW= -0.2514- 0.070 log
[Total PCBs]) using the homogenized concentration of 512 mg/kg, we obtain an MDPR of -
0.0421 and a AMW of -0.0614, substantially lower than the length-weighted average values.
Based on the high resolution core regression line, the AMW value obtained for the calculated low
resolution core concentration corresponds to a concentration of 1,814 mg/kg (ppm) or about 3.5
times higher than the calculated concentration. Note that this increase is in accordance with the
difference between the high resolution core and the low resolution core regressions shown in
Figure 3-9.
For High Resolution Core 21 (Figure 3-11) the length-weighted average would be 66
mg/kg (ppm) with a AMW of 0.1639 and an MDPR of 0.7325. This is to be contrasted against
the conditions of the PCB maximum concentration layer of 260 mg/kg (ppm) with a AMW of
0.1945 and an MDPR of 0.8682. Calculating the MDPR and AMW from the original high
resolution core regression line using the homogenized concentration of 66 mg/kg, we obtain an
MDPR of -0.2628 and a AMW of -0.1236, substantially lower than the length-weighted average
values. Based on the high resolution core regression line, the AMW value obtained for the
3-11
TAMS
-------
calculated low resolution core corresponds to a total PCB concentration of 845 mg/kg or about
13 times higher than the calculated concentration.
In fact, this calculation was done for all the high resolution cores of the Upper Hudson.
The results are illustrated in Figure 3-12. In this diagram, it is evident that if the high resolution
cores had been collected as low resolution cores, the resulting relationship between the degree of
dechlorination as measured by AMW or MDPR and the total PCB concentration would be shifted
to the left by a factor of five- to eight-fold relative to the high resolution sediment core regression
line. That is, concentrations would be five to eight times lower to achieve a given level of
dechlorination. Based on the relationships determined for the high resolution and low resolution
coring programs, it is evident that any coring study which fails to consider the degree of vertical
variability within the sediment will overestimate the extent of dechlorination that can be
anticipated for a given sediment concentration.
The impact of vertical homogenization has different impacts depending upon the
concentration range in the sediments being collected. Areas of high sediment variability and high
sediment inventory, such as hot spot areas, will yield a greater degree of dechlorination than
would be predicted from the measured concentration over a large vertical interval. This is
because the most concentrated layers with correspondingly high levels of dechlorination will be
diluted by the sampling process while the internal measures of dechlorination (MDPR and AMW)
will not be affected. Conversely, in sediments of relatively low contamination, such as those
found in the extensive areas of coarse sediments, the sampling process will have little effect and
these samples will match the trend obtained from the high resolution cores. This is evident in the
extent of scatter in Figure 3-9. Specifically, the range of the low resolution core results includes
the mean trend (the regression line) from the high resolution cores. In addition, the scatter
associated with the low resolution cores (R2 = 0.65 for AMW and total PCBs) is greater than that
for the high resolution cores (R2 = 0.73), as would be expected due to sample homogenization.
In summary, the low resolution core results are consistent with the conclusions drawn from
the high resolution cores concerning dechlorination and PCB mass. The low resolution cores
closely replicate the relationship between AMW and MDPR, confirming the occurrence of meta-
3-12
TAMS
-------
and para-dechlorination and the absence of ortho-dechlorination. Due to several issues concerning
low resolution sediment core collection, specifically vertical homogenization and potential cross-
contamination, the relationship between dechlorination and total PCB mass appears somewhat
different in the low resolution core results. However, when these issues are factored in to the
examination of the data, the conclusions drawn from the high resolution cores concerning
dechlorination are confirmed by the low resolution core results.
3.2 Interpretation of the Relationships Among the Low Resolution Core
Parameters
This section describes the relationships found among total PCBs and other measured
parameters. The main purpose of this examination is to assess which parameters may be useful
in predicting sediment PCB inventories. In addition, the examination will assess the degree to
which PCBs found in Hudson River sediments conform to general expectations of PCB behavior
in the environment. The analyses in this section are expected to show limited relationships because
PCBs, as well as other parameters, most likely vary over narrower depth segments than sampled
in the low resolution coring program.
The type and number of parameters collected during the low resolution coring program
are outlined in Chapter 2. The parameters themselves can be classified according to type as
shown in Table 3-1. Nearly all possible parameter pair relationships were examined for this
report. The notable exceptions are the congener-specific and homologue data which were
excluded from this analysis. This exclusion was made so as to maintain the major focus of this
report, i.e., an examination of sediment PCB inventories warranting use of total PCB
concentrations. PCB congener pattern matching or 'fingerprinting' was discussed at length in the
previous Phase 2 DEIR (TAMS et al., 1997) and is not covered here.
The relationships among most of the parameters in Table 3-1 were initially examined on
a regression-basis, that is, regression plots and regression coefficients were generated for the
various parameter pairs. These results aided in the selection of parameter pairs for further
3-13
TAMS
-------
exploration. In addition, several parameters were examined based on their expected influences
on PCB concentration.
Using selected samples based on criteria described in Section 3.1, the total PCB
concentration more than doubled to 31.8 mg/kg (Table 3-2). This increase resulted from focusing
on hot spots and more contaminated sediments. Despite removing lowest level samples, which
resulted in an increase in the mean PCB concentration, the 12 percent mass loss by dechlorination
remained constant because cross-contaminated samples were scattered over the range of AMW.
Cross-contamination occurred when recent deposition was shallow (0 to 12-inches) and deeper
slices were obtained, independent of the level of PCB contamination in the sediment itself.
The results of the initial regression analyses are presented in Tables 3-3 to 3-8. The first
table represents a correlation matrix for the entire suite of Laser grain-size distribution
parameters. This large table demonstrates the strong correlations among many of these
parameters, as would be expected since many of the parameters represent similar properties. For
example, silt % is strongly correlated with the phi 4.5 to 7 (r greater than 0.8). These phi
fractions represent sediments in the silt classification (diameters 44 to 7 nm, respectively).
Similarly silt % is inversely correlated with fine sand % as might be expected since as the fraction
of silt % increases, the fraction of other classes would be expected to decrease. The strong
correlations among many of these parameters indicates that only a limited suite of them are
required to represent the grain-size characteristics. The next four tables summarize the
relationships between three PCB measures (total PCB concentration, change in molecular weight
relative to Aroclor 1242 (AMW), and MDPR) and Laser grain-size distribution parameters,
ASTM grain-size distribution parameters, radionuclide parameters, and bulk sediment properties
(Tables 3-4, 3-5, 3-6, and 3-7, respectively). The number of samples represented in a table is
equal to the number of samples that remained in each analysis after outliers were excluded using
a Mahalanobis analysis (Mahalanobis, 1930). For example, in Table 3-4, the 170 top core
segments were examined for grain-size distribution parameters, using Laser grain-size
methodology, and for PCBs, but only 136 to 149 of the samples were considered acceptable.
Similarly, of the 143 core segments taken from various depths and run for ASTM (sieve) analysis
between 122 to 130 were considered to be within acceptable ranges (Table 3-5). In each table,
3-14
TAMS
-------
regression coefficients for each parameter vs total PCBs, AMW, and MDPR are presented. In
general, parameters describing the fine-grained properties of the sediment showed the strongest
correlations with total PCBs although few exhibited strong predictive power (r greater than 0.5).
The notable exceptions were the sediment density-related properties which had r values around
0.6 for shallow sediments, as shown in Table 3-7. Also notable is the generally stronger
correlations seen between total PCBs and the Laser analysis parameters relative to those for the
ASTM analyses. This was attributed to the fact that the Laser analyses were performed on the
shallow sediments exclusively, where the majority of the sediment PCB inventory was found. The
parameters AMW and MDPR are included in each table to examine the correlation between the
extent of dechlorination and parameters other than total PCB concentration. The relationship
between total PCB concentration and the degree of dechlorination as measured by AMW and
MDPR was discussed extensively in the previous section.
In the last regression result table (Table 3-8), the more highly correlative variables from
the previous four tables are compared with the two integrated PCB parameters. Specifically, PCB
mass per unit area and the core length-weighted average PCB concentration are compared with
bulk density, percent solids, percent silt, and 137Cs. In these tables, the non-PCB parameters
represent the properties of the shallow (0 to 9-inch) or surficial (0 to 1-inch) sediment layers.
Evident in Tables 3-4 to 3-8 is the generally weak correlation among the PCB
measurements and the other parameters. Regression coefficients are rarely higher than 0.5,
indicating only weak predictive power. The regression of the two integrated variables provided
the highest level of correlation (Table 3-8). In action, PCB inventories tended to be greater
towards the surface, thereby resulting in higher correlation coefficients. Percent silt had the
strongest correlation (0.35) of the major soil classifications, but the mean phi regression
coefficient was also 0.35, with phi 4.5 to phi 6.5 (i.e., samples classified as silts) having the
highest correlations (Table 3-4). None of the D(x) parameters were particularly strong although
the regression coefficients were comparable to the other grain size parameters (Table 3-4). The
D(x) parameters represent the effective diameter in millimeters of a sieve which would retain "x"
percentage of the sample mass. Thus, the d(50) is the approximate median particle diameter and
the d(15) and d(90) represent the coarsest and finest fractions of the sample. Additional
3-15
TAMS
-------
discussions on the relationship between the two different grain size distribution methods used in
the Low Resolution Coring program are presented later in this section. In the discussions and
figures to follow, the median phi (the median diameter on the phi scale) is used to represent the
class of Laser-based parameters.
Confirmation of Regression Analyses
After completing the initial regression analysis, the better regression pairs were examined
in more detail, as discussed below. The comparisons were made by grouping or binning the non-
PCB parameters into equally spaced bins, generally with a sufficient number of values (at least
eight if at all possible) to permit the accurate calculation of a mean and standard deviation. The
mean PCB concentration in logspace for each of the bins were then compared to each other using
a Tukey-Kramer honestly significant difference calculation (Box et al., 1978) at the 95 percent
confidence level to test for differences among the bins and confirm the regression analysis. This
test permitted identification of those trends of PCBs and parameters which were statistically
significant. Log-transformed total PCB concentration data were used in these analyses since the
PCB data were found to be log-normally distributed, as discussed in Chapter 2.
Outliers to the regression analyses were examined when apparent, based on regression
plots of each parameter pair {e.g., total PCB and bulk density). A Mahalanobis test was
performed to identify the outliers which were eliminated from the regression analysis. The
regression results presented in the tables and figures of Section 3-2 do not include the identified
outliers.
The data presented in the figures of this section are represented in a "box and whisker"
format. Essentially, the data for PCBs are binned according to the variable on the x-axis and then
represented by a series of box-and-whisker plots. Each box encloses 50 percent of the data with
the median value of the variable displayed as a line. The top and bottom of the box mark the
limits of the central 50 percent of the variable population. The lines extending from the top and
bottom of each box mark the minimum and maximum values that fall within an acceptable range.
3-16
TAMS
-------
Any value outside of this range, called an outlier, is displayed as an individual point. The box
diagram describes the following statistics for the data set:
Median The data value located halfway between the smallest and
largest values.
Upper Quartile (UQ) The data value located halfway between the median and the
highest data value (top of box).
Lower Quartile (LQ) The data value located halfway between the median and the
lowest data value (bottom of box).
Interquartile Distance (IQD) The distance between the Upper and Lower Quartiles
(UQ - LQ).
Range The minimum and maximum values of the data set that fall
within UQ + 1.5 * IQD and LQ-1.5 * IQD
Outliers Points whose value is either greater than UQ + 1.5 * IQD
or less than LQ - 1.5 * IQD
Note that the identification of outliers by the box-and-whisker plot is not the basis for the
exclusion of points from the regression analysis. Only those points identified by the Mahalanobis
test mentioned above are excluded from the regression analysis. The plots presented in this
section include all points, including the outliers.
For the comparisons of the various parameters with PCB concentration, the vertical axis,
representing total PCB concentration, is always in log-scale. This is based on the analysis
presented previously which showed the PCB concentrations to be log-normally distributed. Thus,
the correlations are all made to the log of the PCB concentration since this is a relatively constant
parameter to which the standard statistical assumptions concerning normality can be applied.
3-17
TAMS
-------
In addition to the box-and-whisker diagrams for each data bin or group, the data bin
median and arithmetic mean are shown. The median corresponds closely to the geometric mean
since the PCB data are log-normally distributed. The arithmetic average is calculated on the
actual PCB concentrations and not on the log-transformed data.
Correlation with Bulk Sediment Properties
Figures 3-13 to 3-18 and Table 3-7 show various comparisons between the PCB measures
and the bulk sediment properties. Total PCB concentrations were found to generally decline as
the bulk density, percent solids, solid specific weight and particle density increase. This would
be expected assuming that the organic carbon content declines as these parameters increase. This
was confirmed for bulk density and percent solids for both shallow sediment samples and all
sediment samples. Most bins from the sediment samples were found to have statistically different
mean values from at least half of the other bins in the diagram. In general, the relationship to
total PCBs was found to be stronger in the shallow sediments. This is attributed to the constant
presence of PCB in this layer. When the deeper layers are included, layers with similar bulk
properties but little PCB mass are added (typically the deepest layers), thus adding scatter to the
data and weakening the correlations.
For solid specific weight (the product of bulk density and percent solids), the trend was
much weaker and typically only the end bins were statistically different from the rest of the set.
Similarly, particle density did not show any statistical significance with respect to total PCB
concentrations based on the binned data. This was somewhat unexpected but is probably the result
of the calculation of particle density which involves both bulk density and percent solids. It is
likely that the combined uncertainty of these two parameters as particle density creates too much
total uncertainty and prevents the detection of a statistically significant trend.
In Figures 3-17a and 3-17b, bulk density is compared with the measures for dechlorination
(AMW and MDPR) for all sediments and shallow sediments, respectively. The AMW and MDPR
both decline with increasing bulk density. The correlation with shallow sediments and measures
of dechlorinations is greater than the correlation with all sediments and is attributable to the
3-18
TAMS
-------
constant presence of PCBs in the shallow layer. Regression coefficients for these parameters are
shown in Table 3-7. This is the same trend found for total PCB mass. However, given the
weakness of the regressions with bulk density, it is likely that the trend is the result of the strong
correlation among total PCBs, AMW and MDPR, and not related to any change in bulk density.
In summary, these results show the dependence of total PCB concentration on the presence
of lighter material in the sediment {e.g., Table 3-6, Regression Coefficients). Presumably this
lighter material represents organic matter. The correlation of AMW and MDPR values is largely
a ramification of the correlation with total PCB. None of the bulk sediment parameters are better
predictors of the extent of dechlorination than total PCB concentration. Among the four bulk
sediment parameters, percent solids had the strongest correlation with total PCBs.
Correlation with Grain Size Distribution
Three different measures were available to describe the grain-size distribution of the
sediment samples: Laser analysis, ASTM (sieve) analysis, and the geologist's visual classification.
In general, the parameters generated by these techniques exhibited weaker correlations for the
laser analysis (Table 3-4) and the ASTM analysis (Table 3-5) than the bulk sediment parameters
(Table 3-7) described above. All three techniques generated a principle grain-size fraction
parameter. The comparison among these results for principle fraction was presented in Section
2.4.3. In comparing the principle fraction to total PCB content, each technique produced a
consistent finding that samples classified with silt (or fines) as the principle fraction had a
substantively higher PCB content relative to the ot^r sediment classes.
Using the geologist's classification and the Laser technique results, the silt group was
statistically higher than all other groups, with approximately a four-fold higher PCB
concentration. The coarser sediment classes, fine-sands and coarse sand/gravel, were not
statistically different from one another, based on the Tukey-Kramer honestly significant difference
at the 95 percent confidence level. Figure 3-18 shows the relationship between principle fraction
by geologist's visual inspection versus total PCB concentration. Diagrams are shown for all
3-19
TAMS
-------
sediments and for shallow sediments. The diagram for all sediments represents all 371
measurements since all samples analyzed for PCBs were also classified by the field geologist.
The PCB results support the widely held theory that PCBs tend to associate with fine-
grained sediments. Also notable in Figure 3-18 are the distinctly low values for clay samples.
These samples were identified as clay by the geologist, although subsequent grain-size analysis
later classified them as silts. However, the nature of these "clay" samples was distinct from fine-
grained sediments found elsewhere in the river. Specifically, the clay samples were distinctly
gray in color with visible banding typically associated with varves. These samples were believed
to represent lake bottom deposits from a glacial-age lake resulting in their low level of PCB
contamination and failure to conform with the expected trend of fine-grained sediment and higher
PCB content.
The ASTM grain-size distribution data, while confirming the general finding of higher
PCB levels in samples whose principle fraction was silt, did not yield as much statistical
separation among the various sediment classes as the two other methods. In addition, as a whole,
the ASTM samples were less contaminated than the shallow sediment Laser samples (geometric
means for total PCBs of 7.0 and 23.4 ppm, respectively). The lessened statistical significance and
the lower average PCB level were attributed to the fact that the ASTM sample data set consisted
of many samples collected at depth (second and third core segments) where PCB concentrations
were typically lower, regardless of sediment type. Presumably, a portion of these sediments
would include PCB-free material deposited prior to 1950 which would exhibit no correlation
between PCB content and sediment type.
Since the silt classification proved to be a statistically significant predictor for PCB, it was
explored further on a quantitative basis as percent silt. This analysis could only be done for the
Laser and ASTM technique results which had quantitative results. Figure 3-19 shows the results
for percent silt vs total PCB concentration for the Laser analysis. This figure represents the data
from all 170 top segments (shallow sediments). The Laser results were chosen for this
presentation because they represent shallow sediments exclusively where most of the PCB mass
is found. Sorting the percent silt data into three groups yielded three statistically different means
3-20
TAMS
-------
showing the predictive power of this parameter. The results show a relatively steady increase to
higher PCB levels with higher silt content. The sediments with silt fractions greater than 66
percent were on average five times higher in PCB content than sediments less than 33 percent silt.
The last parameter to be examined under grain-size distribution is the mean phi (4>). The
parameter phi is a measure of the particle diameter calculated as follows:
<|> = - log2 (diameter in mm) (3-4)
Thus, each unit of phi represents a halving of the particle diameter (e.g., 8 phi equals
0.0039 mm, which equals half of 7 phi at 0.0078 mm). Mean phi is a measure of the center of
the sample's grain-size distribution, unlike percent silt which is more a measure of the sample
mass than a measure of specific diameter. It should be noted that phi is a negative log value and
thus as the phi value increases, the particle diameter decreases. Figure 3-20 groups total PCBs
by mean phi and exhibits a trend to higher PCB levels with higher phi. However, only the
rightmost bin, representing sediment samples whose mean diameter is less than 4.5 phi or 0.044
mm (44 //m; i.e. samples classified as silt), is statistically significant from the others.
In summary the grain-size distribution data showed the expected correlation of PCBs and
fine-grained sediments but little more. The parameters generally showed a slightly weaker
correlation (absolute value of r, |r|, was between 0.41 and 0.47 for phi 4.5 to phi 6) with total
PCBs than the bulk sediment properties (| r | = 0.50 to 0.58). Similarly, the correlations between
AMW, MDPR and the grain-size parameters (|r| = 0.36 to 0.43) were lower than those for
AMW, MDPR and the bulk sediment properties (|r| for MDPR = 0.40 to 0.54) for shallow
sediments. It would appear that correlations between total PCB concentration and both the grain-
size distribution data and the bulk property data are concurrently tracking the fine-grained, rich-in-
organic matter silt content of the sediment and little else.
3-21
TAMS
-------
Chemical and Radionuclide Parameters
Six parameters were examined under this heading. Three, consisting of total organic
carbon (TOC), total kjeldahl nitrogen (TKN), and the carbon-to-nitrogen molar ratio (C/N), were
conducted on a very limited data set (27 samples) and thus are limited with regard to correlation
analysis. The C/N ratio is the molar ratio of the TOC and TKN analyses. As noted in Section
2.3.3, TOC and TKN were obtained from a very limited sample set. The samples were randomly
selected from the set of core segment samples and were originally intended to merely confirm the
results of the total carbon/total nitrogen analyses which were never completed. The other three
parameters, surficial I37Cs, surficial 7Be, and bottom segment 137Cs, were not measured on samples
run for PCB analysis and cannot be considered on an individual sample basis. These parameters
are descriptive of the core as a whole, without a direct connection to any individual sample. For
this reason, the 137Cs and 7Be results were compared to the entire core inventory as PCBs per unit
area. In addition, surficial 137Cs and 7Be were also compared with the corresponding core shallow
segments since they partly overlapped. The regression results for these parameters vs total PCB,
AMW, and MDPR are summarized in Table 3-6.
The TOC, TKN, and C/N ratio generally exhibited only weak correlations with total
PCBs, AMW, and MDPR. PCB concentration showed a positive correlation with TOC as would
be expected although the correlation was weak (r = 0.39). The relationship between TOC and
total PCB is shown in Figure 3-21. Most of the response between TOC and total PCB
concentration occurs between those with samples less than three percent TOC and those greater
than three percent TOC, although none of the groups displayed in Figure 3-21 were statistically
different from the others. Few samples had concentrations below three percent, which made it
more difficult to determine statistical significance.
TKN exhibited a slightly weaker correlation with total PCB than TOC and consequently
was not examined further. The C/N ratio exhibited a weaker relationship with total PCB (r =
0.29) than either of the individual measurements. Interestingly, the C/N ratio showed a relatively
strong correlation with the dechlorination measures AMW and MDPR (r = 0.36 for both) as
compared to the individual TOC and TKN relationships (| r | <0.2). Changes in the C/N ratio
3-22 TAMS
00091
-------
are believed to be reflective of increases in the wood cellulose content of the sample, which would
increase the molar C/N ratio. As discussed previously, soil and algal-based organics would be
expected to have C/N ratios around 10 while wood-based materials can values greater than 100.
Wood cellulose would be presumably derived from wood chips. The suggestion that the degree
of dechlorination increases with the C/N ratio would indicate that the presence of woody materials
correlate with high PCB concentrations in the sediment. Unfortunately, the visible presence of
wood chips did not correlate with the C/N ratio, as noted in Section 2.4.3, so no visual clue could
be confirmed for the possible occurrence of dechlorination. However, this data set is limited in
the number of samples represented and so further exploration of these issues is not possible.
The radionuclide results showed several significant correlations with the PCB parameters,
such as between 137Cs in surficial sediments and PCB concentrations in the corresponding shallow
sediments (r = 0.45). ^ measurements in surficial sediments were shown to have a statistically
significant relationship with shallow sediment PCB concentrations when grouped into 7Be detected
and non-detected groups. The median total PCB concentrations were about three times higher for
the sediments when 7Be was detected, indicating that PCB concentrations in shallow sediments
were higher in areas showing recent deposition. No correlations between 7Be concentrations and
sediment PCB concentrations were found, other than the detect/non-detect groupings. Similar but
weaker results were found when the data for PCB mass per unit area were examined (see Figure
3-22 and Table 3-6). Although the presence of 7Be was shown to correlate with higher PCB
concentrations, this should not be assumed to result from recent deposition of PCBs from recent
GE releases, effectively increasing the inventory. As will be discussed in Section 4.1 the majority
of sites studied exhibited PCB mass loss relative »o previous studies. If recent deposition is
relatively contaminated as compared to prior deposition, its impact is undiscernible given the
vertical resolution obtained for these cores and the estimated net PCB losses. Thus, 7Be presence
is correlated with relatively higher PCB inventories but is not necessarily evidence of recent
deposits of highly contaminated material. Nonetheless, since all locations had measurable levels
of 137Cs and PCBs, the following inferences can be made:
1. The presence of 137Cs and PCBs indicates that all sites cored were depositional at
least at one point in the last 40 years.
3-23
TAMS
-------
2. The sites without 7Be present are not currently accumulating sediment and
therefore are probably not being buried.
3. The lower PCB concentrations in the 7Be non-detect cores indicate that either less
PCBs were deposited at these sites to begin with or that PCB loss, potentially via
scour, has occurred, or both.
While the inverse correlation between 7Be absence and PCB concentration does not offer
proof of sediment scour, it does suggest that burial of contaminated sediments with "clean"
sediments is not occurring in at least 30 percent of the coring sites since there is no evidence of
recent burial.
137Cs was measured in both surficial and bottom core sediments. Both measurements were
compared with total PCB concentration and inventory. 137Cs in core bottoms had no statistical
correlation with sediment PCB concentration or inventory. This suggests that the incomplete
cores described in Chapter 2 can probably be considered good estimates of the current PCB
inventory at their coring sites. Notable exceptions are discussed in Section 4.2 However, as a
predictor for PCB concentration or the degree of dechlorination, 137Cs was not useful.
137Cs in surficial sediments exhibited the strongest correlations with PCBs of any of the
radionuclide measurements. Table 3-6 contains the regression coefficients for 137Cs in surficial
sediments. The relatively strong positive correlation of 137Cs with shallow sediment PCB
concentration is shown in Figure 3-23. The data can be grouped into three statistically
significantly different groups 0 - 500 pCi/kg, 500 - 1,000 pCi/kg and >1,000 pCi/kg. The two
highest 137Cs bins (1,000 - 1,500 pCi/kg and 1,500 - 8,000 pCi/kg) are statistically
indistinguishable. These data support the anticipated association of PCB and 137Cs. Although they
have different release histories, their association with fine-grained sediments yields a positive
correlation between them. In addition, the thick low resolution core segments tend to span a
sufficient depositional interval (roughly 30 to 40 years), so as to average-out their different
histories and yield a positive correlation. Interestingly, the three-fold range in 137Cs concentration
yields an order of magnitude increase in the PCB concentration in surface sediments. The range
3-24
TAMS
-------
of 137Cs obtained for the low resolution core surficial sediments (44 - 8700 pCi/kg) is greater than
that obtained for the high resolution core surficial sediments (250 - 1300 pCi/kg) but only three
samples exceeded 2000 pCi/kg. In fact, the median value for the low resolution surficial sediments
was 715 pCi/kg, well within the range of the surficial sediments for the high resolution cores.
Given the varied sediment types and varied depositional environments sampled under the low
resolution coring program, which would serve to deposit, accumulate and rework sediments of
various 137Cs levels, the range of the I37Cs levels measured in the surficial sediments does not
appear extraordinary.
137Cs was also positively correlated with the sediment PCB inventory as would be expected
given the correlation with surface concentration. The PCB inventory data exhibited more
variability with surface 137Cs but still showed at least an order of magnitude increase across the
range of 137Cs values.
137Cs in surficial sediments was also compared with the MDPR and AMW to examine the
relationship between 137Cs and dechlorination. Both the shallow sediment MDPR and AMW as
well as a core average MDPR and AMW were examined against 137Cs. The core averages of
MDPR and AMW were calculated using length-weighted average congener concentrations. All
dechlorination measures were positively correlated with 137Cs. The AMW and MDPR for shallow
sediments are plotted in Figure 3-24. Although positively correlated, the regression coefficients
are weaker than those between PCB concentration and surface 137Cs, indicating that the correlation
between 137Cs and dechlorination is merely a result of the correlation between 137Cs and total PCB
concentration and the far stronger correlation between dechlorination and sediment PCB
concentration. As in all the other variables examined, the correlations using the entire core were
weaker than those for the shallow core segments alone.
Summary
Correlation analysis among the various parameters and the PCB measures generally
confirmed the expected relationships for PCBs in the environment. The strongest predictors for
log-transformed, length-averaged PCB concentration among the parameters examined were percent
3-25
TAMS
-------
solids, solid specific weight and bulk density, all of which were negatively correlated with
concentration. This was anticipated, given that at lower density, more porous sediments tend to
be richer in organic matter and fine-grained sediments, which also have an affinity for PCBs.
PCB concentrations were positively correlated with fine-grained sediment as measured by the
principle sediment fraction, the percent silt, or the median diameter. Several phi fractions, phi
4.5,phi 5, phi 5.5, phi 6 and phi 6.5, exhibited stronger correlations than the three more general
sediment grain-size parameters just listed. These phi fractions correspond to particle sizes in the
range of 11 to 44 ,um, which are all contained within the silt fraction (5 to 75 ,um). Correlation
coefficients were in the range of 0.41 to 0.47, which were still substantively lower than those for
the bulk sediment properties (|r| = 0.56 to 0.65). The other phi fractions and the fractional
diameters, D(10) to D(99), did not prove any more useful in predicting PCB concentrations than
did the more general grain-size parameters such as percent silt and median diameter.
PCB concentrations were also positively correlated with organic carbon (TOC) although
the data set was quite limited, preventing a more thorough examination of the relationship. The
degree of dechlorination, as measured by AMW, was positively correlated with the C/N ratio,
suggesting that sediments high in wood materials exhibit higher degrees of dechlorination. This
correlation may be the result of the higher PCB concentrations which have been demonstrated for
wood chip-bearing layers (Bopp and Simpson, 1989).
Correlation analysis between PCBs and the radionuclide measures yielded some interesting
findings. Although the I37Cs measurement on the bottom-most core segment did not yield any
statistical correlations, both surficial 7Be and l37Cs did. 7Be was used on a detect/non-detect basis
and PCB concentrations were higher by a factor of three in samples in which 7Be was detected,
as compared to samples where 7Be was not detected. The absence of 7Be in approximately one-
third (30%) of the cores indicates the lack of sediment burial in these coring sites and their
potential as active scouring sites.
Surficial 137Cs was shown to positively correlate with the shallow sediment PCB
concentration, the sediment PCB inventory, the degree of dechlorination in shallow-sediments
(AMW and MDPR) and the degree of dechlorination in the entire core. The first two correlations
3-26
TAMS
-------
were expected, given that both PCBs and 137Cs preferentially adsorb to the same sediment type.
The latter two correlations are believed to stem from the strong correlation between dechlorination
and PCB concentration and do not represent an independent correlation between dechlorination
and 137Cs.
In all of the correlation analyses, shallow sediments always provided stronger correlations
between the variables than did the results for the entire core. This was attributed to the fact that
most of the cores' PCB maxima occurred in the shallow sediment (top) segment, and most of the
PCB inventory was found in the upper 9 inches of sediments. In addition, the inclusion of the
deeper layers in the correlation analysis meant the likely inclusion of sediments deposited prior
to the PCB discharges by GE, sediments whose properties would have no correlation with PCBs.
Thus, the whole core correlations were weakened relative to shallow sediments.
The correlation analyses confirmed the expected behavior of PCBs in the Hudson River
environment. PCBs are most concentrated in fine-grained, relatively low density, probably
organic-rich sediments with relatively high levels of 137Cs. Dechlorination correlates with a
number of variables but none are as useful for predicting dechlorination as the PCB concentration
itself.
3.3 Interpretation of the Low Resolution Core and the Side-Scan Sonar
Results
An important intended use of the low resolution coring results was their application to the
side-scan sonar data collected in 1992 and initially reported in the Phase 2 Volume 2C Data
Evaluation and Interpretation Report (TAMS et al., 1997). In this section the application of both
the grain-size distribution data and the sediment PCB concentrations obtained during the low
resolution core sampling program are compared with the prior results reported in the DEIR.
While a large concurrent grain-size distribution data set (the confirmatory sediment samples) was
originally available to calibrate the side-scan sonar data, the PCB data in the DEIR analysis were
not concurrent, having been collected in 1984 by NYSDEC. Thus, the analysis of the low
resolution core sediment PCB data will provide a relationship for current conditions. In addition,
3-27 TAMS
-------
the low resolution core data set extends below the TI Dam, and so provides data on current
conditions over a longer distance than the original 1984 data set.
Generation of the Digital Number Values (DNs) for Low Resolution Coring Sites
To apply the low resolution core data to the side-scan sonar results, it was first necessary
to obtain the arithmetic mean acoustic response, or digital number, for each of the low resolution
sampling sites. The arithmetic mean digital number (DN) was obtained for both 10- and 50-foot
circles circumscribed about each sampling point based on the 500 kilohertz (kHz) acoustic signal
images, the primary side-scan sonar results. In the analysis of the side-scan sonar data in the
DEIR, both the arithmetic mean and the median DN50 were examined and shown to yield similar
results. In this analysis only the arithmetic mean was examined since it was found to yield slightly
better correlations relative to the median. The DN 10 and DN50 represents the average of all
acoustic results contained in the 10- and 50-foot circles, approximately 82 and 1900 points per
circle, respectively. The results for the 50-foot circle (DN50) were used in all of the subsequent
analyses described here. Both 50-foot and 10-foot circles were used in the analyses reported in
the DEIR. Figure 3-25 shows the general relationship between the DN50 and the DN10 for the
low resolution sites selected for use in this analysis. Although some scatter is evident in the
relationship between the two DN values, a regression on the data is relatively close to the ideal
match. At the upper end of the range, the difference between the DN50 and DN10 is only 15
percent. The selection criteria for the low resolution core sites is described immediately below.
As part of the calculation of the DN50, the quality of the acoustic image around each
sampling point had to be reviewed. This was accomplished using a geographical information
system to coordinate sample locations, the 50-foot circles, and the side-scan sonar images
themselves. Plates 3-1 to 3-19 are the side-scan sonar image maps showing the locations of the
low resolution core sites and their associated circles. Of the 170 core sites, only 119 were located
in areas covered by the side-scan sonar images. Of the 119 locations, 26 locations had to be
excluded due to image quality issues. Many points had to be excluded due to proximity to the
edge of the image. The entire 50-foot circle had to be within the sonar coverage or else false
values would be included in the DN50 calculation. Other locations were excluded simply due to
3-28
TAMS
-------
poor image quality, typically the occurrence of the ship's track within the 50-foot circle. To
confirm the visual choices of poor image quality, the DN50 and DN10 values were compared.
Good quality images would have similar values for both DNs. This is shown in Figure 3-25
which is a plot of the DN50 vs DN10 for the final point selection: Poor quality images would fall
far from the ideal line shown in Figure 3-25. The selection process yielded 93 locations whose
associated images, within the circle, were acceptable. These locations are marked on Plates 3-1
to 3-19 by a 50-foot circle. Excluded locations are marked with a circle overwritten with an 'X'.
It is important to note that in the production of these plates, some image quality was sacrificed in
order to fit the images on the 8-1/2 by 14-inch sheets. In selecting acceptable locations, the
original side-scan sonar images were examined at their original size and scale.
Application of the Low Resolution Core Grain Size Distribution Data to the Side-Scan Sonar
Images
The DN50 values obtained for the low resolution coring sites were then compared with the
i Laser grain-size distribution parameters obtained on top (shallow) core segments using least •-
squares linear regressions. These regression results were then compared with the confirmatory
sampling program results on the basis of the correlation coefficient (r) term (See Figure 3-26). In
general, both the low resolution and confirmatory sampling data sets show relatively strong
correlations (r > 0.38) for the same grain-size distribution parameters. Nonetheless, the low
resolution core results generally yielded lower r values. For example, the r value for the mean
phi () (geometric mean diameter) for the low resolution coring results was 0.68 as compared to
0.73 for the confirmatory cores. Similarly, the correlation with percent silt plus clay at 0.64 was
less than that for the confirmatory samples at 0.73. Only percent sand and percent gravel had
higher correlations for the low resolution cores relative to the confirmatory cores. The precise
reason for the general decrease in correlation is not known; however, it is likely that it stems from
several differences between the two programs. In particular, the low resolution core sediment
grain-size distribution parameters were obtained for shallow sediments, nominally from 0 to 9-
inches in depth. Confirmatory sediment samples were obtained from core tops from 0 to 2-inches
in depth and from shallow grab samples approximately 4 to 6-inches in depth. Thus, the
confirmatory samples would more closely measure the actual sediments (i.e., surface sediments)
3-29
TAMS
-------
responsible for the acoustic signal at a given location. In addition to the difference in sampling
technique, the orientation of the two programs was different. For the confirmatory sampling
programs, the intent was to obtain samples from many different sediment types so as to calibrate
the side-scan sonar results. The low resolution cores were intended to characterize sediments in
a limited number of areas focusing on regions of high PCB inventory. In this manner, the low
resolution program did not obtain many coarser-grained sediment samples. The lack of these data
might serve to weaken the strength of the regressions. Lastly, it is also possible that there were
some changes in sediment type at some locations. However, as will be discussed below, there
does not appear to be any extensive changes in sediment type.
In addition to the regression comparisons, the low resolution core data were also compared
with the side-scan sonar images on the basis of sediment type. For each of the top low resolution
core segments, a principal sediment class was identified by both visual inspection and by Laser
analysis. Similarly, the image areas had been classified as to the general sediment type based on
the image itself. The comparison between the two sediment classification techniques and the side-
scan sonar classifications are shown in Figure 3-27. It is evident from the two diagrams that the
results for the Laser analysis correspond better than the visual inspection with the side-scan sonar
interpretation, although both techniques agree well with the side-scan sonar classifications. Note
for example, the greater correspondence of the silt samples by Laser analyses with the fine-
sediments classification by side-scan sonar. The reason for the better agreement between side-
scan sonar and Laser analysis is not known but probably results from the more accurate
classification by the Laser technique. Also, the original side-scan sonar interpretations were based
on Laser analysis of the confirmatory sediment samples. For this reason, the remaining discussion
on the correspondence between sediment sample classifications and side-scan sonar results will
discuss the Laser results exclusively. It should be noted that all of the low resolution core sites
covered by the side-scan sonar images, a total of 126 locations, are represented in Figure 3-27.
This interpretation and discussion is not limited by the local image quality unlike the previous
numerical regressions on digital number (DN50).
In general, low resolution cores whose principal fraction was silt were obtained from
image areas classified as fine or finer sediments (88 percent of samples and areas matched).
3-30
TAMS
-------
Similarly, cores whose principal fraction was classified as medium-sand were typically obtained
from coarse or coarser sediment areas (86 percent of samples and areas matched). Fine-sands
were found to be equally distributed between the fine and coarse sediment areas as might be
expected given the limited classification types for the side-scan sonar interpretation. The few
gravel samples were also split between fine and coarse areas which was not anticipated.
However, this set of samples is believed to be too small and unrepresentative to infer any
relationship between gravel and the side-scan sonar images.
The agreement between the three main sediment types by Laser analysis and the side-scan
sonar images is considered to be quite good in that the descriptions and sample locations agree in
the vast majority of cases. These results also indicate that the sediment types have remained the
same in the areas studied by the low resolution program. Given that the low resolution core sites
were obtained throughout the area of side-scan sonar coverage, it is likely that the sediment area
classifications have remained relatively constant during the Phase 2 study period (1992 to 1994).
Figure 3-28 presents a different perspective on the correspondence between the low
resolution coring grain size results and the side-scan sonar data. This figure presents the DN50
values grouped by principal sediment fraction. This figure was constructed in the same fashion
as the similar box- and-whisker plots presented in Section 3.2. The figure shows the steady
increase in the DN50 value from silts to fine-sand to medium-sand plus gravel. Note that the fine-
sand box-and-whisker plot straddles the range of DN50 from 40 to 60, which was the approximate
range used to define fine- and coarse-grained sediment in the original side-scan sonar
interpretations. This would explain the nearly eqir! distribution of fine-sand between the two
sonar classifications seen in the previous figure (Figure 3-27). Note that this diagram is again
limited to the 93 low resolution sites for which DN50 values could be defined.
Using the Tukey-Kramer honestly significant difference test, all three groups shown in
Figure 3-28 were shown to be statistically significant and different from each other at the 95
percent confidence level. This is further confirmation of the correspondence between the side-
scan sonar response and sediment type as classified by a standard technique.
3-31
TAMS
-------
The last figure to be presented under the discussion of grain-size distribution and side-scan
sonar results is Figure 3-29. This figure shows the correspondence between the mean grain size
on the phi scale and the DN50. Also shown is the relationship for the mean DN50 and mean phi
for the confirmatory samples. This diagram and regression are nearly identical to the original
diagram for the confirmatory samples presented in Chapter 4 of the DEIR (TAMS et al., 1997)
which is based on the median DN50 as opposed to the mean DN50. As is evident in the large
diagram in Figure 3-29, the relationship between the mean DN50 and the mean diameter for the
low resolution cores is not the same as that found previously. Nonetheless, both curves show a
- substantive decline in DN50 as the mean particle diameter decreases (phi increases). Although
the regression curves are relatively close, the resulting low resolution regression yields a slightly
better fit between DN50 and mean diameter after the exclusion of 4 points, based on a
Mahalanobis analysis. The reason for the difference between the two curves is not known but it
is likely that it is related to the following issues. Specifically, the vast majority of low resolution
core samples were obtained from fine-grained sediment areas and the data set probably does not
contain a sufficient number of values to constrain the regression at higher DN and lower phi
(larger diameters). In addition, as mentioned previously, the low resolution sediments were
obtained from substantially lower depths in the sediment (0 to 9-inches for the low resolution top
segment versus 0 to 2-inches for the confirmatory cores and less than 4 to 6-inehes for
confirmatory grab samples). The collection of deeper sediments may alter the underlying
relationship. Lastly the conditions may have changed between 1992 and 1994, but this seems
unlikely in light of the sediment classification evidence presented earlier.
In summary, the grain-size distribution data reaffirm the general conclusions concerning
the use of the side-scan sonar data, which are:
• The side-scan sonar data can be used to classify large areas of the river bottom in
terms of their sediment properties. This was confirmed by the samples collected
which closely matched the sediment types identified by side-scan sonar.
3-32
TAMS
-------
• The acoustic signal as represented by the DN50 decreases as the sediment becomes
more fine grained. This was confirmed by both the principal fraction correlation
and the direct comparison of DN50 and mean diameter.
Comparison Between the Side-Scan Sonar Results and PCB Levels in Shallow Sediments
In the DEIR, a comparison was made between the DN10 and the 1984 NYSDEC sediment
survey results. The comparison showed that the PCB levels in shallow sediments had a
statistically significant inverse correlation with the DN10. The main limitation of the comparison
was the length of time between the 1984 NYSDEC sediment survey and the 1992 side-scan sonar
survey. As part of the analysis of the low resolution core results, the 1994 shallow sediment PCB
levels were similarly compared to the 1992 side-scan sonar results eliminating much of the issue
over time elapsed. In this instance, the DN50 was used instead of the DN10 since it has been the
main parameter for comparison throughout this section. As shown in Figure 3-25, use of the
DN50 relative to the DN10 does not appear to introduce any substantive error, given the inherent
variability of these parameters.
In the original analysis in the DEIR, the DN values were grouped into bins of about five
units (e.g., DN10 of 40 to 45). The small bin intervals were usable for this analysis because of
the large number of samples (about 1,200). The low resolution cores were far fewer in number
(93) and thus would not support such narrow bins. For this reason, just three groupings of the
total PCB data were made, based roughly on the important transitions seen in the original analysis
using the 1984 data. These bins were defined as follows:
• DN50 greater than 60;
• DN50 less than or equal to 60 and greater than 30; and
• DN50 less than or equal to 30 and greater than 10.
No DN50 values were obtained lower than ten. The value of 60 corresponds to a marked
change in the relationship between DN10 and the total PCB concentration in the original analysis.
3-33
TAMS
-------
This value also represents an approximate sediment texture boundary based on the DN50 and the
Laser grain size distribution for the confirmatory sediment samples.
The results for total PCB concentration in shallow sediments grouped by DN50 are shown
in Figure 3-30. Evident in the figure is the marked increase in PCB concentration across the
DN50 value of 60. A smaller increase occurs across the DN50 value of 30. When these bins
were tested for statistical significance using the Tukey-Kramer honestly significant difference test,
the first bin (DN50 greater than 60) was found to be statistically different from the other two bins.
The latter two bins were not statistically different.
The results as presented in Figure 3-30 confirm the original findings of the DEIR
concerning the acoustic signal as represented by the DN50. Essentially, coarse-grained sediments,
typical of highly-reflective side-scan sonar areas (DN50 greater than 60) are distinctly less
contaminated than the finer-grained, less-reflective areas. Based on the average or the median
conditions, less-reflective finer-grained sediments are about ten to twenty times more contaminated
than the coarser-grained areas. The results of the low resolution coring program in fact yield a
larger difference in PCB concentration across the DN50 threshold of 60 (about a seven-fold
increase) than did the original analysis using 1984 data, although the change in this difference may
not be statistically significant given the various uncertainties.
A similar comparison between low resolution core mass per unit area and DN50 was made
using the same DN50 bins. The comparison is represented in Figure 3-31. The relationship
covering the three DN50 groups in Figure 3-31 matches the results seen in Figure 3-30 almost
exactly. This would be expected since so much of the PCB inventory is located in the shallow
sediments as measured by the top core segments. The figure also reveals an approximate seven-
fold increase in mean and median PCB inventory across the DN50 value of 60. This result
implies that the side-scan sonar images can be used to estimate both shallow sediment PCB
concentrations and sediment PCB inventories.
As a part of the DEIR, areas covered by the side-scan sonar images and interpreted as fine-
grained sediments were considered to be areas of potentially high PCB contamination. This
3-34
TAMS
-------
association was based on the interpretation of the 1984 sediment PCB data. The low resolution
core data set supports this interpretation of the side-scan sonar images by again demonstrating a
statistically significant correspondence between the image data and a more current sediment PCB
inventory.
3.4 Summary of Chapter 3
This chapter has focused on the interpretation of the low resolution cores in the context
of the conclusions from the high resolution cores presented in the DEIR (TAMS et al., 1997). In
this perspective, the results of the low resolution cores confirm the relationship between the
changes in molecular weight relative to Aroclor 1242 (AMW) and the molar ratio of the sum of
BZ# 1, 4, 8, 10, and 19 to the total sample (MDPR). Confirmation of this relationship also
confirms the occurrence of meta- and para-dechlorination and the absence of ortho-chlorine loss
in the sediments. This represents an important confirmation since the low resolution cores can
be considered more spatially representative of Upper Hudson conditions, thereby showing the
consistency of the dechlorination process in the Upper Hudson. This confirmation also means that
the AMW term can be used as a surrogate for the estimation of mass loss by this process.
The low resolution cores yielded a similar average degree of dechlorination (AMW =
0.10) to that obtained from the high resolution cores (AMW = 0.08). Given that the low
resolution core sites are focused on some of the most contaminated areas of the Upper Hudson and
yet only an average mass loss of 12 percent (AMW = 0.10) has occurred, it can be inferred that
the rest of the Hudson has experienced an average mass loss of less than ten percent due to
dechlorination, as was concluded from the high resolution cores.
The low resolution cores also showed that the degree of dechlorination in the sediments
was proportional to the level of contamination. However, these results could only be shown to
be consistent with the high resolution cores results and could not confirm the results due to factors
inherent in the "vibra-coring" sample collection procedures and the sample homogenization
process. Nonetheless, this consistency between programs is still an important result.
3-35
TAMS
-------
A close examination of the correlation between PCB concentration and the degree of
dechlorination showed that the low resolution cores achieved an apparent degree of dechlorination
equivalent to a high resolution sediment core at 3.5 times the low resolution core sample total
PCB concentration. However, this 3.5-fold increase in response was attributed to the
homogenization process wherein concentrated sediment layers would be mixed with low-level
sediment layers, effectively lowering the measured concentration but leaving the degree of
dechlorination in the concentrated layer intact. This was demonstrated by calculating this effect
with the high resolution core results which yielded a similar shift in response.
Potentially cross-contaminated samples were ferreted out using the 137Cs measurement at
the core bottom and the PCB profile in the core to highlight those core segments most at risk for
this problem. After screening the original 371 samples, 229 samples were kept to finalize the
AMW to total PCB relationship for the low resolution cores. Cross-contamination was an
important issue in this chapter because of its differing effects on PCB mass and PCB ratios. The
level of cross-contamination found in the cores will not represent a significant issue for the
sediment-mass inventory estimates presented in the next chapter since the actual mass represented
in the potentially cross-contaminated segments is so small.
The geometric mean total PCB concentration for the final data set of 229 samples was 30.8
mg/kg with a AMW of 0.10 and an MDPR of 0.54.
A large number of ancillary parameters were examined for potential correlation with total
PCBs, AMW, and MDPR. Most parameters that could be related in some fashion to total organic
carbon or silt content were found to exhibit some correlation with total PCB concentration. By
and large, no strong correlations (i.e., |r| >0.8) were seen, although some were better than
others. Among these, percent solids and bulk density had the strongest correlation coefficients
(R = -0.72 and R = -0.61, respectively; Table 3-8) on log-transformed, length-averaged PCB
concentrations.
Weak but anticipated relationships were also found with percent silt, median phi (4>), and
principal sediment class when compared to individual PCB sample results on a log-transformed
3-36
TAMS
-------
basis. Percent solids, solid specific weight and bulk density were the most highly correlated
parameters for this comparison. Several individual grain-size parameters which represented the
silt fraction of the sediment, specifically phi 4.5 to phi 6.5, exhibited the strongest correlations
among the grain size parameters (0.41 to 0.47) and shallow sediment PCB levels. The TOC plus
TKN data set was limited but yielded a weak correlation between total PCBs and both TOC and
TKN across all depths. In addition, a weak but notable correlation was seen between the C/N ratio
and AMW. This suggests that maximum dechlorination occurs in wood-bearing layers, although
a correlation between C/N and the visual presence of wood chips was not apparent (Section 2.3).
Most likely, these layers represent highly concentrated, wood chip bearing sediments produced
by the Fort Edward Dam removal in 1973 and the subsequent washout of the channel sediments
in 1974 to 1976. Evidence of wood chips in Hudson River sediments was extensively noted by
NYSDEC during the 1976-1978 surveys (Normandeau, 1977, Tofflemire and Quinn, 1979). In
addition, Bopp et al., 1985, noted the presence of high PCB concentrations and woody material
in Hudson River sediments.
The radionuclide 7Be was somewhat predictive, with 7Be-bearing cores having statistically
higher PCB levels than cores without 7Be. Presumably the 7Be marks areas of recent deposition
while the absence of 7Be indicates areas that are likely to experience scour and subsequent PCB
loss. Surficial (0 to 1-inch) 137Cs showed a stronger positive correlation (r = 0.48) with PCBs,
although still not predictive. The fairly strong correlation between total PCBs and 137Cs was
expected given the affinity of both analytes for fine-grained sediment and the thick low resolution
core segments that served to integrate their different deposition histories. Shallow sediments
correlated better than the core inventory (PCB/unit area) (r = 0.34), as might be expected since
most of the PCBs, and probably most of the 137Cs, was found in this layer.
I37Cs also correlated with AMW, albeit weakly (r = 0.36). It is likely that the correlation
between 137Cs and AMW is simply the result of each parameter's correlation with total PCBs.
Similar weak correlations were found for many of the ancillary parameters. None had the level
of correlation exhibited among the PCB parameters themselves.
3-37
TAMS
-------
The last part of the chapter explored the relationship between the side-scan sonar images
expressed as a digital value (DN50) and the low resolution core results. The DN50 values showed
similar correlations to the Laser grain-size distribution parameters for shallow sediment when
compared to the confirmatory sediment samples originally used to interpret the side-scan sonar
images. Most regression coefficients (r) for DN50 and the low resolution grain-size distribution
parameters decreased relative to the confirmatory sample results. The weakened correlations were
attributed a number of factors, with the confirmatory samples considered more representative of
the surface sediments mapped by the side-scan sonar.
The strongest correlation for DN50 and grain-size distribution parameters were seen for
mean phi (), median diameter (d[50], and percent silt plus clay). The first two parameters both
track the center of the grain-size distribution on a log basis inferring that the images are related
to the general sediment texture as expected. Other diameter measures, near the center of the grain-
size distribution (e.g., d(40) and d(70) were similarly correlated with the DN50. The percent silt
and clay tracked the proportion of fine-grained material and was inversely proportional to the
DN50 as expected.
The classified side-scan sonar image areas (e.g., fine sediment), were found to match the
principal sediment fraction (e.g., silt) very well for silt and medium-sand (85 percent accuracy
or better). The samples classified with a principal fraction of fine-sand straddled the two side-scan
sonar classifications of fine sediment and coarse sediment. However, this was consistent with the
mapping approach since a DN50 value of about 50 to 60 was generally used as a boundary
between the two sonar classifications. This range of DN50 values also fell within the middle of
the DN50 values for fine-sand so that splitting of the samples between the two classifications
would be expected. Taken together, these results all confirmed the ability of the side-scan sonar
images and interpretation to characterize sediments on a coarse/fine basis for the entire survey
area.
A regression was performed on the DN50 and the mean phi as was done with the
confirmatory samples in the DEIR. The low resolution core regression relationship was similar
but not identical to the previous regression. This was attributed to two factors, the greater depth
3-38
TAMS
-------
of the top low resolution core segments and the lesser number of low resolution core samples with
relatively coarse sediments. Thus the low resolution core results were considered to be consistent
with the previous work in this regard but could not be used to confirm the previously measured
trend.
Lastly, the relation between DN50 and PCB concentration was examined. The prior
examination completed for the DEIR showed higher PCB levels at low DN50 values. The results
from the 1984 survey, as analyzed in the DEIR, showed a five- to ten-fold increase in total PCB
concentration for sediments with DN50 less than 60 relative to those with DN50s greater than 60.
The low resolution cores confirmed this result quite well with a seven-fold increase between the
finer sediments (DN50 < 60) and the coarser sediment (DN50 > 60). A similar increase based
on the DN50 threshold value of 60 was noted for the sediment PCB inventory as mass per unit
area.
These results verily the usefulness of the side-scan sonar images for both the classification
of sediment types as well as for the identification of areas of potentially high PCB contamination
(greater than 50 ppm). In total, the low resolution coring results support conclusions of the
previous Phase 2 investigations and interpretations. The correlations seen among the various low
resolution core parameters are consistent with those seen earlier and affirm the application of the
previous study conclusions to the entire Upper Hudson.
3-39
TAMS
-------
AN EXAMINATION OF HUDSON RIVER
SEDIMENT PCB INVENTORIES
£
s
-------
4. An Examination of Hudson River Sediment PCB
Inventories: Past and Present
A major goal of the low resolution sediment coring program was the estimation of
sediment inventories in the areas of study. In this chapter, sediment inventories estimated from the
low resolution cores are compared with historical studies of the sediments conducted by NYSDEC
in 1976 to 1978 and 1984. In the T1 Pool, low resolution cores results are compared on a
point-by-point basis with the NYSDEC 1984 results. Below the TI Dam, low resolution core results
are compared on a spatial basis with the 1976-1978 NYSDEC hot spot inventories. This chapter also
contains a discussion of the near-shore sample results and compares them with the 1984 NYSDEC
samples collected from the same region of the river. The last section of the chapter provides a
summary of the analysis presented.
4.1 Sediment Inventories of the Thompson Island Pool
The TI Pool has been the subject of several large sediment surveys, each of which attempted
to map sediment PCB inventories and areas of concentrated contamination. In 1976 to 1978, the first
major survey of the TI Pool and the Upper Hudson was completed by NYSDEC. This survey was
used to identify forty areas of highly contaminated sediments (hot spots), twenty of which were
located in the TI Pool. In 1984, NYSDEC completed a second, more intensive survey of the TI Pool.
On the basis of this survey, NYSDEC (Brown et al., 1988) and Malcolm Pirnie (MPI, 1992)
identified areas or polygons of elevated sediment contamination.
Because of the scale and coverage of the 1984 survey, it has been considered a benchmark
in attempting to assess and understand sediment PCB inventories in the Upper Hudson. As a part
of the Phase 2 investigation, the low resolution coring program was intended to assess the
applicability of the 1984 survey to recent PCB inventories. This was to be accomplished by
reoccupying selected 1984 sampling locations and collecting new cores to form a basis for
comparison. As discussed in Chapter 2, the low resolution coring program was quite successful in
4-1
TAMS
-------
reoccupying these sites. This section compares in detail the 1984 and 1994 surveys and discusses
the ramifications.
PCB loss or gain from the sediment can take many forms. Scour, diffusion, groundwater
advection, and biological activity can all potentially remove PCBs from a given location. Biological
activity in the form of anaerobic microbial dechlorination can also serve to decrease PCB
concentration in the sediments. PCB inventories can be increased chiefly by deposition, either with
sediment contaminated by newly released PCBs or with redeposited sediments from other
contaminated locations. Up until 1997 when GE brought nearly all PCB discharge from the Hudson
Falls facility under control, it is likely that sediment deposition involved both fresh and redeposited
materia] (General Electric Corp., 1991-1997). Tracing and estimating all of the various fluxes
represent a daunting task made all the more difficult by inherent spatial and temporal variations. The
low resolution coring program provided an alternate means of assessing these fluxes by using the
PCB inventories found in the sediments to explain removal and deposition processes.
In the TI Pool, 63 sites originally sampled in 1984 were selected and reoccupied in 1994,
providing a ten-year period of integration. The premise for analysis is then, has the sediment
inventory of PCBs increased or decreased during the intervening ten years? While the premise itself
is simple enough, there was concern that sediment heterogeneity, differing sedimentation rates,
analytical technique differences, and other issues would confound the ability to discern true changes
in the sediment inventory.
Concern over sediment heterogeneity factored directly into the low resolution site selections.
Coring sites were selected in clusters of similar concentration and sediment texture in an attempt to
minimize sediment heterogeneity of the sampling areas. Concern over differing sedimentation rates
was addressed by individually assessing each core for completeness via the use of ,37Cs (Chapter 2).
Analytical techniques were extensively reviewed to assess differences, as discussed below.
Before assessing changes in sediment inventory from 1984 to 1994, it was first necessary to
establish a common measurement basis for comparison. Because of concern over differing
4-2
TAMS
-------
deposition rates and variations in core segment intervals, cores were integrated over their lengths to
yield mass per unit area (MPA) estimates as well as length-weighed averages (LWA).
The MPA was calculated as follows:
MPAl
g
m
n
i=i
>»g
* Ljtcm) * SSW{
J SDW ^ 1 kg ^ 1g +
\ cc J 1000^ 1000m;
100— (4 1-D
m I
where C,
L,
the PCB concentration in the core segment in mg/kg diy weight (ppm)
the length of the core segment in cm
SSW, = the mass of dry solids per unit wet core volume in g dry-weight
number of core segments analyzed for PCBs
The LWA was calculated as follows:
Y C *L
I !
LWA
i = i
£i.
cc
(4.1-2)
where C, and L, are defined as above. The number of core segments in this calculation (m) may be
the same or less than the total number of segments analyzed for PCBs (n) above. This distinction
is made because LWA is most useful over the interval of recent sediments, rather than over the entire
core length since a core may contain a considerable thickness of pre-1954 sediment. The selection
of segments was based on each individual core profile. When the difference in PCB concentration
between two adjacent segments was greater than an order of magnitude with the upper segment
containing the higher concentration, the lower segment and all segments below it were excluded
from the LWA.
Both parameters (MPA and LWA) can provide some insight concerning the change in
sediment inventory from 1984 to 1994. The MPA provides the more intuitive interpretation. Its
4-3
TAMS
-------
interpretation requires no knowledge of sediment depth so long as cores are considered complete or
nearly complete. However, the inclusion of the solid specific weight (SSW) in the calculation adds
another level of uncertainty to the parameter. Estimation of the solid specific weight for the 1984
survey was discussed extensively in the DEIR (TAMS et al, 1997). The mean SSW value obtained
for the low resolution core segments in the TI Pool (1.09 g/cc) compares well with the mean value
of the 1984 core segments (1.06 g/cc) but the comparison of the shallow (top-most) core segments
shows a difference of about 10 percent (1.05 g/cc for 1994 vs 0.92 g/cc for 1984). This difference
would yield a higher mass per unit area (by 10 percent) for the 1994 core given identical sediment
concentrations. Lacking any basis on which to correct or select the SSW data means that this
uncertainty must be incorporated into the MPA comparison.
The LWA concentration avoids this issue since SSW is not involved. However, the LWA
adds a different uncertainty since core length has a direct bearing on its value. It is possible for two
cores to have identical LWA but represent very different inventories simply on the basis of their
length. Nonetheless, because of the general consistency of core lengths in both programs (roughly
15 to 30 inches), the length-weighted average provides an alternate basis for evaluating the change
in sediment inventory. In the subsequent discussions, the MPA will be used primarily although
information on LWA will also be provided.
Establishing a core integration scheme was only part of the basis needed to compare the 1984
and 1994 (Phase 2) survey results for PCBs. A second issue regarding analytical techniques had to
be resolved before a direct comparison could be made. Specifically, analytical techniques for PCB
quantitation had changed markedly both in congener resolution and on a sample reporting basis
between 1984 and 1994. In particular, the 1984 data were quantitated and reported on an Aroclor,
rather than congener-specific basis.
Three Aroclor results, representing Aroclors 1242,1254, and 1260, were used in estimating
the total PCB of the sample by their sample sum. (Brown et al, 1988; TAMS et al., 1997). The
1994 data were based on congener-specific standards, with no inherent tie to any Aroclor. Hence,
the total PCB concentration by the Phase 2 method represents a true total.
4-4
TAMS
-------
As part of the Phase 2 investigation, a study was made of the differences between the two
techniques. This is documented in Appendix E, which describes the quantitation issues relating the
1994 Phase 2 and 1984 NYSDEC PCB data. The recommendation of this analysis was to use the
1984 quantitation of total PCB as representative of the sum of congeners in the trichloro through
decachloro homologue groups after applying the following corrections:
2 Trichloro to Decachloro homologues (mg/kg) = 0.934 x 1984 Total PCB Concentration (mg/kg)
(4.1-3)
During review of this analysis while preparing Appendix E, the correction factor was refined to
0.944. That value differs from the value actually used in the body of this report (0.934), but the one
percent correction is minor and does not affect the comparison between 1984 and 1994 Thompson
Island Pool inventories or the conclusions drawn from this comparison.
These results can then be compared with the same sum based on the low resolution core
analyses. This interpretation of the 1984 PCB data also indicates that the 1984 analyses would have
ignored the lightest congeners and so cannot yield any information on dechlorination. Both the
trichloro to decachloro homologue sum and the total PCB value are used in the subsequent
discussions. Emphasis is placed on the trichloro to decachloro homologue sum, however. From this
point in the discussion, the trichloro to decachloro homologue sum is represented by ETri+.
Before beginning the analysis of the core data, it is useful to see the coring locations in the
context of the river sediment classifications. Plate 4-1 is a key map to the locations of Plates 4-2 to
4-9. These plates consist of side-scan sonar images on which the river sediment properties have been
mapped. In addition, each coring site is denoted along with the 1984 and 1994 MPA and LWA
values. The 1984 values are located to the left of each core marker. The 1994 values are located to
the right. Similarly, the LWA values are located above each marker and the MPA values are located
below. Symbols for both the 1984 (solid circle) and 1994 (hollow circle) samples show the close
agreement of the paired survey locations. The core symbols are also scaled according to the MPA
value, providing a visual clue to the size of the inventory at a given location. The symbols also
provide information on the extent of change at a location. The larger the difference in the circles, the
4-5
TAMS
-------
greater the inventory change. The sediment classifications as derived from the side-scan sonar
analysis are also shown on the map along the sedimentological boundaries. Evident in these figures
are the general trends of mass loss at a majority of sites as well as the association of the greatest
inventories in areas of fine-grained sediment.
A second set of plates provides more direct information concerning sediment inventory
changes on a mass basis. Plate 4-10 is the key location map to Plates 4-11 to 4-19. These plates
cover the same areas as Plates 4-2 to 4-9 but they present only one circle per sampling site. This
symbol is scaled to the absolute difference in total PCB inventory (MPA) between the 1984 and
1994 surveys. In addition, the percent change between the two surveys is given numerically at the
upper right of each symbol. The percent change (Delta) was calculated for each coring site as
follows:
A =
' MPA9A - MPA
MPAU
* 100% (4'-*>
Thus Delta is the absolute difference in total PCB inventory divided by the original 1984
inventory. Note that in Plates 4-2 to 4-19, all numbers for 1984 reflect the reported total PCB
concentrations and were not corrected to ETri+ as discussed above. The corrected data will be
presented and discussed later in this section.
Calculation of Delta requires a pair of NYSDEC and low resolution core results. Sixty-four
of the 76 coring locations in the TI Pool had an associated NYSDEC core or grab. Four of these
locations had grab samples which were only screened via a GC/MS technique which did not provide
an absolute measure of contamination (see Brown, et al. 1988 and TAMS, et al., 1997). An estimate
of the level of contamination for these samples was assigned as part of the kriging analysis presented
in TAMS, etal. (1997) but these values are not used here due to their large uncertainty. This left just
sixty locations where it was possible to compare PCB quantitation directly. Of the sixty locations
45 were cores and 15 were grab samples.
4-6
TAMS
-------
In assessing sediment inventory changes, cores are preferable to grabs since the actual depth
of a sediment sample is known. The depth associated with a grab sample is not well known. In
addition, the grab samples were extrapolated to a depth beyond their expected penetration in an
attempt to represent all PCB contamination at the location. This approach is generally acceptable
when averaging large areas together to assess inventory on an average basis but it is not as useful
on an individual sample basis. The lack of good depth control adds to the uncertainty associated with
core to grab comparisons. Nonetheless, the core -grab pairs can still provide some useful information
on the change in sediment inventory.
4.1.1 A Comparison of 1984 and 1994 Conditions
As a first step in assessing the difference between the 1984 and 1994 conditions, the
individual core and core-grab pair profiles were compared. Appendix C contains diagrams for all low
resolution cores collected in the TT Pool. Also shown on these diagrams are the profiles for the
associated 1984 NYSDEC core or grab sample when available. Figure 4-1 contains several example
profiles of cores from the TI Pool. In assessing the core profiles, the major concern was the change
in sediment inventory from 1984 to 1994. Differences in slicing intervals between 1984 and 1994
were considered, so that the comparable thicknesses were compared. The sediment core
classifications were not meant to be the final assessment on inventory change but were an attempt
to examine the profiles in detail, rather than simply rely on the mathematical integration of the
results. Cores were classified as to apparent, inventory loss, represented by lower PCB concentrations
as well as typically shallower contaminated profiles. Cores exhibiting inventory gain were
characterized with higher sediment concentrations as well as deeper profiles. Some cores appeared
to exhibit little change and were classified as such. Lastly, two cores were unclassified due to
incomplete profiles. Table 4-1 is a list of the classified cores.
To a large degree, the interpretation of these cores was based on the profiles obtained during
the high resolution coring work. Specifically, a core site undergoing steady deposition and burial
would look similar to a high resolution core, allowing for thicker slices. This would be expected to
generate the profile seen for core 5D in Figure 4-1. This can be compared to the high resolution core
shape seen in Figure 4-2, representing core 19 collected near the TI Dam.
4-7
TAMS
-------
The reasons for the apparent changes in inventory between 1984 and 1994 were not always
readily apparent. Sediment inventory losses could result from several processes as discussed
previously. Sediment inventory gains presumably result from additional deposition at the coring site
since 1984. Sediment heterogeneity, although minimized by the sampling site selection and
collection process, was certainly a factor in yielding apparent changes in some cores. In a limited
number of cores, however, there was enough information contained in the core profiles to
definitively suggest sediment scour. In these instances, both the NYSDEC and the low resolution
cores contained enough slicing intervals so as to indicate the depth of recent contamination at the
time of collection. When the low resolution core had a substantively shallower recent sediment
profile than the earlier NYSDEC core, it was classified as exhibiting scour. The underlying
assumption here is that the apparent upward movement in the sediment profile is the result of
removal of the overlying layers by scour since it is unlikely that any other process could move PCB
contamination from deeper layers and completely purge the deeper layers in the process. Of the 30
core sites where mass loss was evident, 8 profiles contained enough information to suggest scour as
the mechanism for mass loss. Figure 4-3 shows a clear case of sediment scour at two of these sites.
In both these cores, the absence of ,37Cs in the bottom-most layer provides the basis for the
conclusion that no undetected PCBs remain below the 1992 core and thus that the sediment profile
has become shorter over time, implicating scour as the removal mechanism.
After classifying the cores in this manner, the classifications were compared to the integrated
core results as shown in Figure 4-4. This diagram represents the mass per unit area (MPA) estimates
for total PCBs for the 1984 and 1994 site pairs. Sites with unchanged inventories would fall along
the diagonal line through the center of the diagram. As can be seen in the figure, few points fall very
close to the line. Note as well that the axes are log-scale. Nonetheless, also evident in the diagram
is general agreement of the visual core classifications and the absolute core inventories. Core pairs
showing loss fall below the line and vice versa, as expected. It is also apparent in this diagram, that
the majority of points fall below the diagonal line, indicating a greater number of sites with
inventory loss than gain. The discussions to follow will attempt to resolve the significance and cause
of this trend.
4-8
TAMS
-------
Figure 4-5 contains a second representation of this data set. In the upper half of the figure,
the 1984 and 1994 total PCB MPA values are graphed against each other. Also shown is the diagonal
line mentioned previously as well as a regression line fit to the data. The data do not provide a linear
representation but the trend is statistically significant and suggests that PCB inventory loss occurs
more frequently at higher inventories. Also shown in Figure 4-5 is the MPA for just the trichloro and
higher homologues (ETri+). In this instance, the 1984 trichloro and higher homologue
concentrations are substantially higher than those for 1994, indicating extensive loss of those
homologues from the sediment.
This loss becomes even more striking when the loss itself is plotted against the MPA for
2Tri+ for 1984. This is shown in the upper diagram of Figure 4-6. Evident in this plot is a linear
relationship between ETri+ mass loss and the original ETri+ inventory. This should be contrasted
with the lower diagram in the figure which shows the relationship for the total PCB mass loss against
the 1984 ETri+ inventory. The MPA for the ETri+ will be written as MPA3+for the remainder of this
section. These diagrams, as well as those of Figure 4-5, show that the total sediment inventory does
appear to decline but not nearly in the manner seen for the ETri+ between 1984 and 1994. Part of
the "goodness of fit" seen in the top diagram of Figure 4-6 stems from the presence of the ETri+
term in the variables on both axes. Thus, errors in ETri+ tend to be correlated and an overestimate
of the degree of correlation is obtained. Nonetheless, there is a difference between the two diagrams
in Figure 4-6. The MPAJ+ loss consistently increases with increasing 1984 MPA3+while total mass
inventory (MPA) shows a less consistent trend, with many sites exhibiting inventory increases. This
suggests that some of the loss seen in Figure 4-6 results from dechlorination, where the loss of
ETri+ would not yield an extensive loss of mass. Rather this process serves to convert the congeners
represented in the ETri+ to lighter congeners, specifically BZ# 1,4, 8,10, and 19. In this process,
PCB moles are largely conserved while the PCB mass of the entire mixture can decline up to 26
percent. As was discussed in Chapter 3, no samples were found indicating PCB mass loss beyond
24 percent in the low resolution sediment cores based on the change in molecular weight relative to
Aroclor 1242 (AMW) (Figure 3-3). Thus the differences seen in the two diagrams represent the same
processes in different ways. The plot of change in MPAJ+ vs the 1984 MPA3„ tracks the combined
effects of dechlorination and PCB loss to the overlying water column since both processes can lower
the ETri+ inventory to essentially zero. The plot of change in total PCBs as MPA vs 1984 MPA3+
4-9
TAMS
-------
is much less sensitive to the dechlorination process since this process can only lower the total PCB
inventory by 26 percent as opposed to loss to the water column which can potentially reduce the
inventory to zero.
The difference in these processes and their effects on the sediment inventory can be seen
more readily in log scale as shown in Figure 4-7. In the upper diagram, the loss in MPA3. vs the
1984 MPAj. is plotted. The diagram shows that the majority of ETri+ loss from the sediment is
relatively minor in magnitude although nearly all samples exhibit some loss. This suggests that either
no sediment PCB inventories have increased or else that increased inventories (presumably by
deposition) were subsequently dechlorinated, resulting in no net increase in the ETri+ inventory. The
latter scenario is supported by the lower diagram in Figure 4-7. This diagram shows the change in
the total PCB inventory (MPA) vs the 1984 MPA}+ In this diagram, both positive and negative
inventory changes are evident, indicating that some locations have seen gain despite their decrease
in ETri+. Note as well that the level of gains are in some cases similar in scale to the larger losses.
Since PCB moles are largely conserved by dechlorination and lost by re-release to the water
column, it should be possible to separate this process by tracking the PCBs on a molar basis. This
requires an estimate of PCB molar inventory for both 1984 and 1994. The mole concentrations and
moles per unit area can be calculated directly from the 1994 data set since it is congener specific.
Calculation of the moles present in 1984 is a little less straightforward.
Knowing that the 1984 measurements represent the mass of ETri+, what is needed is an
estimate of the mean molecular weight of this fraction at the time of sediment deposition to permit
the calculation of the moles present at the time. This can be estimated in two ways. First, the mean
molecular weight of the ETri+ fraction can be obtained from Aroclor 1242 directly since the
congener composition of Aroclor 1242 was measured as part of the Phase 2 investigation. Based on
this analysis, a mean molecular weight of 275 grams per mole (g/mole) is obtained for the ETri+
fraction. Alternatively, this value can be estimated from the recently deposited sediments of the high
resolution cores. Figure 4-8 shows the relationship between the mean molecular weight for the ETri+
fraction and the AMW of the sample for the entire set of Upper Hudson high resolution core
samples. Also plotted on the diagram are the results for the low resolution cores, although these are
4-10
TAMS
-------
not used in the regression. Given that dechlorination is restricted to the sediments, it is likely that
depositing sediment will be similar to Aroclor 1242 in molecular weight since this was the major
form of PCB released. That water column contamination was derived from Aroclor 1242 or its
altered form in the sediments was demonstrated in the DEIR (TAMS, et al., 1997). In fact, because
of the nature of PCB partitioning, suspended matter appears more Aroclor 1242-like than the
associated whole water sample, even when the water column load is highly altered. Thus
extrapolating the trend of molecular weight for ETri+ vs AMW back to a value for AMW of zero
yields an alternate estimate for the molecular weight at the time of deposition. As shown in Figure
4-8, a regression line through the high resolution core data yields a value of 281 g/mole for the
molecular weight at a AMW of zero. Alternatively, the high resolution core data themselves appear
to cross the AMW value of zero at roughly 282 to 285 g/mole, providing a third estimate. Note in
the diagram that few samples exceed the value of 285 g/mole, suggesting this may be close to the
actual molecular weight at time of deposition. The range of molecular weights based on these three
values is 275 to 285 g/mole, or less than a 4 percent range. Thus the choice of any of these will have
little effect on the mole estimates. The value obtained from the zero intercept of the curve in Figure
4-8, i.e., 281 g/mole was used as the molecular weight estimate for ETri+ at the time of deposition.
Having a molecular weight estimate for the PCBs in sediments in 1984, the change in mole
content of the sediments can be calculated using a few additional assumptions. Since the 1984
measurements best describe the ETri+ inventory, no additional data are available to define the
concentration of lighter congeners. Changes in moles between 1984 and 1994 must then be
constrained to only the PCB molecules measured in 1984, i.e., the trichloro and higher homologues.
However, in 1994, not all of the molecules present as trichloro and higher homologues will still be
in their original form. Many will have been dechlorinated during their residence in the sediments and
converted to the congeners BZ# 1,4,8,10 and 19. Thus to completely account for the original ETri+
molecules, the 1994 mole tally must include the current inventory of ETri+ as well as the
dechlorination product congeners. However, the occurrence of lighter congeners in the sediment in
1984 would also be included in the mole tally for 1994 if the congeners are still present in the
sediment at the time of low resolution core collection. Since no information is available to constrain
the estimate of lighter congeners, their contribution to the current sediment inventory cannot be
calculated and will be assumed to be zero. This clearly represents an underestimation but it will serve
4-11
TAMS
-------
to provide a minimum estimate or lower bound on the number of moles lost from the sediment
between the 1984 and 1994 sediment surveys.
It should be noted here that the ETri+ sum for the 1994 sediment samples represents all
trichloro and higher congeners excluding BZ#19. This was done to separate the dechlorination
products from those potentially alterable. This difference could not be made for the 1984 sediments
since no congener specific data were available. This represents a minor error if the sediments were
unaltered since BZ#19 is less than a few percent in Aroclor 1242. If the sediments were extensively
altered, this error simply becomes part of the larger uncertainty associated with the assumption that
no lighter congeners were present. In any event, the result is still a minimum estimate for the moles
lost.
The range of mole loss and the range of loss in ETri+ as MPAJ+ loss in moles is represented
in Figure 4-9. This figure shows the definitive loss in the ETri+ relative to the mole differences
which are centered more closely to zero. When the 60 low resolution core/1984 core and grab pairs
were considered as a whole, the distribution for the mass change was found to be statistically
different from zero while the molar change was not. This figure again demonstrates that loss in
ETri+ does not directly correspond to loss in PCB moles.
These results are presented graphically on Plate 4-20 as well. This plate is a map of the TI
Pool showing the coring locations considered in this analysis. At each matched coring location, a
circle is plotted whose diameter is proportional to the change in the ETri+ mass due solely to molar
loss or gain. Red circles indicate loss while black circles indicate gain. Losses represented on this
map are exclusive of dechlorination and instead represent sites exhibiting mass loss to the overlying
water column via one or more transport mechanisms. Gains are presumably due to post-1984
deposition. The distribution of red and black circles on Plate 4-20 show some areas to be
experiencing consistent losses (e.g. hot spots 10 and 16), others gains (e.g. northern end of hot spot
8) while most exhibit a mixture of losses and gains. Overall there are more locations exhibiting loss
than gain. The next section of this report examines the relationships within the data to aid in the
understanding of the distribution of losses and gains seen in Plate 4-20.
^nni51
4-12
TAMS
-------
4.1.2 Assessment of Sediment Inventory Change Based on the Original 1984 ETri+ Sediment
Inventory
With the measures of sediment inventory (MP A and MP A,.) and the estimates of molar
content, it is possible to assess the changes in the sediment inventories due to PCB loss or gain from
the water column as well as to dechlorination. The goal of the analysis presented here is to find
useful groupings of the data which can provide a basis for applying the low resolution core results
to larger areas of the TI Pool. As a first step, the PCB mass and PCB molar differences were
compared statistically. The results were grouped based on the 1984 ETri+ sediment inventories
(MPAj.)- Specifically, the data were grouped into cores from areas less than 10 g/m2 and greater than
10 g/m1. There were too few samples in the lower concentrations to further subdivide the groups
without losing much of the statistical power of the tests. The selection of this split point was based
on the regression line shown in Figure 4-5 which crosses the diagonal near this value. A value of
10 g/m2 is generally characteristic of the more contaminated sediments of the Upper Hudson. The
length-weighted average concentration corresponding to 10 g/mJ is approximately 12 mg/kg. Peak
concentrations for these cores are 50 mg/kg. Thus the greater-than-10-g/mJ group corresponds to
sediments typically found in hot spot areas. To assess these differences, the results were compared
on both the absolute difference in inventory as well as a relative difference. The relative difference
for MPA was calculated as follows:
1994 MP A - 1984 MPA^
&PCB = 1 (41'5)
PCB 1984 MPA3
where: ApcB is the relative or fractional change in mass per unit area (Delta^)
1994 MPA is the total PCB mass per unit area, for 1994 results; and
1984 MPA3, is the STri+ mass per unit area for 1984 results.
The relative difference for the molar inventory was similarly calculated as:
4-13
TAMS
-------
1994 Total Moles 1984 27W + Moles
m 2 m 2
AW ffl
= (4.1-6)
1984 S7W + Moles
m 2
m
where: AM is the relative or fractional change in moles per unit area (Delta,^
1994 Total Moles/m2 is the molar sum of the trichloro and higher homologues plus the
moles of BZ# 1,4, 8,10, and 19 on a unit area basis; and
1984 2Tri+ moles/m2 is the molar sum of the trichloro and higher homologues on a unit area
basis.
The delta function is not a normally distributed one as can be seen in the first diagram in
Figure 4-10. Specifically, the loss range is between 0 and 1 while the gain can range to infinity. The
log-transform of this data helps to remove some of the skew but it is still not normal. Since a
log-transform of a negative number is not defined, a value of 2 was added to all Delta results before
performing the log transform. Although the log-transformed data are still not normal, the distribution
is sufficiently close to normal to permit to the use of parametric tests to assess change.
Non-parametric tests will be used to confirm the statistical significance determined from the
parametric tests.
The first analysis presented for the >10 g/m2, <10 g/m2 MPA data grouping is shown in
Figure 4-11. This figure shows the distributions for the absolute difference in molar inventory for
the two groups. Declines in this inventory represent molecular loss from the sediment, indicating
PCB release from the sediment to the overlying water, or could result from destruction of PCBs.
Increases suggest additional deposition of PCB-bearing sediments. Evident in the figure is the
extensive overlap between the two groups. This overlap is confirmed by the Tukey-Kramer circles
shown on the right of the figure. Statistically significant differences between groups are indicated
when the circles are separate or touch only slightly. In this instance, the circles indicate that these
groups are not significantly different for the molar inventory change. This analysis indicates no trend
in absolute molar inventory change with 1984 PCB sediment inventory. Further explanation of the
4-14
TAMS
-------
Tukey-Kramer circles can be found in the key sheets to Appendix F. A definition of the Tukey
Kramer honestly significant difference (HSD) on which these circles are based is given the glossary.
In Figure 4-12, the absolute mass change is examined as a function of these two groups.
Again the groups show substantive overlap which is confirmed by the Tukey-Kramer circles. This
analysis also indicates the absence of a trend in absolute mass difference with 1984 PCB sediment
inventory. It should be noted in the data represented in both Figure 4-11 and 4-12 that, although the
two groups were not statistically different from each other, the >10 g/m2was statistically less than
zero, confirming the occurrence of net PCB loss from the sediment for this group.
Examining the differences in PCB inventory on an absolute basis was not particularly
fruitful, in part because the magnitude of the PCB change can represent both analytical variability
as well as real change. By examining the relative changes in inventory using the delta functions
described above, much of the analytical variability can be diminished in importance relative to real
change since analytical variability is typically small relative to the absolute measurement values.
Figure 4-13 presents the analysis for the relative change in molar inventory (Deltas,) as a function
of the two inventory groups. Note that the results are based on the function:
log(DeltaM + 2)
since the log of values less than or equal to zero is undefined. In this analysis, the groups are found
to be statistically different, with sediments >10 g/m2 exhibiting a large relative loss of approximately
28 percent.
This result indicates that on average, higher concentration sediments have shown a
substantive, statistically significant loss which cannot be accounted for by dechlorination. While the
mechanism(s) responsible for this loss are not known, some cores clearly suggest scour as a possible
loss mechanism. Other transport processes may also be important, such as preferential flux from the
sediments to water column of monochlorobiphenyl and dichlorobiphenyl dechlorination products,
as suggested by the seasonal PCB release patterns noted in recent GE water column data. Most
importantly, though is the clear loss of PCB molecules from sediments of higher inventories,
4-15
TAMS
-------
presumably to the water column. This result confirms the findings of the DEIR wherein water
column PCB contamination was shown to originate within the TI Pool. The analysis presented
documents a likely candidate for this source, i.e., the sediments of relatively higher contamination.
This is not to suggest that all Upper Hudson water column contamination is associated with the TI
Pool sediments. As will be shown later in this report, sediments below the TI Dam also show PCB
loss. This is also consistent with the DEIR which indicated the presence of additional PCB input to
the water column downstream of the TI Dam during a spring sampling event (Transect 3) and
possibly during summer sampling as well.
This analysis also indicated the occurrence of a net inventory gain for the sediments less than
10 g/m2. In this instance the average gain was 104 percent, representing a potential doubling of the
sediment inventory. However, this gain must be considered in light of the way it was derived.
Specifically, the estimate of the 1984 molar inventory represents a lower bound estimate for the
inventory due to the fact that no information was available concerning the lighter congeners (i.e.,
monos and dis). This premise serves to minimize any molar loss estimated between 1984 and 1994.
Thus statistically significant losses have a high probability of being true and in fact larger than
estimated. However, gains have the opposite concern. Because the 1984 estimate is a lower bound,
an apparent gain would be expected based on undercounting of monochloro- and
dichlorohomologues. Therefore, statistically significant gains may be much less than estimated and
have the potential not to be real. Thus from the analysis presented above, it can be concluded that
definitive PCB losses have occurred for sediments greater than 10 g/m2 to a degree greater than that
estimated here, presumably via loss to the overlying water column. In addition, the data suggest
inventory increases for sediments less than 10 g/m2, although to a degree less than that estimated
here.
As a final part of the analysis presented here, two non-parametric tests were conducted to
confirm these findings. Specifically, a rank sums test and a median test were performed on these
data. These results confirmed the statistically significant difference between the two data groups at
a greater than 99 percent probability. These results are included along with the rest of the statistics
for this analysis in Appendix F.
4-16
TAMS
-------
The next analysis examined the relative change in sediment mass (Delta^) and is shown in
Figure 4-14. Note that the results are again based on the function:
log(DeltapcB + 2)
since the log of values less than or equal to zero is undefined. This analysis also indicated a
statistically significant decline in sediment inventory for sediments greater than 10 g/m2. In this
instance the decrease was estimated to be 39 percent. The decrease in mass per unit area for these
sediments represents the sum of both the dechlorination loss and the loss to the overlying water
column. In fact, this loss is quite consistent with the other independent determinations of loss.
Specifically, the previous analysis of molar change suggested an inventory decrease of 28 percent
via loss to the water column. Assuming that all molecules are equally likely to be lost as would be
the case in a scour scenario, this 28 percent molar loss translates to a 28 percent mass loss. In the
discussions in Chapter 3, a mean dechlorination loss of 12 percent was found for sediments of the
Upper Hudson based on the change in molecular weight (AMW). These two processes yield an
approximate mass loss of 40 percent which compares closely with the direct measure of mass loss
estimated here.
The Deltapce also yielded an inventory gain for sediments less than 10 g/m2. This was
estimated to be 87 percent. Like the molar inventory gain, however, this estimate represents an upper
bound on the actual inventory gain. These results were confirmed using the non-parametric tests
mentioned previously, which yielded a greater than 99 percent probability for significant difference
between the groups. The test results are contained in Appendix F.
The 1984 grab samples were examined as an additional sub-grouping of the MPA groups
(Figure 4-15) to evaluate the accuracy of the grab samples relative to the cores in regard to sediment
inventory. The grab samples appear to yield potentially lower values for DeltaM but no statistical
difference was found. To some degree this analysis is limited by the number of grab samples
available (15).
4-17
TAMS
-------
4.1.3 Assessment of Other Potentially Important Characteristics
Two other properties were used as criteria to group the Delta values and look for statistically
significant relationships. Specifically Tie and a cohesive/noncohesive sediment properties were used
to group the sediments. The 7Be results from the surficial core layer (0-1 in.) were used on a
presence/absence basis to group the Delta data. This yielded two statistically significant groups, with
the Tie absent group yielding a lower value for Delta*,. This may indicate that the absence of 7Be is
a relatively good indicator for sediment PCB loss, presumably to the water column. This is a
somewhat different result than that found for sediments below the TI Dam as discussed in the next
section. Part of this difference probably results from the lack of one-to-one sample correspondence
between the earlier and current study for samples collected below the TI Dam. A statistical summary
sheet is provided in Appendix F for this analysis.
The data were also grouped based on a cohesive/noncohesive sediment classification
developed by Limno-Tech and reported in the Preliminary Model Calibration Report (LTI, 1996).
This classification was largely based on the side-scan sonar results. In this analysis, a general trend
toward higher inventory losses was seen for cohesive relative to noncohesive sediment but it was
only significant at the 90 percent confidence level. Statistical summary sheets are provided in
Appendix F for both the 7Be and cohesive/noncohesive analyses.
4.1.4 Implications of the Inventory Assessment
The implications of the inventory changes noted above are important in assessing the current
status of the TI Pool sediments. Specifically, the inventory losses represented by DeltaM are
presumed to apply to all similar sediments. This implies that most hot spots within the TI Pool have
undergone substantive losses to the water column via scour or some other release mechanism. The
estimate for the degree of mass loss via this process is 28 percent although this is a lower bound
estimate. Given that the inventory for these hot spots is on the scale of ten or more metric tons, this
mass loss is consistent with some of the recent estimates of release to the water column of 0.5 to 1
kg/day. Over ten years, this water column load would represent 1825 to 3650 kg. This is not to imply
that there have been no additions to the PCB inventory of the TI Pool. As noted in the DEIR,
4-18
TAMS
-------
upstream releases may have served to add PCBs to the TI Pool sediments. The point here is to denote
that, nonetheless, sediment inventories from the more contaminated areas have declined consistent
with a sediment release process. Again dechlorination, while present, is limited in its ability to cause
PCB inventory decline and is clearly overshadowed by the scale of the sediment PCB losses to the
water column. Figure 4-16 shows the distribution of the nearly the entire set of PCB mass/area
samples for the trichloro and higher homologues, based on the 1984 survey. Roughly one third of
these samples represent sediments which would be expected to see this level of PCB loss.
The inventory assessment also yielded an upper bound on the expected gain that might be
seen for the low level (<10 g/m2) sediment. While some of the inventory gain would be expected to
result from the upstream releases from the GE facilities, it also is probable that some portion of the
gain in low-level sediment is from the redistribution of PCBs from more heavily contaminated areas.
4.2 Sediment Inventories of the Upper Hudson Below the Thompson Island
Dam
As part of the low resolution coring program, six hot spots below the TI Dam were selected
to be surveyed with some additional coring in a limited number of areas. During sampling the field
crew unintentionally extended the coverage around Hot Spot 34 to include Hot Spot 35, bringing the
number of hot spots covered to seven. These hot spots were originally chosen since they represented
the majority of hot spot contamination (74 percent) below the TI Dam (Tofflemire and Quinn, 1979).
Table 4-2 provides summary information on these hot spots.
The original premise for the low resolution coring program below the TI Dam was to provide
independent estimates of sediment inventories in a number of hot spot areas to establish current
conditions and to compare with the 1976-1978 NYSDEC survey. Some additional coring was also
planned to explore other areas of potential contamination. To meet these objectives, 5 to 14 cores
were placed within each of the six main hot spots as well as several just beyond each hot spot
boundary. The cores within each hot spot area were averaged to obtain an estimate of the
concentration in mass per unit area for the hot spot. The coring locations outside the hot spot
boundary were used to confirm the boundary itself. Cores placed outside the boundary were
4-19
TAMS
-------
expected to yield substantially lower concentrations per unit area. Additional exploratory cores were
placed in Hot Spot 35 (4 cores) as well as in three additional study areas.
Plates 4-21 to 4-28 show the locations of the hot spots surveyed by the low resolution coring
program, the original 1976-1978 NYSDEC sampling sites, and the sediment classifications
developed from the side-scan sonar for the areas covered. Plate 4-29 is a key to the locations of
Plates 4-21 to 4-28. Original side-scan sonar images for these locations were presented in Plates
3-11 to 3-20. Plates 4-21 to 4-28 also provide the LWA concentrations for shallow sediments at 0
to 12-inches for both surveys. The location markers are coded to indicate 7Be absence or presence
for the 1994 low resolution core survey and to indicate a core or grab sample for the 1976-1978
NYSDEC survey.
4.2.1 Calculation of the Length-Weighted Average Concentration (LWA) and Mass Per Unit
Area (MPA) for Sediment Samples Below the TI Dam
Before estimating current sediment inventories or comparing previous and current conditions,
it is first necessary to establish a consistent measurement basis for sediment PCB contamination.
The issues here are similar to those raised for the 1984 to 1994 survey comparisons discussed in
Section 4.1. While estimation of current inventories can be done in a similar fashion to the TI Pool
cores, the comparison to the previous studies requires resolution of analytical techniques and
sampling procedures between the 1994 low resolution core survey and the 1976-1978 NYSDEC
survey.
The 1976-1978 sampling program results were examined by a number of investigators who
translated the results into area estimates and inventories (Tofflemire and Quinn, 1979; MPI, 1978;
Brown et al, 1988; and MPI, 1992). In particular, the 1992 study by Malcolm Pirnie, Inc.
established sediment inventories for a series of potential dredge areas as part of a study undertaken
for NYSDEC. This study was never finalized but a draft report was made available to the USEPA.
The data organization and presentations in the 1992 MPI report were used extensively in this report,
as were the original hot spot estimates made by Tofflemire and Quinn, (1979) and MPI (1978).
4-20
TAMS
-------
Analytically, the 1994 and 1976-1978 sampling programs are based on different procedures.
The low resolution core survey PCB analyses were based on congener-specific capillary column gas
chromatography. The 1976-1978 NYSDEC survey in contrast used packed column gas
chromatography with Aroclor-based standards, representing Aroclors 1221, 1016, and 1254. These
Aroclor mixtures have relatively few common congeners among them so "double-counting" of
congener concentrations is unlikely. The simple sum of the reported Aroclor values yields an
estimate for total PCB concentration, which is probably the best that can be done to create a value
for comparison to the low resolution coring results. Although this sum is a best estimate for total
PCBs, the estimate is still limited in its quantitation of the lightest congeners. Specifically BZ #1
and # 4 were not directly quantitated and may have been present at elevated levels relative to Aroclor
1221 due to dechlorination processes. Thus, the sum of Aroclors may represent a lower bound on
the actual PCB concentration in the sediments in 1976-1978.
Sampling techniques were also different between the 1976-1978 and 1994 surveys. In 1976
to 1978 both core and grab samples were collected by NYSDEC. Cores were typically 12 to
16-inches in depth and subdivided into two to three segments for analysis. Most PCB contamination
(greater than 80 percent) was found within the top 12-inches (MPI, 1992). Grab samples were
assumed to penetrate to about 4-inches (Tofflemire and Quinn, 1979; MPI, 1992). In calculating
sediment inventories, core results were simply integrated by length. Grab sample results were
extrapolated to 12-inches, based on a factor developed from the coring results. The factor for the
grab samples was developed by Malcolm Pimie, Inc. (MPI, 1992) for NYSDEC based on the ratio
of surface sediment PCB levels (0 to 4-inches) to those of the entire core.
To calculate the sediment PCB inventory and to compare 1976-1978 conditions with those
in 1994, PCB results from both programs were reduced to LWA and MPA values. The calculation
of the LWA values was relatively straightforward since most cores were analyzed to a depth of at
least 12-inches and the grab sample data had already been corrected to this depth by MPI. LWA was
generated for shallow sediments, at 0 to 12-inches deep, for both 1976-1978 and 1994 data sets. For
the 1994 data when the top-most segment ended above the 12-inch mark (e.g., a nine-inch top
segment), the remaining inches were included in an LWA by using the concentration of the next
deepest layer for just the needed inches. When the top-most segment was greater than or equal to
4-21
TAMS
-------
12-inches, the reported concentration for the segment was used without modification.
Length-weighted averages for 0 to 12-inches were previously determined by Malcolm Pirnie, Inc.
(MPI, 1992) for the 1976-1978 core and grab samples, as mentioned previously. Due to
discrepancies between the values reported by MPI and those contained in the electronic NYSDEC
database, the results could not be cross-checked. As a result, the MPI results, which were already
in the desired form, were used in this report.
Determination of the MP A for the 1976-1978 data set was more problematic due to the lack
of density data for this study. Although it is never explicitly stated in the original report by
Tofflemire and Quinn (1979) nor in the analytical report by O'Brien and Gere (1978), it can be
inferred from the way the data are used that the original PCB analyses were reported on a dry-weight
of sediment basis. Therefore, to calculate the PCB mass per unit area from these results, the solid
specific weight (weight of dry sediments per unit volume of wet sediments) is required. Since this
is an essential part of the calculation, estimated values for the solid specific weight (SSW) were
obtained based on the strong correlation between total PCBs and SSW for the low resolution coring
program as shown in Figure 3-15. When this plot is reversed by using the total PCBs for the
grouping (bin) values, the median and mean SSW values for each PCB range can be obtained from
the relationship as shown in Figure 4-17. Utilizing the 1976-1978 total PCB concentrations, SSW
values were obtained for the NYSDEC core and grab samples based on the bin mean values listed
in Table 4-3. With this parameter, the MPA values for the 1976-1978 sampling locations could be
calculated by the formula described in Section 4.1. Note this correction for SSW represents a
potentially large (20 to 50 percent) correction to the original MPA estimates generated by MPI
(1992) and Tofflemire and Quinn (1979). Specifically, the earlier estimates assumed a SSW of 1
g/cc for all sediments. In light of the current low resolution coring work, this assumption was found
to be an oversimplification. At the PCB levels reported by MPI (1992), SSW ranged from 0.5 to
0.79 g/cc with the lowest SSW values at the highest concentration. When using a constant SSW of
unity, the contributions by the most contaminated sediments will be overestimated. For the
discussions on MPA which follow, the 1976-1978 MPA values have been calculated using the low
resolution core SSW to PCB concentration relationship given in Table 4-3. These results will be
contrasted against the original MPI (1992) and Tofflemire and Quinn (1979) estimates under the
discussion of hot spot PCB mass later in this section.
4-22
TAMS
-------
The 1976-1978 PCB results in the TAMS database and in this report were obtained directly
from NYSDEC (Bopp, 1990), as well as from the MPI (1992) report. In particular, the extrapolated
grab sample results and the length-weighted average core values were taken directly from the MPI
report. Visual sediment classification data and grain-size distribution data were obtained
electronically from NYSDEC as well as from Normandeau (1977).
4.2.2 Comparison of 1976-1978 Sediment Classifications and the Side-Scan Sonar
Interpretation
Before beginning the discussion of PCB inventories for 1994 and 1976-1978 data sets, it is
useful to establish the comparability of the sediment classification data available for the two data
sets. A good level of agreement between the 1976-1978 and 1994 Phase 2 conditions serves to
support the use of the historical data set as a good characterization of the river, as well as to further
confirm the internal consistency of the various sediment classification data sets. Consistency
between 1976-1978 sediment classification data and the current conditions would also imply that
river-bottom conditions have remained relatively constant over time, that is, areas with fine-grained
sediment tend to remain fine-grained and coarse-grained sediments tend to remain coarse-grained.
This comparison was made by contrasting the principal fraction classifications used by NYSDEC
and its consultants for the 1976-1978 samples with the classification developed from the side-scan
sonar images. This is similar to the analysis presented in Chapter 3 for the low resolution core
grain-size distribution data and the side-scan sonar classifications. However, consistency in
sediment classification cannot be used to infer that the sediments do not move or that their PCB
inventories are unchanged. As was discussed in Section 4.1 and presented below, PCB inventories
have substantively changed in many instances.
Sediment classification data were obtained for a total of 493 locations throughout the area
of side-scan sonar coverage. These locations were obtained for both the Tl Pool and for areas below
the TI Dam. The sediment classification data for the 1976-1978 survey was obtained from two
sources: a report by Normandeau Associates, Inc. (1977) and an electronic data file transmitted to
the USEPA (Bopp, 1990). The data from the Normandeau report represented standard grain-size
distribution data reported on a phi scale plus silt and clay fractions. The data from both sources were
4-23
TAMS
-------
converted to a principal fraction by summing across phi bins as shown in Table 4-4. The largest
percent fraction among the classifications developed for each sample, i.e., coarse-sand, medium
sand, fine-sand, silt, and clay, was defined as the principal fraction. Of the 405 samples reported by
Normandeau (1977), 278 were located within the areas covered by side-scan sonar.
Data for other locations were contained in the electronic data files obtained from NYSDEC.
These data contained a numerical code describing the sediment texture. The original sediment
description for these codes was contained in Tofflemire and Quinn (1979). These codes were used
to assign the principal fraction classification as shown in Table 4-5. Using this translation, the
electronic files provided an additional 215 locations within the side-scan sonar coverage, bringing
the total number of 1976-1978 survey locations to 493. A new sediment classification, "muck", was
used for the description data set only. The precise sedimentological classification for these "muck"
samples is not known, but they are believed to represent organic-rich silts and fine-sands.
The comparison between the NYSDEC classifications and those by side-scan sonar
interpretation is given in Figure 4-18. This figure shows the sample data organized by NYSDEC
principal fraction and side-scan sonar classification. The primary sediment category collected during
the 1976-1978 survey was fine-sand. The agreement between the NYSDEC and side-scan sonar
classification appears quite good. Beginning with the silt and muck classifications, it is apparent that
the majority of the silt and muck samples were obtained from areas classified as fine-grained.
Similarly, medium-sand and coarse-sand samples by the NYSDEC survey were generally mapped
to coarse-grained sediments by the side-scan sonar interpretation. As a group, medium-sand and
coarse-sand samples were mapped as coarse sediments roughly 72 percent of the time. For silt and
muck, the agreement was also good but not as accurate, with about 57 percent of the samples
mapping as fine-grained sediment.
All NYSDEC sediment classifications except for clay had a small but non-trivial number of
sites which were mapped as rocky locations by the side-scan sonar. These sites are attributed to
pockets of sediment located in and among the rocks or rocky outcrops on the river bottom. Although
rocks would be the most prevalent kind of material in these areas, the standard sampling procedures
used by NYSDEC (i.e., coring tubes and relatively small grab samplers) would not reflect this.
4-24
TAMS
-------
Instead, these samplers would focus on the coarse-sands to fine sediments found among the rocks
which are more readily collected by these samplers. Thus, rocky areas yield a variety of sample
types. When rocky areas are excluded from the tally, the fine-grained fractions of silt and muck
yield a 66 percent rate of agreement with side-scan sonar. Similarly, the coarser fractions, medium-
sand and coarse-sand as classified by NYSDEC, yield an improved agreement of 83 percent when
the rocky locations are excluded.
Only seven NYSDEC sampling locations were classified as clay, yielding a very small group
to assess. Nonetheless, only two of the seven samples were obtained from fine-grained sediment
areas as defined by the side-scan sonar. The reason for this is unclear but may represent small
pockets of glacial lake clays located in among coarse-grained sediment areas. During the Phase 2
confirmatory sampling, all clay samples collected were associated with historical glacial lake
deposits.
Fine-sands yielded the greatest number of samples in the NYSDEC data set. The results map
out as approximately 55 percent coarse-grained sediment and 45 percent fine-grained sediment based
on the side-scan sonar, when rocky locations and the other minor areas are excluded. This split in
area type is very consistent with the results obtained for the low resolution cores, as shown in Figure
3-28. Note that Figure 3-28 uses bins based on the side-scan sonar assignments and maps the
grain-size classification whereas Figure 4-18 uses bins based on the NYSDEC classifications and
maps the side-scan sonar assignments. Fine-sand samples are approximately evenly split (52 percent
coarse-grained and 48 percent fine-grained) using the side-scan sonar classification and the
NYSDEC results (upper diagram of Figure 3-28). These results are consistent with the resolution
afforded by the side-scan sonar images, in that the acoustic signal (DN50) value used to separate
fine-grained and coarse-grained sediments (55 to 60) roughly corresponds to the middle of the range
of DN50 values obtained for fine-sands, as shown in Figure 3-30. Thus an even split of fine-sand
samples among fine-grained and coarse-grained sediment areas would be expected for both the low
resolution core sites and the NYSDEC sampling locations.
Although the rates of agreement between the NYSDEC classification and the side-scan sonar
classifications are not as good as that for the low resolution core samples, the agreement is still
4-25
TAMS
-------
considered acceptable. The poorer agreement is due in part to the use of visual classifications, as
well as standard grain-size distribution analysis in this comparison. There is also the possibility of
changes in sediment type in a limited number of cases. Overall, the historical 1976-1978 NYSDEC
grain-size distribution results appear quite consistent with the current Phase 2 side-scan sonar
classifications and the Phase 2 sediment grain-size distribution results. Historical silt, fine-sand, and
medium-to-coarse-sand samples all map onto the side-scan sonar classifications in a manner
consistent with the low resolution core data. This result supports the contention that Hudson
sediment classifications have remained relatively constant over the last 15 years for large areas of
the river bottom. That is, large areas of fine-grained sediments as classified in 1976-1978 are still
areas of fine-grained sediments. Similarly, large areas of coarse-grained sediments also remain as
originally classified. However, these results cannot be used to imply that the PCB inventories
remain intact. The status of the PCB inventories below the TI Dam is discussed in the next section.
4:2.3 Comparison of Sediment PCB Inventories: NYSDEC 1976-1978 Estimates versus 1994
Low Resolution Core Estimates
The 1976-1978 and 1994 data taken below the TI Dam represent focused sampling programs
centered on the areas of greatest PCB contamination. In particular, the low resolution coring
program was centered on a limited number of hot spots, so as to provide a basis for comparison to
the earlier survey without duplicating the scale of the effort.
In order to make a comparison between the two surveys, a common sampling basis had to
be established. Clearly defined hot spot areas were needed to divide samples into those outside and
inside the hot spot areas. The original report (Tofflemire and Quinn, 1979) did not provide maps
with sufficient resolution so as to permit a clear definition of the hot spot boundaries. The 1992 MP I
report defined a series of smaller areas, called dredge locations, whose borders approximated those
of the original hot spots. Twelve dredge locations were used to represent the seven hot spots
surveyed during the low resolution coring program. Using these borders, both NYSDEC and Phase
2 sampling locations were designated as internal or external to the area.
4-26
TAMS
-------
This procedure was used to designate sample association for Hot Spots 25,28,31,34,35. 37,
and 39, the seven hot spots studied by the low resolution coring program. Incidentally, one of the
areas covered by the exploratory cores (TAMS location 42) also coincided with MPI dredge location
182. This area was added to the evaluation of sediment inventories, bringing the total number of
dredge locations to 13. Within these areas, MPI identified 111 points. However, upon review of the
maps supplied in MPI (1992), two additional points were noted in Hot Spot 37, yielding a total of
113 sampling locations from 1976-1978. The low resolution coring study occupied 64 coring
locations from within these areas. Plates 4-21 to 4-28 show the 1992 MPI dredge location areas as
well as the sampling locations for both sampling programs.
To compare the PCB levels within these areas, arithmetic and geometric means were
calculated. Because of the log-normal nature of the data distribution for both data sets, the geometric
mean and its standard error provide the best statistical basis to assess change in the sediment
inventories over time. The log-normal nature of the entire 1976-1978 data set was originally
established by Tofflemire and Quinn (1979). The subset of 113 NYSDEC samples was also
log-normally distributed, as seen in Figures 4.2-3 and 4.2-4. These figures show that both the
one-foot length-weighted averages (LWA) and the SSW-corrected PCB mass per unit area estimates
(MPA) are log-normally distributed. Similarly, Figures 4.2-3 and 4.2-4 show the LWA and MPA
distributions for the subset of 64 low resolution cores from the seven study areas below the TI Dam
as well as for all low resolution core results below the TI Dam. These results were determined to
be log-normally distributed using the Shapiro-Wilk W test for normality (Table 4-6).
Geometric means were compared in log space to determine the statistical significance of
changes in the sediment inventories. That is, the average of the log-transformed values for each hot
spot or dredge area for the 1976-1978 survey was compared to the average of the log-transformed
values for the same hot spot or dredge area based on the 1994 low resolution coring study. The
geometric mean (CG) was calculated as follows:
4-27
TAMS
-------
where: C,
n , = i
ln(Q =
n
L WA or MPA for each sampling location in a hot spot or dredge area;
number of samples in the hot spot or dredge area;
mean log for the hot spot or dredge location; and
natural log of Q.
The comparisons were made using the Tukey-Kramer honestly significant difference test
(JMP, 1994) which tests for the statistical significance of the difference between means of sample
populations when the populations are normally distributed. In this instance, data were normalized
by performing a log-transformation. Non-parametric tests (i.e., tests that do not require a normal
distribution) were also used to test for differences between populations.
Although the geometric mean and log-transformed data provide the best basis for establishing
the occurrence of change in the sediment PCB inventory, the arithmetic mean must be used if the
degree of change is needed. The arithmetic mean for the LWA values can be determined in two
ways as follows:
1.
Arithmetic Mean
(4.2-2)
n> = l
where: C,
LWA or MPA
n
number of samples in study areas;
Arithmetic Mean (C/A) =e
i A
- £ in (c,)
(4.2-3)
2.
where: C„ n, and In (Cj) are defined as above,
Sy2 = variance of the sample log values, and
*Pn(t) = an infinite series as follows:
4-28
TAMS
-------
(0 = 1+ ("~1)r + C"-1^2 + ("-'A3 + (n-l)V
n 2ln 2(« + l) 3!wJ(w + l) (n-3) 4!rt4(«+l)(«+3)(n+5)
(4.2-4)
where:
2
The first formula represents the simple arithmetic mean of the data. Given the log-normal
distribution of the data set and the limited number of samples for each hot spot or dredge location,
this formula does not provide an unbiased estimate of the true arithmetic mean. However, it is
possible to take advantage of the knowledge that the underlying data set is log-normal. The second
formula represents a minimum variance, unbiased estimation of the arithmetic mean, given that the
underlying distribution is log-normally distributed (Gilbert, 1987). It is also possible to obtain the
standard error of the estimator, given the same assumptions about the data set. The standard error
of the arithmetic mean is calculated as a'A.
where: C„ n, 1„, t and Sy2 defined as above. Table 4-7 provides a summation of the geometric and
arithmetic means.
The relationship among the three estimates can be expressed as CG < CA < C'A. The
geometric mean is less than the simple arithmetic mean which is less than or equal to the minimum
variance unbiased estimate of the arithmetic mean. This relationship holds in all hot spots for both
the length-weighted average (LWA) and the PCB mass per unit area (MPA) except for Hot Spots
25 and 39 for the 1976-1978 data and Hot Spots 34 and 35 for the 1994 data. In each of these cases,
the unbiased estimate (C\) was roughly 5 to 10 percent lower than the simple arithmetic mean (CA)
The geometric mean (CG) was substantially smaller than the arithmetic mean, which is typically
characteristic of log-normal distributions.
—/
O
W
n
(4.2-5)
4-29
TAMS
-------
Using the log-trans formed data, the 13 dredge locations were evaluated for statistically
significant changes in the PCB levels between the 1976-1978 and 1994 surveys. This evaluation is
summarized in Figure 4-21. The figure presents the geometric mean values with error bars
representing two standard errors about the mean. Each paired set was tested for statistically
significant difference using the Tukey-Kramer honestly significant difference (JMP, 1994).
Statistically different values are represented by larger, solid markers. In this evaluation, only three
of the thirteen dredge locations were found to be different based on LWA, and only four differed
based on MPA. This lack of statistical significance was due in part to the very limited data set
available for each dredge location. However, when grouped on a hot spot basis, a larger fraction of
the areas was shown to be statistically different. These results are summarized in Figure 4-22.
(Appendix F provides summary statistical data which form the basis for the information displayed
in Figure 4-22).
Five of the eight study areas were statistically different for LWA, the length-weighted
average PCB concentration for shallow sediments. Of these, four showed decreases in concentration
from 1976-1978 to 1994 conditions while one indicated increased PCB concentrations. The
remaining three areas were unchanged. Similarly, four of the eight study areas, with three declining
and one increasing, were statistically different for MPA. These four areas corresponded to four of
the five areas with significant changes in LWA.
The implications of the changes in MPA are straightforward. Decreases in MPA correspond
directly to decreases in the inventory of PCBs within the sediment and vice versa. The interpretation
of the changes in MPA is summarized in Table 4-8 for each of the hot spot areas. The tables provide
the surface area, number of samples, MPA, and PCB inventory in kilograms for each of the hot spots
under 1976-1978 and 1994 conditions. The MPA given in Table 4-8 is the unbiased minimum
variance estimator of the arithmetic mean and not the geometric mean represented in Figure 4-22.
As noted previously, the arithmetic mean is needed to estimate the inventory while the geometric
mean is used to discern statistically significant changes. It should also be noted that the 1976-1978
conditions are calculated using the solid specific weight correction described earlier. The original
MPI (1992) inventory estimates are provided in Table 4-8, in addition to the 1976-1978 sediment
inventories recalculated for this report.
4-30
TAMS
-------
The PCB inventories were calculated for both the 1976-1978 and 1994 results using the same
area and the minimum variance unbiased estimator for the arithmetic mean on an individual hot spot
basis. In this manner, the magnitude of the change in PCB inventory in each hot spot between the
two sampling events could be estimated.
Table 4-8 provides important information on the scale of the hot spot inventory changes. All
of the statistically significant changes represent differences of a factor of two or higher, i.e.,
statistically significant increases represent at least a doubling of the 1976-1978 sediment PCB
inventory, while decreases represent at least a halving of the 1976-1978 sediment PCB inventory.
In fact, most of these changes are even greater. Specifically, the inventories at Hot Spots 31, 34, and
37 have declined by approximately five-fold, two-fold, and three-fold, respectively, while the
inventory at Hot Spot 28 has increased eleven-fold. The remaining four hot spot inventories appear
unchanged with relatively good agreement (less than 10 percent difference) between 1976-1978 and
1994 for three of the four areas. The fourth area, Hot Spot 39, had a large, but not significant,
increase in PCB inventory. Overall, the surveyed areas increased in PCB inventory over the study
period, as shown at the bottom of Table 4-8. However, the nature of this gain warrants further
discussion.
The procedures to estimate the MPA and sediment inventory differed between the 1976-1978
and 1994 sampling programs. The earlier program relied on extrapolation to calculate the total PCB
inventory, while the 1994 program did not. The 1976-1978 program used a limited number of long
cores that completely encompassed the entire PCB inventory as a basis to vertically extrapolate the
large number of grab samples obtained. These grab samples were assumed to represent sediments
from 0 to 4-inches in depth. Extrapolation factors for the grab samples were developed separately
for each reach by both Tofflemire and Quinn (1979) and MPI (1992). Table 4-9 shows the
distribution of grab and core samples for the 1976-1978 data set. There is an inherent uncertainty
associated with the estimation process, since the correct extrapolation for each individual grab
sample is not known. However, the individual cores do provide a basis for a realistic assessment of
this uncertainty.
4-31
TAMS
-------
Figure 4-23, derived from MPI (1992) and the Hudson River Database Release 3.5,
represents the relationships among the 0 to 4-inch section, the 0 to 12-inch section, and the entire
core for the 1976-1978 data. The data is grouped by location above or below Lock 5. The intercept
in each regression was not statistically different from zero, so the lines were forced through zero.
From these diagrams, it is apparent that the extrapolation of the 0-4-inch section to the 0-12-inch
section should on average be within 6 to 8 percent of the true value when groups of samples are
considered. The diagram also indicates that no systematic bias exists. The error on individual cores
is larger (42 mg/kg above Lock 5 and 16 mg/kg below Lock 5) based on the intercepts of the
individual 95% confidence limits.
The MP A for the 0-12-inch section of the entire core yields slopes >1 in both instances
indicating that a portion of the mass is unaccounted for in the 1976-1978 estimates. The error in the
individual points and the regression line is small for the cores below Lock 5, 2 g/m2 on individual
cores and 10 percent on the regression line. For points above Lock 5, the error on both the regression
line (30 percent) and on individual cores (61 g/m2) is much greater.
For the 1994 low resolution core collection and interpretation, extrapolation was not
necessary, since most cores collected were complete. l37Cs was used to assess core completeness
as discussed previously in Chapter 2. Although some cores were defined as incomplete, these cores
are typically scattered throughout the data set and do not greatly impact the vertical integration of
PCB mass, i.e., nearby complete cores of comparable thickness suggest that little PCB inventory was
missed by the incomplete cores. Table 4-9 presents the number of complete and incomplete cores
for each hot spot for the 1994 data set. Additional factors indicating near-completeness in a core
include the PCB profile itself (Chapter 2) and the trend of 137Cs between surface (0 to 1-inch) and
bottom core segments.
Given the relatively well-known 137Cs impact to the Hudson watershed as recorded in the
various high resolution cores, as well as by various authors (e.g., Bopp and Simpson, 1989), the
trend in I37Cs levels between surface and deep sediments can yield useful information about the core.
As can be seen in Figure 4-24, both 137Cs and PCB have seen substantive maximum concentrations
relative to the oldest and most recent deposition. In ,37Cs in particular, the surface sediments are
4-32
TAMS
-------
substantially higher in concentration than the sediments deposited in the mid-1950s. Assuming that
this kind of l37Cs distribution applies in a rough manner to all Hudson River locations, if surface
l37Cs levels are higher than those at the bottom of the core, then it is highly likely that the core
encompasses the entire section of recent deposition. This is because 137Cs levels have not yet dropped
to the levels seen in the mid-1950s. Thus, the core bottom is close to the early 1950s horizon and
little post-1950 deposition lies below it.
Conversely, when ,37Cs increases from top to bottom in a low resolution core, it is unclear
how much recent sediment lies below the core since higher 137Cs levels relative to the surface layer
exist from 1992 depositions back to the early 1960s. Ideally in an incomplete core, falling 137Cs
levels from surface to bottom indicate that little recent deposition lies below the core and that the
core is nearly complete. Rising l37Cs levels with depth indicate that much recent deposition may lie
below the core bottom. Table 4-9 provides a summary of the l37Cs trends for the incomplete cores.
Hot Spots 25, 28, and 31 have a large majority of complete cores and so these inventory estimates
should be accurate. Hot Spots 34, 35, 37, and dredge location 182 have a relatively large proportion
of incomplete cores, but most are characterized by falling ,37Cs, indicating near-completeness in the
cores and subsequently relatively accurate inventories. Only Hot Spot 39 has a large proportion of
incomplete cores with rising l37Cs. Estimation of an accurate inventory is difficult here but it is still
possible to assess the change at Hot Spot 39, as discussed later in this section.
Returning to the discussion of hot spot inventory changes, it is evident that several of the
smaller hot spots have lost a substantial portion of the 1976-1978 PCB inventory. These hot spots
are shown in Plates 4-23,4-24, and 4-25. More than five metric tons (5,040 kg) were represented in
Hot Spots 31, 34, and 37 in 1976 to 1978, based on the MPA calculations presented in this report
(Table 4-8). This substantial mass is still less than the MPI (1992) estimate of almost seven metric
tons (6,830 kg) based on a constant SSW of 1 g/cc (Table 4-8). In 1994, the best estimate of 5,040
kg in these areas was reduced to only 1,881 kg, a loss of 63 percent. This loss is well beyond any
dechlorination process (maximum mass loss of 26 percent) and must therefore, represent the re-
release of PCBs from the sediment to the water column via one or more processes. The exact
mechanism or mechanisms responsible for this release are not well known but the effect in these
areas is quite profound.
4-33
TAMS
-------
Given the relatively thin depth of recent (post-1954) deposition in these hot spots (less than
12-inches; see Table 4-10 and Appendix D), resuspension of river sediment is certainly a viable
mechanism. Ground water advection or biologically driven losses are also possible. A portion of
this apparent loss may be due to differences in quantitation techniques, but it is highly unlikely that
quantitation differences could account for such a substantial loss.
Of particular note among these three areas is Hot Spot 37. Although the samples contained
strictly within the dredge boundaries yielded a statistically significant loss, a core collected just
downstream yielded a very high MP A of 125 g/m2 as compared to 17 g/m2 for the hot spot itself in
1994. This core appears to be within the original hot spot boundary defined in Tofflemire and Quinn
(1979). These results suggest that this area may be a source of the large PCB load increase described
in the DEIR (TAMS et al., 1997) for water column Transect 3. A high flow event in the Hoosic
River, the mouth of which is immediately upstream of Hot Spot 37, may have remobilized material
from the upstream end of Hot Spot 37. The frequency of similar high flow events in Hoosic River
suggests that this Hot Spot may continue to be a source of PCBs downstream in the water column.
Hot Spot 28 near the downstream entrance to Lock 6 (see Plate 4-22) exhibited a statistically
significant gain in PCB inventory. PCB inventory increased from just under two metric tons (1,850
kg) to over 20 metric tons (20,390 kg). The 1994 core profiles for Hot Spot 28 typically exhibited
two forms, as shown in Figure 4-25. In Profile 1, the coring site has apparently seen substantial
deposition over time, burying the more contaminated PCB layers. In this profile, peak concentrations
reach 526 mg/kg, but are under and overlain by sediments at lower concentrations. The PCB
concentration in the thickly sliced core in Profile 1 matches the better resolved high resolution cores
collected in 1992 (Figure 4-24), but the shallow sediments (0 to 12-inches) still represent a
substantial PCB concentration (145 mg/kg). Therefore, although burial of the highest sediment
concentrations is occurring, the more recently deposited material bears a substantial PCB
contamination level of its own and is not considered "clean". Core profiles LH-28F, LH-28I, and
LH-28K in Appendix D have conditions similar to Profile 1.
In Profile 2, the overlying lower PCB level sediments are missing and apparently peak PCB
concentrations are at or very near the surface. This type of profile can occur when sediment
4-34
TAMS
-------
deposition effectively stops at a given location, leaving highly contaminated sediments at the
surface. This profile can also be found in areas undergoing scour, where the upper layer in Profile
1 has been removed and the river is now scouring peak PCB concentration at the coring location. In
either case, sediment burial is not occurring. Consequently the potential for PCB re-release at these
sites is very high and may be on-going. Cores LH-28E, LH-28G, LH-28J, and LH-28N in Appendix
D have similar profiles. It was noted that the 7Be levels were detected in surficial sediments at all
of these locations, indicating recent deposition. This evidence for recent deposition apparently
contradicts the strong PCB profile evidence for long-term sediment loss or lack of burial. One likely
explanation is that the area is subject to sporadic scour events followed by periods of deposition.
This would provide a thin layer of 7Be-bearing material at the surface which would be removed
periodically along with some of the underlying material during scour events.
Although the PCB inventory has apparently increased at Hot Spot 28, its long term storage
there is clearly not assured. In addition, the burial process to date has clearly not yielded clean
overlying sediments. More than likely, the PCB contamination seen in the shallow sediments of
Profile 1 is the result of PCB re-release from sediments further upstream and from PCB released
from the GE Hudson Falls facility as well as from local re-release from the hot spot itself. Hot Spot
28 does not suggest a steady rate of deposition so as to sequester contaminated sediment, instead the
burial, subsequent re-working, and re-release of the contaminated sediment by the river appears to
be continuous and on-going.
It is most likely that the apparent increase in total inventory is the result of an underestimate
of PCB inventory in 1976-1978 derived from cores of insufficient length and incorrect assumptions
about the total depth of PCB contamination. The percent mass deposited between 1977 and 1994 can
be estimated using the dated high resolution cores shown in Figure 4-24. These cores are considered
recorders of river PCB loads, as described in the DEIR (TAMS, et al, 1997). In these and essentially
all other dated sediment cores from the Hudson, the sediment record shows a substantial decline over
time in the PCB loads carried by the river. Based on these core profiles, only 2 to 5 percent of the
cumulative PCB load was transported in the period after 1977. Thus the river was not carrying the
volume of PCBs which would be required to substantially raise sediment inventories between 1977
and 1994. Since at the time of the 1976-1978 surveys the river had already transported at least 95
4-35
TAMS
-------
percent of its total PCB load, it is highly unlikely that the remaining 2 to 5 percent to be transported
in the post-1978 period could yield the eleven-fold increase in inventory found in Hot Spot 28. Thus,
it is unlikely that a true substantive increase in PCB inventory has occurred at Hot Spot 28 since
1976-1978. Rather, it is likely that the 1976-1978 inventory was badly underestimated.
At the same time, the on-going reworking and re-release process indicated by the core
profiles is born out by the changes in the shallow-sediment concentrations (LWA). Both the total
sediment inventory and the shallow sediment inventory as represented by LWA have substantially
increased (see Table 4-10). Given that the PCB input history to the sediment is similar to that
recorded in the high resolution cores as shown in Figure 4-24, an increasing shallow sediment
inventory is inconsistent with an increasing total sediment inventory. Specifically, the 1976-1978
cores were collected just after the PCB maximum seen in the high resolution cores. Thus, after the
1976-1978 period, PCB concentrations in sediment have been declining. In locations where true
burial occurs between 1976 to 1978, the total sediment inventory should increase as additional
contaminated sediment is deposited above the mid-1970s peak concentration. However, the shallow
sediment concentration (LWA) should decline as the less contaminated sediments overlay the
sediments previously deposited. This is not the case in Hot Spot 28 where shallow sediment
concentrations have increased six-fold.
While it is also true that upstream inventories of PCBs have also declined and could thus
provide some additional material to Hot Spot 28, it is unlikely that this process is responsible for the
eleven-fold increase noted. Specifically, upstream resuspended sediment would be mixed with the
sediment load already carried by the river, thus downstream transport serves to dilute the sediment
concentrations. Yet shallow sediment concentrations are comparable or higher than many of the core
locations occupied upstream. Also, it is likely that any major transport event sufficient to
substantially increase the sediment inventory at Hot Spot 28 to this degree would be recorded in the
nearby high resolution cores. High resolution core 18 was collected near the downstream end of Hot
Spot 28 at Rm 185.8. This core clearly shows that the majority of PCB transport occurred prior to
1978 and that deposited sediment concentrations are declining over time. This contradicts the
observation that shallow sediment concentrations have increased at this hot spot since 1978.
4-36
TAMS
-------
Both the Phase 2 water column and high resolution core analysis demonstrated the absence
of local sources in the vicinity of Hot Spot 28. In the absence of local PCB input, increases in both
shallow sediment and total sediment inventories can only occur when both PCB deposition and PCB
re-release (probably via scour) are occurring or have occurred during the intervening period. Based
on the number of cores similar to Profile 1 in Figure 4-25, it appears that slightly less than half (4
of 10) of the coring sites in Hot Spot 28 are undergoing burial with less contaminated, but hardly
clean sediments. However, a similar fraction of sites (5 of 10) is undergoing scour, thus bringing the
peak PCB concentrations back to the surface.
Conditions at Hot Spot 28 can be contrasted with those at Hot Spot 39. At this hot spot the
shallow sediment inventory has shown statistically significant decline of about 30 percent (1.4-fold
decrease). This is matched by a three-fold increase in mass which is not statistically significant. Hot
Spot 39 contained only six complete cores out of a total of 14. Of the six complete cores, five yield
profiles similar to Profile 1 in Figure 4-25, indicating steady deposition and burial. Four incomplete
cores yield profiles with rising l37Cs and rising PCB, indicating that a potentially significant portion
of recent sediment and PCB inventory is present below the coring depth. Only two cores yield
profiles similar to Profile 2 in Figure 4-25 and the remaining cores indicate little activity, with the
PCB inventory contained almost entirely in the shallow sediments. Thus nine of 14 cores indicate
that burial of PCB bearing sediments is the predominant process at Hot Spot 39. Shallow sediment
concentrations indicate that the depositing sediments are not clean, with PCB concentrations in the
range of 4 to 62 mg/kg with a median of about 5 mg/kg. Hot Spot 39 represents the only hot spot
study area to exhibit a consistent PCB burial process. Unfortunately, an accurate estimate of the
inventory gain could not be made, due in part to the large number of incomplete cores. Also, the
1976-1977 estimates are based on short cores and grabs which would underestimate the PCB mass.
Based on the results available, the apparent gain is at least a three-fold increase.
The remaining three study areas, Hot Spots 25 and 35, and dredge location 182 were
characterized as unchanged. Plates 4-21,4-24, and 4-27 show the sampling locations associated with
each area. Two of these areas are particularly limited in sample coverage. Specifically Hot Spot 35
and dredge location 182 have only four and two samples collected, respectively, in 1994 as
compared to 11 and six samples, respectively, collected in 1976 to 1978. Thus, statistical analyses
4-37
TAMS
-------
of these two areas are quite limited, especially for the dredge location 182. However, taken together,
these areas still provide useful information on relatively unchanged areas.
As shown in Table 4-8, the PCB inventory estimates for 1976-1978 and 1994 are quite close
in the areas characterized as unchanged, within 10 percent of each other in each case. The apparent
lack of change is borne out by the core profiles shown in Appendix D. Figure 4-26 shows two core
profiles typical of these areas. In general, these areas are characterized by profiles where little PCB
mass occurs below the top core segment. These profiles appear somewhat similar to Profile 2 in
Figure 4-25, but there is an important distinction. Cores from Hot Spots 25, 35, and dredge location
182 show a much greater level of change between the shallow sediment segment and the underlying
layer. Typically, the PCB inventory in the underlying layer is at least five times less than the
shallow sediment layer and is frequently zero. This would suggest that only a shallow PCB
inventory exists at these hot spots. No cores from within these areas exhibited the PCB burial profile
typically seen in Hot Spot 39. Although several cores in these areas were not complete (see Table
4-9), all incomplete cores were characterized with falling 137Cs levels, thus the majority of recent
sediment was captured in the core. The finding that in many hot spots the current PCB inventory
is in shallow sediments indicates that:
• Burial of these hot spots is not occurring;
• Opportunity for remobilization of PCBs is present; and
• There are little or few overlying clean sediments to remove to get to contaminated
sediments.
4.2.4 7Be in Surface Sediments
It was originally hoped that the presence or absence of 7Be in the 0 to 1-inch layer of a low
resolution core would provide some information on the occurrence of sediment scour, since 7Be is
generally detected in more recently deposited sediments. While the absence of 7Be does not preclude
the presence of sediment less than six months old, little direct correlation could be found between
7Be presence/absence and sediment deposition or scour. This is evident in the four core profiles
presented in Figures 4-25 and 4-26 that all have 7Be present, yet are distinctly different. These
4-38
TAMS
-------
profiles characterize deposition (burial), scour, and no net change. On an individual basis, 7Be does
not appear to be able to separate these environments, but this does not imply that the 7Be data are
of no use. When viewed on a hot spot area basis, it is apparent that Hot Spot 37, which has
undergone PCB inventory loss, has a distinctly lower detection frequency for Tie (3 of 11 core sites
or 36 percent) as compared to the depositional environment around Hot Spot 39 (13 of 14 sites or
93 percent). (See Plates 4-25 and 4-26/
On average, the three hot spots exhibiting PCB loss had a lower average frequency of 7Be
detection (0.54) than the two hot spots with mass increases (0.87). Although these data sets are too
small to perform statistical analysis, the average values support this general correlation. The three
hot spots with unchanged inventories yielded a wide range of 7Be detection frequencies. However,
the number of points available for these areas is quite limited and so they were not examined further
for 7Be. All the TSe data in terms of presence/absence for the hot spots sampled as shown in Plates
4-21 through 4-28.
Because of the nature of its geochemistry and short half life, 7Be still provides the potential
to prove the presence of less than six-month old sediments. However, presence of six-month old
sediment at time of collection does not guarantee a complete depositional record in the core. In
many instances 7Be was present in cores from hot spots with declining PCB inventories such as Hot
Spots 31 and 35, indicating that it is not a perfect predictor for PCB loss environments. Taken on an
area basis however, 7Be frequency of detection was lowest for areas of PCB inventory loss and
highest for areas of inventory gain, as might be expected.
4.2.5 Hot Spot Boundaries
A subset of cores was collected just beyond the original hot spot boundary so as to assess the
accuracy of the boundary as drawn. While in most instances PCB inventories were lower in the
cores external to the hot spot or dredge location, there were some important exceptions including the
four TAMS exploration areas (cores labeled LH-41,42,43, and 44). Twenty-eight cores were sited
outside hot spot or dredge locations. Of these, 21 yielded relatively low PCB concentrations. The
range of the LWA values (representing sediments from 0 to 12-inches) for the 21 cores was 1 to 27
4-39
TAMS
-------
mg/kg with a median value of 12 mg/kg. Two of these cores had larger inventories at depth and
were associated with Hot Spot 39 and dredge location 182. The remaining seven cores with elevated
concentrations were associated with Hot Spots 28, 31, 34, 37, dredge location 182, and TAMS
location 41. The range of LWA for these cores was 46 to 709 mg/kg with a median value of 126
mg/kg. For Hot Spots 28,31, and 34, the elevated sediment concentrations were found outside the
hot spot boundary but still within the same sediment type as classified by the side-scan sonar
interpretation. These cores would suggest that the dredge location boundaries may need to be
expanded to encompass more of the same sediment type in the vicinity of the dredge location.
4.2.6 Comparison of the 1994 Hot Spot Inventories with Other 1977 Estimates
The estimates of total PCB quantities in the hot spots must be presented on a consistent basis
for the comparisons to be meaningful. Table 4-11 presents the estimates of Malcolm Pirnie (1979),
Tofflemire and Quinn (1979), Malcolm Pirnie (1992), and an estimate derived by TAMS from the
data provided in Malcolm Pirnie, 1992. The appropriate low resolution core total PCBs quantities
are derived by multiplying the hot spot area used in the 1976-1978 estimate by the 1994 mean MP A.
The 1976-1978 estimates of the inventory are presented and corrected using a solid specific weight
(SSW) based on the mean total PCB concentration for the hot spot. These SSW values are obtained
from Table 4-3 rather than accept the original author's assumption of an SSW of 1 g/cc. Results are
compared via a Delta function, where:
A, =
1994 Low Resolution PCB Quantity - (1976-197&) Revised PCB Quantity ^
(\916~\91%)RevisedPCBQuantity
* 100% (4.2-6)
In this manner, Deltaj (a,) is negative for losses and positive for gain. A Delta; of 100% represents
a doubling of the 1976-1978 PCB inventory. A Deltaj of -50% represents a halving of that inventory.
Differences between the 1976-1978 and 1994 estimates of more than a factor two are
considered significant and likely to be beyond the level of uncertainty. Hot Spots 25 and 35 lack
4-40
TAMS
-------
significant differences for all inventory estimate comparisons. Hot Spots 28, 31, 34, and 37 have
significant differences for each estimate with only 28 showing a gain and the others loss. The losses
are considered definitive because the 1976-1978 estimates may be biased low due to insufficient
depth of cores as discussed previously in Section 4.2.3. Thus the estimated 1994-1977 mass loss
represents a minimum estimate of the actual mass loss. Overall, six of the seven hot spots inventories
have a consistent level and direction of change for all four comparisons. That is Hot Spots 31, 34,
and 37 show losses greater than 50 percent of the 1976-1978 inventory, Hot Spots 25 and 35 appear
to be unchanged or to have lost no more than a 40 percent and Hot Spot 28 has had a large apparent
increase.
Only Hot Spot 39 shows an insignificant loss in the Malcolm Pirnie (1978) estimate and a
significant gain in both estimates using the Malcolm Pirnie (1992) data. This hot spot shows
considerable levels of PCB contamination at more than two feet of depth including eight incomplete
cores. Of these eight, five have rising l37Cs levels strongly indicating that more contamination lies
below. The Malcolm Pirnie (1992) estimate considered only the shallow sediments and therefore,
underestimated the inventory. The MPI (1978) estimate includes the values at depth and yields no
significant change in inventory for Hot Spot 39 to 1994. However, the MPI (1978) estimate is based
on only about one-quarter of the surface area for the hot spot, i.e., 26, 400 m2 for the MPI (1978)
estimate and 105,700 m2 for the MPI (1992) estimate. Thus, the absolute inventories given by MPI
(1978) and other 1976-1978 estimates are not very different. Although the MPI (1978) estimate was
based on deeper samples, it is still an underestimate of the current inventory as given in the last
section of Table 4-11.
The inventory of Hot Spot 28 appears to have increased between 1.5 and 10 times between
1976-1978 and 1994. The 1994 estimates range from 18-20 metric tons of total PCBs in Hot Spot
28 which is comparable in scale to the 23.2 metric ton estimate for the entire TI Pool inventory given
by M. Brown et al. (1988). Like Hot Spot 39, the inventory may have been present but missed by
incomplete cores in the 1976-1978 sediment survey. The 1976-1978 estimates are probably biased
low due to insufficient core depth causing the true gain to be less than measured due to failure to
core the entire profile.
4-41
TAMS
-------
As discussed in Section 4.2.3, the issue of sediment inventory gain is particularly sensitive
to the accuracy of the original core collection. If earlier cores were not collected from a sufficient
depth, then the 1994 results will show an apparent inventory gain which is not real. An examination
of the high resolution cores presented in Figure 4-23 shows that most PCB deposition occurred prior
to 1977. In fact, based on the two cores shown which were collected from this area of the Hudson,
roughly 95 percent or more of the sediment inventory was deposited prior to 1977. Thus the reported
inventory gains of 194 to 1040 percent for Hot Spots 28 and 39 seem highly unlikely, particularly
given the observed PCB peak concentration which is in the hundreds of mg/kg. The high resolution
cores indicate that sediments of this level of contamination were last deposited in the 1970s. Nothing
since 1980 has exceeded the 100 mg/kg level.
In summary, this analysis of hot spot inventories shows a consistent pattern of loss of at least
50 percent of the 1977 inventories for Hot Spot areas 31, 34, and 37 regardless of the source of the
original estimate. Hot Spots 25 and 35 appear to be basically intact or at least have lost no more than
40 percent, again regardless of the source of the original estimate. Inventories for Hot Spots 28 and
39 in 1994 are distinctly higher than the 1977 estimates. A small portion of this gain may be due to
post-1977 deposition. However, it appears that the majority of this gain is likely due to inaccurate
estimates derived from cores of insufficient length to capture the post-1954 recent sediment profile.
The best estimate for the 1994 sediment inventories is given in Table 4-12. A total of 27.3 metric
tons of PCBs resides in these hot spots with Hot Spot 28 representing the vast, previously
undocumented, majority of this material.
4.3 Sediment Contamination in the Near-Shore Environment
As part of the low resolution coring program in the Thompson Island Pool, several clusters
of samples were located in near-shore areas, approximately within 50 ft of shore. Part of this
program was to assess the level of PCB contamination in areas where direct sediment exposure to
local populations was likely. As a part of this program, cores were collected from 4 cluster locations
with 1 or five cores per tocation. The specific clusters were numbers 16, 17, 18 and 19. These
locations can be seen on Plate 2-1. On the plate, the clusters are denoted as "LR-xx" where xx is the
cluster number. Three of the clusters (16,18, and 19) are located around Griffin Island and one (17)
4-42
TAMS
-------
is located about 1 mile upstream across from hot spot 8. Two of these cluster locations (16 and 18)
had a single previous NYSDEC coring location which became the center of the cluster when locating
the coring locations.
The reference point for comparison with these data was the estimated PCB concentration
derived for the Phase 1 Report (TAMS/Gradient, 1991) which used a value of 66 mg/kg as the
exposure point concentration for human exposure in the preliminary risk assessment. This value
represented the 95 percent confidence interval of the arithmetic mean of the shallow sediment
concentrations based on the 1984 NYSDEC sediment collection effort. Concern with this estimate
stemmed from its basis on the entire sediment 1984 data set, potentially underestimating near-shore
exposure concentrations.
To address this issue to a limited degree, the low resolution sediment coring program
obtained cores from 4 near-shore areas in the TI Pool. Due to some limitations in the accessibility
of the shoreline area, not all near-shore cluster cores were located within 50 ft of shore. On this basis,
a subset of the original 16 near-shore cores was analyzed based on their proximity to the shoreline.
A distance of 50 ft was chosen since it approximated a water depth of 4 ft, a likely water depth for
wading and swimming. This yielded 11 low resolution cores from these clusters. The results from
this subset are shown in Table 4-13. The 95 percent confidence limit value (151) is substantially
greater than the original estimate used in the Phase 1 report. Its usefulness may be limited, however,
due to the small sample size. Also represented in the table are the minimum, maximum, geometric
mean, arithmetic mean and an unbiased, minimum variance estimator (MVUE) of the arithmetic
mean. The MVUE represents the best estimate of the arithmetic mean, given that the underlying data
distribution is log-normal. This property of the PCB data was discussed previously in this report and
is not repeated here. The 95 percent confidence interval is the value typically used in risk assessment
calculations.
A second approach to estimating the near-shore sediment concentrations for the TI Pool was
made by considering all near-shore, fine-grained, low resolution TI Pool cores within 50 ft of shore.
This yielded a larger data set (19 cores) as well as a higher 95 percent confidence limit (264 mg/kg).
The assignment of sediment classification was based on the classifications obtained from the side-
4-43
TAMS
-------
scan sonar results described previously in this report and the DEIR. The results from the two low
resolution data groupings were not statistically different as might be expected since many sample
locations were common to both groups.
As a last approach, the 1984 NYSDEC data set was examined on the same basis, i.e., fine-
grained sediment cores as classified by side-scan sonar within 50 ft of shore. These results yielded
values similar to those obtained from the low resolution cores.
In each approach, it is apparent that the original estimate of 66 mg/kg for near-shore sediment
exposure in the preliminary risk assessment was too low and that a value in the range of 135 to 264
would be more appropriate. Given that the original risk estimate did not exceed the USEPA
acceptable range, it is unlikely that the revised range will represent an unacceptable risk. These
results will be reviewed and incorporated in the Phase 2 risk assessment as part of the human health
risk assessment.
4.4 Summary and Conclusions
4.4.1 Sediment and PCB Inventories in the TI Pool
Sediment core profiles obtained in 1994 exhibited a range of characteristics from inventory
gain to loss relative to the original 1984 core profiles, with sediment inventory loss being the most
common. Some 1984-1994 core pairs strongly suggested sediment scour, based on the apparent
upward vertical movement of the PCB contamination horizon.
The 1984 sediment survey results are best characterized as the sum of trichloro and higher
chlorinated homologues when a correction factor of0.934 is applied. The 1994 data showed a linear
decline in the £Tri+ inventory when plotted against the original 1984 inventory. This loss was
attributed to both dechlorination and to re-release to the water column
4-44
TAMS
-------
Using two different estimates for molecular weight, an estimate for the 1984 sediment £Tri+
molar inventory was obtained. This estimate was used to track PCB loss to the water column from
the sediment relative to dechlorination loss over the period 1984 to 1994.
When sediment inventories were grouped by the 1984 £Tri+ inventory as greater than or less
than 10 g/m2, a statistically significant trend in the relative inventory change was found. Specifically,
sediments greater than 10g/m2 yielded inventory losses of about 39 percent. Of this 28 percent was
attributed to PCB re-release to the water column and 12 percent was attributed to dechlorination. The
estimates of 28 and 39 percent represented lower bounds on the actual PCB losses.
The scale of the PCB loss from these sediments is of similar magnitude to the water column
inventory gains reported in the DEIR. These results confirm and support the contention put forth in
the DEIR that TI Pool sediments contribute directly and substantively to the water column load. For
sediments less than 10g/m2, an upper bound on the sediment gain was obtained, at between 87 and
104 percent. Presumably these gains, if real, are due to deposition of PCBs from sediment re-release
as well as upstream inputs.
Based on these trends, the river is apparently re-distributing the sediment PCB inventory.
Little evidence for extensive sediment burial was found in the TI Pool. More core sites exhibited loss
than gain or burial. Instead, it appears that the PCB inventories of the more contaminated sediments
are being redistributed, ensuring that the burial of contaminated sediments with clean sediments is
not occurring. These same processes would also serve to deliver PCBs to the water column and
produce the loads recorded by both USEPA and GE monitoring.
4.4.2 Sediments and PCB Inventories Below the TI Dam
• Comparison of the texture type indicated by the sonar images with the 1976-1978 NYSDEC
sediment survey grain-size data found good agreement. This suggests that the river bottom
depositional types remain constant.
4-45
TAMS
-------
Malcolm Pirnie, Inc. (MPI, 1978), Tofflemire and Quinn (1979), and MPI (1992) provided
estimates of the 1976-1978 total PCB inventory in hot spots located in the Upper Hudson.
Data from the MPI (1992) assessment was used to compare the 1976-1978 and 1994
inventories in this analysis because of the availability of PCB concentration data and hot spot
area definitions (as scale drawings).
The historical analyses assumed a solid specific weight (SSW) of 1 g/cc. Based on the low
resolution core relationship between solid specific weight and total PCB concentration,
estimates for SSW ranged from 0.5 to 0.79 g/cc for most 1976-1978 hot spot sample
locations. Applying a SSW based on length-weighted average concentrations for 1976-1978
yielded about a 20 to 30 percent decrease in the original PCB inventory estimates.
Sediment PCB data were shown to be log-normally distributed for both the 1976-1978 and
1994 data. The geometric mean, arithmetic mean, minimum variance, and unbiased
estimates of the arithmetic mean were calculated to estimate and compare PCB inventories.
Low resolution core samples were placed to define the hot spot area properties, not the
smaller dredge location areas defined in MPI, 1992. The hot spots are the minimum area of
PCB contamination for comparison between 1976-1978 and 1994 inventories.
The low resolution core data provides the best estimate of the 1994 total PCB inventory. For
the 1976-1978 inventory, the most accurate estimate is based on NYSDEC data provided in
MPI, 1992. A comparison of these inventories indicate statistically significant loss of 50 to
80 percent for three hot spots (31, 34 and 37) relative to the 1976-1978 estimate, while Hot
Spot 28 exhibited an apparent gain of 18 metric tons of total PCBs between 1976-1978 and
1994.
Examination of the core profiles at Hot Spot 28 shows less than half of the sample locations
have undergone deposition (burial). The remaining sites are either unchanged or have
undergone scour based on the presence of the maximum total PCB concentration in the
shallow sediment layer, as well as a net increase in shallow sediment concentration (six-fold)
4-46
TAMS
-------
from 1976-1978 to 1994. The deposition history recorded by the high resolution cores
indicates that this type of rise in shallow sediments can only be caused by scour. Between
two and five percent of PCB mass was deposited between 1976-1978 and 1991 based on two
dated high resolution cores collected from the upper Hudson below the TI Dam. The small
increase in inventory in these cores makes such a large gain in inventory unlikely at Hot Spot
28. Since the high resolution cores are believed to represent continuous deposition, most of
the gain at Hot Spot 28 probably results from an initial inaccurate assessment of the Hot Spot
by the 1976-1977 sediment survey caused by too many shallow cores and grabs.
Hot Spot 39 exhibits burial with total PCB concentrations much greater at depth. Compared
to the Malcolm Pirnie, 1978 estimate (which used the entire core) there is not a significant
change in inventory.
Three areas appear unchanged (Hot Spots 25, 35 and dredge location 182) but only one (Hot
Spot 25) has a sufficient number of samples to confirm the lack of change. Of the two
remaining areas, Hot Spot 35 is better characterized.
Deposition thickness varied considerably and consistently with PCB inventory gain and loss.
The average depth of the maximum total PCB concentration for the low resolution cores is
18.7-inches (46.8-cm) in the hot spots exhibiting an apparent PCB inventory increase and
10.6-inches (26.5-cm) in the hot spots exhibiting loss. The difference in mean depth between
areas of PCB loss and gain is statistically significant.
Hot Spots 28 and 39 had large PCB inventories of 20 and 4 metric tons, respectively. The
combined total is equivalent to the inventory of the entire Thompson Island Pool, estimated
to be 23 metric tons by M. Brown et al., 1988. Most of this mass (about 95 percent) was
deposited prior to 1977, based on deposition histories developed from the high resolution
cores.
Because the low resolution cores can be assessed for completeness using both the core
profiles and radionuclide information, the occurrence of sediment PCB inventory losses from
4-47
TAMS
-------
the 1976-1978 sampling to the 1994 low resolution coring can be ascertained more
definitively than can gains. Specifically, because the 1994 inventories are from cores that
penetrated all recent sediments, it is probable that the entire current inventory has been
captured. Therefore, the 1978 inventories may be underestimates. Thus, observed losses are
minimum estimates and observed gains are maximum estimates.
l37Cs proved to be invaluable in assessing core completeness or near completeness, because
the absence of l37Cs in the bottom of the core was a reliable indicator that the core was
complete.
7Be occurrence showed a correlation with PCB inventory gain and loss on an area basis, but
could not be proven to have a statistically significant relationship based on the 1976-1978
to 1994 inventory comparison.
Hot spot boundaries appeared accurate, although in some instances hot spot areas needed to
be increased to include all nearby areas of high contamination. Sonar image interpretation
appears to provide some guidance as to regions of fine grain, higher contamination
sediments.
Comparison of the 1994 inventory with other estimates of the 1976-1978 inventory yield
similar magnitude and direction of mass change in most instances. The MPI (1978) and
Tofflemire and Quinn (1979) estimates took more of the deeper sediments into account.
Thus, losses tended to be greater and gains smaller as compared to the MPI (1992) estimates
of the 1976-1978 inventory. The mass change in Hot Spot 39 became an insignificant loss
using the MPI (1978) estimates showing that some mass at depth was detected in the 1976-
1978 sediment survey for this region. However, because the area used by MPI (1978) was
much smaller than that for subsequent studies, the actual PCB quantity estimated was still
much smaller than the 1994 estimate based on a larger hot spot area.
4-48
TAMS
-------
• The large apparent increases in PCB inventory in Hot Spots 28 and 39 are probably more
related to underestimated PCB inventories in 1976 to 1978 (MPI, 1992) than to actual
increases in PCB inventory since that time.
• Overall, sediments below the Thompson Island Dam exhibit both losses and gains. Losses
totaled a minimum of 3.2 metric tons. Gains were estimated at 18.6 metric tons, but the
actual gain in inventory is probably much less. Poor assessment of Hot Spots areas 28 and
39 in the 1976-1978 sediment survey caused by failure to capture what was apparently in
place at the time of surveying yielded badly underestimated PCB inventories at these
locations. Based on high resolution cores collected in the same area of the Upper Hudson,
only a small percent of the 1994 mass was deposited between 1977 and 1991.
• These results show that the stability of the sediment deposits cannot be assured. It is likely
that PCBs will continue to be released from the sediments. Burial of contaminated sediment
by cleaner material is not occurring in most hot spot areas below the TI Dam. Burial of more
PCB-contaminated sediment by less contaminated sediment has occurred in Hot Spot 39 and
to a limited degree in Hot Spot 28. However, this process is limited and much of the other
hot spot inventories has been re-released to the environment.
4.4.3 Sediment Contamination in the Near-Shore Environment
Sediments in the near shore environment were found to have higher PCB concentrations than
originally estimated in the Phase 1 report. (TAMS/Gradient, 1991) Sediment concentrations upper
95% confidence interval for sediment concentrations based on the Low Resolution Coring samples
ranged from 151 to 264 as compared to Phase 1 value of 66 mg/kg. The applicability of these results
may be limited due to the small number of samples. These data will be considered in the preparation
of the final human health risk assessment in Phase 2.
4-49
TAMS
-------
4.4.4 Summary
In this chapter, 1994 sediment inventories for the Upper Hudson have been compared with
historical studies of the same river areas. Comparisons between the 1984 and 1994 sediment surveys
for the Thompson Island Pool demonstrated the occurrence of statistically significant PCB losses
from the sediment over the period 1984 to 1994. During this time approximately 39 percent of the
inventory of sediments with PCB concentrations greater than 10 g/m2 was lost. Of this loss,
approximately 28 percent was the result of the re-release of PCB contamination from the sediment
via scour or other sediment release processes. Both these estimates represent minimum estimates of
the actual PCB losses. The remaining portion of the loss was attributed to dechlorination. The level
of dechlorination noted was consistent with that obtained from the change in molecular weight of
the PCB mixture as given in Chapter 3. Sediment inventories greater than 10 g/m2 are associated
with sediments with an average PCB concentration of 12 mg/kg and higher and peak concentrations
of 50 mg/kg and higher.
Concurrent with this loss was the gain in PCB inventory among the less contaminated
sediments (less than 10 g/mJ). Inventories were estimated to have increased by 87 to 104 percent.
This gain was attributed to input from upstream as well as the re-release of PCBs from the more
contaminated sediment.
Examination of individually paired 1984-1994 sediment profiles found sites exhibiting
inventory loss, inventory increase and lack of change. Inventory losses were the most commonly
occurring condition, with some sites presenting indications of sediment scour. General burial of
contaminated sediment by clean sediment was not apparent.
The comparison of the 1976-1978 and 1994 surveys for the areas below the TI Dam yielded
similar conclusions. Of eight hot spot or dredge location areas studied, three exhibited inventory
losses representing between 50 and 80 percent of the original inventory. Changes in three other areas
could not be discerned. The remaining two areas exhibited apparent large gains which appeared to
be the result of badly underestimated 1976-1978 PCB inventories and not the result of post-1978
deposition. One of these areas exhibited sediment burial although shallow sediments were still
4-50
TAMS
-------
contaminated at the 5 mg/kg level. The other, more massive inventory associated with Hot Spot 28
exhibited many core profiles suggesting PCB loss, probably via scour, based on the occurrence of
the highest PCB levels in the shallowest sediments. Thus, although the 1992 estimate for the PCB
inventory at Hot Spot 28 was eleven-fold higher than the original 1976-1978 inventory estimate,
coring evidence suggests that this area has been losing PCB inventory over time and that the 1976-
1978 inventory was probably higher than the 1992 estimate.
In total, the low resolution coring survey achieved its goal of assessing current sediment PCB
inventories and estimating inventory change since previous surveys. The results provide unequivocal
evidence for the loss of PCBs from the sediment to the water column, supporting the contentions of
the DEIR (TAMS, et al., 1997). These results indicate that since 1984, a sediment PCB inventory
loss has occurred, representing 28 percent of the PCB inventory from the more contaminated areas
of the TI Pool. Similarly, since the 1976-1978 survey, a loss of about 3200 kg of PCBs has occurred
from the hot spots below the TI Dam. In the meantime, the possibility of inventory gains in less
contaminated areas of the river suggest that at least a portion of the re-release is serving to
recontaminate other areas. Lastly, little evidence was found to suggest that burial of more
contaminated materials with clean sediment was occurring on a widespread basis.
4-51
TAMS
-------
REFERENCES
Ballschmiter, K.ard M. Zell. 1980. Analysis of Polychlorinated Biphenyls (PCB) by Glass Capillary Gas
Chromatography. FreseniusZ. Anal. Chem. 302,20-31.
Bopp, R.F. 1990. Transmittal memo from R. F. Bopp at NYSDEC to D. Merrill at Gradient dated November 19, 1990.
Re: 1976-1978 sediment survey results in electronic form.
Bopp, R.F. 1979. The Geochemistry of Polychlorinated Biphenyls in the Hudson River. Ph.D. Dissertation, Columbia
University, New York, New York.
Bopp, R.F. and H.J. Simpson. 1989. Contamination of the Hudson River, the Sediment Record, pp. 401-416 in
Contaminated Marine Sediments - Assessment and Remediation. National Academy Press; Washington, DC.
Bopp. R.F., H.J. Simpson and B.L. Deck. 1985. Release Of Polychlorinated Biphenyls From Contaminated Hudson
River Sediments. Report to the NYS Dept. of Environmental Conservation, June 30, I98S.
Box, G.E.P., W.G. Hunter, J.S. Hunter. 1978. Statics for Experimenters. An Introduction to Design Data Analysis, and
Model Building. John Wiley &Sons ,New York. 204-205.
Brown Jr., J.F., R.E. Wagner, and D.L. Bedard. 1984. PCB Transformations in Upper Hudson Sediments. Northeast.
Environ Set. 3: 184-189.
Brown, M.P., M.B. Werner, C.R. Canisone and 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.
NYSDEC, Albany, New York.
Butcher, J.B. 1998a. Low Res vs High Res MDPR/DMW Stats. Memorandum to Claire Hunt and Ed Garvey
(TAMS/NJ). January 19, 1998.
Butcher, J.B. 1998b. Low Res vs CC DN-50/Mean Phi stats. Memorandum to Ed Garvey (TAMS/NJ). February 17,
1998.
Butcher, J.B. 1996. Co-Kriging to Incorporate Screening Data: Hudson River Sediment PCBs. Journal of the American
Water Resources Association, Vol. 32, No. 2, April 1996, pp. 349-356.
Fisher, F. 1970. Test of equality between sets of coefficients in two linear regressions. an expository note.
Econometrica 3*: 361-366.
General Electric Corp., 1991-1997. Remnant Deposit Monitoring Program, Monthly Reports. Albany, New York..
Gilbert, R.O. 1978. Statistical Methods for Environmental Pollution Monitoring, Van Nostrand Reinhold, New York,
158-159.
Kennedy, P. 1979. A Guide to Econometrics, The MIT Press, Cambridge, MA.
Limburg.K.E. 1984. Environmental Impact Assessment of the PCB Problem: A Review. Northeastern Environmental
Science 3(3/4): 122-136.
Mahalanobis, P.C. 1930. On Tests and Measures of Group Divergence. Journal and Proceedings of Asiatic Society
of Bengal, 26:541-588.
R-l
TAMS
-------
Malcolm Pimie, Inc. 1978. Phase I Engineering Report Dredging of PCB Contaminated Hot Spots in the Upper Hudson.
Report to NYSDEC. Albany, New York.
Malcolm Pimie, Inc. 1992. Proposed Dredging Locations for The Thompson Island Pool, Lock 5 and 6 Pools, and Lock
2, 3, 4 Pools. Report to NYSDEC. Albany, New York.
Normandeau Associates, Inc. 1977. Hudson River Survey: 1976-1977. NYSDEC, Albany, New York.
NUS Corporation. 1984. Feasibility Study: Hudson River PCBs Site, New York. April 1984. USEPA, New York, New
York.
Pindyck, R.S. and D.L. Rubinfeld. 1981. Econometric Models and Economic Forecasts (2"d edition). McGraw-Hill,
New York.
SAS Institute, Inc. 1994. JMP® Statistics and Graphics Guide Version 3. Cary, North Carolina.
Shapiro, S.S., and M.B. Wilk. 1965. An analysis of Variance Test for Normality (complete samples), Biometrika
52:591-611.
Shillabeer, N„ B. Hart, and A. M. Riddle. 1992. The Use of a Mathematical Model to Compare Particle Size Data
Derived by Dry-Sieving and Laser Analysis. Estuarine, Coastal and Self Science (1992). 35: 105-111.
Sttderlund, R. And B. H. Svensson. 1975. The Global Nitrogen Cycle. Nitrogen, Phosphorus, and Sulphur - Global
Cycles. Ecological Bulletins NFR 22. Swedish SCOPE Committee of the Royal Swedish Academy of Sciences.
Orsundsbro, Sweden. December 14-18. pp 23-73.
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.
TAMS Consultants, Inc., The Cadmus Group, Inc., and Gradient Corporation. 1997. Further Site Characterization and
Analysis, Volume 2C- Data Evaluation and Interpretation Report, Hudson River PCBs Reassessment Rl/FS. February
1997.
TAMS Consultants, Inc. and Gradient Corporation. 1995. Further Site Characterization and Analysis Database Report.
Phase 2 Report - Review Copy. Hudson River PCBs Reassessment RI/FS. October 1995.
TAMS Consultants, Inc. and Gradient Corporation. 1994. Sampling and Analysis/Quality Assurance Project Plan,
Volume 4: Low Resolution Sediment Coring.Hudson River PCBs Reassessment Rl/FS. May 17, 1994.
TAMS Consultants, Inc. and Gradient Corporation. 1991. Phase I Report - Review Copy Interim Characterization and
Evaluation, Volume I. Hudson River PCBs Reassessment Rl/FS. August 1991.
Theil, H. 1961. Economic Forecasts and Policy. North-Holland, Amsterdam.
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.
R-2
TAMS
-------
F
I
s
H
APPENDICES
-------
APPENDIX A
-------
APPENDIX A
Data Usability Report for Pcb Congeners
Low Resolution Sediment Coring Study
-------
Table of Contents
Page
A.l Introduction A-l
A.2 Field Sampling Program A-2
A.3 Analytical Chemistry Program A-3
A.3.1 Lab Selection and Oversight A-3
A.3.2 Analytical Protocols for PCB Congeners A-4
A.4 Data Validation A-7
A.5 Data Usability A-9
A.5.1 Approach A-9
A.5.2 Usability - General Issues A-ll
A.5.3 Usability - Accuracy, Precision, Representativeness, and Sensitivity ... A-16
A.5.3.1 Accuracy A-l7
A.5.3.2 Precision A-22
A.5.3.3 Representativeness A-22
A.5.3.4 Sensitivity A-23
A.5.4 Usability - Principal Congeners A-24
A.6 Conclusions A-27
References A-28
TABLES
A-l List of 126 Phase 2 Target and Non-Target PCB Congeners Used in Low Resolution
Sediment Coring Study Report
A-2 Data Qualification Codes
A-3 Low Resolution Sediment PCB Field Co-located Samples
A-4 PCB Detects Changed to Non-Detects
A-5 Low Resolution Coring PCB Sample Analysis Summary
FIGURE
A-l Low Resolution Sediment Core Preparation
-------
A.l Introduction
The usability of data relates directly to the data quality objectives of the environmental
investigation (Maney and Wait, 1991; USEPA, 1993, 1994). The Hudson River PCB congener
chemistry program required sophisticated, high resolution gas chromatography analyses with
stringent quality control criteria. In addition, various inorganic and physical parameters were
analyzed to define the chemical context within which the PCB congeners exist. This approach
was necessary to delineate the concentration of PCB congeners within the context of
geochemical and biological processes occurring in the river.This report focuses on the usability
of the PCB data generated by the Low Resolution Sediment Coring Study, one of several studies
including the High Resolution Sediment Coring Study and the Ecological Study, that when taken
together constitute the overall program. The data usability assessment was done in a manner
consistent with that used during the assessment of the PCB data generated during the High
Resolution Sediment Coring Study.
TAMS/Gradient selected a total of 90 PCB congeners as target congeners based on their
significance in environmental samples and the availability of calibration standards at the start of
the overall program fi e., the high resolution sediment coring study). As the program evolved,
Aquatec obtained qualitative and quantitative information for additional PCB congeners (non-
target congeners) from each sediment sample analysis using relative retention time information
detailed in the literature, and more recently verified with actual standards. For the low resolution
sediment coring study, data for 126 different PCB congeners were utilized; these congeners are
listed on Table A-l. Included in this group of 126 congeners are 12 for which Aquatec calibrated
on a daily basis, listed as "No-Cal" on Table A-l. Also included in the 126 congeners is one
pair, BZ #101 and BZ #90, which coeluted and could not be quantitated separately. Therefore,
the database of 126 congeners consists of 125 data points per sample.
Certain target congeners are of particular importance in evaluating geochemical and
biological processes within the Hudson River sediments. These are the 12 "principal" target
congeners, which consist of BZ #1, 4, 8, 10, 18, 19, 28, 52, 101, 118, 138, and 180. The focus of
this report will be on the usability of the analytical data for these 12 principal congeners.
A-l
TAMS/Gradient
-------
This report serves as ait overall evaluation of the PCB congener analyses performed for
the Hudson River low resolution sediment coring study. The evaluation is based on the
assessment of data quality relative to the objectives of the study. This report will first provide a
synopsis and assessment of the field sampling, analytical chemistry and data validation
programs, and then evaluate data usability for the 126 congeners for which data was used in the
low resolution sediment report, with particular emphasis on the 12 principal target congeners. A
data usability report assessing the non-PCB chemical and physical analyses for the low
resolution sediment samples is provided separately (Appendix B).
It should be noted that the data generated during the course of the low resolution
sediment coring program included more than the 126 congeners discussed in this usability report.
The usability of the data for additional congeners is provided in the usability reports associated
with the part of the overall program in which the data from these additional congeners is used.
However, for consistency with the high resolution sediment coring program, only the 126
congeners that are in common between the low and high resolution coring programs are utilized.
A.2 Field Sampling Program
TAMS/Gradient designed the low resolution sediment coring study to examine the long-
term inventory of PCB in the sediment of the Thompson Island pool; to refine the PCB mass
estimates for six hot spots below the Thompson Island pool; and to explore several areas in
which little was known with regard to PCB distribution. TAMS/Gradient described the low
resolution sediment collection program, sampling procedures, analytical protocols, and quality
control/quality assurance requirements in Volume 4 of the "Phase 2B Sampling and Analysis
Plan/Quality Assurance Project Plan - Hudson River PCB Reassessment RI/FS"
(TAMS/Gradient, June 1994; referred to in this report as the Phase 2B SAP/QAPP).
TAMS/Gradient collected cores using a vibrating coring device (vibra-coring). Three to five
cores were collected at each station. Once the cores were returned to shore, the sampling team
extruded and aliquoted sediments from the cores in a manner described in the Phase 2B
SAP/QAPP, and illustrated in Figure A-1. For most samples, this procedure involved reserving
the lowest portion of the core (approximately a 3-inch thick slice from the bottom) for
A-2
TAMS/Gradient
-------
radionuclide (137Cs) analysis, then dividing the remainder of the core into three slices of equal
thickness, with a 1 -inch thick portion of the top slice of the core also being designated for
radionuclide (I37Cs and 7Be) analysis. The sampling team aliquoted each slice into appropriate
containers and submitted the samples to a contract laboratory for analysis.
Scientists from TAMS and their subcontractors performed sampling for the low
resolution sediment coring study from July 13, 1994 through August 12, 1994. The sampling
team collected a total of 371 sediment samples (excluding duplicates and co-located samples)
from 170 sampling cores in the Thompson Island pool and at various locations downstream from
the Thompson Island pool. Aquatec allocated these samples into 20 sample delivery groups
(SDGs). The TAMS/Gradient Program Quality Assurance Officer (QAO) conducted a field
sampling audit on July 21, 1994 to assess compliance of the sampling procedures with the Phase
2B SAP/QAPP. The audit findings indicate that the sampling program was being conducted in a
technically acceptable manner consistent with the Phase 2B SAP/QAPP (Wait, 1994).
A.3 Analytical Chemistry Program
A.3.1 Laboratory Selection and Oversight
TAMS/Gradient retained a number of analytical laboratories to perform the analyses
required for this program. To verify that the selected laboratories had the capacity, capabilities,
and expertise to perform sample analyses in strict accordance with the specified methodologies,
each qualifying laboratory underwent an extensive audit by TAMS/Gradient's senior chemists.
TAMS/Gradient retained Aquatec Laboratories, a division of Inchcape Testing Service located in
Colchester, Vermont to perform the low resolution sediment sample PCB congener, total organic
carbon (TOC), and total kjeldahl nitrogen (TKN) analyses for the Hudson River RI/FS program.
Aquatec was the sole analytical laboratory which conducted the PCB congener analyses for the
entire program, including the high resolution sediment study and the ecological study, thus
maximizing the comparability of the PCB data across these programs.
TAMS/Gradient conducted routine laboratory audits during the low resolution sediment
A-3
TAMS/Gradient
-------
coring study to verify compliance of Aquatec with the Phase 2B SAP/QAPP requirements.
Unique requirements of the PCB congener method necessitated refinements of previously
published methods. In conjunction with these changes, Aquatec conducted Method Detection
Limit (MDL) studies and Extraction Efficiency (EE) studies for the sediments to evaluate the
adequacy of the methods. To conduct these studies, TAMS/Gradient collected seven replicate
Hudson River sediment samples. For the MDL studies, TAMS/Gradient collected the samples
upstream from the zone of major PCB contamination. TAMS/Gradient collected samples used
for the EE study from within the zone of major PCB contamination. A synopsis of the MDL/EE
studies is provided in a TAMS/Gradient memorandum dated July 12, 1993 (Cook, 1993). The
TAMS/Gradient Program Quality Assurance Officer oversaw and approved the method
refinements throughout the process.
A.3.2 Analytical Protocols for PCB Congeners
The method used by TAMS/Gradient for the determination of PCB congeners in Phase
2B is a program-specific method, essentially the same as that used in the high resolution
sediment coring program except as noted herein, and was based on NYSDEC's Analytical
Services Protocol Method 91-11 (NYSDEC, 1989) for PCB congeners. Appendix A4 of the
Phase 2A SAP/QAPP describes procedures for the calibration, analysis, and quantitation of PCB
congeners by fused silica capillary column gas chromatography with electron capture detection
(GC/ECD). The method is applicable to samples containing PCBs as single congeners or as
complex mixtures, such as commercial Aroclors. Aquatec extracted sediment samples with
hexane, and performed applicable cleanup procedures prior to analysis by GC/ECD, as detailed
in Appendix A3 of the Phase 2A SAP/QAPP. Aquatec analyzed hexane extracts for PCB
congeners on a dual capillary-column GC/ECD, as detailed in Appendix A4 of the Phase 2A
SAP/QAPP and identified PCB congeners using comparative retention times on two independent
capillary columns of different polarity.
Aquatec used calibration standards for each target congener to define retention times. In
addition, Aquatec routinely analyzed Aroclor standards and mixtures of Aroclor standards to
n n
A-4
T AMS/Gradient
-------
verify identification and quantitation of the primary calibration standards. Because of the non-
linear nature of the ECD over any significant calibration range (for this project 1 to 100 ppb in
extract), Aquatec generated the calibration curves used for quantitation from a quadratic
weighted least squares regression model where the correlation coefficient is greater than 0.99
(McCarty, 1995; USEPA, 1986 - Method 8000B, proposed 1995 update; promulgated in Update
III, December 1996).
For each PCB congener which elutes as a single congener on each GC column, Aquatec
reported the result as the lower of the two values. Although this quantitation scheme is in
compliance with USEPA CLP guidelines for dual-column analyses (USEPA, 1991), it may
introduce a slightly low bias when calculating homologue and total PCB sums. TAMS/Gradient
compared data in the database relative to absolute results on both columns and found the bias
was usually negligible, and on a worst-case basis, may be as low as 2% to 10% low. For
situations where coelution occurred on one column, Aquatec quantitated the result from the
column not displaying coelution. When only coelution results were available, Aquatec
performed a calculation to decipher concentrations using response factors derived by Mullen
(1984). Five of the 12 principal congeners (BZ #1, 18, 28, 52, and 180) were eluted as a single
congener peak on both GC columns. Six principal congeners (BZ #4, 8, 10, 19, 118, and 138)
were eluted as a single congener peak on one column and coeluted on the other column. One
congener, BZ #101, was coeluted on both columns and always reported with BZ #90.
Approximately 10% of all samples analyzed by GC/ECD also underwent additional
analysis using a GC-ion trap detector (ITD) as an additional means of confirming PCB congener
identifications, as detailed in Appendix A5 of the Phase 2A SAP/QAPP. When possible,
Aquatec selected samples with the highest concentrations of PCB congeners for confirmation
analysis by GC/ITD. Usually, Aquatec performed two GC/ITD analyses per SDG, even if
congener concentrations were minimal throughout the SDG.
At the start of the Phase 2B sampling and analysis program, TAMS/Gradient and
Aquatec selected 90 target PCB congeners. These target congeners are listed in Table A-l
(identified by "yes" in the "Target Congener" column) and identified by BZ number
A-5
T A M S /Gradient
-------
(Ballschmiter and Zell, 1980). TAMS/Gradient and Aquatec based the selection of these 90 PCB
congeners on their significance in environmental samples and the commercial availability of
calibration standards. TAMS/Gradient referred to PCB congeners for which calibration
standards were available as "target congeners". To verify that congener response for these
calibration standards was reproducible over time, TAMS/Gradient examined calibration data
from November 1992 and October 1993. TAMS/Gradient found temporal consistency to be
acceptable on both GC columns (the RTX-5 and the SB-Octyl 50 columns) (Bonvell, 1994a).
The high resolution column chromatography techniques employed by Aquatec produced
an acceptable PCB resolution for numerous congeners not contained in the target congener
calibration standards. Thus, TAMS/Gradient decided during method refinement to report
approximately 50 additional PCB congeners. The laboratory identified these additional PCB
congeners based upon the relative retention times reported in the published literature (Mullen,
1984; Schulz, 1989; Fischer and Ballschmiter, 1988, 1989). Aquatec calibrated these additional
"non-target" congeners using the calibration curve for target congener BZ #52. Aquatec chose
BZ #52 because it eluted as a single congener peak in the middle region of the chromatogram for
both GC columns and is a major component of Aroclor 1242, the Aroclor anticipated in Hudson
River samples. Using additional congener calibration standards which became commercially
available by August 1993, Aquatec performed analyses to verify and refine the historical relative
retention times, and to determine individual congener calibration parameters. These analyses
confirmed a majority (36) of the historical non-target congener relative retention times. For all
analyses performed prior to August 1993, the results for 14 non-target congeners were not
confirmed by this analysis; thus TAMS/Gradient considered them unusable and deleted them
from the database, leaving a database of 126 congeners. A review of high resolution sediment
data indicated that the 36 confirmed non-target congeners represent a significant percentage, up
to 25 percent, of the total PCB mass. Therefore, TAMS/Gradient decided to include the non-
target congener results to calculate homologue and total PCB masses in the Hudson River. If
TAMS/Gradient did not include these non-target congener results, the resulting calculations for
homologue and total PCBs would have been significantly biased low. Since the non-target
congener results were to be included in the calculations of homologue and total PCB mass,
TAMS/Gradient applied an individual correction factor to each congener's results based on the
A-6
T A M S/Gradient
-------
analysis of the additional congener standards. The application of these correction factors served
to minimize the uncertainty associated with quantitation of non-target congeners. A series of
TAMS/Gradient memoranda describe the method for deriving these calibration correction factors
(Bonvell, 1993a, b, c). A listing of the derived calibration correction factors is provided in a
TAMS/Gradient memorandum (Bonvell, 1994b).
To establish a method of quantitating total Aroclor concentrations from PCB congener
data, Aquatec performed duplicate analyses of seven Aroclor standards (Aroclors 1016,1221,
1232, 1242, 1248, 1254, and 1260). TAMS/Gradient defined the quantitation of an Aroclor for
this program as the sum of all congeners present in the standard Aroclor mixture at a
concentration greater that 0.1% of the total Aroclor mass. The percentage of the total mass
represented by such congeners was then compared to the actual (prepared) concentrations of each
Aroclor standard. The results produced the following yields for the seven Aroclor standards:
Aroclor 1016=93.3%, Aroclor 1221=86.8%, Aroclor 1232=91.0%, Aroclor 1242=90.6%,
Aroclor 1248=89.2%, Aroclor 1254=95.8%, and Aroclor 1260=87.0%. Thus, in each case, the
90 target and 36 non-target congeners represented more than 87% of the original Aroclor mass.
For those Aroclors most important to the Hudson River based on General Electric's reported
usage (Brown et al., 1984), these congeners represented more than 90% of the Aroclor mass {i.e.,
Aroclors 1242, 1254, and 1016).
A.4 Data Validation
An essential aspect of understanding the uncertainties of the Phase 2B sediment data is
understanding the significance of the qualifiers associated with the results. Each result may have
an associated qualifier. Qualifiers denote certain limitations or conditions that apply to the
associated result. Initially, the analytical laboratories applied qualifiers to the results, and then
the data validators modified the qualifiers, as necessary, based on the established validation
protocols. Data reporting and validation qualifiers direct the data users concerning the use of
each analytical result. TAMS/Gradient used two sets of qualifiers in the database, one set for
PCB congener data, and a second set for non-PCB chemical and physical data. Aquatec
developed an extensive list of data reporting qualifiers to be applied to the PCB congener data.
A-7
T MAS! Gradient
-------
The list is based on standard USEPA qualifiers used for organic analyses, with additional
qualifiers provided to note unique issues concerning PCB congener analysis, e.g., the
quantitation scheme. The data reporting qualifiers for PCB congener data, as applied by
Aquatec, are defined in detail in Table A-2. Qualifiers for non-PCB data are discussed in a
separate document (Appendix B).
During validation, the validators made modifications to the data qualifiers which are
reflected in the database. CDM Federal Programs Corporation and their subcontractors, under a
separate USEPA contract, performed data validation for the low resolution sediment coring
study. Validation procedures employed by CDM for GC/ECD analyses for the low resolution
sediment coring study were the same as for the high resolution coring study except as noted
below. These procedures are detailed in Appendix A6 of the Phase 2A SAP/QAPP, and
validation guidelines for GC/ITD analyses are provided in Appendix A7 of the Phase 2A
SAP/QAPP. TAMS/Gradient devised the validation procedures to reflect the data quality
objectives of the program, as well as to conform with USEPA (1988, 1992a) standards as
appropriate. USEPA Region II concurred with these method-specific validation protocols. In
addition, TAMS/Gradient designed comprehensive data validation templates to facilitate
consistency of approach and actions during validation. Prior to validation of the PCB data,
Gradient conducted a training workshop to aid CDM in properly performing the validation.
Gradient reviewed and commented on the initial CDM validation reports and provided real-time
QA oversight.
The initial data validation efforts for the low resolution sediment samples were completed
in August 21, 1995. The results were subsequently incorporated into the TAMS/Gradient
database and were available for review in August 1996. The issues encountered during review of
PCB data from the high resolution sediment coring study regarding the inappropriate application
of blank data during validation were resolved prior to TAMS/Gradient's review of the low
resolution sediment coring data.
As an overall assessment of data quality, the TAMS/Gradient Program QAO reviewed
pertinent aspects of the sampling and analysis program {e.g., historical data, implementation of
A-8
T A M S/Gradient
-------
sampling protocols, laboratory performance) relative to the data quality objectives. Decisions on
data usability sometimes overrode data qualification codes, as justified in this report. All
qualifier changes made by the TAMS/Gradient Program QAO, as reflected in this data usability
report, are noted in the final database (code "Y" in the QA Comment field of database). For the
low resolution sediment coring study, TAMS/Gradient Program QAO modified 349 qualifiers
out of 46,375 PCB congener data records (125 data points [126 congeners] for 371 samples) as a
result of data usability issues, representing less than 0.8% of the data. Specifically,
TAMS/Gradient Program QAO restored the rejected data to usable status for three reasons. First,
octachloronaphthalene (OCN) was deemed to be an unacceptable surrogate standard (see Section
A.5.2), and therefore, TAMS/Gradient Program QAO restored any sample results rejected solely
due to poor OCN recoveries. Second, CDM rejected certain positive BZ #18 detects due to poor
dual column precision. The TAMS/Gradient Program QAO changed the rejection qualifier (R)
to estimated and presumptively present (JN). The TAMS/Gradient Program QAO based this
decision on the routine presence of BZ #18 in historical sediment samples containing PCBs, the
consistent PCB congener pattern distribution present throughout the Hudson River sediments,
and the confirmation of the presence and concentration of BZ #18 by the GC/ITD analysis on the
samples analyzed. Both the preponderance of BZ # 18 retention time data and BZ # 18
identification verification by GC/ITD for most ITD-confirmed samples warrants inclusion of this
principal congener in the database. Third, certain rejections due to retention time shifts were
restored because validators noted that shifts were documented in associated QC samples, and
thus, adjusted retention time windows could be used for accurate congener identification.
A.5 Data Usability
A.5.1 Approach
Most previous studies of PCB chemistry in Hudson River sediments have focused on the
concentration of specific Aroclors, total PCBs and/or the distribution of PCB homologues. The
current assessment of PCB fate and distribution in the Hudson River required TAMS/Gradient
scientists to implement sophisticated equilibrium chemistry and transport modeling studies
requiring concentration ratios of certain PCB congeners. As noted previously (Section A.l), 12
A-9
TAMS/Gradient
-------
target congeners are of particular importance. The usability of these 12 "principal" congeners is
the focus of this low resolution sediment coring study data assessment.
Principal congeners will be employed in the following studies by the data users:
• Molar dechlorination product ratio (MDPR) - The molar sum of BZ #1, 4, 8, 10, and
19 are compared to the molar sum of all 126 congeners analyzed. This ratio is then
compared to a similar index for Aroclor 1242 to assess, calculate, and evaluate the
extent of dechlorination.
• Transport modeling - BZ #4, 28, 52, 101, and 138 are considered
independently as compounds to model PCB transport.
• Aroclor 1016 and 1242 - BZ #18 is used to estimate the potential contribution of
Aroclor 1016 and 1242 to Hudson River sediments.
• Aroclor 1254 - BZ #118 is used to estimate the potential contribution of Aroclor 1254
to Hudson River sediments.
• Aroclor 1260 - BZ #180 is used to estimate the potential contribution of Aroclor 1260
to Hudson River sediments.
Thus, 12 principal congeners (BZ #1,4, 8, 10, 18, 19, 28, 52, 101, 118, 138, and 180) are the
focus of this usability report. However, the remaining target and non-target congeners have
important implications to the low resolution sediment coring study as well. TAMS/Gradient
used these congeners to calculate the concentrations of total PCBs, PCB homologues, and
Aroclor mixtures, as well as for congener pattern analysis.
A-10
TAMS/Gradient
-------
A.5.2 Usability - General Issues
The data quality objectives for the Hudson River low resolution sediment coring study
required the development of a sensitive program-specific gas chromatography method.
Available standard agency methods were not adequate to achieve the congener-specific
identifications and detection limits needed for the project. TAMS/Gradient based the method
utilized on a modified NYSDEC ASP Method 91-11 (1989) protocol encompassing information
published in the literature, as well as in-house research conducted by Aquatec. This research
included Method Detection Limit (MDL) studies and Extraction Efficiency (EE) studies
conducted in accordance with USEPA (1984, 1986) guidance. During the course of these
studies, and the inception of the first study of the overall program (high resolution sediment
coring); TAMS/Gradient and Aquatec noted various nuances to the methods that required
refinement. As such, TAMS/Gradient and Aquatec made modifications to some of the original
protocols. This section will discuss some of the more significant changes and ramifications of
those changes.
Additional Calibrated Congeners
Aquatec increased the number of PCB congeners contained in the calibration standards
from the original 90 target congeners selected by TAMS/Gradient to include an additional 18
congeners, 12 of which are included in the 126 congeners utilized for the low resolution coring
study. The 12 of these additional congeners which are utilized in the low resolution coring study
are as follows: BZ#17, 20, 33, 42, 45, 74, 110, 135, 143, 156, 174, and 178. Aquatec selected
these additional congeners for daily calibration due to their presence in Aroclor mixtures and
potential significance for the ecological study. This change occurred before the analysis of the
low resolution and ecological studies, but after analysis of the high resolution core, water column
and transect studies. These 12 congeners are reported in all data sets. Use of the data for six
additional calibrated non-target congeners (BZ#59, 72, 165, 168, 176, and 179) should be limited
since they are not consistently quantitated for all data sets. Comparison of the concentrations of
these congeners between the low resolution sediment coring study and the previous studies is not
A-li
J MAS! Gradient
-------
appropriate as the two methods of quantitation are not comparable; therefore, these six congeners
are not included in the discussions of data in the low resolution report. None of these six
additional congeners were selected as principal congeners, and therefore, the data analyses efforts
should not be affected.
Identification of Non-Target Congeners
At the beginning of the overall program, Aquatec identified non-target congeners based
on historical relative retention times reported in the literature. In August 1993, Aquatec analyzed
calibration standards for each of the non-target congeners. Using these additional calibration
standards, Aquatec performed analyses to confirm historical relative retention times. Though
these analyses verified a majority of the historical non-target congener relative retention times,
some of the historical relative retention times used to identify non-target congeners did not match
the relative retention times determined by the analyses of the non-target congener standards. At
that time, TAMS/Gradient deleted 14 non-target congeners from the database for all analyses
performed prior to August 1993 due to these unconfirmed identifications. The 14 non-target
congeners deleted were: BZ #35,39,46, 100,104, 130,131,132, 134,162,165,173, 176, and
179. Aquatec identified and confirmed these 14 congeners based on the current laboratory-
derived relative retention times for samples analyzed during and after August 1993, which
includes all the low resolution sediment analyses. Therefore, the results for these 14 non-target
congeners will remain in the database for all samples analyzed during and after August 1993;
however, the data are not utilized in the low resolution coring study report and are not included
in this data usability discussion. Use of these non-target congener data has been limited since
they are not consistently available for all data sets. If a situation arises where information for the
deleted non-target congeners is critical to a data user, an in-depth review of the chromatograms
and re-calculation of the concentrations could potentially produce usable results for some of
these congeners.
A-12
TAMS/Gradient
-------
Quantitation of Non-Target Congeners
The laboratory originally quantitated non-target congeners using the calibration curve
determined for BZ#52. Since the non-target congener results were to be included in the
calculations of homologue and total PCB mass, TAMS/Gradient desired a more accurate method
of quantifying the non-target congeners. Aquatec analyzed calibration standards for the non-
target congeners in September 1993, and again in April 1994, for the determination of congener-
specific response factors. Based on this information, TAMS/Gradient calculated correction
factors for each non-target congener and applied these to the laboratory data within the database
(Bonvell, 1994b).
GC Column Change
Initially, Aquatec used a HP-5 (or RTX-5) column and a SB-octyl-50 GC column for
PCB congener analyses. In November 1993, Aquatec obtained new SB-octyl-50 columns for
pending analyses of Phase 2 biological samples. Each of the new SB-octyl-50 columns showed
signs of column degradation resulting in severe peak retention time shifts. Due to the concern
that an acceptable SB-octyl-50 column would not be obtainable, TAMS/Gradient solicited
approval from USEPA Region II for a replacement column, ApiezonL. TAMS/Gradient was
concerned about data comparability for the overall program, but had no alternative. USEPA
Region II concurred with the replacement of the SB-octyl-50 column with the Apiezon L
column in December 1993. The Apiezon L column was selected for the following reasons:
• The Apiezon L column phase is similar to the SB-octyl-50 column phase.
• The Apiezon L column provides PCB congener separations similar to the SB-
octyl-50 column.
• The PCB congener retention times on the Apiezon L column are more stable than
on the SB-octyl-50 column.
A-13
T MAS! Gradient
-------
• The NYSDEC analytical laboratory performing Hudson River PCB congener
analyses was using the ApiezonL column successfully for fish samples.
In February 1994, Aquatec performed a comparison study for the two column sets, HP-
5/SB-octyl-50 and HP-5/Apiezon_L (Cook, 1994). Aquatec analyzed four Phase 2 pilot fish
samples on both the HP-5/SB-octyl-50 column combination and also the RTX-5/Apiezon_L
column combination. The PCB congener results compared well qualitatively and quantitatively
with a few exceptions. The results for BZ #15 and 37 were consistently 2 to 10 times higher on
the SB-octyl-50 column pair. Data users are cautioned that the results for BZ #15 and 37
reported through March 1994 and the same congeners reported after March 1994 are not
comparable due to differences in the method of quantitation. For example, comparisons of
sediment data between the high resolution sediment coring study and the low resolution sediment
coring study are not appropriate for BZ #15 and 37. All of the low resolution sediment samples
were collected and analyzed after March 1994.
Lower Column Concentration Bias
The USEPA CLP protocol specifies that for dual column GC analyses, the lower of the
two values from each column will be reported (USEPA, 1991). TAMS/Gradient incorporated
this same quantitation scheme into this program. This quantitative method may introduce a
slight low bias when calculating homologue and total PCB sums. TAMS/Gradient determined
that this bias was usually negligible, and on a worst-case basis, may be as much as 2 to 10% low.
Therefore, the data user should consider these totals as usable, but estimated values, due to the
uncertainties of the individual results which are summed to form these values.
Surrogate Spike Compound
At the inception of the high resolution sediment coring study, TAMS/Gradient and
Aquatec employed two surrogates, tetrachloro-m-xylene (TCMX) and octachloronaphthalene
(OCN). Aquatec noted, soon after the program began, that OCN recoveries were a problem. For
many of the sediment samples. OCN recoveries were less than 10% and sometimes 0% although
A-14
J kMSi Gradient
-------
the TCMX and matrix spike/matrix spike duplicate results for these same samples were usually
acceptable. Re-extraction and re-analysis of the same samples produced similar results. The
purpose of surrogate spike analyses is to evaluate the performance of the extraction procedure.
TAMS/Gradient and Aquatec determined that OCN was an inappropriate surrogate for this
program. Research by Aquatec suggested that OCN was breaking down to
heptachloronaphthalene and hexachloronaphthalene. This information was known before the
analysis of the low resolution sediment coring samples and therefore BZ #192 was used as a
surrogate compound as well. During the validation process, CDM did not, in general, reject data
that had OCN recoveries below 10%, but when they did, the TAMS/Gradient Program QAO
considered these results to be usable and changed the "R" qualifier (rejected data) to a "J"
qualifier (estimated value) for any result which had been rejected solely due to poor OCN
recoveries.
Confirmation by GC/ITD
Aquatec analyzed approximately 10% of all samples analyzed by GC/ECD by GC/ITD to
provide an additional mechanism to verify congener identification and, as a secondary objective,
quantitation of congeners. The ITD is not as sensitive as the ECD (approximately an order of
magnitude less sensitive); therefore, when possible, samples with the highest concentration of
PCBs were selected for GC/ITD confirmation. Although this may result in a program bias for
only confirming high concentration samples, the overall effect does not impair data usability.
One unanticipated effect of selecting high concentration samples is that they were often
diluted for the GC/ECD analysis to a greater extent than the GC/ITD analysis. Consequently, the
sample-specific quantitation limit for the GC/ECD was often greater than that of the GC/ITD
analysis. In some cases, congeners were detected by the GC/ITD at concentrations less than the
GC/ECD quantitation limit and thus were not detected by the GC/ECD analysis. CDM qualified
such congeners with "M" during data validation, even though, the results from the two analyses
were consistent. TAMS/Gradient converted 46 of the "M" qualifiers which met this criterion to
"UJ".
A-15
J Gradient
-------
In addition, there is the potential for some quantitative bias associated with the GC/ITD
results relative to the GE/ECD results. Aquatec quantified each congener detected in the
GC/ITD analysis using an average response factor for each level of chlorination (i.e., homologue
group) rather than using response factors determined specifically for each individual congener.
As such, potential bias, which will vary for each congener within a chlorination homologue
group, is present with the GC/ITD results.
A.5.3 Usability - Accuracy, Precision, Representativeness, and Sensitivity
TAMS/Gradient established a quality assurance system for this program to monitor and
evaluate the accuracy, precision, representativeness, and sensitivity of the results relative to the
data quality objectives. These are all important elements in evaluating data usability (e.g.,
USEPA, 1992b, 1993). Accuracy is a measure of how a result compares to a true value.
Precision indicates the reproducibility of generating a value. Representativeness is the degree to
which a measurement(s) is indicative of the characteristics of a larger population. Sensitivity is
the limit of detection of the analytical method.
This section will evaluate each of these parameters for the low resolution sediment coring
study. TAMS/Gradient assessed accuracy using holding times, instrument performance and
calibrations for both the GC/ECD and GC/ITD, internal standard performance for the GC/ITD.
surrogate criteria for both the GC/ECD and GC/ITD, spike recoveries, matrix spike/matrix spike
duplicate recovery results, and compared identification results. TAMS/Gradient assessed
precision by comparing matrix spike and matrix spike duplicate results. TAMS/Gradient
evaluated representativeness by comparing field duplicate results, and assessed sensitivity using
blank results and the sample-specific quantitation limits achieved.
Comparability and completeness are two other important data quality attributes.
Comparability expresses the confidence with which data are considered to be equivalent to other
data sets (USEPA, 1992b). Comparable data allowed for the ability to combine the analytical
results obtained from this study with previous Hudson River studies. An in-depth discussion of
data comparability was provided in Chapter 3 of the report on the high resolution sediment
A-16
TAMS/Gradient
-------
coring program. In addition, Gauthier (1994) has provided Aroclor translation procedures for
Hudson River capillary column GC data relative to previous packed column GC studies.
Completeness is a measure of the amount of usable data resulting from a data collection activity
(USEPA, 1992b). For this program, a 95% completeness goal was established. A discussion of
completeness for the low resolution sediment coring study is provided in the conclusions section
of this report.
A.5.3.1 Accuracy
Holding Times
Exceedance of holding times may indicate a possible loss of PCB congeners due to
volatilization, chemical reactions, and/or biological alterations. Due to the persistent nature of
PCBs, only severe exceedance should be considered deleterious to quantitative accuracy. For the
sediment samples, TAMS/Gradient established an extraction holding time of 7 days from
sampling, followed by an analysis holding time of 40 days from extraction.
Aquatec missed the extraction holding times for four sediment samples and four sediment
sample re-extractions by 2 to 22 days and 72 to 90 days, respectively. Aquatec missed the
analytical holding times for 10 primary sample analyses and 6 dilution analyses by 16 to 62 days.
CDM appropriately qualified as associated results for these samples as estimated. Aquatec has
routinely demonstrated the stability of all PCB congener standards in solvent is at least six
months. The TAMS/Gradient Program QAO considered all data qualified as estimated due to
analytical holding time violations to be usable as estimated values.
GC/ECD Instrument Performance
Adequate chromatographic resolution and retention time stability throughout an
analytical sequence are essential attributes for qualitative identification of congeners on a GC.
TAMS/Gradient defined criteria for congener resolution and retention time windows in the Phase
2A SAP/QAPP and these were applied to the low resolution sediment coring program. The data
A-17
TAMS/Gradient
-------
validation reports appropriately noted exceedances according to these criteria and qualified the
data affected data as estimated. There were few qualifications based on resolution or retention
time windows exceedances. Aquatec initially established retention time windows for both
columns at ±0.3% relative to the average initial calibration retention times for all target
congeners and surrogates. For data validation purposes, EPA Region II agreed to allow
expanded retention time windows of ±0.5%
GC/ECD Calibration
Instrument calibration requirements were established to verify the production of
acceptable quantitative data. Initial calibrations (IC) using 5-level standard concentration curves
demonstrate an instrument is capable of acceptable performance prior to sample analysis. The IC
criteria is 20% relative standard concentration error (%RSCE) for monochlorobiphenyl and 15%
RSCE for all remaining PCB congeners, as well as a correlation coefficient > 0.995. Continuing
calibration standards document maintenance of satisfactory performance over time. The data
validation reports appropriately noted any deviation from these criteria. Deviations from the
criteria were not significant. TAMS/Gradient noted no significant continuing calibration
problems.
Surrogate Spike Recoveries
Aquatec spiked surrogate compounds into all sediment samples prior to extraction to
monitor recoveries. Recoveries may be indicative of either laboratory performance or sample
matrix effects. For the low resolution sediment coring study, Aquatec used TCMX, OCN, and
BZ #192 as surrogates. As previously discussed, OCN did not perform properly as a
representative surrogate, therefore, only TCMX and BZ #192 recoveries provided useful
information. The TAMS/Gradient Program QAO considered data which had been rejected solely
because of poor OCN recoveries to be usable as estimated values. Data was restored to usable
status for six sediment samples including 39B0008, 39D0814, 39F1222, 10C0009, 10D0009, and
11A1019.
A-18
T AMS/Gradient
-------
Matrix Spike/Matrix Spike Duplicate Recoveries
Within each SDG, two aliquots of a representative sediment sample were spiked with a
suite of 20 congeners (BZ #8, 18, 28, 44, 52, 66, 77, 101, 105, 118, 126, 128, 138, 153, 170, 180,
187, 195, 206, and 209). The purpose of the spikes were, in part, to evaluate the accuracy of the
analytical method relative to laboratory performance and specific sample matrix. The advisory
limits for spiked congener recoveries are 60-150%. TAMS/Gradient noted no significant spike
recovery problems for the low resolution sediment cores. Matrix spike/matrix spike duplicate
analyses were analyzed for 22 low resolution sediment core samples. This represents a
frequency of 5.9%, which exceeds the 5% requirement stipulated in Phase 2B SAP/QAPP.
Compound Identification
TAMS/Gradient established qualitative criteria to minimize erroneous identification of
congeners. An erroneous identification can be either a false positive (reporting a compound
present when it is not) or a false negative (not reporting a compound that is present). The
calculated concentrations for congeners detected in both columns should not differ by more than
25% between columns (%D < 25%). This criterion applies to only those congeners which can be
resolved as individual congeners on both columns. If the %D for the results between the two
columns is > 25% but <50%, the results were estimated. If the %D was > 50% but < 90%, the
results were estimated and presumptively present (GN). If the %D between columns was > 90%,
the results were unusable (R).
TAMS/Gradient noted problems with congener identifications as a result of dual column
imprecision for numerous SDGs. The majority of the estimated and rejected data for the low
resolution sediment coring study were a result of dual GC column imprecision. CDM qualified
the following congeners as rejected at frequencies greater than 10% as a result of dual column
imprecision: BZ #2 (14%), BZ #3 (23%), BZ #12 (19%), BZ #137 (14%), and BZ #194 (10%).
With the level of background organic material present in Hudson sediments, resultant
interferences, particularly for congeners with low concentrations, likely caused these differences
A-19
1MASI Gradient
-------
between the dual GC column results.
As previously mentioned, the QAO restored BZ #18 data had been rejected because of
dual column imprecision. This change was made for 67 samples. The QAO based this decision
on the routine presence of BZ #18 in Hudson River sediments, the consistent PCB congener
pattern distribution present throughout the sediments, and the confirmation of the presence and
concentration of BZ #18 by the GC/ITD analysis of the samples so analyzed. This treatment of
the data is consistent with the approach taken in the high resolution sediment coring study.
GC/ITD Instrument Performance
Verifying proper GC/ITD performance required evaluating GC column resolution, ion
trap detector sensitivity, and ion trap calibration. The GC resolution criteria required baseline
separation of BZ #87 from BZ #154 and BZ #77. The ion trap sensitivity requires the
signal/noise ratio to be m/z 499 for BZ #209 and m/z 241 for chrysene-di2 lo be greater than 5.
For ion trap calibration, the abundance of m/z 500 relative to m/z 498 for BZ #209 must be >
70% but <95%. CDM appropriately qualified GC/ITD exceedances of these parameters during
validation. The criteria were met and the GC/ITD results were useful in confirming GC/ECD
results. In general, TAMS/Gradient noted no significant ITD performance problems for samples
analyzed during the low resolution sediment coring study.
GC/ITD Calibration
The initial calibration criteria for acceptable quantitative data for GC/ITD analyses
required percent relative standard deviations (% RSD) of the congener relative response factor
(RRF) to be less than 20%. For continuing calibration, the RRF for each congener must be
within 20% of the mean calibration factor from the 5-level calibration at the beginning and end
of each calibration sequence. For the low resolution sediment coring study, TAMS/Gradient
noted no significant GC/ITD calibration problems.
A-20
1 AMSi Gradient
-------
GC/ITD Internal Standard Performance
To demonstrate the stability of the ITD, internal standard performance criteria were
monitored. Internal standard area counts must not vary by more than 30% from the most recent
calibration or by more than 50% from the initial calibration. In addition, the absolute retention
time of the internal standard must be within 10 seconds of the retention time in the most recent
calibration, and ion abundance criteria must be met for chrysene-di2 and phenanthrene-d 10. For
the low resolution sediment coring study, TAMS/Gradient noted no significant internal standard
problems.
Confirmation by GC/ITD
CDM qualified all positive GC/ITD results that had signal/noise ratios of less than 3 as
not detected due to uncertainty in the identification. TAMS/Gradient considered these results to
be usable as undetected data at the reported quantitation limits.
Aquatec analyzed approximately 10% of all samples analyzed by GC/ECD by GC/ITD to
provide an additional mechanism to verify congener identification and, as a secondary objective,
quantitation of congeners. Since the ITD method was not designed to be a primary quantitative
tool, some variations in quantitative results were expected. TAMS/Gradient considered
quantitative differences between the GC/ITD and GC/ECD results less than a factor of five to be
acceptable, while differences greater than five times were considered unacceptable. CDM
qualified GC/ECD results that were detected at concentrations above the GC/ITD quantitation
limit but that were not confirmed by GC/ITD with a "Q". TAMS/Gradient converted all "Q"
qualifiers to "JN" due to the potential of reporting false positive results. CDM qualified 47
sediment results with "Q" qualifiers (of which one was a principal congener); TAMS/Gradient
considered these results to indicate the presumptive presence of the affected congener. CDM
qualified GC/ECD results that were not detected or were less than one-fifth the GC/ITD results
with an "M". TAMS/Gradient converted these "M" qualifiers to "R" as the nondetect GC/ECD
may be a false negative or the GC/ECD result may be significantly biased low. Of the 458
sediment results which CDM qualified with "M" (of which 21 were principal congeners);
A-21
MAS>! Gradient
-------
TAMS/Gradient considered 412 of these results to be unusable. As noted previously (Section
A.5.2), the other 46 "M" qualified data points were changed to "UJ".
*
A.53.2 Precision
Matrix Spike/Matrix Spike Duplicate Comparison
The analysis of matrix spike (MS) and matrix spike duplicate (MSD) samples can also
provide valuable information regarding method precision relative to laboratory performance and
specific sample matrix. The advisory limit for relative percent difference (RPD) of spiked
congeners in a MS/MSD pair is 40%, and for nonspiked congeners, the precision criterion is 40%
Relative Standard Deviation (RSD).
Overall, the MS/MSD performance for the low resolution sediment coring study was
good.
A.5.3.3 Representativeness
Field Duplicate Results
Analysis of field duplicate samples provides an indication of the overall precision of the
sampling and analysis program. These analyses measure both field and laboratory precision;
therefore, the results will likely have more variability than laboratory duplicates and MS/MSD
samples, which only measure laboratory precision. Data validators used a 50% RPD criterion for
evaluating field duplicate precision. Any congener precision greater than 50% RPD was
qualified as estimated ("J").
A total of 21 field duplicate samples were analyzed for the low resolution sediment
coring study. This represents a frequency of 5.7%, which exceeds the 5% requirement stipulated
in the Phase 2B SAP/QAPP. Overall, field duplicate precision was acceptable; especially in the
context of river sediments, which are typically heterogeneous. Table A-3 summarizes the
A-22
TAMS/Gradient
-------
duplicate precision results for the 12 principal congeners for each field co-located sample.
Typically a few congeners for each pair of co-located sediments exceeded the precision criterion.
CDM appropriately qualified the results for these results as estimated. TAMS/Gradient
considered these data to be usable as estimated values.
A.5.3.4 Sensitivity
Blanks
An important data quality objective associated with the low resolution sediment coring
study was to obtain detection limits as low as the analytical method could produce. Due to the
low detection limits achieved, low concentration blank contamination was detected during the
preparation and analysis of the sediments. As a result, numerous congeners in all samples in all
SDGs required qualification due to blank contamination. TAMS/Gradient reviewed the
distribution of blank contaminants and found most contamination associated with the
monochlorobiphenyls, particularly with BZ #2. Blank levels for BZ #2 usually ranged from 20
to 80 ppb in extract. Since BZ #2 is not a dechlorination product, a major Aroclor component, or
a principal congener, TAMS/Gradient did not consider this to be a serious data quality problem.
CDM qualified principal congeners in several samples due to blank contamination including: BZ
#1(15 results); BZ #4 (10 results); BZ #8 (8 results); BZ #10 (30 results); BZ #18 (14 results);
BZ #19 (9 results); BZ #28 (11 results); BZ #52 (9 results); BZ #101 with BZ #90 (3 results); BZ
#118 (16 results); BZ #138 (3 results); and BZ #180 (9 results). TAMS/Gradient considered
these results to be usable as non-detects.
CDM qualified results during data validations with a "B", which indicated that the result
was within 5 times of the blank action level (i.e., the highest concentration in a blank associated
with that sample result). TAMS/Gradient converted all "B" qualified results in the database to
nondetect results due to uncertainty in this detection. Table A-4 summarizes the congener
detects changed to non-detects for the sediment samples. TAMS/Gradient considered these
results to be usable as non-detects at the reported quantitation limit.
A-23
TAMS/Gradient
-------
Quantitation Limits
Evaluating dechlorination processes and modeling transport pathways of PCB congeners
in sediments necessitated obtaining low detection limits. TAMS/Gradient and Aquatec devised
analytical methods to enhance lower detection limits. This, in part, required employing
sample/extract cleanup methods to remove matrix interferences, and maximizing sample size
when possible. For the low resolution coring study, TAMS/Gradient defined optimum detection
limits as 1 fig/kg for monochlorobiphenyls, 0.5 ng/kg for dichlorobiphenyls through
hexachlorobiphenyls, and 0.5-1 fig/kg for heptachlorobiphenyls through decachlorobiphenyl.
Results of the MDL study necessitated raising the detection limit for BZ #2 (a
monochlorobiphenyl) significantly above these requirements (approximately a factor of 3).
In general, achieving appropriate detection limits for the sediment samples was not a
problem. Whenever TAMS/Gradient noted elevated detection limits, the affected samples
contained high organic content; specifically, the presence of PCBs. The relative ratio of
congeners detected within each high-concentration sample remained reasonably consistent,
therefore the elevated detection limit for non-detected congeners did not affect data usability.
A.5.4 Usability - Principal Congeners
The 12 principal target congeners employed in the high resolution sediment coring study
are key to delineating PCB geochemistry in the Hudson River. The following synopsis will
provide data users with the strengths and weaknesses of the principal target congener data within
the context of this study:
BZ #1. The reported results for BZ #1 met the data quality objectives of the
program. Results for BZ #1 in 10 sediment samples were rejected (out of
371 samples) based on quality control exceedances. Analytically, BZ #1
eluted as a single peak on both GC columns. Detection limits for BZ # 1. a
monochlorobiphenyl, were generally 1 to 6 ppb, which were acceptable.
BZ #4. All reported results for BZ #4 met the data quality objectives of the
program and are usable for project decisions. Analytically, BZ #4 eluted as
A-24
T AMS/Gradient
-------
a single peak on one GC column, and coeluted with BZ #10, another
principal congener, on the other GC column. Data for both BZ #4 and BZ
#10 were considered usable. With regard to detection limits, a goal of 0.5
ppb was established. In general, this goal was met, however, there were
many samples with associated blank levels of 10 to 20 ppb of BZ #4
in the extract, which required raising the detection limit. This did not
affect data usability.
BZ #8. All reported results for BZ #8 met the data quality objective of the
program and are usable for project decisions. Analytically, BZ #8 eluted
as a single peak on one GC column and coeluted with BZ #5 on the other
GC column, which was acceptable for the purposes of this program. The
detection limit goal of 0.5 ppb was met for nearly all samples. Matrix
spike results for BZ #8 further indicated that the method was successful.
BZ#10. The usability assessment for BZ #10 is similar to that for BZ #4. BZ#10
eluted as a single peak on one GC column and coeluted with BZ #4 on the
other GC column. All results that were reported for both BZ #4 and BZ
#10 were considered usable. In general, the detection limit goal of 0.5 ppb
was met.
BZ #18. Numerous results for BZ #18 were initially rejected by the data validator
due to poor dual column precision. The TAMS/Gradient Program QAO
changed the rejection qualifier to a presumptively present qualifier based
on the presence of BZ #18 in historical sediment samples containing
PCBs, the consistent PCB congener pattern distribution present throughout
the Hudson River sediment, and GC/ITD confirmational analysis on about
10% of the data. Detailed review of the affected BZ#18 data suggested an
interferant causing the high %D values. Analytically, BZ # 18 eluted as a
single peak on both GC columns. The detection limit goal of 0.5 ppb was
met for nearly all samples. Matrix spike results for BZ #18 further
indicated that the method was successful. As such, all reported results for
BZ #18 met the data quality objectives of the program.
BZ #19. All reported results for BZ #19 met the data quality objectives of the
program. Analytically, BZ #19 eluted as a single peak on one GC column
and coeluted on the other. The detection limit goal of 0.5 ppb was met for
nearly all samples.
BZ #28. The reported results for BZ #28 met the data quality objectives of the
program. The BZ #28 result for one sediment samples was rejected due to
dual GC column imprecision. Analytically, BZ #28 eluted as a single
congener peak on both GC columns. The detection limit goal of 0.5 ppb
was met for nearly all samples. Matrix spike results for BZ #28 further
indicates the method was successful.
A-25
TA M S/Gradient
-------
BZ #52. All reported results for BZ #52 met the data quality objectives of the
program and are usable for project decisions. Analytically. BZ #52 eluted
as a single congener peak on both GC columns. The detection limit goal of
0.5 ppb was met for nearly all samples. Matrix spike recovery for BZ #52
further indicated that the method was successful.
BZ #101. Data users should be aware that BZ #101 always coeluted with BZ #90 (on
both GC columns), and therefore was always reported with BZ #90. For
all reported results, all other QA/QC requirements were met, and
therefore, these results are usable for project decisions. The detection limit
goal of 0.5 ppb was met for nearly all samples. Matrix spike results for
BZ #101 further indicated that the method was successful.
BZ #118. The reported results for BZ #118 met the data quality objectives of the
program in most samples. BZ #118 results in 9 sediment samples were
rejected due to dual column imprecision. Analytically, BZ #118 eluted as
a single peak on one GC columns and coeluted with BZ #122 on the other
GC column. The detection limit goal of 0.5 ppb was met for nearly all
samples. Matrix spike results for BZ #118 further indicated that the
method was successful.
BZ #138. The reported results for BZ #138 met the data quality objectives of the
program for most samples. BZ #138 results in 11 sediment samples were
rejected due to dual column imprecision. Analytically, BZ #138 eluted as
a single peak on one GC column and coeluted on the other GC column.
The detection limit goal of 0.5 ppb was met for nearly all samples. Matrix
spike results for BZ #138 further indicated that the method was successful.
BZ #180. The reported (valid) results for BZ #180 met the data quality objectives of
the program. BZ #180 results in 32 sediment samples were rejected due to
dual column imprecision.The 32 rejections (8.6%) exceeds the 5%
unusable data DQO (data is less than 95% complete), so the completeness
objective was not met for BZ#180. Analytically, BZ #180 eluted as a
single peak on both GC columns. The detection limit goal of 0.5 ppb was
met for nearly all samples. Matrix spike results for BZ #180 further
indicated that the method was successful.
Typically, rejection of parameters occurred randomly. In no single sample were
all principal target parameters rejected. Rejection of one or more parameters does
not signify rejection of the entire sample or the entire core. Total PCB and total tri
and higher chlorinated congeners was calculated for each sample despite rejected
parameters, because the contribution of mass for a single congener to the total
PCB mass in a sample is small (approximately 1-2%) for the majority of samples.
A-26
1 \MS>!Gradient
-------
A.6 Conclusions
The analytical chemistry program implemented by TAMS/Gradient for the Hudson River
low resolution sediment coring study was extremely sophisticated, requiring the use of state-of-
the-art GC methodology. Data for 126 congeners were utilized from a total of 371 sediment
samples analyzed (excluding 21 field duplicate samples). (The low resolution database also
contains data for an additional 20 non-target congeners which were not used in the low resolution
sediment coring study report.) Considering the complexity of the program, TAMS/Gradient
considers the outcome of the analytical chemistry program to have been successful.
A summary of the number of qualifiers applied to each PCB congener is tabulated in
Table A-5. For the low resolution sediment coring study, 46,375 congener measurements were
recorded, of which 1,228 values were rejected. Congeners most often rejected include BZ #2
(14%), BZ #3 (23%), BZ #12 (19%), BZ #137 (14%) and BZ #194 (10%). The reason for most
of these rejections was the imprecision between the GC columns. A 97.4% overall completeness
rate was achieved for the low resolution sediment coring analytical program, which successfully
exceeded the 95% completeness objective. The only principal congener which did not meet the
completeness objective was BZ #180 (91% completeness), however, this did not impair the
overall integrity of the program.
A majority (54%) of all congener results (both detects and nondetects) were qualified as
estimated or as estimated and presumptively present. Again, the main reason for most of the
qualifications was detection at concentrations below the calibrated quantitation limit and/or
exceedance in the dual GC column precision criteria. Numerous congeners for nearly all SDGs
had calculated concentrations on each GC column which differed by more than 25%, but less
than 50%, which warranted qualification as estimated values. With the level of background
organic material present in Hudson sediments, resultant interferences, particularly for congeners
with low concentrations, likely caused these differences between the GC columns. Other
problems contributing to data qualification included missed holding times, and some GC/ECD
calibration criteria exceedances. Data users should consider all detect and non-detected results
which were estimated to be usable relative to the data quality objectives of the program.
A-27
T Gradient
-------
References
Ballschmiter, K. and M. Zell. 1980. "Analysis of Polychlorinated Biphenyls (PCB) by Glass
Capillary Gas Chromatography. Composition of Technical and Aroclor and Clophen PCB
Mixtures." Fresenius Z. Anal. Chem., 302:20-31.
Bonvell, S. 1993a, b, c. Congener Calibration. TAMS/Gradient memoranda, dated August 26,
September 17, and December 29.
Bonvell, S. 1994a. Congener Calibration - Temporal Consistency. TAMS/Gradient
memorandum, dated March 7.
Bonvell, S. 1994b. Calibration of Non-Target Congeners. TAMS/Gradient memorandum, dated
June 22.
Brown, J.F., R.E. Wagner, D.L. Bedard, M.J. Brennan, J.C. Carnahan, and R.J. May. 1984.
"PCB Transformation in Upper Hudson Sediments." Northeast Environ. Sci. 3:167-179.
Cook, L.L. 1993. Sediment Method Detection Limit/Extraction Efficiency Determination.
TAMS/Gradient memorandum, dated July 12.
Cook, L.L. 1994. Apiezon L column study. TAMS/Gradient memorandum, dated August 4.
Fischer, R. and K. Ballschmiter. 1988. "Ortho-substituent Correlated Retention of
Polychlorinated Biphenyls on a 50% n-octyl Methylpolysiloxane Stationary Phase by
HRGC/MSD." Fresenius Z. Anal. Chem., 332:441-446.
Fischer, R. and K. Ballschmiter. 1989. "Congener-specific Identification of Technical PCB
Mixtures by Capillary Gas Chromatography on a n-octyl-methyl Silicon Phase (SB-octyl-50)
with Electron Capture and Mass-selective Detection." Fresenius Z. Anal. Chem., 335:457-463.
Garvey, E. 1995. Letter to Scott Graber and Jennifer Oxford of CDM Federal Programs Corp.
from TAMS, dated May 9.
Gauthier, T. 1994. Aroclor Translation Procedures. TAMS/Gradient memorandum, dated July
7.
Maney, J. and D. Wait. 1991. "The Importance of Measurement Integrity." Environ. Lab,
3(5):20-25.
McCarty, H. and B. Lesnik. 1995. "Approaches to Quality Control of Non-linear Calibration
Relationships for SW-846 Chromatographic Methods." In Proceeding of the Eleventh Annual
Waste Testing & Quality Assurance Symposium. American Chemical Society and U.S.
Environmental Protection Agency, Washington, DC, July 23-28, pp 203-208.
Mullen, M. 1984. "High Resolution PCB Analysis: Synthesis and Chromatographic Properties
A-28
TAMS/Gradient
-------
of all 209 PCB Congeners." Environ. Sci. Technol., 18:468-475.
NYSDEC. 1989. Analytical Service Protocols. Issued September 1989, revised December 1991
and September 1993, Method 91-11, pp D-XXVIII, 5-59. New York State Department of
Environmental Conservation, Bureau of Technical Services and Research, Albany, New York.
Schulz, D. 1989. "Complete Characterization of Polychlorinated Biphenyl Congeners in
Commercial Aroclor and Clophen Mixtures by Multidimensional Gas Chromatography-electron
Capture Detection." Environ. Sci. Technol., 23:852-859.
TAMS/Gradient. 1992. "Phase 2A Sampling and Analysis Plan/Quality Assurance Project Plan
- Hudson River PCB Reassessment RJ/FS." EPA Contract No. 68-S9-2001.
TAMS/Gradient. 1994. "Phase 2B Sampling and Analysis Plan/Quality Assurance Project Plan
- Hudson River PCB Reassessment Ri/FS." EPA Work Assignment No. 013-2N84.
USEPA. 1984. "Definition and Procedure for the Determination of the Method Detection Limit -
Revision 1.11." Federal Register, 49(209): 198-199.
USEPA. 1986. Test Methods for Evaluating Solid Waste. U.S. Environmental Protection
Agency, Office of Solid Waste and Emergency Response, Washington, DC, SW-846, Third
Edition, Chapter 1.
USEPA. 1988. "Laboratory Data Validation. Functional Guidelines for Evaluating Organics
Analyses." U.S. Environmental Protection Agency, Hazardous Site Evaluation Division,
Contract Laboratory Program, Washington, DC.
USEPA. 1991. "Statement of Work for Organic Analysis, Multi-Media, Multi-Concentration."
Document Number OLM 01.0, including revisions through OLM 01.8, August 1991, pp 55.
U.S. Environmental Protection Agency, Washington, DC.
USEPA. 1992a. "CLP Organics Data Review and Preliminary Review." SOP No. HW-6, Rev
No. 8, U.S. Environmental Protection Agency Region II, Edison, New Jersey, January.
USEPA. 1992b. "Guidance for Data Usability in Risk Assessment." U.S. Environmental
Protection Agency, Office of Emergency and Remedial Response, Washington, DC, EPA PB92-
963356, Publication 9285.7-09A.
USEPA. 1993. "Data Quality Objectives Process for Superfund." U.S. Environmental
Protection Agency, Office of Solid Waste and Emergency Response, Washington, DC, EPA 540-
R-93-071, Publication 9355.9-01.
USEPA. 1994. "Guidance for the Data Quality Objectives Process." U.S. Environmental
Protection Agency, Quality Assurance Management Staff, Washington, DC, EPA QA/G-4.
Wait, A.D. 1994. Field Quality Assurance Audit Report, dated July 21.
A-29
1 MAS!Gradient
-------
Table A-l
List of 126 Phase 2 Target and Non-Target PCB Congeners Used in
Low Resolution Sediment Coring Study Report
Congener Homologue Congener Target
Number Group Name Congener*
BZ #1
Mono
2-Chlorobiphenyl
Yes
BZ #2
Mono
3-Chlorobiphenyl
Yes
BZ #3
Mono
4-Ch!orobiphenyl
Yes
BZ #4
Di
2,2'-Dichlorobiphenyl
Yes
BZ #5
Di
2,3-Dichlorobiphenyl
Yes
BZ #6
Di
2,3'-Dichlorobiphenyl
Yes
BZ #7
Di
2,4-Dichlorobiphenyl
Yes
BZ #8
Di
2,4'-Dichlorobiphenyl
Yes
BZ #9
Di
2,5-Dichlorobiphenyl
Yes
BZ #10
Di
2,6-Dichlorobiphenyl
Yes
BZ #12
Di
3,4-Dichlorobiphenyl
Yes
BZ #15
Di
4,4'-Dichlorobiphenyl
Yes
BZ #16
Tri
2,2',3-Trichlorobiphenyl
Yes
BZ #17
Tri
2,2',4-T richlorobiphenyl
No - Cal
BZ #18
Tri
2,2',5-Trichlorobiphenyl
Yes
BZ #19
Tri
2,2',6-T richlorobiphenyl
Yes
BZ #20
Tri
2,3,3'-Trichlorobiphenyl
No - Cal
BZ #22
Tri
2,3,4'-T richlorobiphenyl
Yes
BZ #23
Tri
2,3,5-Trichlorobiphenyl
No
BZ #24
Tri
2,3,6-Trichlorobiphenyl
No
BZ #25
Tri
2,3',4-Trichlorobiphenyl
Yes
BZ #26
Tri
2,3',5-Trichlorobiphenyl
Yes
BZ #27
Tri
2,3',6-Trichlorobiphenyl
Yes
BZ #28
Tri
2,4,4-Trichlorobiphenyl
Yes
BZ #29
Tri
2,4,5-T richlorobiphenyl
Yes
BZ #31
Tri
2,4',5-Trichlorobiphenyl
Yes
BZ #32
Tri
2,4',6-Trichlorobiphenyl
No
BZ #33
Tri
2',3,4-Trichlorobiphenyl
No - Cal
BZ #34
Tri
2',3,5-Trichlorobiphenyl
No
BZ #37
Tri
3,4,4'-Trichlorobiphenyl
Yes
BZ #40
Tetra
2,2',3,3'-Tetrachlorobiphenyl
Yes
BZ #41
Tetra
2,2',3,4-Tetrachlorobiphenyl
Yes
BZ #42
Tetra
2,2'.3,4'-T etrachlorobiphenyl
No - Cal
BZ #44
Tetra
2,2',3,5'-T etrachlorobiphenyl
Yes
BZ #45
Tetra
2,2',3,6-T etrachlorobiphenyl
No - Cal
BZ #47
Tetra
2,2',4,4'-Tetrachlorobiphenyl
Yes
CD
N
%
00
Tetra
2,2',4,5-Tetrachlorobiphenyl
No
BZ #49
Tetra
2,2',4,5'-Tetrachlorobiphenyl
Yes
BZ #51
Tetra
2,2',4,6'-T etrachlorobiphenyl
No
BZ #52
Tetra
2,2',5,5'-Tetrachlorobiphenyl
Yes
TAMS/Gradient
-------
TABLE A-l (continued)
BZ #53
Tetra
2.2',5,6'-Tetrachlorobiphenyl
Yes
BZ #56
Tetra
2.3,3',4'-Tetrachlorobiphenyl
Yes
BZ #58
Tetra
2,3,3',5'-Tetrachlorobiphenyl
No
BZ #60
Tetra
2,3,4,4'-Tetrachlorobiphenyl
No
BZ #63
Tetra
2,3,4',5-TetrachIorobiphenyI
No
BZ #64
Tetra
2.3,4',6-T etrachlorobipheny 1
No
BZ #66
Tetra
2,3',4,4-T etrach lorobipheny 1
Yes
BZ #67
Tetra
2,3',4,5-Tetrachlorobiphenyl
No
BZ #69
Tetra
2,3',4,6-Tetrachlorobiphenyl
No
BZ #70
Tetra
2,3',4',5-Tetrachlorobiphenyl
Yes
BZ #74
Tetra
2,4,4',5-Tetrachlorobiphenyl
No - Cal
BZ #75
Tetra
2,4,4',6-T etrach lorobipheny I
Yes
BZ #77
Tetra
3,3',4,4'-Tetrachlorobiphenyl
Yes
BZ #82
Penta
2,2',3,3',4-Pentachlorobiphenyl
Yes
BZ #83
Penta
2,2',3,3',5-Pentachlorobiphenyl
Yes
BZ #84
Penta
2,2',3,3',6-PentachIorobiphenyl
Yes
BZ #85
Penta
2,2',3,4,4'-Pentachlorobiphenyl
Yes
BZ #87
Penta
2,2',3,4,5'-Pentachlorobiphenyl
Yes
BZ #90 (w/BZ#I01)
Penta
2,2',3,4',5-Pentachlorobiphenyl
No
BZ #91
Penta
2,2',3,4',6-Pentachlorobiphenyl
Yes
BZ #92
Penta
2,2',3,5,5'-Pentachlorobiphenyl
Yes
BZ #95
Penta
2,2',3,5',6-Pentachlorobiphenyl
Yes
BZ #96
Penta
2,2',3,6,6'-Pentachlorobiphenyl
No
BZ #97
Penta
2,2',3',4,5-Pentachlorobiphenyl
Yes
BZ #99
Penta
2,2',4,4',5-Pentachlorobiphenyl
Yes
BZ # 101 (w/BZ#90)
Penta
2,2',4,5,5'-Pentachlorobiphenyl
Yes
BZ #105
Penta
2,3,3',4,4'-Pentachlorobiphenyl
Yes
BZ # 107
Penta
2,3,3',4,5'-Pentachlorobiphenyl
Yes
BZ # 110
Penta
2,3,3'.4',6-Pentachlorobiphenyl
No - Cal
BZ #114
Penta
2,3,4,4',5-Pentachlorobiphenyl
No
BZ "115
Penta
2,3,4,4',6-Pentachlorobiphenyl
Yes
BZ #118
Penta
2,3'.4,4',5-Pen(ach!orobiphenyl
Yes
BZ #119
Penta
2,3',4,4',6-Pentachlorobiphenyl
Yes
BZ #122
Penta
2',3.3',4,5-Pentachlorobiphenyl
Yes
BZ #123
Penta
2'.3,4,4',5-Pentachlorobiphenyl
Yes
BZ #126
Penta
3.3'.4,4'.5-Pentachlorobiphenyl
Yes
BZ #128
Hexa
2,2\3.3',4,4'-Hexachlorobiphenyl
Yes
BZ #129
Hexa
2.2',3,3',4,5-Hexachlorobiphenyl
Yes
BZ #135
Hexa
2,2'.3,3',5,6'-Hexachlorobiphenyl
No - Cal
BZ #136
Hexa
2.2\3.3',6,6'-Hexachlorobiphenyl
Yes
BZ #137
Hexa
2,2',3,4,4',5-Hexachlorobiphenyl
Yes
BZ #138
Hexa
2.2',3,4,4',5-Hexachlorobiphenyl
Yes
BZ #140
Hexa
2,2',3,4,4',6'-HexachlorobiphenyI
No
BZ #141
Hexa
2,2',3,4,5,5'-Hexachlorobiphenyl
Yes
BZ #143
Hexa
2.2',3.4.5,6-Hexachlorobiphenyl
No - Cal
BZ #144
Hexa
2.2.3.4.5',6-Hexachlorobiphenyl
No
BZ #146
Hexa
2.2'.3.4'.5,5'-Hexach!orobiphenyl
No
BZ #149
Hexa
2.2'.3,4',5',6-Hexachlorobiphenyl
Yes
BZ #151
Hexa
2,2',3,5,5',6-Hexachlorobiphenyl
Yes
T A M S/Gradient
-------
TABLE A-l (continued)
BZ#15 3 Hexa 2,2',4,4',5,5'-Hexachlorobiphenyl Yes
BZ#156 Hexa 2,3,3',4,4',5-Hexachlorobiphenyl No - Cal
BZ#157 Hexa 2,3,3\4,4',5'-Hexachlorobiphenyl Yes
BZ#158 Hexa 2,3,3'A4',6-Hexachlorobiphenyl Yes
Hexa 2,3',4,4\5,5'-Hexachlorobiphenyl Yes
BZ#167 Hexa 3,3',4,4',5,5'-Hexachlorobiphenyl No
BZ#169 Hepta 2,2',3,3',4,4',5-Heptachlorobiphenyl Yes
BZ#170 Hepta 2,2',3,3',4,4',6-Heptachlorobiphenyl Yes
BZ #171
BZ#172 Hepta 2,2',3,3\4,5,5'-Heptachlorobiphenyl No
BZ#174 Hepta 2,2\3,3\4,5,6'-Heptachlorobiphenyl No-Cal
BZ#175 Hepta 2,2',3,3',4,5',6-Heptachlorobiphenyl No
BZ#177 Hepta 2,2',3,3',4',5,6-Heptachlorobiphenyl Yes
BZ#I78 Hepta 2,2',3,3',5,5',6-Heptachlorobiphenyl No-Cal
BZ#180 Hepta 2,2',3,4,4',5,5'-Heptachlorobiphenyl Yes
BZ#183 Hepta 2,2',3,4,4',5',6-Heptachlorobiphenyl Yes
BZ#184 Hepta 2,2',3,4,4\6,6'-Heptachlorobiphenyl No
BZ#185 Hepta 2,2',3,4,5,5',6-Heptachlorobiphenyl Yes
BZ#187 Hepta 2,2',3,4',5,5',6-Heptachlorobiphenyl Yes
BZ#189 Hepta 2.3,3',4,4',5,5'-Heptachlorobiphenyl Yes
BZ#190 Hepta 2,3,3',4,4',5,6-Heptachlorobiphenyl Yes
BZ#191 Hepta 2,3,3',4,4',5',6-Heptachlorobiphenyl Yes
BZ#193 Hepta 2,3,3',4',5,5',6-Heptachlorobiphenyl Yes
BZ#194 Octa 2,2\3,3\4,4',5,5'-Octachlorobiphenyl Yes
BZ#195 Octa 2,2',3,3',4,4',5,6-Octachlorobiphenyl Yes
BZ#196 Octa 2,2',3,3\4,4',5\6-Octachlorobiphenyl Yes
BZ#197 Octa 2,2',3,3',4,4',6,6'-Octachlorobiphenyl No
BZ#I98 Octa 2,2',3,3',4,5,5',6-Octachlorobiphenyl Yes
BZ#199 Octa 2,2\3,3',4,5,6,6'-Octachlorobiphenyl Yes
BZ #200 Octa 2,2',3,3\4,5',6,6'-Octachlorobiphenyl Yes
BZ#20I Octa 2,2',3,3',4',5,5',6-Octachlorobiphenyl Yes
BZ #202 Octa 2.2',3,3\5,5\6,6'-Octachlorobiphenyl Yes
BZ #203 Octa 2,2'.3,4,4\5,5',6-Octachlorobiphenyl No
BZ #205 Octa 2,3,3',4.4',5,5',6-Octachlorobiphenyl Yes
BZ #206 Nona 2,2',3,3',4,4',5,5',6-Nonachlorobiphenyl Yes
BZ #207 Nona 2,2',3,3',4,4\5,6,6'-Nonachlorobiphenyl Yes
BZ #208 Nona 2,2',3,3',4,5,5\6,6'-Nonachlorobiphenyl Yes
BZ #209 Deca 2,2'.3,3',4,4',5,5',6,6'-Decachlorobiphenyl Yes
T A M S/Gradient
-------
TABLE A-l (continued)
Homologue Group Congener Ratiob
Mono
3:3
Di
9:12
Tri
18:24
Tetra
23:42
Penta
23:46
Hexa
19:42
Hepta
16:24
Octa
11:12
Nona
3:3
Deca
1:1
Sum
126:209
Sotes Yes: Target: So: Son-target: So - Cal Calibrated non-target
hRatio of number of congeners used to total number of congeners in homologue group.
T A M S/Gradient
-------
Table A-2
Data Qualification Codes
Source of
Qualifier
Definition of Qualifier Code
Data Validation/ Database
Assessment Qualifier
Qualifier Code Code
Laboratory Compound not detected above reporting limit of 0.1 ppb in extract
for all PCB congeners (0.5 ppb in extract for the monochiorinated
biphenyls). The reported value is the quantitation limit (QL).
Laboratory Compound detected above reporting limit, but below calibration
range.
This qualifier is applied to any positive result that is less than the
lowest calibration standard. The reported result is an estimated
value, due to uncertainty in the reported value near the quantitation
limit.
Laboratory Compound concentration exceeds the calibration range.
This qualifier is applied to any positive result that exceeds the
calibration range. The laboratory may report some congeners with
concentrations up to twice the concentration in the highest
calibration standard, in order to report some very low concentrations
and low quantitation limits. The reported result is an estimated
value, due to uncertainty in the quantitation above the calibrated
range of the instrument.
Laboratory Specific column result used for quantitation due to confirmation
column coelution.
U
u
Laboratory
This qualifier designates congeners whose results are always
quantitated from a specific column due to coelution with congeners
or surrogates on the other column. The reported result should be
considered an estimated value, due to inability to confirm the
concentration of the result because of coelution on the other column.
The S qualifier precludes the P qualifier since a %Difference (%D)
between columns is excepted to be greater than 25% due to
coelution on one column.
Tentative identification, specific column result used with no
confirmation information.
JN
Laboratory
This qualifier designates congeners which could not be confirmed
due to an interferant (or surrogate) peak, however, there is good
reason to believe its presence. The reported value should be
considered an estimated value, due to inability to confirm reported
concentrations.
Estimated concentration due to coelution on both columns.
X
This qualifier designates congeners which coelute with congeners or
surrogates on both analytical columns. In order to report a
concentration for the congener of interest, the concentrations of the
coeluting congeners are subtracted from it. Therefore, the reported
result is an estimated value.
1AMS/Gradient
-------
Table A-2 (Continued)
Data Qualification Codes
Laboratory Confirmation column result exceeds reported result by more than
25%.
Laboratory
This qualifier is applied to a congener result if the concentration on
the quantitation and confirmation columns exceed the percent
difference (%D) criteria of 25. The reported result is an estimated
value, due to poor precision of results between columns.
Specific column or estimated result exceeds confirmation result by
more than 25% despite expected confirmation coelution.
Data
Validation
Data
Validation
Data
Validation
Data
Validation
Data
Validation
This qualifier is applied to a congener result if the result from the
quantitation column exceeds the confirmation result by more than
25 %D. even though the confirmation column result was expected
to be greater due to coelution on the confirmation column.
Therefore, the reported result should be considered an estimated
value, bias high.
Estimated data due to exceeded quality control criteria.
This qualifier is applied to data if problems with data quality are
noted and estimation of the data is deemed necessary. Justification
for qualification are given in the data validation report.
Reject data due to exceeded quality control criteria.
This qualifier is applied to data if serious problems with data quality
are noted and rejection of the data is deemed necessary.
Justification for rejection of data are given in the data validation
report. Rejected data are not usable and do not meet the data quality
objectives of the program. No numerical value is reported.
The compound was also detected in associated blank(s).
This qualifier is applied to GC/ECD results that are within five times
the concentration detected in the associated blanks. The reported
result may be considered not detected; a false positive is suspected
due to blank contamination.
GC/ECD result at concentration within GC/ITD calibration range,
but not confirmed by GC/ITD analysis.
This qualifier is applied to GC/ECD results that are not confirmed
by GC/ITD analysis, even though the results are at sufficient
concentration to be detected by GC/ITD. The reported result is
suspect as it may be a false positive.
Positive GC/ITD result was not detected by GC/ECD analysis or
greater than five times GC/ECD result.
JN
M
This qualifier is applied to GC/ECD results if the concentration of
the GC/ITD results are greater than five times the GC/ECD results.
Also the non-detected GC/F.CD result is qualified if a congener is
detected by GC/ITD and not detected by GC/ECD. The reported
result is suspect as it may be a false negative or a misidentification.
TAMS /Gradient
-------
Table A-2 (Continued)
Data Qualification Codes
Data Presumptive evidence for the presence of a material.
Validation
This qualifier is applied to GC/ECD results that exceeded the
compound identification criteria. The reported result is suspect as
it may be a false positive.
Data Results generated by decoupling BZ #4 and 10 using regression
Management analysis.
Data Results updated by Aquatec due to revisions in GC column
Management performance.
Data Results requalified by QAO due to decisions made during data
Management usability assessment.
L
K
Y
MAS! Gradient
-------
Table A-3
Low Resolution Sediment PCB Field Co-located Samples
Hudson River RI/FS PCB Reassessment
Sample Result and Duplicate Result and RPD
TAMS ID
BZ
Parameter
Units
Qualifier
Qualifier
(%)
LH-28C-00I5
1
BZ#1
ug/Kg DW
42400
37000
14
LH-28C-0015
4
BZ#4
ug/Kg DW
50800
J
42300
J
18
LH-28C-0015
8
BZ#8
ug/Kg DW
8630
J
6920
J
22
LH-28C-0015
10
BZ#10
ug/Kg DW
7520
J
6270
J
18
LH-28C-0015
18
BZ#18
ug/Kg DW
3010
J
2550
17
LH-28C-0015
19
BZ#19
ug/Kg DW
9120
J
7290
J
22
LH-28C-00I5
28
BZ#28
ug/Kg DW
1440
1080
29
LH-28C-0015
52
BZ#52
ug/Kg DW
3350
2630
24
LH-28C-00I5
101
BZ# 101 with BZ#[90]
ug/Kg DW
428
J
367
J
15
LH-28C-0015
118
BZ#118
ug/Kg DW
158
JN
148
J
7
LH-28C-0015
138
BZ#138
ug/Kg DW
35.9
JN
204
J
-140
LH-28C-0015
180
BZ#180
ug/Kg DW
352
U
56.7
J
145
LH-28C-1530
1
BZ#1
ug/Kg DW
120000
85500
34
LH-28C-1530
4
BZ#4
ug/Kg DW
170000
J
123000
J
32
LH-28C-1530
8
BZ#8
ug/Kg DW
85100
J
69400
J
20
LH-28C-I530
10
BZ#10
ug/Kg DW
16900
J
11100
J
41
LH-28C-I530
18
BZ#18
ug/Kg DW
16300
12800
24
LH-28C-I530
19
BZ#19
ug/Kg DW
28400
J
19600
J
37
'-28C-1530
28
BZ#28
ug/Kg DW
8810
8990
-2
""EH-28C-1530
52
BZ#52
ug/Kg DW
14000
11100
23
LH-28C-1530
101
BZ# 101 with BZ#[90]
ug/Kg DW
1040
J
721
J
36
LH-28C-1530
118
BZ#118
ug/Kg DW
1280
U
728
U
NC
LH-28C-1530
138
BZ#138
ug/Kg DW
829
J
671
J
21
LH-28C-1530
180
BZ#180
ug/Kg DW
270
J
189
J
35
LH-28C-3046
1
BZ#1
ug/Kg DW
9890
J
8540
15
LH-28C-3046
4
BZ#4
ug/Kg DW
31800
J
27000
J
16
LH-28C-3046
8
BZ#8
ug/Kg DW
41400
J
36300
J
13
LH-28C-3046
10
BZ# 10
ug/Kg DW
585
J
482
J
19
LH-28C-3046
18
BZ#18
ug/Kg DW
18900
J
19100
-1
LH-28C-3046
19
BZ#I9
ug/Kg DW
5600
J
4760
J
16
LH-28C-3046
28
BZ#28
ug/Kg DW
10300
J
10500
-2
LH-28C-3046
52
BZ#52
ug/Kg DW
13500
J
14200
-5
LH-28C-3046
101
BZ#101 with BZ#{90]
ug/Kg DW
259
J
182
J
35
LH-28C-3046
118
BZ#118
ug/Kg DW
149
J
114
J
27
LH-28C-3046
138
BZ#138
uglCg DW
730
J
723
J
1
LH-28C-3046
180
BZ#180
ug/Kg DW
90.7
J
88
J
3
LH-39M-0008
1
BZ#1
ug/Kg DW
5680
J
4490
23
LH-39M-0008
4
BZ#4
ug/Kg DW
7530
J
7210
J
4
LH-39M-0008
8
BZ#8
ug/Kg DW
4250
J
3900
J
9
LH-39M-0008
10
BZ#10
ug/Kg DW
794
J
776
J
2
LH-39M-0008
18
BZ#18
ug/Kg DW
1120
J
1030
J
8
' M-39M-0008
19
BZ#19
ug/Kg DW
1430
J
1300
J
10
s_,-39M-0008
28
BZ#28
ug/Kg DW
646
596
8
LH-39M-0008
52
BZ#52
ug/Kg DW
1070
1030
4
x :\hudson\lowresre\appendix\ALRSTATS XLS
A-3 5
X AWStGradienl
-------
Table A-3
Low Resolution Sediment PCB Field Co-located Samples
Hudson River RI/FS PCB Reassessment
Sample Result and Duplicate Result and RPD
TAMS ID
BZ
Parameter
Units
Qualifier
Qualifier
(%)
LH-39M-0008
101
BZ#101 with BZ#[90J
ug/Kg DW
87.8 UJ
109 J
-22
LH-39M-0008
118
BZ#118
ug/Kg DW
54.9 J
71.2 J
-26
LH-39M-0008
138
BZ#138
ug'Kg DW
26.6 JN
41.7 J
-44
LH-39M-0008
180
BZ#180
ug/Kg DW
119 U
25 J
131
x:\hudson\lowresreVappendix\Al.RSTATS XI.S
A-36
JAMS/Gradient
-------
Table A-4
PCB Detects Changed to Non-detects
Low Resolution Sediment Samples
Hudson River RI/FS PCB Reassessment
Congener Name
Number of results
considered nondetect*
Total number of
results
Percentage of results
considered nondetect
BZ#I
54
371
15
BZ#2
12
371
3
BZ#3
122
371
33
BZ#4
37
371
10
BZ#6
214
371
58
BZ#7
33
371
9
BZ#8
30
371
8
BZ#9
27
371
7
BZ#10
111
371
30
BZ#12
26
371
7
BZ#15
46
371
12
BZ#16
114
371
31
BZ#17
57
371
15
BZ#18
53
371
14
BZ#19
32
371
9
BZ#20
203
371
55
BZ#22
27
371
7
BZ#23NT
32
371
9
BZK25
48
371
13
BZ#26
67
371
18
BZ#27
34
371
9
BZ#28
39
371
11
BZWI
22
371
6
BZ#32NT
11
371
3
BZ*33
75
371
20
BZ#37
46
371
12
BZ«40
21
371
6
BZ#41
36
371
10
BZ#42
87
371
23
BZ«44
71
371
19
BZ#45
15
371
4
BZ#47
37
371
10
BZ#49
172
371
46
BZ#52
34
371
9
BZ#53
59
371
16
BZ#56
35
371
9
BZU66
57
371
15
Bzm
39
371
11
BZ#74
24
371
6
BZ#75
199
371
54
BZ#77
105
371
28
B7M2
1
371
0
BZttm
10
371
3
BZ«84
15
371
4
x:\hudson\lowresre\appcndixALRSTATS XLS
Page A-37
T AM S / Gradient
-------
Table A-4
PCB Detects Changed to Non-detects
Low Resolution Sediment Samples
Hudson River Rl/FS PCB Reassessment
Number of results
Total number of
Percentage c
Congener Name
considered nondetect*
results
considered n
BZ#85
119
371
32
BZ#87
87
371
23
BZ#91
8
371
2
BZ#92
13
371
4
BZ#95
38
371
10
BZ#97
62
371
17
BZ#99
11
371
3
BZ#10I with BZ#[90]
12
371
3
BZ#105
123
371
33
BZ#107
20
371
5
BZ#110
61
371
16
BZ#115
92
371
25
BZ#118
58
371
16
BZ#119
104
371
28
BZ#122
5
371
1
BZ#123
19
371
5
BZ#126
30
371
8
BZ#128
108
371
29
BZ#129
34
371
9
BZ#135
26
371
7
BZ#136
5
371
1
BZ#137
9
371
2
BZ#138
12
371
3
BZ#I41
59
371
16
BZ#143
18
371
5
BZ#149
38
371
10
BZ»# 151
18
371
5
BZ#I53
53
371
14
BZ*156
44
371
12
BZ#157
51
371
14
BZ#158
1
371
0
BZ#167
9
371
2
BZ#170
87
371
23
BZ#171
17
371
5
BZ#174
40
371
11
BZ#177
6
371
2
BZ#178
31
371
8
o
00
N
33
371
9
BZ#I83
80
371
22
BZ#I85
18
371
5
BZ*187
53
371
14
BZ#190
57
371
15
BZ#194
126
371
34
BZ#195
35
371
9
udson\lowresre\appendixAl.RSTATS XLS
Page A-38
TAMS/Gradient
-------
Table A-4
PCB Detects Changed to Non-detects
Low Resolution Sediment Samples
Hudson River RI/FS PCB Reassessment
Number of results
Total number of
Percentage i
Congener Name
considered nondetect*
results
considered i
BZ#196
53
371
14
BZ#198
145
371
39
BZ#199
14
371
4
BZ#200
43
371
12
BZ#201
67
371
18
BZ#202
24
371
6
BZ#205
24
371
6
BZ#206
98
371
26
BZ#207
4
371
1
BZ#208
10
371
3
BZ#209
14
371
4
[Not specified by Gradient]
x:\hudson\lowresre\appendixALRSTATS XLS
Page A-39
TAMSIGradient
-------
Table A-5
Low Resolution Coring Sample PCB Analysis Summary
Hudson River RI/FS PCB Reassessment
Convener Name' To,al Number Unqualified Estimated Unqualified Estimated Rejected „ .
Congener Name of Results Nondetects Nondetects Detects Detects Presumed Present /.Rejected
Detects
BZttl
371
32
24
205
87
13
10
3%
BZ#2
371
235
84
0
0
0
52
14%
BZ#3
37 f
90
193
0
0
1
87
23%
BZ#4
371
1
38
0
332
0
0
0%
BZ#5
37!
280
81
0
0
0
10
3%
BZ#6
371
136
88
102
39
0
6
2%
BZ#7
371
114
62
1
185
0
9
2%
BZ#8
371
0
52
0
319
0
0
0%
B7J9
371
13
28
0
330
0
0
0%
BZ#I0
371
2
112
0
252
5
0
0%
BZ#I2
371
204
69
4
9
15
70
19%
BZ# 15
371
2
49
0
320
0
0
0%
BZ#I6
371
89
140
0
118
0
24
6%
BZ#I7
371
3
59
0
309
0
0
0%
BZ#18
371
24
46
159
83
59
0
0%
BZ#I9
371
0
32
0
327
12
0
0%
BZ#20
371
40
217
0
107
0
7
2%
B7J22
371
9
28
198
126
7
3
1%
BZ023NT
371
94
0
0
277
0
0
0%
BZ024NT
371
79
0
0
292
0
0
0%
BZ#25
371
10
40
206
no
3
2
1%
BZ»26
371
2
67
0
301
1
0
0%
BZH27
371
0
34
0
337
0
0
0%
Bzm
371
13
26
217
114
0
1
0%
azm
371
292
68
0
0
0
11
3%
BZM 1
371
0
22
0
348
1
0
0%
BZ#32NT
371
12
0
0
8
351
0
0%
BZ#33
371
18
82
0
267
0
4
1%
BZM4NT
371
71
0
0
238
62
0
0%
BZ«37
371
9
48
0
314
0
0
0%
v:\hud •;onM(mrcsrcUippcm!i\\Al,RST ATS
XI .s
Page 1 of 5
TAMS/Gruifoii!
-------
Table A-5
Low Resolution Coring Sample PCB Analysis Summary
Hudson River RI/FS PCB Reassessment
Congener Name1
Total Number
of Results
Unqualified
Nondetects
Estimated
Nondetects
Unqualified
Detects
Estimated
Detects
Estimated and
Presumed Present
Detects
Rejected
Results
% Rejected
BZ#40
371
117
53
0
170
0
31
8%
BZ#41
371
41
50
0
265
0
15
4%
BZ#42
371
17
90
0
263
0
1
0%
BZ#44
371
29
59
164
102
6
U
3%
BZ#45
371
11
14
189
149
6
2
1%
BZ#47
371
2
37
0
332
0
0
0%
BZ#48NT
371
160
0
0
211
0
0
0%
BZ#49
371
2
172
0
188
9
0
0%
BZ#51 NT
371
12
0
0
106
253
0
0%
BZ#52
371
24
10
240
96
1
0
0%
BZ#53
371
9
63
0
298
0
1
0%
BZ056
371
31
45
0
285
0
10
3%
BZ058NT
371
365
0
0
3
3
0
0%
BZS60NT
371
104
0
0
258
9
0
0%
BZ#63NT
37!
62
0
0
180
129
0
0%
BZ#64NT
371
5
0
0
67
299
0
0%
BZ#66
371
8
58
0
303
0
2
1%
BZ#67NT
371
196
0
0
135
40
0
0%
BZ#69NT
371
360
0
0
11
0
0
0%
BZ#70
371
13
43
179
124
3
9
2%
BZ#74
371
21
27
157
151
6
9
2%
BZ#75
371
38
206
0
126
0
1
0%
BZ#77
371
29
112
0
220
6
4
1%
BZ#82
37!
116
26
28
159
23
19
5%
BZ#83
37!
96
34
17
182
17
25
7%
BZ#84
371
16
23
45
277
7
3
1%
B7J85
371
56
132
0
178
0
5
1%
BZ#87
371
56
97
0
207
0
11
3%
BZ#9I
371
16
12
40
299
2
2
1%
BZ#92
371
16
17
144
184
5
5
1%
x:\hudson\lowresrc\appcndix\Al
.RSTATS.XI.S
Page 2 of 5
T AMS/Gradienl
-------
Table A-5
Low Resolution Coring Sample PCB Analysis Summary
Hudson River Rl/FS PCB Reassessment
Congener Name1
Total Number
of Results
Unqualified
Nondetects
Estimated
Nondetects
Unqualified
Detects
Estimated
Detects
Estimated and
Presumed Present
Detects
Rejected
Results
% Rejected
BZ#95
371
14
40
0
317
0
0
0%
BZ#96NT
371
208
0
0
157
6
0
0%
BZ#97
371
100
69
63
108
4
27
7%
BZ#99
371
31
17
122
144
36
21
6%
BZ#I01 with BZ#(90]
371
12
14
0
345
0
0
0%
BZ#I05
371
49
136
0
176
0
10
3%
BZ#107
371
137
45
0
169
0
20
5%
BZ#110
371
4
63
0
304
0
0
0%
B7JI14NT
371
252
0
0
110
9
0
0%
BZ#115
371
174
131
0
52
0
14
4%
BZ#118
371
30
66
0
263
3
9
2%
BZ#1I9
371
155
153
0
37
4
22
6%
BZ#I22
371
284
75
0
1
0
11
3%
BZ#123
371
227
72
0
56
0
16
4%
BZ#126
371
245
81
0
31
0
14
4%
BZ# 128
371
100
124
10
115
6
16
4%
BZ#129
371
214
85
0
63
0
9
2%
BZ#135
371
57
42
0
263
0
9
2%
BZ#I36
371
90
48
2
214
1
16
4%
BZ#I37
371
213
49
0
37
20
52
14%
BZ#I38
371
28
18
1
259
54
11
3%
BZ#I40NT
371
362
0
0
9
0
0
0%
B7J141
371
154
92
0
116
0
9
2%
BZ#143
371
267
77
0
6
0
21
6%
B7J144NT
371
326
0
0
42
3
0
0%
BZ# 146NT
371
120
0
0
184
67
0
0%
BZ«149
371
40
49
0
273
0
9
2%
BZ# 151
371
43
33
0
289
0
6
2%
BZ#153
37!
33
64
0
268
0
6
2%
BZ#156
371
147
75
0
129
10
10
3%
x \hudson>Utwresre .appendix
A! KM A I S XI S
I'AMS,'Gradient
l ¦' I'l -
/
-------
Table A-5
Low Resolution Coring Sample PCB Analysis Summary
Hudson River RI/FS PCB Reassessment
Congener Name1
Total Number
of Results
Unqualified
Nondetects
Estimated
Nondetects
Unqualified
Detects
Estimated
Detects
Estimated and
Presumed Present ,fJ ,
Detects Resu,,s
% Rejected
BZ#I57
371
240
110
0
8
0
13
4%
BZ#158
37!
174
40
0
146
0
11
3%
BZ#167
37!
231
59
0
65
3
13
4%
BZ# 169NT
371
369
0
0
2
0
0
0%
BZ#I70
371
93
102
0
170
0
6
2%
BZ#171
371
247
72
0
43
0
9
2%
BZ#!72NT
371
316
0
0
0
55
0
0%
BZ#174
371
159
74
0
125
0
13
4%
BZ#175NT
371
367
0
0
3
1
0
0%
BZ#177
371
126
35
3
183
7
17
5%
BZ#178
371
108
59
0
194
0
10
3%
BZ#180
371
78
46
44
144
27
32
9%
BZ#183
371
168
125
0
66
0
12
3%
BZ#184NT
371
210
0
0
146
15
0
0%
BZ#I85
371
250
80
0
31
0
10
3%
BZ#187
371
56
68
45
171
13
18
5%
BZ#189
371
289
67
0
1
0
14
4%
BZ#190
371
173
94
0
95
0
9
2%
BZ# 191
371
292
64
0
2
2
11
3%
BZ#193
371
291
69
0
0
0
11
3%
BZ#I94
371
139
162
2
24
8
36
10%
BZ#!95
371
228
86
0
35
2
20
5%
BZ#196
37)
174
93
0
94
0
10
3%
BZ# 197NT
371
371
0
0
0
0
0
0%
BZ# 198
371
170
190
0
1
0
10
3%
BZ#I99
371
276
72
0
12
0
11
3%
BZ#200
371
248
97
0
16
0
10
3%
BZ#201
371
147
98
0
116
0
10
3%
BZ#202
371
246
76
0
36
0
13
4%
BZ#203NT
371
208
0
0
146
17
0
0%
x;\hudson\Iowresrc\appcndixlAl.
.RSTATS Xl.S
Page 4 of 5
TA M SI Gradient
-------
Table A-5
Low Resolution Coring Sample PCB Analysis Summary
Hudson River RI/FS PCB Reassessment
Congener Name'
Total Number
of Results
Unqualified
Nondetects
Estimated
Nondetects
Unqualified
Detects
Estimated
Detects
Estimated and
Presumed Present
Detects
Rejected
Results
% Rejected
BZ8205
371
260
93
0
0
0
18
5%
BZ#206
371
152
129
9
39
15
27
7%
BZ#207
371
279
71
0
2
6
13
4%
BZ#208
371
238
63
2
42
2
24
6%
BZ#209
371
260
71
2
16
5
17
5%
TOTALS
46,375
15,651
7,352
2,600
17,789
1,755
1,228
2.6%
Notes:
I. NT in the congener name stands for non-target indicating a congener added to the program in addition to the original target 90 congeners.
See text for discussion.
\ \hudsy" wicsft'wippemlKlAI.RS I'AIS XI.S
Page 5 of5
TAMS/(/VW/cv«
-------
Radionuclides
(?Be + 137Cs)
Grain size (ASTM)
Notes:
1. TOC/TKN sample frequency at 7%.
2. Grain size distribution analysis
based on the ASTM technique were
performed at least once per core for
approximately 68% of the cores
collected.
3. The segment thicknesses shown are
median values for four segment cores.
Sediment/water
interface
PCBs, bulk density,
%-solids, grain size
distribution (laser),
TOC1, TKN1
PCBs, bulk density,
%-solids, grain size
distribution (ASTM)2,
TOC1, TKN1
PCBs, bulk density,
%-solids, grain size
distribution (ASTM)2,
TOC1, TKN1
3 in. Radionuclides ('^Cs)
Not Used
Legend:
TOC - Total Organic Carbon Analysis
TKN - Total Kjeldahl Nitrogen Analysis
¦Source TAMS/Gradient Database, Release 3 5
TAMS
Figure A-l
Low Resolution Sediment Core Preparation
-------
TAMS
Appendix B
-------
APPENDIX B
Data Usability Report for Non-pcb Chemical
And
Physical Data Low Resolution Sediment Coring Study
-------
Table of Contents
Page
B.l Introduction B-l
B.2 Data Usability Approach B-l
B.3 Grain-Size Distribution Data B-3
B.3.1 Sieve/Hydrometer Grain-Size Distribution Data B-4
B.3.1.1 Accuracy B-4
B.3.1.2 Precision B-5
B.3.1.3 Sensitivity B-5
B.3.1.4 Representativeness B-6
B.3.1.5 Summary of Data Usability B-6
B.3.2 Laser Grain-Size Distribution Data B-6
B.3.2.1 Accuracy B-6
B.3.2.2 Precision B-7
B.3.2.3 Sensitivity B-7
B.3.2.4 Representativeness B-7
B.3.2.5 Summary of Usability B-7
B.3.3 Overall Grain-Size Usability B-7
B.4 Total Kjeldahl Nitrogen (TKN) Data B-10
B.4.1 Accuracy B-10
B.4.2 Precision B-l 1
B.4.3 Sensitivity B-l 1
B.4.4 Representativeness B-l 1
B.4.5 Summary of Usability B-l 2
B.5 Total Organic Carbon (TOC) Data B-l2
B.5.1 Accuracy B-l 2
B.5.2 Precision B-12
B.5.3 Sensitivity B-l3
B.5.4 Representativeness B-13
B.5.5 Summary of Data Usability B-13
B.6 Radionuclide Analyses B-l4
B.6.1 Accuracy B-l 5
B.6.2 Precision B-l6
B.6.3 Sensitivity B-l6
B.6.4 Representativeness B-l7
B.6.5 Summary of Data Usability Assessment B-l 7
References B-l8
TABLES
B-l Low-Resolution Sediment Non-PCB Sample Analysis Summary
B-2 Low-Resolution Sediment Sieve Grain-Size Sample Analysis Summary
B-3 Low-Resolution Sediment Radionuclide Sample Analysis Summary
-------
FIGURES
B-1 Classification of Shallow Sediment Samples - Comparison of Visual Inspection
and Laser Grain-Size Analytical Technique
B-2 Classification of Sediment Samples - Comparison of Visual Inspection and
ASTM Grain-Size Analytical Techniques
B-3 Classification of Sediment Samples - Comparison of Grain-Size Analytical
Techniques (ASTM and Laser Methods)
-------
B.l Introduction
The usability of data is highly dependent on the data quality objectives (DQOs) defined for
an environmental investigation. Throughout its duration, the Hudson River PCB congener chemistry
program has required stringent quality control criteria to maintain data usability for all of the
analytical parameters performed in support of the project. For the Phase 2B low resolution sediment
coring study, various non-PCB chemical and physical parameters were analyzed to aid in defining
the context within which the PCB congeners exist. These parameters helped to delineate the
concentration of the PCB congeners within the context of geochemical and biological processes
occurring in the Hudson River.
This report serves as an overall evaluation of the data usability for the Hudson River Phase
2B Low Resolution Sediment Coring Study non-PCB analyses based upon criteria set forth by
TAMS/Gradient. The low resolution field sampling program, analytical protocols, and quality
control/quality assurance requirements are described in Appendix A. The data usability reports
assessing the PCB congeners for the low resolution sediment coring study are also provided in
Appendix A.
B.2 Data Usability Approach
Data validation of the non-PCB parameters was performed by CDM based upon the specific
method criteria listed in the Appendices of the "Phase 2B Sampling and Analysis Plan/Quality
Assurance Project Plan, Volume 4: Low Resolution Sediment Coring, Hudson River PCB
Reassessment Rl/FS" (TAMS/Gradient, 1992a, referred to hereafter in this report as the Phase 2B
SAP/QAPP), and the USEPA Region II validation guidelines (USEPA, 1992a), where applicable.
The non-PCB chemical and physical data for the low resolution sampling program included grain-
size (particle size) distribution, total organic carbon (TOC), total kjeldahl nitrogen (TKN), and
radionuclide (l37Cs and 7Be) analyses.
TAMS/Gradient determined the usability of the data based upon an evaluation of the data
validation reports in conjunction with historical or expected results and the program dataquality
objectives (DQOs) as defined in the Phase 2B SAP/QAPP for the low resolution sediment coring
B-i
TAMS/Gradient
-------
study. Additionally, TAMS/Gradient based the usability evaluation upon the intended use(s) of the
data, consistency with other data sets (both internal, i.e., from the Hudson River PCB Reassessment
RI/FS and external, i.e., historical data or data gathered from the literature), and professional
judgment.
Criteria used, in part, to evaluate usability include accuracy, precision, representativeness,
sensitivity, and completeness. Accuracy is a measure of how a result compares to a true value.
Precision indicates the reproducibility of generating a value. Representativeness is the degree to
which a measurement(s) is indicative of the characteristics of a larger population. Sensitivity is
represented by the limit of detection of the analytical method. Completeness is a measure of the
amount of usable data resulting from a data collection activity (USEPA, 1992b). For this program,
a 95% completeness goal was established. These criteria are discussed in detail in Appendix A as
well as the Phase 2B SAP/QAPP.
Accuracy was evaluated for TOC, TKN, and radionuclide analyses through the assessment
of quality control samples, including initial and continuing calibration verification (ICV and CCV,
respectively), laboratory control samples (LCS), and/or matrix spikes. Precision was evaluated for
grain-size analyses, TOC and TKN through the assessment of laboratory duplicate analyses.
Sensitivity was evaluated for all parameters based upon the assessment of blanks and/or detection
levels. Representativeness was evaluated for grain-size, TOC and TKN analyses through the
assessment of field duplicate results.
During the usability assessment, the final qualifications of the data presented in the Hudson
River low resolution sampling project database were determined. In most cases, TAMS/Gradient
maintained the qualifications added during validation and interpreted these qualifications in terms
of usability of the results relative to project objectives. In cases where the qualification of the data
was changed from the validation actions, details of the technical justification for these changes, and
the resultant usability of the data, are presented in this appendix for all non-PCB results.
An essential aspect of understanding the uncertainties of the Phase 2B chemical and physical
data is understanding the qualifiers associated with the results. Initially, the analytical laboratories
applied qualifiers to the results. The data validators then modified these qualifiers, as necessary,
B-2
TAMS/Gradient
-------
using established validation protocols from the USEPA Region II standard operating procedure
(SOP) for data validation (USEPA, 1992a), where applicable, the specific DQOs and quality control
(QC) criteria established for the non-PCB analyses in the SAP/QAPP, and professional judgment.
The data were validated using protocols established by TAMS/Gradient and all data validation was
performed by CDM. The validation qualifiers were further modified during the usability assessment
to direct the data users concerning the use of each result, if required. Specifically, the data were
evaluated in accordance to the Special Analytical Services (SAS) request and the Phase 2B
SAP/QAPP, adherence to technical specifications of the analytical method, and achievement of
precision and accuracy objectives. The definition of the final qualification flags that appear in the
database for non-PCB results are based upon USEPA data validation guidance (USEPA, 1992a) and
are listed below:
Qualifiers for Non-PCB Data
U The chemical or parameter was analyzed for, but was not detected above the level of the associated value. The
associated value is the sample quantitation limit. The associated value is usable as a nondetect at the reported
detection level.
J The associated value is an estimated quantity due to QA/QC exceedance(s). The estimated value may be
inaccurate or imprecise. The associated value is an estimated result.
UJ The chemical or parameter was analyzed for, but was not detected above the level of the associated value. The
associated value is an estimated sample quantitation limit and may be inaccurate or imprecise. The value is
usable as a nondetect value with an estimated detection limit.
R The value (result) was rejected due to significant errors or QA/QC exceedance(s). The result is not usable
for project objectives.
A complete list of result qualifiers for both the PCB and non-PCB data can be found in the "Qualify
Table" of the project database. Table B-l presents a summary of data usability statistics for laser
grain-size, TKN, and TOC analyses. Tables B-2 and B-3 present summary statistics for the sieve
grain-size and radionuclide analyses, respectively.
B.3 Grain-Size Distribution Data
Grain-size distribution was determined for all low resolution sediment core sections to
classify the type of sediment collected. Grain-size results are used for interpreting sediment PCB
chronologies and degradation, particularly where important geochemical features correspond to
B-3
TA MS/Gradient
-------
changes in sediment texture. Due to the limited sample sizes for the low resolution top sediments
and the need to classify the entire grain-size distribution on the same basis, a laser particle technique
was used to measure grain-size in the top core slices. These cores were also analyzed by a sieve
and hydrometer method (hereafter, sieve/hydrometer), in addition to the field (visual) classification.
Grain-size distribution for the top sediment core slices was determined mathematically by
combining the laser method and sieve/hydrometer method results. Additionally, the remaining low
resolution sediment core slices, with the exception of the bottom slices, were measured using
standard sieve/hydrometer methodologies for grain-size distribution. Low resolution sediment core
slices were collected and analyzed for grain-size distribution by Midwest Laboratories, Inc. (150
samples, including seven field duplicates) using a sieve and hydrometer method (ASTM Methods
D-421 and D-422) and by GeoSea Consulting, Ltd. (179 samples, including nine field duplicates)
using a combined sieving method (ASTM D421-85 equivalent, to remove the particles greater than
1 mm) and laser methodology (for the particle size distribution under 2 mm). Data were validated
by CDM and were subsequently evaluated for usability by the TAMS/Gradient team. QC samples
(field duplicates) were collected and analyzed for grain-size distribution at a frequency of greater
than or equal to 5%. The interpretation of the QC results and the accuracy and representativeness
of the grain-size data are evaluated in this section.
B.3.1 Sieve/Hydrometer Grain-Size Distribution Data
B.3.1.1 Accuracy
At the commencement of the low resolution core study, sample bins were incorrectly labeled
by Midwest Laboratories. In order to have reporting bins which were consistent with previous
sampling rounds and so that the laser grain-size analyses results would be comparable, the bins
were re-labeled under the direction of TAMS. Data quality was unaffected by the re-labeling of the
bins.
Accuracy was compromised for the sieve/hydrometer results due to inappropriate method
procedures. The method requires that after hydrometer analysis, the sample soil suspension must
be transferred to a No. 200 (75 |im) sieve. The material remaining on the sieve is then dried and
sifted through the remaining sieves. Instead of transferring the suspension to the appropriate sieve,
B-4
TAMS/Gradient
-------
the laboratory dried the sample prior to hydrometer analysis, destroying the true sand/silt split. This
changed the natural distribution of the soil sample for all intervals below 75 jim. As a result of this
method deviation, the grain-size data from the less than 75 ^m fraction is not accurate. Therefore,
all of the low resolution sample sieve/hydrometer data from the less than 75 |im fraction were
considered estimated (qualified J) due to lack of differentiation between the sand and silt fractions.
The results are usable as estimated values for which uncertainty exists.
During validation, all sieve/hydrometer grain-size results were qualified as estimated ("J")
because a number of samples were not analyzed within the 35 day Verified Time of Sample Receipt
(VTSR) limit. In addition, the validator chose to qualify ("J") all results because the 2.8 mm
fraction was not analyzed by the laboratory. Since the 35 VTSR holding time criterion was
established solely for project management reasons, exceedance of this holding time criterion did
not affect overall data quality or compromise comparability of the data to previous sampling events.
In addition, the lack of the 2.8 mm fraction analysis is not critical because this fraction was
bracketed by other analytical intervals. During data usability assessment, the TAMS/Gradient team
reversed the validator's decision to qualify as estimated (i.e., the "J" qualifier was removed) the
sieve/hydrometer grain-size results because overall data quality and accuracy were not effected by
either of these issues.
B.3.1.2 Precision
Eight laboratory duplicate pairs were analyzed for sieve/hydrometer grain-size, exceeding
the 5% minimum frequency stipulated in the Phase 2B SAP/QAPP. Overall precision of the
sieve/hydrometer data was acceptable based upon results for the eight laboratory duplicates.
Duplicate precison was assessed by a percent similarity criterion developed specifically for
evaluating grain-size data (Shilabeer et al, 1992), with a percent similarity precision objective of
80% or greater established in the Phase 2B SAP/QAPP. All laboratory duplicate analyses achieved
a percent similarity of > 80%.
B.3.1.3 Sensitivity
There were no issues affecting sensitivity of the grain-size analyses.
B-5
T AMS/Gradient
-------
B.3.1.4 Representativeness
Seven field duplicates pairs were analyzed in association with the 143 sieve/hydrometer
grain-size samples,a frequency of 4.9%, slightly less than the 5% frequency stipulated in the Phase
2B SAP/QAPP. Overall precision of the sieve/hydrometer data was acceptable based upon results
for the seven field duplicate pairs, as all duplicate analyses achieved a percent similarity of > 80%.
Based upon the method deviation performed by the laboratory (described in section B.3.1.1),
data users are cautioned that the grain-size distribution for the less than 75 (am fraction does not
represent the true sand and silt split.
B.3.1.5 Summary of Data Usability
All Midwest Laboratories, Inc. sieve/hydrometer data are usable for general geochemical
classifications and ratios of fractions. A total of 13% of the results were qualified as estimated (J)
due to uncertainty in the <75 |am fraction. The completeness for these data was 100%. The
summary statistics for these analyses are presented in Table B-2.
B.3.2 Laser Grain-Size Distribution Data
B.3.2.1 Accuracy
During data validation, laser/sieve results for 64 of the 179 samples were qualified as
estimated ("J") because the samples were not analyzed within the 35 VTSR holding time criterion.
The validator also estimated these data because the laboratory did not analyze particle size intervals
2.25 mm, 3.75 mm, and 7.75 mm. As with the sieve/hydrometer analyses, the 35 VTSR criterion
was established solely for project management reasons. Thus, holding time exceedances do not
affect the quality of the grain-size distribution results. In addition, the lack of the three particle size
intervals does not impact the overall quality of the data or the comparability of the laser/sieve data
to previous sampling events because these intervals were bracketed by the other sieve sizes
analyzed. The TAMS/Gradient team reversed the qualification of the data during the data usability
assessment. Therefore, there were no issues affecting the accuracy of the laser/sieve results.
B-6
T A M S/G" r ad tent
-------
B.3.2.2 Precision
Ten laboratory duplicate pairs were analyzed in association with the laser/sieve grain-size
samples. This exceeded the 5% frequency required by the Phase 2B SAP/QAPP. Overall precision
of the sieve/hydrometer data was acceptable based upon results for the ten laboratory duplicates (all
duplicate analyses achieved a percent similarity of > 80%).
B.3.2.3 Sensitivity
There were no issues affecting sensitivity of the laser/hydrometer grain-size analyses.
B.3.2.4 Representativeness
Overall precision of the laser/sieve data was acceptable based upon results for nine field
duplicate pairs (all duplicate analyses achieved a percent similarity of > 80%). Field duplicates
were analyzed at the required frequency of 5%.
B.3.2.5 Summary of Usability
All of the low resolution sample laser/sieve data are considered acceptable without
qualification. The GeoSea Consulting LTD (Canada) laser/sieve data are usable without
qualification for general geotechnical classifications and rations of fractions. Completeness of
100% was achieved for these analyses. Summary statistics for these analyses are presented in Table
B-l.
B.3.3 Overall Grain-Size Usability
In addition to the field classification, low resolution sediments were classified by two
laboratory techniques discussed above:
• combined sieve and laser particle analysis (Laser); and
• combined sieve and hydrometer analysis (ASTM).
B-7
T kWS>! Gradient
-------
Results from these techniques are summarized in Tables B-l and B-2. Both Laser and ASTM
techniques were applied to a large subset of the samples collected. Visual field inspections were
performed for every sediment sample.
Evident in all three data sets is the predominance of samples classified as silt (fines in the
case of the ASTM results). The predominance of this fraction reflects the orientation of the
sampling program, i.e., to obtain cores from areas of substantive PCB contamination, generally
areas of fine-grained sediments. In general, the three methods yield similar results for most
samples. The results of these methods are compared by principal fraction in Figures B-l to B-3.
In Figure B-l the results of the visual and Laser classifications are compared for the
shallow sediments only, {i.e., just the top slice of each of 169 cores). The uppermost diagram
shows the coincidence between principle fraction by visual inspection versus that obtained by the
Laser technique. The two lower diagrams represent the distribution of matched samples as
classified by each method. In most instances, the two methods agree on the principal fraction
for samples classified as silt and fine sand, effectively verifying the subjective visual
classification. When the two methods disagree, it is usually by only one class (i.e., fine sand by
visual inspection is assigned silt by the Laser technique). In most of these instances, the actual
fractions are very close (e.g., 35% silt and 32% fine sand). The coarser materials, i.e., medium
or coarse sand and gravel, were not as constant as silt and fine sand for the two methods. In
particular, the medium sand as classified by visual inspection could be found in every class by
the Laser method. This is indicative of the poor sorting of the coarse sediments, which made
visual classification more difficult.
In Figure B-2, the visual inspection results are compared with the ASTM method for
samples (n = 143) from a range of depths and locations, as opposed to the shallow sediment
samples presented in Figure B-l. Again, the two methods generally agree for silt and fine sand;
however, the coarser fractions are more problematic. As discussed above, this is attributed to
the poorly sorted nature of the sample materials.
Figure B-3 compares the results for the Laser and ASTM methods directly for the 69
shallow sediment samples run by both methods. The top diagram shows the agreement of the
B-8
TAMS>I Gradient
-------
principal fractions between the two methods. Although the methods agree for most fines, the
Laser method characterizes more samples as silt than does the ASTM method. This trend is
apparent for all sediment classes, with the Laser method tending to characterize more samples
into smaller fractions than the ASTM method. The lower half of Figure B-3 is a histogram of
the percent similarity calculated for each Laser-ASTM measurement pair. Percent similarity is
calculated by summing the smallest value in each of the sediment classes for a pair of
measurements as shown below:
Sediment and Class Fraction
Silt
Fine
Medium
Coarse
Gravel
Sand
Sand
Sand
Laser Analysis
45
28
12
15
0
= 100%
of Sample 1
ASTM Analysis
35
32
18
12
3
= 100%
of Sample 1
Similarity (%)
35
28
12
12
0
= 87%
Similarity
The range of percent similarity for this data set is 34% to 98% with a mean value of 76%. This is quite
similar to the work of Shillabeer, et al, 1992, where a set of 406 sediment sample pairs was analyzed by
both Laser and sieve techniques. A mean similarity of 79% and a range of 55% to 97% similarity was
obtained, with the Laser technique consistently predicting larger fractions of the finer sediments. This
matches the results obtained for the low resolution coring program quite well. The authors attributed the
difference to the way the techniques measure particles. Essentially the Laser technique reports the
particle-size distribution by volume while the ASTM (sieve) method is sensitive to particle diameter and
shape.
Thus, the two methods report different distributions for the same sample. Since the primary goal
of these analyses was to classify sediments in a qualitative sense for potential PCB contamination, this
difference is unlikely to be important. In particular, the Laser results can be applied directly to the
B-9
J MAS! Gradient
-------
existing Phase 2 database, to expand and confirm the correlations seen between the side-scan sonar and
the confirmatory samples (TAMS et al., 1997). This application is presented later in the low resolution
sediemtn coring report.
In summary, for the low-resolution sediment core samples, all grain-size data are usable for both
qualitative and quantitative analyses. The laser analysis of the fine-grained material is a more accurate
representation of the particle size distribution of the fraction below 75 ^m. Uncertainty exists for the
sieve/hydrometer results for particle size intervals less than 75 (am due to method deviations.
B.4 Total Kjeldahl Nitrogen (TKN) Data
Total Kjeldahl Nitrogen (TKN) is defined as the sum of free-ammonia and organic nitrogen
compounds. The project objective for this measurement, along with the total organic carbon (TOC)
measurement, was to determine the importance of the carbon-to-nitrogen ratio in the sediment. According
to the Phase 2B SAP/QAPP, low resolution sediment coring samples were to be collected and analyzed
for total carbon/total nitrogen (TC/TN). Approximately 10% of the TC/TN samples were to be analyzed
for TOC/TKN to verify that negligible amounts of inorganic carbon and inorganic nitrogen were present
in the samples and to verify the assumption that the TOC/TKN analyses from previous sampling events
are comparable to the current TC/TN data. However, due to a problem with procuring an analytical
laboratory, the TC/TN analyses were excluded from the low resolution sampling program.
A total of 28 sediment samples, of which one was a field duplicate, were collected for TKN
analysis during the low resolution sediment coring program. All TKN analyses were performed by
Aquatec under the requirements of the USEPA Special Analytical Services (SAS) program. The samples
were prepared and analyzed for TKN using USEPA Method 351.2 from Methods for the Chemical
Analysis of Water and Wastes (USEPA, Revised 1983). Data are reported on a dry-weight basis in units
of mg/kg.
B.4.1 Accuracy
Accuracy, as measured by holding times, calibration QC (initial and continuing calibration checks
and blanks), matrix QC (matrix spike samples), and laboratory control samples (LCSs) met acceptance
B-10
J MAS! Gradient
-------
criteria as set forth in the SAS request with the following exception. Two matrix spikes exceeded the
upper limit for percent recovery (125%) as stipulated in the Phase 2B SAP/QAPP. Therefore, the TKN
results for the four samples associated with these matrix spikes were qualified as estimated ("J") based
upon the high recoveries of the associated spike analyses. The results are usable as estimated values that
may be biased high.
B.4.2 Precision
Six laboratory duplicate pair analyses were performed. All duplicate TKN measurements met the
laboratory split (duplicate) precision criterion of relative percent difference (RPD) < 20%, as stipulated
in the Phase 2A SAP/QAPP.
B.4.3 Sensitivity
Blanks were analyzed as required by the method. All blank concentrations were below the method
detection limit (MDL). Therefore, all sensitivity criteria were met for TKN analyses.
B.4.4 Representativeness
One field duplicate pair was associated with the 28 sediment samples. During validation, CDM
determined that the representativeness of the TKN results was compromised for 3 of the 28 samples due
to poor field duplicate precision. The TKN results associated with the field duplicate were estimated
(qualified "J"). According to the data validation guidelines, for results > 5xMDL (results were 4420
mg/kg and 4090 mg/kg) , the relative percent difference (RPD) should be used to evaluate precision.
CDM had evaluated precision using the absolute difference between results. Since the RPD for the
analysis was 7.4%, precision criteria were met and no actions were required. Therefore, TAMS/Gradient
reversed the decision to qualify these data and the "J" qualifier was removed from the affected samples.
The frequency criterion of 5% for field duplicate analyses was not met for TKN. (The actual
frequency of one duplicate pair in 27 environmental samples was 3.7%.) No actions were taken because
precision evaluation was made possible through the review of laboratory duplicate analyses.
B-l 1
T MAS!Gradient
-------
B.4.5 Summary of Usability
The overall data quality was acceptable and all TKN results are usable for project objectives. A
total of 15% of the TKN results were qualified as estimated ("J") due to high matrix spike recoveries.
The overall completeness for TKN was 100%, meeting the project DQO for completeness. Summary
statistics for TKN are presented in Table B-l.
B.5 Total Organic Carbon (TOC) Data
A total of 28 sediment samples (including one field duplicate) were collected for TOC analyses
during the low resolution sediment coring program. The TOC analyses were performed by Aquatec. All
samples were prepared and analyzed for TOC analysis using the 1986 version of the Lloyd Kahn TOC
in Sediment Method, rather than the 1988 version. Since the 1986 version of the method was used, the
TOC data were validated based on duplicate relative percent differences rather than on criteria related to
the initial laboratory establishment of precision as well as quadruplicate precision as defined in the
February 18, 1994 memorandum from TAMS. The overall quality of the data was not compromised by
the using the 1986 method criteria.
B.5.1 Accuracy
Accuracy, as measured by holding times, calibration QC (initial and continuing calibration checks
and blanks), method blanks, LCSs, and matrix QC (matrix spike samples) met acceptance criteria as set
forth in the SAS request with the following exceptions. Approximately 25% of the TOC results were
qualified as estimated ("J") due to potential sample degradation as a result of exceeding the recommended
analysis holding time. The affected TOC results are usable as estimated values that may be biased low.
In addition, a continuing calibration verification (CCV) exceeded the upper limit o f the recovery criteria
range (80 tol20%). Therefore, approximately 14% of the TOC results were qualified as estimated ("J").
The affected TOC results are usable as estimated values that may be biased high.
B.5.2 Precision
Laboratory duplicate analyses were not performed for TOC analyses. Precision evaluation was
still made possible because all samples were analyzed in duplicate as required by the 1986 version of the
B-12
T A M SI Gradient
-------
Lloyd Kahn method. Quality control criteria for these duplicate analyses were set forth in a memorandum
from TAMS Consultants dated February 18, 1994.
The precision of the TOC results was compromised for approximately 18% of the results due to
poor replicate precision (RPDs were > 25% but < 100%). The affected TOC results were usable as
estimated values, but a bias could not be determined. One TOC result was rejected (R) because
uncertainty in quantitation existed based upon extremely poor replicate precision (RPD > 100%). The
result is unusable for project objectives.
B.5.3 Sensitivity
Sensitivity issues affecting the TOC analyses, in terms of blank evaluation and detection limits,
were not noted during the usability assessment. All blank results were < 0.01% TOC.
B.5.4 Representativeness
One field duplicate sample was associated with the TOC analyses. The precision criterion of RPD
< 100% was met for this analysis; the RPD for the duplicate pair analyzed was 25.5%. Frequency criteria
for field duplicate analyses were not met, but, since all samples were analyzed in duplicate, precision
evaluation was still possible.
B.5.5 Summary of Data Usability
Approximately 48% of the TOC results were qualified as estimated ("J") due to QC exceedances
including holding time exceedances, high CCV recovery, and laboratory duplicate imprecision. The
results are usable as estimated values. All TOC results are usable with the exception of one result, which
was considered unusable (rejected) due to severely poor replicate precision. Therefore, overall
completeness for low resolution sediment core TOC analyses is 96.3% meeting the project DQO for
completeness. Summary statistics for TOC are presented in Table B-l.
B-13
1AMS! Gradient
-------
B.6 Radionuclide Analyses
Radionuclide analysis was performed on ail low resolution sediment core sections to establish
sediment core chronology. Dried and homogenized sediment aliquots were analyzed for several principal
radionuclides by B&W Nuclear Environmental Services, Inc., in Parks Township, PA and Lynchburg,
VA. For the Phase 2B investigation, only beryllium-7 (7Be) and cesium-137 (137Cs) were validated and
assessed for data usability. The top sediment core slices were only analyzed for 7Be. All sediment core
slices were analyzed for ,37Cs.
Several issues may have affected the overall usability of the radionuclide data: small sample size;
exceeded holding times for Tie; sample density and geometry differences; the presence of wood chips in
the samples; and blank and background corrections.
The first core samples submitted for radionuclide analysis were of limited sample size, which
affected the statistical counting error for the 7Be results. The limited size and low 7Be activity in the core
samples resulted in statistical errors greater than the acceptable 10% maximum error specified in the
QAPP. To reduce this error, the time of sample analysis was increased to up to 60 hours. (The QAPP
stated that the samples were to be counted for 8 hours or until the statistical error was less than or equal
to 10%.) As a result of the increased counting time, the holding times for the 7Be samples were
potentially compromised. Therefore, to produce usable data, TAMS/Gradient established an approach
to the analysis of 7Be (August 30, 1994). The samples could be counted for He for 8 to 24 hours as long
as the statistical error was less than 40%. Otherwise, the samples were counted for 36 to a maximum of
48 hours to achieve a statistical error of less than 50%.
The calibration curves established for the radionuclide analysis were produced using Allegheny
River sediment. Since B&W generated the calibration curve based on weights of the sediment in the cans
rather than on the percent full, there was some concern that the Allegheny River sediment density was
not comparable to the Hudson River sediment's density. In order to produce accurate results the
geometries of the calibration standards and samples need to be comparable. B&W analyzed a Hudson
River LCS to determine if the calibrations generated were acceptable. The study showed that there was
no significant difference between the Allegheny River and Hudson River Sediments in the 59.5 Kev to
the 898 Kev range (B&W, 1994). The study also indicated that there was no difference in matrix density.
B-14
T A MS/Gradient
-------
The presence of wood chips in the samples could dilute the radionuclide activity by affecting the
geometry of the sample; therefore, wood chips were removed from most of the samples prior to counting.
Some of the initial samples received by B&W were prepared and analyzed with the wood chips retained
in the sample. This issue is further addressed in the accuracy section, B.6.1, below.
The radionuclide method requires that activities (results) be corrected for background, blanks, the
radionuclide branching ratio, the efficiency geometry of the detector, and for radionuclide specific decay.
TAMS/Gradient established validation criteria for radionuclides to verify that sample results were
accurate.
For the low resolution coring program, a total of 178 sediment samples (including 9 duplicates)
were analyzed for radionuclides, generating 980 records (including field and laboratory duplicates). A
total of 169 (178 less the 9 duplicates) validated samples (a total of 338 records for both 7Be and l37Cs)
were reported in the project database.
B.6.1 Accuracy
The validator qualified as estimated ("J") sample results in a number of SDGs due to the lack of
an associated Laboratory Control Sample (LCS) analyses for both 7Be and137 Cs. TAMS/Gradient
concluded that this was not a technically appropriate reason to qualify the associated results because the
lack of the associated QC sample did not impact the overall quality of the data. Therefore, the decision
to qualify results due to lack of LCS analyses was reversed during the usability assessment.
The accuracy of some low resolution core samples was compromised due to the presence of wood
chips in the samples. Approximately 20% of the low resolution core samples contained wood chips in
a range of 10% to 90% by volume. The presence of wood chips could dilute the radionuclide activity by
affecting the geometry of the sample; therefore, the wood chips were removed prior to counting. Some
of the initial samples received by B&W were prepared and analyzed with the wood chips retained in the
sample. The radionuclide activity results for the sample containing wood chips may be biased low
compared to those samples in which the majority of the wood chips were removed prior to counting. No
qualifications were made to the data during this usability assessment due to the qualitative nature of the
results. Data from samples containing wood chips are clearly indicated as such in the project database.
B-15
Gradient
-------
There were no other issues affecting accuracy noted during the data usability assessment.
B.6.2 Precision
Precision, in terms of laboratory duplicate analyses, was met for all 7Be and 137Cs radionuclide
analyses with the exception of four 7Be laboratory duplicare analyses. This affected approximately 9%
of the 7Be samples, which were already estimated (J) due to statistical error exceedances. During
validation some 137Cs results were estimated because laboratory duplicate frequency was not met. During
the usability assessment the TAMS/Gradient team reversed this decision. Therefore, no qualifications
were made due to the lack of an associated laboratory duplicate.
B.6.3 Sensitivity
For radionuclide analyses, measured background counts were subtracted from sample counts prior
to calculation of concentrations. In some cases, this resulted in negative concentration values, which
should be considered zero for purposes of data interpretation. Low-level activities, for which the counting
statistics show a high relative error (counting error of greater than 50% of the reported result), are also
considered not significantly different from background. These evaluations were applied to the data during
validation; therefore, some low-level positive values were considered to be not detected, i.e., no activity,
following data validation. Following background correction, of the 169 total records for each
radionuclide, 70% and 16% of the 7Be and 137 Cs radionuclide results^ respectively, had activities
significantly greater than background. These results were considered estimated (qualified J) due to
statistical counting errors between 10% and 50%. Approximately 12% and 18% of the radionuclide
results for 7Be and 137Cs, respectively, did not have low-level activities that were significantly above
background due to statistical counting errors greater than 50%. Thus, these results were considered to
be estimated and comparable to background activity (qualified UJ). The statistical counting errors,
representing one standard deviation, have been maintained in the database to give the data user additional
information regarding the uncertainty of the reported radionuclide activities.
In addition to the radionuclide results that were reported with activities and statistical errors,
approximately 18% of the 7Be and 52% of the 13ts results were qualified by the laboratory with a "LLD",
meaning lower level of detection. During the assessment, the TAMS/Gradient team determined that these
B-16
TAMS/Gradient
-------
radionuclide results did not have reportable activities above background and thus were considered to be
detection limits (qualified U).
B.6.4 Representativeness
Field duplicate pairs were collected for radionuclide analyses. However, representativeness for
these data is a qualitative indicator for radionuclide analyses, rather than a quantitative indicator.
Therefore, the field duplicate data were reviewed for consistency, i.e., to verify that radionuclides which
were not detected in samples were also not detected in the field duplicate sample. All of the nine 7Be and
the eight ,37Cs field duplicate pairs exhibited consistent results, with the exception of LH-39M-0001. In
that sample, 7Be was detected (412 pCi/Kg) but was not detected in the field duplicate sample (less than
the LLD [458 pCi/Kg]), and qualified "U"). No actions were required, since the the results are
comparable.
B.6.5 Summary of Data Usability Assessment
Based upon QA oversight during analysis and review of radionuclide calibrations, data packages,
and data validation reports, all 7Be and137 Cs results were considered usable by TAMS/Gradient.
Approximately 82% of the 7Be and 17% of the l37Cs results were qualified (estimated J) due to statistical
counting errors and imprecision. The results are usable as estimated values and detection limits. No 7Be
or l37Cs radionuclide results were rejected (qualified R) during data validation or this data usability
assessment. Therefore, completeness of 100% was achieved for these analyses, meeting the project DQO
for completeness. Summary statistics for these analyses are presented in Table B-3.
B-17
1 MAS! Gradient
-------
References
B&W Nuclear Environmental Services. 1994. Letter to Susan Chapnick, Gradient, from William Stagg, B&\
Nuclear Environmental Services. Hudson River PCB Reassessment Rl/FS USEPA Work Assignment 01.
2N84, TAMS Project No. 5213-14, B&WNESI-NEL Project No. 431-1017-00. OctoberE21, 1994.
Shillabeer, N., B. Hart, and A. M. Riddle. 1992. "The Use of a Mathematical Model to Compare Particle Si2
Data Derived by Dry-Sieving and Laser Analysis." Estuarine, Coastal and Shelf Science, Vol. 35, p{
105-111.
TAMS. 1994. Memorandum to Jennifer Oxford, CDM, from Allen Burton, TAMS Consultants. Hudson Rive
PCB Rl/FS - Criteria for Evaluating Data Generated Under 1986 Lloyd Khan TOC in Sediment, Februar
18, 1994.
TAMS/Gradient. 1992. "Phase 2A Sampling and Analysis Plan/Quality Assurance Project Plan - Hudson Riv<
PCB Reassessment Rl/FS." EPA Contract No. 68-SP-2001.
TAMS/Gradient. 1994. "Phase 2B Sampling and Analysis Plan/Quality Assurance Project Plan - Hudson Rive
PCB Reassessment Rl/FS." EPA Work Assignment No. 013-S9-2N84.
USEPA. Revised 1983. Methods for the Chemical Analysis of Water and Wastes. U.S. Environmental Protectio
Agency, Environmental Monitoring Support Laboratory. EPA/600/4-79-020. 1979; Revised Marc
1983.
USEPA. 1992a. Evaluation of Metals Data for the Contract Laboratory Program (CLP) based on SOW 3/90 (SO
No. HW-2, January 1992)
USEPA. 1992b. "Guidance for Data Usability in Risk Assessment." U.S. Environmental Protection Agency
Office of Emergency and Remedial Response, Washington, DC. EPA PB92-963356, Publication 9285.'
09A.
B-18
T MAS'Gradient
-------
Table B-l
Low Resolution Sediment Non-PCB Sample Analysis Summary
Hudson River RI/FS PCB Reassessment
Parameter
Total Number
of Results
Unqualified
Nondetects
Estimated
Nondetects
Unqualified
Detects
Estimated
Detects
Rejected
Results
% Rejected
l'la>% (1 user)
170
0
0
170
0
0
0%
Coarse Sand°<> (Laser)
170
0
0
170
0
0
0%
Fine Sand% (l.aser)
170
0
0
170
0
0
0%
Geometric Mean Diameler
170
0
0
170
0
0
0%
<.iravel% (l.aser)
170
0
0
170
0
0
0%
Median Diameler
170
0
0
170
0
0
0%
Medium Sand'!<>(l.aser)
170
0
0
170
0
0
0%
Sill®»(l.aser)
170
0
0
170
0
0
0%
Skeuness (l.aser)
170
0
0
170
0
0
0"o
Sorting (1 aser)
170
0
0
170
0
0
0%
TKN
27
0
0
23
4
0
0%
TOC
27
0
0
13
13
1
4%
Totals
1754
0
0
1736
17
i
0%
\\|huIm>i>'li'wrcsic;i|tpciuli\ II! KSIAIS XI S
TAMS/(
-------
Table B-2
Low Resolution Sediment Sieve Grain Size Sample Analysis Summary
Hudson River RI/FS PCB Reassessment
Parameter
Total Number
or Results
Unqualified
Nondcteets
Estimated
Nondetects
Unqualified
Detects
Estimated
Detects
Rejected
Results
% Rejected
' 0.075 mm
14.3
0
0
0
143
0
0%
•0 075 mm
143
0
0
143
0
0
0%
0.150 mm
143
0
0
143
0
0
0%
-'0.425 mm
143
0
0
143
0
0
0%
1.0 mm
143
0
0
143
0
0
0%
> 1.4 mm
143
0
0
143
0
0
0%
>2.0 mm
143
0
0
143
0
0
0%
>4.0 mm
143
0
0
143
0
0
0%
• 4.75 mm
143
0
0
143
0
0
0%
Coarse Sand % (Sieve)
143
0
0
143
0
0
0%
Fine Sand % (Sieve)
143
0
0
143
0
0
0%
Fines % (Sieve)
143
0
0
0
143
0
0%
Gravel % (Sieve)
143
0
0
143
0
0
0%
Largest >4.75 mm
143
0
0
143
0
0
0%
Medium Sand % (Sieve)
143
0
0
143
0
0
0%
Totals 2145 0 0 1859 286 0 0%
xAhiulson'iim-icMc'iiipiKnilixVHI KS'I'A I S XI S
TAMS/(irmliciit
-------
Table B-3
Low Resolution .Sediment Radionuclide Sample Analysis Summary
Hudson River RI/FS PC'B Reassessment
Total Number Unqualified Estimated Unqualified estimated Rejected .
1 .ir.imeiiT ()f Rt,suus Nondetects Nondctects Detects Detects Results " 4 'l
|k.7 16') 30 20 0 119 0 0%
CS-|;,7 I6«) 88 31 23 27 0 0%
Totals 338 118 51 23 146 0 0%
vhmKoii Ioukhk ,tp|Kmlt\ HI KM A IN XI N
TAMS'(/V(i^c/i/
-------
Visual Inspection and Laser Grain-Size Distribution Analysis
Compared by Principal Fraction
120'
100 _
80-
60'
3
O
O
40'
20~
-. • O • Oj oy* O; O • > 0
V/ '/*' '. -
Sample Count by Las«r Analysts
Silt
HI
Principal Fraction by Visual Inspection
Sample Count by
Visual Inspection:
5 0 Clay/Organic
99 ~ Silt
31 0 Fine Sand
33 0 Coarser Sand
l ¦ Fine-Medium Gravel
' ' -71
. r - V rf* . r. r r
• o» ^ %\« o • n* ¦ vs •
»*. **f •*, **, * '>**» >f* *** »/*
'• sV« Oj»V« sN ¦ O; *> -
Fine Sand
ji
J
Medium Sand
i)
Gravel
5
Principal Fraction by Laser Grain-Size Distribution Analysis
c
3
O
O
o
U
120
100 -
so
60-
40 j:
20 -
0
Distribution of Samples Classified by Both Methods
Clay'
"sTTT
Principal Fraction by
Visual Inspection
Fine
Sand
Coarser
Sand
F/M
Gravel
Source: TAMS/Gradient Database. Release 3.5 TAV1S
Figure B-l
Classification of Shallow Sediment Samples
Comparison of Visual Inspection and Laser Grain-Size Analytical Technique
-------
Visual Inspection and ASTM Grain-Size Distribution Analysis
Compared by Principal Fraction
80 ~T"
70 " ••
60"
50 '
Principal Fraction by
Visual Inspection
Sample Count b\
Visual Inspection
r
! 66 ~
: 23 0
44 S
Cla>
Silt
Fine Sand
Coarser Sand
Fine-Medium Gravel
Fines Fine Sand Medium Sand
Sample Count b> ASTM Analysis 77 42 22
Principal Fraction by ASTM Sieve Method
Gravel
c
3
o
o
U
80
70
60
50
40
30
20
10
0
80
70
60
50
40
30
20
10
0
Distribution of Samples Classified by Both Methods
Fine
Sand
Coarser
Sand
F'-M
Gravel
source: TAMS/Gradient Database. Release v5 TAMS
Figure B-2
Classification of Sediment Samples
Comparison of Visual Inspection and ASTM Grain-Size Analytical Techniques
-------
ASTM and Laser Grain-Size Distribution Analysis
Compared by Principal Fraction
c
3
O
U
40 —
35
30 '
25
-0 "j
Principal Fraction by Laser Method
Sample Count byi ^ j
l aser AnaUsi>
i 11
3
4
: 0
Fines
Fine Sand
Medium Sand
Coarse Sand
Gravel
101
1
¦i
5 ~>.
0
h ines
Sample Count bv -\S IM Anal> si?» 39
__i__ i XI A .
Fine Sand Medium Sand Coarse Sand
22 7 0
Principal Fraction by ASTM Sieve Method
Oravt
Comparison Between Laser and ASTM Grain-Size Methods
6
4
*>
0 -
% Similarity.
12 •
Minimum
*4 3
Maximum
48 4
10-
Points
64
Mean
76 1
= 8 •
Median
76 4
p
Std IX'viation
13 6
30
40
50
60 70
% Similarity
80
90
100
Source: lAMSCiraiiieni Database. Release 3.5
TAMS
Figure B-3
Classification of Sediment Samples
Comparison of Grain-Size Analytical Techniques (ASTM and Laser Methods)
-------
APPENDIX C
r^!
-------
APPENDIX C
1994 Low Resolution Core And 1984 Nysdec Core Profiles
For The Thompson Island Pool
-------
m
9
OS
J
E
« 3
u
I S*
<8
»
E ^
.2 a
U CL.
m a
? 5
as oc
S
H
o
D
ffi
9
• ~
L8
(~*r>
I
u
£, i
as cd 1
J £ 1
- S1
O
©J
<-< t
£
<
; 9
! M
I -J
a
.2
a.
<
©
ac
a
k
a
<
©
at
# ~
ao
<*
o £
© «/>
m Q
hm
o
8 %
rs ffl
O
-------
o
t
N>
o
VO
It-
s'
*
SO
3
©,
Ef
n
o
•n
n
»
so
oo
c/s
©
W
O
o
o
<3
10
•n
0
r>
2
01
«¦
sr
n
o
s
T3
w
©
9
»
B
O.
*0
©
©
LR-02A
Total PCB (mg/kg)
0 100 200 300 400 500 600 700 800
0 ¦
10
20
c
s. 30
V
Q
40
50
]
60 --I
20
S. 30
s
40
50
60
• LR-02A Be-7
~ LR-02A Cesium
I.R-02A Phase 2 Total PCBs
I.R02A DEC Total PCBs
500 1000 1500 2000
Be7 or Cs"7 (pCi/kg)
LR-02C
Total PCB (mg/kg)
0 50 100 150 200 250 300 350 400
0 .... 1. I ¦ -4
10
2500
LR-02C Phase 2 Total PCBs
LR-02C DEC Toul PCBs
O-
«s
10
20
u
I
LR-02B
Total PCB (mg/kg)
50 100 150 200 250 300 350
E. 30
40
50 -
60
LR-02B Be-7
LR-02B Cesium
LR-02B Phase 2 Toul PCBs
LR-02B DEC Total PCBs
600
200 400
Be7 or Cs137 (pCi/kg)
800
1000
-------
0
1
vo
i
50
e
s
n
e
«
v©
00
E/J
©
tfl
o
n
e
¦*
o
5
5"
M
3*
"t
n>
H
sr
o
3
¦o
w
e
s
w
9
O.
*e
©
e
Q.
4i
Q
x
5.
LR-03A
Total PCB (mg/kg)
0 10 20 30 40 50 60 70
80
10
20
5. 30
40
50
60
i
i
i
i
LR-03A Be-7
I.R-03A Cesium
I.R-OJA Phase 2 !ol«l PCB I
LR-03A Di C lotal PCBs
I '
20
10
20
30
40
50
60
40 60 80 100 120 140
Be7 or Cs"7(pCi/kg)
LR-03C
Total PCB (mg/kg)
10 20 30 40 50 60 70 80
« i.i . . . . I . . i ». I ¦ • ' ^ * * j- L.i .
I
I
i
• I.R-03C Be-7
~ LR-03C Cesium
LR-03C Phase 2 Tola! PCBs
[.R-0JC DfcC Tuial PCBs
100 200 300 400 500 600 700 800
Be' or€sn'(pCi/kg)
20
S. 30
5
40 -
50
60
20
40
LR-03B
Total PCB (mg/kg)
80 100
60
^ i .
l.R-WB Be-7
LR-03B Cesium
I.R-0JB Phase 2 Tola) PCBs
I.R 03U DEC Total PCBs
120
i
140
500 1000 1500 2000 2500 3000 3500 4C
Be7 or Cs'" (pCi/kg)
-------
o
-U
v©
v©
50
n
v>
©
o
s
n
©
¦t
rs
»
3
D.
v©
00
C/3
©
tn
o
o
0
1
o
M
©
•n
©
3
¦o
w
O
5
6
•«
o
e
LR-04A
Total PCB (mg/kg)
10 -
20
S. 30
<3
40 -
50
60
20
¦ ' ¦
40
60
80 100 120 140 160
i
10 -
_ 20
5
S. 30 -|
6
40 -I
50 -
60
_j_
-h
• LR-04A Be-7
~ LR-04A Cesium
L.R-04A Ph«s« 2 Toul PCB*
LR-04A DEC TouJ PCBs
foo ioio'' ioo 4oo 'ioo 'too 7oo
Be7 orCs"' (pCi/kg)
LR-04C
Total PCB (mg/kg)
o
r
200
400
600
800
- I
1000
1200
• LR-04C Be-7
~ I.R-04C Cesium
LR-04C Phase 2 Toul PCB«
LR-04C DEC Toul PCB»
^00 I000 ' (500
Be7 or Cs137 (pC'i/kg)
2000
10
20
E. 30
&
I
40 -
50
60
100
¦ ' ¦
LR-04B
Total PCB (mg/kg)
200
300
• ' ¦
400
~
500
» J ^
LR-04B Be-7
LR-04B Cesium
LR-04B Phase 2 Toul PCBs
I LR-04B DEC Toul PCBs
I
ioo
1000
1500
Be or Cs (pCi/kg)
600
000
10
20 -
50
. I ,
LR-04D
Total PCB (mg/kg)
100 150 200 250
?
300 350
JB
a.
&
S. 30
40 -
50 -
60
• LR-04D Be-7
~ LR-04D Cesium
LR-04D Phuc 2 Toul PCB.
I LR-04D DEC Toul PCB.
I
ioo ' 4oo ' ' ioo doo
Be7 or Cs137 (pCi/kg)
1000
-------
LR-05A
50
Total PCB (mg/kg)
100 150 200
250
300
~ i
• LR-05A Be-7
~ LR-05A Cesium
| I.R-05A Phase 2 tola I PCBs
I
I I.R-05A 1)11 1 ola) PCIts
I
100 200 300 400 500 600 700 800
Be7 or Cs"7 (pCi/kg)
50
LR-05C
Total PCB (mg/kg)
150 200 250 300 350
1
•
*
-J
LR-05C Be 7
LR-05C Cesium
I.R-05C Phase 2 Total PCBs
LR-05C and l.R-05fc
DEC Tottl PCBs
200 400 600 800
Be7 or Cs1" (t)Ci/ke)
1000 1200
LR-05B
Total PCB (mg/kg)
SI
Cl
3
10
20
30
40
50
60
50
100
150 200
~
250
¦ ¦ ' ¦
• I.R-05B Be 7
~ LR-05B Cesium
I 1.R-05B Phase 2 loul PCBs
« I.R-05B I)I C Tolal PCBs
I
300
X _^.l. 1
350
200 400 600 800 1000 1200 1400
Be7 or Cs137 (pCi/kg)
_c
o.
LR-05D
Total PCB (mg/kg)
0 200 400 600 800 1000 1200
10
20
30
40
50
60 -
I
1 r -
i
LR-05D Be-7
I.R-05D Cesium
l.R-OSD Phase 2 Total PCBs
1 R-05D Di:C Tout PCBs
0 200 400 600 800 1000 1200 1400
Be7 or Cs"7 (pCi/kg)
-------
u
in
0
I
01
_3
8-
t 8-
03
u
± o
« ?
o
(-
E
T 1
i> o
fiS U
UJ —
• ~
0£ ffi
— o
9 O
2 5
fs ^
t-oo aa
u
¦ o c.
I 03
is
o
o
( ui)qjd3Q
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the Thompson Island Pool
C-6
-------
00
E
VO
o
I
Qt
m
-J u
¦¦ a.
- si
o
H
a u
s =
S 8
ci d
• ~
r 2 ea
hs >£
^ * l/J
I L§ ^
...Ji
L8
CO
To
( U|) qidaa
o
©
"3
<
so
o
I
X
J
i>
00
CO
E
(2 *
c
¦"2
k
2
«
s!
<*
D
<«
C
E
3
n
u
-
:§
c.
y.
U
«»
"~rn
r-
a
O
o.
(/)
<
<
<
<
u
£
QC
s
ek
£
(X
s
AC
i-8
r
-------
LR-07A
Total PCB (mg/kg)
20
40 60
^.,1 ^
80 100 120 140 160
-1 - *" •
• IR-Q7A fle-7
~ LR-07A Ccs»um
| I R-07A Phase 2 lolil KBs
I
I I.R-07A UK lotal PCBs
I
200' 400 600 800 1000 1200 1400
Be7 or Cs'" (pCi/kg)
LR-07C
Total PCB (mg/kg)
50
-.i
100
'I
•
I
I
150
' *
~
LR-07C Be-7
t.R-07C Cesium
I.R-07C" Pha«r 2 Toul I'CBs
LR-07C DEC lol»i Klis
200
200
400
600
800
1000
Be7 or Cs"7 (pCi/kg)
LR-07B
Total PCB (mg/kg)
20
40
60
20
60
80
100
i
c
*w
i
JC
i
S. 30
V
i
Q
i
# 1.R-07B Bc-7
40
~ I.R-07B Cesium
I i R-07B Phase ? Total PCBs
50
1
1 1R-U7B l>n Total PCBs
1 I
100 200 300 400 500 600 7
-------
CO
oo
?
OS
-J
so
v-U
00 —
E
cc
u
a. _
£
CD
o a.
a
u
2
a
• ~
f o
I ©
r
TO
00
c
Q.
4J
CO
00
e
i
cc
J
a
u
a.
"«
o
H
o ^
CD
2
£
a
£
id
c
3
' ©
roo
Li *
~ §
V
m
• ~
'8
O
O
O
o
(N
o
TT
O
>0
( UI) qidaa
( uj) qjdsa
<
00
o
I
a
_i
^lb
_E
CQ
a.
"ra
(2
a
<
• ~
U
u.
O
V
CD
U
00
9
i
OS
J
0-.
00 I
00
^co
U
a.
o
H
a
k
• ~
s
fS
i-r->
i.
KnO
r
-------
( uDqidsa
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the Thompson Island Pool
C-10
-------
LR-09A
Total PCB (mg/kg)
l 10 20 30 40 50 60 7
, j >. l. i,. j i . .1 >. f * ' ^' » 4 < * ' L a. . X .
• LR-09A Bc-7
~ LR-WA C esium
| LR-(WAPhasc3 lolaiPCBs
I I.R-UOA 1)1 C loull't'lls
200
400
600
Be'or Csl17(pCi/kg)
800
LR-09C
Tolal PCB (mg/kg)
20 40 60
1 i. !• « ^ -
I
I
I
80
100
• LR-09C Bc-7
~ LR-O^C Cerium
LR-OTC Chaw: 2 Total K Bs
I.R-09C D1:C tolal PClis
100 200 300 400 500 600
Be7 or Cs"' (pCi/kg)
LR-09B
Total PCB (mg/kg)
5 10 15 20 25 30
L.i.1 i , *..1.^*.*. L_»
35
¦4-
40
•5
o.
<3
10
20
30
40
50
60
• LR-09B Be 7
~ LR-09B Cesium
1.R-0911 Phase 2 Total PCB«
I I.R-CWH 1)K* r«»l PCBs
I
10
20
30
40
Be7 or Cs "7 (pCi/kg)
50
E
<3
10
20
30
40
50
60
LR-Q9D
Total PCB (mg/kg)
o 50 100
150 , 200
250
• I.R-09D Be?
~ I.R-09D Cesium
I ROTO Phase 2 Total PCBs
LR 09D DI C Toul PCB5
300
500 1000 1500
Be' or Cs'37 (pCi/kg)
2000
2500
-------
I ©
i ©
00
U. ^
£ g»
o E.
Q£
-J
o
H
oj
* Si
$ S
ce i
U
o.
(/>
u
rg o
r-.
O
CQ
o
rs
o
TT
o
NO
( U!) qjdaa
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the Thompson Island Pool
C-12
-------
n
¦
ve
\o
s:
*
so
!S1
©,
E"
5'
9
n
o
"1
r»
»
9
CL
so
oe
cn
O
W
0
n
©
1
r#
»r
©
n
t/i
H
sr
©
3
T3
w»
©
9
»
9
a.
©
©
10
20
t 30
•40
50
60
20
LR-10A
Total PCB (mg/kg)
50 100 150 200 250 300
¦ —1- 1 ' -1' J-1
i
350
~
LR-I0A Be 7
1.R-I0A Cesium
I.R-IOA Phase lolal l'( Ds
I.R-I0A l)IC lolal I'Clls
200 400 600 800 1000 1200
Be7 or Cs'17 (pCi/kg)
LR-10C
Total PCB (mg/kg)
100 200 300
^ i
1400
4(H)
' * •'
500
c
i i
-c , . <
~
5. 30
Q
•
40
~
1 1
50
| 1
1
1
|
60
0
I.R-I0C Be 7
I.R-I0C Cesium
l.RIOC I'hasc 2 lolal PCBs
1.K-I0C HI t lolal I'CHs
500 1000 1500
Be7 or Cs"7 (pCi/kg)
2000
10
20
S. 30
u
a
40
50
60
50
LR-10B
Total PCB (mg/kg)
200
100
150
250 300
i
i
i
• LR-IOB Be-7
~ I.R-I0B Cesium
| I.R-I0I1 Phase 2 Total PCBs
I
¦ I R- I0U 1)1 C: lolal PCBs
200
400
600
800
Be orCs (pCi/kg)
1000
10
20 -
H 30
40
50
60
LR-10D
Total PCB (mg/kg)
50 100 150
200
I"1
I.R-I0D Be-7
I.R-I0D Cesium
I R-10D I'hasc 2 Total PCBs
1.R-10D DI C Total PCBs
100 200 300 400 500
Be7 or Cs"7 (pCi/kg)
600
-------
s
a
00
U
Cl
• ~
CO
1- 2
o
fN
O
^r
o
sO
(ui) md^a
2-J
fS
-r-
;§
oo
— CO
S s
0C
-1
CS C"
o
H
§-
< <
Of X
m ~
O0
Li
u
CD
00
u -
mm 00
T it
a: m
o
H
CQ
*
£
• ~
a
u
a
o
s «
-v> «*
c
C.
F:1 s
fg «
o
©
^r
( UI) qidba
(•ui) qidsa
1994 Low Resolution Core and 19S4 NYSDEC Core Profiles for the Thompson Island Pool
C-14
-------
0
1
»-—•
LSI
\o
VO
r
o
*
90
re
o
n
o
rs
to
s
a.
\o
0©
4k
c/5
a
M
n
n
o
n
*0
o
n
<*>
©
rt
H
sr
o
3
•o
M
o
9
te
s
Q.
O
o
c
JC.
10
20
5. 30
u
Q
•40
50
60
20
6. 30
&
40
50
60
LR-12A
Total PCB (mg/kg)
5
10
15
20
i . L
200
I.R-I2A Bc-7
I.R-I2A Cesium
I.R-I2A Phase 2 Tolll PCBs
1.R-I2A Dt C lotal PCBs
400 600 800
Be7 or Cs"7 (pCi/kg)
LR-12C
Total PCB (mg/kg)
I.R-I2C Be-7
[ R-12C Cesium
1.R-I2C Phase 2 loUl PCBs
LR I2CIHC lolal PCBs
200 400 600 800
Be7 or Cs'17 (pCi/kg)
25
VI
1000 1200
10
I
1000 1200
LR-12B
Total PCB (mg/kg)
4 6
_w—.1 -
8
200
10
'f L J
12
I.R-I2B Bc-7
LR-I2B Cesium
I.R-I2B Phase 2 Toul PCBs
1.R-I2B 1)1 C Tout PCBs
400 600 800 1000 12(H)
Be7 or Cs"7 (pCi/kg)
14
LR-12D
Total PCB (mg/kg)
10
20
30
40
50
• LR-12D Be-7
~ LR- I2D Cesium
I I.R-12D Phase 2 Toul PCBs
I.R-I2D DFC Total PCB$
0 500 1000 1500 2000
Be7 or Cs137 (pCi/kg)
-------
00
W €
2 E
a: ~
CD -
O
Cu
O
s
k
3
2 3
UJ -L.
• ~
a
k!
00
flQ
©
T
©
O
>o
(•ui)qvfca
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the Thompson Island Pool
C-16
-------
ve
NO
r
o
3
U9
o,
5"
o*
s
O
e
to
9
O.
\o
oo
C/3
©
PI
n
n
o
re
©
a
«T
w
y
•*
H
BP
e
3
"O
v>
e
B
£
5"
D
Q.
*0
©
©
10
20
5. 30
&>
Q
40
50
60
10
20
S. 30
u
Q
40
50
60
LR-13A
Total PC'B (mg/kg)
10 15 20 25
30
35
• I.R-I3A Bc-7
~ LR-13A Ccsiuin
| LK-I3A I'lusc- 2 Total PCBs
I
I I.RI3A I>l ( lolal l'( lis
100 200 300 400 500 600 700
Be7 or Cs"7 (pCi/kg)
LR-13C
Total PC'B (mg/kg)
10 15
20
25 30
35
V
• LR-I3C Bc-7
~ l.R-1 3C Cesium
j t.R-13C Phase 2 l olal PCBs
I
I I.R-I3CDI.C lolal PCBs
200 400 600 800 1000 1200 1400
Be7 or Cs'" (pCi/kg)
10
20
a. 30
<3
40
50
60
LR-13B
Total PCB (mg/kg)
5
_ i.
10
15
20
• 1.R-I3B Bc-7
~ I.R-I3B Cesium
I.R-I3H Phase 2 Tottl PCBs
I.R-I3B DI:C Tottl PCBs
25
~
100 200 300 400 500 600 700
Be7 or Cs137 (pCi/kg)
-------
0
1
oo
VO
£
*
n
v>
©
o
s
o
o
•n
ft
»
S
Q.
N-l
vo
0©
•ii.
C/5
©
W
o
n
o
3
•t
o
»
5T
CM
o>
H
sr
o
B
•o
VI
O
9
E
sr
9
Q.
o
O
LR-14A
Total PCB (mg/kg)
20
10
20
5. 30
u
Q
40
50
40
i_j
60
80
100
_L .I...*..,
120
60
• LR-I4A Be-7
~ LR-I4A Cesium
I.R-I4A Phase 2 Total PCBs
I I.R-I4A DliC Total PCBs
200 400 600 800
Be7 or Cs'37 (pC'i/kg)
LR-14C
Total PCB (mg/kg)
1000 1200
10
20
30
40
10
20
t 30
50
.-i
60
40
50
60
~
I.R-MC Be-7
1.R-I4C Cesium
LR-I4C Phase 2 Total PCBs
1.R-I4C DliC Total PCBs
T
200 ^00 600 800 1000 1200 1400
Be7 or Cs137 (pCi/kg)
10
20
a.
o
Q
30 -
40
50
60 --
LR-14B
Total PCB (mg/kg)
10
15 20
25
30
35
40
—-V—
1
1
• I.R-I4B Be-7
~ I.R-I4B Cesium
I.R-I4B Phase 2 Total PCBs
I I.R-I4B DliC Total PCBs
I
500 1000 1500 2000
Be7 or Cs137 (pCi/kg)
2500
_c
£
H.
u
Q
10 -
20
30
40
50 -
60
LR-14D
Total PCB (mg/kg)
10
,._i..
15
I.R-I4D Be-7
1.R-I4D Cesium
LR-J4D Phase 2 Total PCBs
LR-I4D DliC Total PCBs
200 400~ 600 800
Be7 or Cs137 (pCi/kg)
20
, i
25
1000 1200
-------
n
«o
VO
\o
4k
r
1
8?
8
e
e
*¦(
mm
n
e
»
s
&
¥"rik
v©
oc
Cfl
©
w
n
n
o
"J
ID
•t
O
E
WJ
<7
*t
#*¦
fS
H
y
o
3
•n
t/>
e
s
t?
iBBMf
m
a
a
*0
e
p
10
20
c
f 30
u
Q
40
50
60
5
"p1-* *
i
LR-15A
Total PCB (mg/kg)
15 20
10
. ..i.. ¦
25
30
• LR-I5A Be-7
~ LR-I5A Cesium
I LR-I5A Phase 2 Total PCBs
I LR-15ADCC: Toutl'CUs
I
500
1000
1500
2000
2500
Be7 or Csn'(pCi/kg)
LR-15C
Total PCB (mg/kg)
10
.. .t.
15
20
25
10
20
c
•C
a. 30
4)
P
40
50
60
• IR-I5C 11(7
~ LR-I5C Cesium
1 LR-I5C Phase 2 Total PCBs
I LR-ISC DEC Total PCUs
I
200
400
600
800
1000
Be1 or Cs"7 (pCi/kg)
0 0.5
0 0......
LR-15B
Total PCB (mg/kg)
I 1.5 2 2.5
a
u
Q
10
20
30
40
50
60
~
LR-I5B Be-7
LR-15B Cesium
I R-1 SB Phase 2 Total PCBs
t.R-150 DEC Total PCBs
3 5
100
200
300
Be7 or Cs"7 (pCi/kg)
400
500
Q.
«
Q
10
20
30
40
50
60
n*
10
~ i
i
i
_ i
LR-15D
Total PCB (mg/kg)
20
i
30
40
50
I.R-I5D Be-7
LR-1 SO Cesium
LR-15D Phase 2 Total PCBs
LR ISO DEC Total PCBs
100
_T
200
~r
300
Be7 or Cs1" (pCi/kg)
60
400
70
Tl
500
-------
LR-I6A
O
¦
to
o
v©
v©
*
£
©
n
o
69
S
a.
SO
oe
4^.
z
*
C/3
o
M
n
n
o
¦n
o
o
n
o
3
-a
v>
O
s
fie
9
a.
"0
o
e
c
j:
10
20
a. 30
40
50
60
10
20
p 30
40
50
60
Total PC[5 (mg/kg)
10 15
20
•'
• LR-I6A Bc-7
~ I.R-I6A t'csruin
I I.R-I6A Phase 2 Tolil PCBs
200 400 600 800
Be7 or Cs117 (pCi/kg)
LR-I6C
Total FCB (mg/kg)
~
LR-I6C Bc-7
I.R-16C Cesium
I.RI6C Phase 2 I'olil PCBs
25
30
~'
1000 1200
10
12
¦...
200
400
600
" 800
1000
Be' or Cs"' (pCi/kg)
10
LR-16B
Total PCB (mg/kg)
20 30 40 50
10
20
a. 30
v
Q
40
50
60
30
60 70
J'
• IRI6BHc7
~ I.R-I6B Cesium
I K I6B Phase 2 Total PCBs
500 1000 1500 2000
Be7 or Cs117 (pCi/kg)
LR-16D
Total FCB (mg/kg)
25(H)
0
5
10
-r 15
c
B. 20
10
25
30
35
40
0
15 20 25
1
30 35
.L.#
• LR-I6D Bc-7
~ I R-I6D Cesium
I.R-I6D Phase 2 Tolal PCBs
I.R-I6D Di e: Toul PCBs
500
1000 TsOO 2000~" 2500 3000
Be7 or Cs'" (pCi/kg)
-------
• ~
w X —
(•ui) mdsa
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the Thompson Island Pool
C-21
-------
£
VI
O
n
o
3
as
s
CL
oe
4k.
C/2
O
W
O
o
o
3
o
9
55*
t«
5»
¦t
B*
rt
O
3
•o
M
o
9
»
9
l
O
10
15
E 20
V
O
25 f
30
15
40
10
- 15
c
f 20
3
25
.10
15
40
LR-17A
Total PCB (mg/kg)
20
40
60
80
-_i >
100
120
• IR-I7A Be-'
~ 1 R I 7A CcMuni
I.R-17A Phase ? 1 <>lal I'lHs
I.R-I7A OIT lolall'Clls
100 200 300 400 500 600 700
Be7 or Cs'17 (pCi/kg)
LR-I7C
Total PCB (mg'kg)
10
15
20
t,Ri7CBe-7
I.R-17C Cesium
I.R-I7C Phase 2 Total PCBs
200
400
600
800
Be7 or Cs"' (pCi/kg)
LR-I7B
Total PCB (mg/kg)
0 20
40
10
15
20
25
30
35
40
60
80
10
15
20 t
25
30
.15
40
100
. .1. ,
120 MO
• t R-17B Be-7
~ IR-17B Cesium
| I.R-I7R Phase 2 Total PCRs
1 l.K-l7RI)i:C Total PCBs
20
1000 2000 3000 4000 5000
Be7 or Cs117 (pCi/kg)
LR-I7D
Total PCB (mg/kg)
40 60 80
100
120
• LR-I7D Be-7
~ I.R-170 Cesium
I R-171) Phase 2 Total PC Bs
500
1000
157 ,
1500
2000
Be or Cs (pCi/kg)
-------
( UI) qidaa
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the Thompson Island Pool
C-23
-------
LR-18A
Total PCB (mg/kg)
50
ino
150
200 250
* '• '0 '
• l,R-l*ABe?
~ I.R-ISA Cesium
l.K-ISA Phase 2 Total PCBs
I R- ISA Dl ( lotal PCIh
2000 4000 6000 8000 I !04
Be7 or Os"7 (pCi/kg)
LR-18C
Total PCB (mg/kg)
[tl 20 30 4(1 50 60 70 80
f
• l.lt-IICBc-7
~ l.R-ISC Cesium
t.R.ISC Phase Dotal PCBs
500
HKK)
1500
2000
¦5
5.
5
0
5
10
15
20 i
25 -
10
35
40
LR-18B
Total PCB (mg/kg)
0 100 200 300 400 500 600 700
.. I ... ^ I ...» I ....... . ._.l . .
0
¦5
s.
0
5
10
15
20
25
30
35
40
• I.K-ISBBc-7
~ I.R-18B Cesium
I.R I SB Phase 2 lotalPCHs
2(K) 400 600 800 1000 1200 1400
Be7 or Cs117 (pCi/kg)
LR-I8D
Total PCB (mg/kg)
2(H) 400 600 800 1000
Re7 or Cs'" (pCi/kg)
• I.R-IIDBc-7
~ IR-ISDCesium
I R-I8D Phase 2 Total PCBs
500 1000 1500
Be7 or Cs"7 (pCi/kg)
2000
-------
fl
r o ^ ©
— — rs fN m -*T
( ui) qidsQ
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the Thompson Island Pool
C-25
-------
( UI) qidaci
1994 Low Resolution Core and 1984 NYSDEC Core Profiles for the Thompson Island Pool
C-26
-------
£
H
APPENDIX D
-------
APPENDIX D
1994 Low Resolution Core Profiles
Below The Thompson Island Pool
-------
Table D-l
Assignment of Low Resolution Cores to Hot Spot Areas
Hat Spot1
COF
IES2
25
LH-25A
LH-25B
LH-25C
LH-25D
LH-25E
LH-25G
LH-25H
LH-251
LH-25J
28
LH-28C
LH-28D
LH-28E
LH-28F
LH-28H
LH-28I
LH-28J
LH-28K
LH-28M
LH-28N
31
LH-3ID
LH-31E
LH-3IF
LH-31G
LH-3II
34
LH-34B
LH-34C
LH-34E
LH-34F
LH-34H
LH-34I
LH-34J
LH-34K
LH-34M
35
LH-35A
LH-35B
LH-35C
LH-35D
37
LH-37A
LH-37B
LH-37C
LH-37D
LH-37E
LH-37G
LH-37H
LH-37J
LH-37K.
LH-37M
LH-37N
LH-370
39
LH-39A
LH-39B
LH-39D
LH-39E
LH-39F
LH-39G
LH-39H
LH-39I
LH-39J
LH-39K
LH-39L
LH-39M
LH-39N
LH-39Q
DL 182
LH-42C
LH-42D
Notes:
1 Hot spot numbers are as assigned by Tofllemire and Quinn (1979) DL 182 represents dredge location 182 from MPI (1992).
2 The cores listed were located within the dredge location boundaries defined by Malcome Pirnie (MPI. 1992).
Typically, hot spots as defined by TotTlemire and Quinn (1979) are represented by I to 4 of these dredge locations.
-------
LH-25A
Total PCB (mg/kg)
4
6
"L-^-
10
20 -I
30
40
50 -J-
• LH-25A Be-7
~ LH-25A Cesium
j LH-25A Total PCBs
500 1000 1500
Be7 or Cs'37 (pCi/kg)
10
20
LH-25C
Total PCB (mg/kg)
30 -
40
50
• LH-2SC Be-7
~ LH-25C Cesium
j LH-25C Total PCBs
200
4oo
600
800
10
c 20
«C
D.
&
30
40 -
50 4
LH-25B
Total PCB (mg/kg)
200
8 10 12 14 16
-»
~
LH-25B Be-7
LH-25B Cesium
LH-25B Total PCBs
400 600 800
Be7 or Cs137 (pCi/kg)
1000
1200
10
c 20
I
Q
30 -
40
50
LH-25D
Total PCB (mg/kg)
12 3 4
• LH-25D Be-7
~ LH-25D Cesium
LII-25D Total PCBs
100 200 300 400
_ 1 _ 137 , .
500
600
-------
20
-4_J_
LH-25E
Total PCB (mg/kg)
40 60 80 100
• LH-25E Be-7
~ LH-25E Cesium
| LH-25E Total PCBs
200 400 600 800
Be7 or Cs'" (pCi/kg)
120
140
1000 1200
0 0,5
o -A—^
10
b 20
30
40
50
LH-25F
Total PCB (mg/kg)
I 1.5 2 2.5
• LH-2SF Be-7
~ LH-2SF Cesium
| LH-2SF Tola! PCBs
3 3.5
100 200 300 400 500
Be7 or Cs137 (pCi/kg)
600 700
20
j 1
LH-25G
Total PCB (mg/kg)
40
60
, L_
• LH-25G Be-7
~ LH-25G Cesium
| LH-250 Total PCBs
200
, f r—r—t r-~r-
400 600
Be7 or Cs1" (pCi/kg)
80
-w-
100
800
1000
10
£ 20
i
30
40
50
100
LH-25H
Total PCB (mg/kg)
200 300 , 400
500
• LH-25H Be-7
~ LH-25H Cesium
| LH-25H ToUl PCBs
~ff
loo
600
4oo
Be7 or Cs'" (pCi/kg)
"J5T
1000
-------
V0
SO
*
%
8
o
e
9
n
o
3
3
S)
5
o*
2.
o"
10
c 20
¦S
Q.
30
40
50
10
LH-2SI
Total PCB (mg/kg)
20
30
40
• LH-251 Bc-7
~ l.H-251 Cesium
I ¦ LH-251 Toul PCBj
50
60
200 400 600
Be7 or Cs"7 (pCi/kg)
800 1000 1200
»
H
Br
o
3
¦o
w
O
B
B
Q.
s
LH-25J
Total PCB (mg/kg)
50 100 150 200 250 300
» -1,—y
• LH-2U Be-7
~ LH-2SJ Cesium
I LH-28J Toul PCBs
200 400 &W 800 1000 1200 1400 I6<
Be7 or Cs"7 (pCi/kg)
-------
0
1
s
w
o,
5"
o*
3
n
e
¦*
5
3
s
I
»¦*
er
«
H
B*
e
S
¦e
«¦
e
s
»
a
a
§
x;
Q.
LH-28B
Total PCB (mg/kg)
10 15 20
c 20
30
40
50
10
c 20
25
_^i_.
30
30
40
50
~
1.H-2II) Be-7
LH-28II Cesium
LH-28H Toul PCBs
100 200 300 400 500 600 700 800
Be7 or Cs"7 (pCi/kg)
LH-28D
Total PCB (mg/kg)
10
15
20
25
• L1I-2ID Be?
~ LH-2JD Cesium
LH-28D ToUl PCBs
200
400
600
800
Be7 or Cs"7 (pCi/kg)
0 100
20
30
40
50
10
20
30
40
50
200
LH-28C
Total PCB (mg/kg)
200 300 400
500 600
~
LH-28C Be-7
LH-2IC Cesium
LH-2IC Toul PCBs
500 1000 1500
Be7 or Cs137 (pCi/kg)
LH-28E
Total PCB (mg/kg)
2000
400
600 800 1000 1200 1400
~
LH-21E Be-7
LH-21E Cesium
LH-2«r: ToUl PCBs
200
400
~T
600
800
1000
Be7 or Cs(pCi/kg)
-------
0
10
20
30
40
50
0
10
20
30
40
50
LH-28F
Total PCB (mg/kg)
200
400
600
• LH-28F Be-7
~ I.H-28F Cesium
I I.H-28F Total PCBs
800
1000
500 1000 1500
Be7 or Cs1" (pCi/kg)
2000
10 -
c 20
t
<3
30 -
40
50
20
LH-28G
Total PCB (mg/kg)
40
60
80
~
100
LH-28G Be-7
I.H-28G Cesium
I.H-28CJ Toul PCBs
100 200 300 400 500
Be7 or Cs"7 (pCi/kg)
300
120
140
~ •
600
700
0 100
*' -
LH28H
Total PCB (mg/kg)
200
300 400
J-*.
500
. . . i .
600 700
• LH-2IH Be-7
~ LH-28H Cesium
I LII-28H Tool PCBs
500
1000
1500
2000
Be7 or Cs"7 (pCi/kg)
30 -
40
50
LH-28I
Total PCB (mg/kg)
• LH-28I Be-7
~ LH-281 Cesium
| LH-281 Total PCBs
200 400 &X)
Be7 or Cs"7 (pCi/kg)
800
1000
-------
50
LH-28J
Total PCB (mg/kg)
100 150 200
250
300
~
LH-211 Bc-7
LH-2IJ Cesium
LH-28J Total PCBs
200 400 600 800 1000 1200 1400 1600
Be7 or Cs"7 (pCi/kg)
LH-28L
Total PCB (mg/kg)
50
100
150
200
• I.H-2IL Be-7
~ I.II-2IL Cesium
I LH-2SL Total PCBs
loo
400
600
Be7 or Cs137 (pCi/kg)
800
1000
10 -
e 20
o.
<3
30
40
50 -
10
20
o. 30 —
40
50 -
60
LH-28K
Total PCB (mg/kg)
50
100
150
t_.
200
250
• LH-28K Bc-7
~ LH-2IK Cesium
I I.H-2SK Total PCBs
200
400
£»
800
• LH-2IM Bc-7
~ LH-2SM Cesium
I LH-2SM Toul PCBs
1000 1200
Be7 or Cs137 (pCi/kg)
LH-28M
Total PCB (mg/kg)
100 200 300 400 • 500 600
700
200 400 600 800 1000 1200
Be7 or Cs137 (pCi/kg)
14
-------
r
11 8j
«"? -E ® 1
X « 1
J a 8-)
— mi
a
(2
m
E £
•7 1 2
as £
z z z
•e «e ae
r« fN
X £ ±
' • ~
00
0
Q.
-------
0
10
20
30
40
50
60
0
10
20
30
40
50
60
LH-31A
Total PCB (mg/kg)
6
10
12 14
~
LH-3IA Rc-7
LH-31A Cesium
I.H-3IA Tolal PCBs
500 1000
Be7 or Cs1,7(pCi/kg)
LH-31D
Total PCB (mg/kg)
1500
10
~
LH-34D Be-7
LH-34D Cesium
LH-34D Toitl PCBs
100 200 300 400
Be7 or Cs"7 (pCi/kg)
500 600
0 SO
10
20
C
"I. 30
a
40
50
60
LH-31B
Total PCB (mg/kg)
100 ISO 200
, i ¦ x , i
250 300
F
• LH-3 IB Be-7
~ LH-3IB Ceiium
LII-3IB Toitl PCBs
0 ^200 ^400 600 ^800 7000 1200 1400 1600
Be7 or Cs137 (pCi/kg)
LH31E
Total PCB (mg/kg)
50
.100
10
20
J
H 30
&
40
50
60
150
~
LH-JIEBe-7
LH-3 IE Cesium
LH-3IEToUl PCBs
500 1000 1500
Be7 or Cs1" (pCi/kg)
2<
-------
10
20 -
30
LH-3IF
Toul PCB (mg/kg)
40
50
60
• LH-3 If Be-7
~ LH-3 IF Cesium
LH-3IK Total PCBs
200 400 600
Be7 or Cs"7 (pCi/kg)
800
0 05
LH-31H
Total PCB (mg/kg)
I 15
10
20
30
40
50
60
• LH-31H Be-7
~ LH-31H Cesium
I LH-31M Totil PCBs
loo
400 600 800
Be7 or Cs137 (pCi/kg)
1000
10
20
H 30
&
40
50 -
60
LH-31G
Toul PCB (mg/kg)
10 20 30
40
50
... . i
~
LH-3IG Be-7
LH-31G Cesium
LH-31G Total PCBs
60 70
500 1000 1500 2000 2500 3000 3500
Be7 or Cs"7 (pCi/kg)
1000
10
20 -
H 30 -|
40
50
60 -4
LH-3 II
Total PCB (mg/kg)
10
¦ ' ¦
15 20
• LH-341 Be-7
~ LH-341 Cesium
I LH-341 Toul PCBs
25
30
100 200 300 400 500 600 700
Be7 or Cs"7 (pCi/kg)
-------
4
•?
S ffi
mi U
J
>_
~
a.
15
©
*
CD
k.
m
L'
r-
E
3
a-
&
U
w
<¦">
t
X
3
zz
r
• ~
06
U
u
©
m
o
CN
o
o
O
v©
(ui) H»d3Q
1994 Low Resolution Core Profiles below the Thompson Island Pool
D-10
-------
ve
4k.
*
$
e
O
o
*v
3
s>
I 3
o"
*
sr
H
13
M
e
a
at
a
a
J
©
-—J
• LH-34E Be-7
~ LH-34E Cesium
I I.H-34E Toul PCBl
100
200
loo'
Tm
Be7 or Cs'" (pCi/kg)
50
+ 91
60
500
600
-------
0
1
to
SO
ve
a.
<
8?
s
s.
£T
n
e
e
3
c
c
5"
20 -
30 -
40
50
60
LH-34F
Total PCB (mg/kg)
0.2
^ L_
0.4
06
08
i
• LH-J4F Be-7
~ LH-J4F Cesium
LH-34F ToUl PCB)
50 100 150 200 250 300 350 400
Be7 or Cs'37 (pCi/kg)
%
H
:r
e
m
©
a»
o
a.
*0
©
©
c
x:
0
0 #
10
20
30
40
50
60
LH-34H
Total PCB (mg/kg)
• LH-MHBe-?
~ LH-34H Cesium
I LH-J4H ToUl PCBj
-I—¦"
200
400
i—¦-
600
10
800~
1000
Be7 or Cs"7 (pCi/kg)
c
10
20
30
40
50
60
LH-34G
Total PCB (mg/kg)
• LH 34C Be-7
~ LH-J4G Cesium
I LH-34G ToUl PCB«
500 1000 1500
Be7 or Cs137 (pCi/kg)
2000 2500
10
20
f 30
40
50
60
LH-341
Toul PCB (mg/kg)
10 1$ 20
25 30
• W
• LH-341 Be-7
~ LH-341 Cesium
I LH-341 ToUl PCBs
—T""r~T'* l""- •¦¦¦* 1
100 200 300 400 500 600 700
Be7 or Cs"7 (pCi/kg)
-------
0
10
20
30
40
50
60
0
10
20
30
40
50
60
LH-34J
Total PCB (rng/kg)
50 100 150 200 250 300 350
1
• LH-34J Be-7
~ LH-34J Cesium
l.H-141 Tout K'Bs
0 500 1000 1500 2000 2500 3000 3500 4000
Be7 or Cs117 (pCi/kg)
10
20
30
40
50 -
60
LH-34K
Total PCB (mg/kg)
10 15 20 25
• LH-34K Bc-7
~ LH-34K Cesium
LH-34K Tolil FCBi
!—¦-
200
400
600
800
Be7 or Cs"7 (pCi/kg)
30
35
1000 1200
10
LH-34L
Total PCB (mg/kg)
20
30
40
50
60
• LH-341. Be 7
~ LH-341, Cesium
LH-341 Toul PCBs
100 200 Too 400 500 600 700
Be7 or Cs"7 (pCi/kg)
n
0
0 ^
io -
.£
_q
20
30
40 -
50
60
LH-34M
Total PCB (mg/kg)
2 3 4 5
• LH-34M Be-7
~ LH-34M Cesium
I LH-34M Toul PCB«
200
400
<>00
Be7 or Cs'" (pCi/kg)
6 7
800
1000
-------
50
LH-35A
Total PCB (mg/kg)
100 150 200
• Bc-7 LII-35A
~ Cesium LH-35A
I Tout PCBs I.H-35A
250
300
500 1000 1500
Be7 or Cs137 (pCi/kg)
2000
LH-35C
Total PCB (mg/kg)
10 15 20
200
25
*'
• Bc-7 LH-3SC
~ Cesium LH-35C
I Tolil Pt Bs LH-35C
600
80cT
400
Be7 or Cs137 (pCi/kg)
1000 1200
10
20
30
40
50
LH-35B
Total PCB (mg/kg)
20
40
60
Be-7 LH-35B
Cesium LH-35B
ToulPCBs LH-35B
200 400 600 800
Be7 or Cs137 (pCi/kg)
SO
i_
100
1000 1200
10
20 -
30 -
40
50 4
LH-35D
Total PCB (mg/kg)
5 10 15 20
L-j
15
.... i.
~
Bc-7 LH-33D
Cesium LH-35D
Toul PCBs LH-33D
200 400 600 800
Be7 or Cs137 (pCi/kg)
1000 1200
-------
V^-4
O J
~ >»J
, J? ° \
& "Si i
VI °P
r> e
X
J
m
U
CL
(2
U*
•r\
tu r»t
>r\ .JL
2 g
r- 2
<2 8
!• ~
£
si
oo
u
L.
o
*ii
GO
O
( ui) qidaa
1994 Low Resolution Core Profiles below the Thompson Island Pool
D-15
-------
0
1
o\
SO
*
t
©
s
O
e
3
*v
3
3
5*
w
W
2,
e
*
ar
w
e
SP
a
O.
*s
o
o
LH-37A
Total PCB (mg/kg)
5 10
15
..L — ^
20
25
10
.£ 20
f
30
40
50
• Be-7 LH-J7A
~ Cesium LH-37A
I Total PCBs IH-37A
200
loo
600
Be'or Cs'37 (pCi/kg)
LH-37C
Total PCB (mg/kg)
5 10
10
c 20
30
40
50
• Be-7 LH-37C
~ Cesium LH-37C
I Toul PCBs I.H-37C
200
400
600
Be7 or Cs'" (pCi/kg)
" 800
1000 1200
15
—J r-
800
1000
10
LH-37B
Total PCB (mg/kg)
4 6 8 10
12
14
20
30
40
50
• Be-7 LH-37B
~ Cesium LH-37B
Toul PCBs LH-37B
200
400
600
Be7 or Cs"7 (pCi/kg)
800
1000
10
LH-37D
Total PCB (mg/kg)
10
15
i_
20
25
20
30
40
50
• Be-7 LI1-37D
~ Cesium LH-37D
I Toul PCBs LH-37D
500 1000 1500 2000
Be7 or Cs"7 (pCi/kg)
2500
-------
LH-37E
Total PCB (mg/kg)
5 10
... ... .
~
• Be-7 LH-J7E
~ Cesium LH-37B
| Tolal PC'Bs 1.II-37E
100 200 ~300 ^ 400 500 600
Be7 or Cs137 (pCi/kg)
LH-37J
Total PCB (mg/kg)
20 40 60 go 100 120 140
i—.1 ' ' f"
• Bc-7 I.H-37J
~ Cesium UI-37J
| Total PC'Bs I.H-37J
200 400 600 800
Be7 or Cs157 (pCi/kg)
10
c 20
•5
o.
(5
30 -
40 -
50 4-
0 5
LH-37F
Total PCB (mg/kg)
1 L
2
25
Bc-7 LH-37F
Cesium LH-37F
Total PCBs LH-37F
3 5
200 400 600 800 1000 1200 1400
Be7 or Cs137 (pCi/kg)
10
c 20
•5
i
30
40
50
LH-37K
Total PCB (mg/kg)
15
10
i—L ...^
• Be-7 LH-37K
~ Cesium LH-37K
I Total PCBi LH-37K
200 400~ 600 800
Be7 or Cs"7 (pCi/kg)
20
25
1000 1200
-------
so
SO
r
W
ft
w
e
©
B
n
e
¦1
3
3
cr
JL
e
H
ar
o
a
*o
M
O
8
te
a
Q.
*e
S
10
c 20
8-
Q
30
40
50
10
c 20
f
30
40
50
LH-37L
Total PCB (mg/kg)
6
_ j_„.
• Be-7 UI-37L
~ Cesium LH-37L
Toul PCBs LH37L
100
300 400 500
200
Be7 or Cs"' (pCi/kg)
LH-370
Total PCB (mg/kg)
400 600
800
Be-7 LH-J70
Cesium LH-370
Toul PCBs LH-370
1000
500
Be7 or Cs157 (pCi/kg)
1500
10
c 20
30
40
50
LH-37N
Total PCB (mg/kg)
2 4 6
• Be-7 I.H-37N
~ Cesium LH-J7N
I Tottl PCBi LH-37N
10
100 200 300 400 500 600 700
Be7 or Cs137 (pCi/kg)
-------
0
10
20
30
40
50
0
10
20
30
40
50
LH-39A
Total PCB (mg/kg)
10 15 20 25 30
loo
20
#•
35
40
~
LH-39A Be-7
LH-J9A Cesium
LH-)«A ruul PC'Bs
1 r-
400
600
Be7 or Cs,n (pCi/kg)
800^
1000
10
2 20
.C
O.
a
30
40
50
LH-39B
ToUl PCB (mg/kg)
02
04
06
i-
08
• LH-39B Bc-7
~ LH-39B Cesium
LH-39B Toul PCB«
100 200 300
Bc'orCs'37 (pCi/kg)
i ¦ ¦
400
12
loo
600
LH-39C
Total PCB (mg/kg)
5 10
l-
• LH19C Be-7
~ Ul 39C Cesium
I UI--M Toul PCBs
500
15
1000
1500
2000
Be7 or Cs'" (pCi/kg)
LH-39D
Total PCB (mg/kg)
0 20 40 60 80 100 120 140 160
10
2 20
i
30
40
50
LH-39D Be-7
I.H-J9DCe»ium
LH-39DToul PCBj
1000 1500 2000 2500 3000 3500
Be7 or Csm (pCi/kg)
-------
s
ST
*
po
e
a
O
e
¦n
re
•-d
cr
2.
o
*
m-
cr
rt>
o
B
e
a
sr
w
B
a.
3
o
10
e 20
j=
o.
&
30
40
50
10
c 20
-c
S.
5
30
40
50
20
LH-39E
Total PCB (mg/kg)
40 60
80
100
• LH-39E Be ?
~ LH-39E Cesium
I LH-39H Toial PCBi
100 200 300 400
Be' or Cs137 (pCi/kg)
LH-39G
Total PCB (mg/kg)
500
600
~
LH-WG Be-7
LH-39G Cesium
UI-39G Tout PCBi
200 400 "&0 ' 80o" H)00 1200 1400
Be7 or Cs'17 (pCi/kg)
50
LH-39F
Tola! PCB (mg/kg)
150
-* t-' **~r-
100
—
• LH-39F Be-7
~ LH-39F Cesium
I LH-39F Tool PCBs
200
250
200 400 600 800 1000 1200 1400
Be7 or Cs15'(pCi/kg)
0 50
LH-39H
Total PCB (mg/kg)
100 150 200
~
LH-39H Be-7
LH-39H Cesium
LH-39II Toul PCBl
1000 2000 3000
Be7 or Cs'" (pCi/kg)
250
t I... *
300
4000
5C
-------
LH-391
Total PCB (mg/kg)
20
. i .
40
60
140
• LH-391 B«7
~ LH 391 Cesium
I LII-391 Toul PCBs
~
i00^ ' 4Oo ' 600 800 1000 ^ 1200
Be7 or Cs117 (pCi/kg)
LH-39K
Total PCB (mg/kg)
10 20 30 40 50 60 70 80
.. -I—• - 1 1 ' -
• LH-39K Bc-7
~ LH-39K Cesium
I LH-39K Total PCBs
500 1000 1500 2000
Be7 or Cs'" (pCi/kg)
10
20 -
30
40
50
LH-39J
Tot«l PCB (mg/kg)
50 100 150
500
• LM-39J B«-7
~ LH-391 Cesium
I U1-39J Total PCBl
1000 1500 2000
Be7 or Cs157 (pCi/kg)
200
2500 3000
10 -
20 -
30
40 -<~
50
20
> I
LH-39L
Total PCB (mg/kg)
40
, 1
60
•_!
80
. 1
100
• LH-391 Be-?
~ LH-391 Cesium
I LH-391 Total PCBs
120
100 200 300 400 500 600 700 800
Be7 or Cs117 (pCi/kg)
-------
(
D
i
ro
to
Id
o
o
to
V0
•u
$
*
£
©
n
o
3
-a
3
n
w
2.
o"
*
H
ET
e
3
¦o
OB
e
to
s
o.
§
•5
a
.<5
LH-39M
Total PCB (mg/kg)
5
10
15 20 25 30 35
4. 4. L... j..i ^ L i ^ —i—i—•- «¦
40
10
c 20
30
40
50
.5 20
¦C
I
30
40
50
• LH-39M Bc-7
~ LH-39M Cesium
I I.H-39M Total PCBs
w
200
300
400
500
600
Be7 or Cs"7 (pCi/kg)
LH-390
Total PCB (mg/kg)
05
I J
•... 1—
• l.H-390 B«-7
~ LH-390 Cesium
I l.H-390 Tolal PCBs
5(T "T5T 150 200
Be7 or Cs'n (pCi/kg)
250
300
50 4
LH-39N
Total PCB (mg/kg)
5 10 15
200 400 ^00
Be7 or Cs137 (pCi/kg)
800
-------
(
u>
o
o
K>
VO
U1
\o
73
n
«
o
n
o
53
>v
n
o
SI
a 5"
K>
oj cr
SL
5"
*
ro
H
BT
e
3
"O
w
e
s
IB
B
a.
2
10
2 20
_c
D.
30
40
50
•s
o.
a
10
c 20
30
40
50
0 1
LH-41A
Total PCB (mg/kg)
0 2 0 3 0 4
L .
05
... i_
06
~
LH-4IA Be-7
LH-4IA Cesium
LH-4IA Vols I PCB*
05
500 1000 1500
Be7 or Cs"7 (pCi/kg)
LH-41C
Total PCB (mg/kg)
I 15 2
2000
25
LH-4ICB«-7
LIMIC Cesium
LH-4IC Toial PCBj
200 ^400 600 800
Be7 or Cs157 (pCi/kg)
1000 1200
10
I 20
JZ
O-
&
J
30 -
40
50 -4-
20
i
LH-41B
Total PCB (mg/kg)
40 60
• LH-4IB Be-7
~ I.H-4IB Cesium
I LH-4IB Tout PCBs
80
100
200 400 600 800 1000
Be7 or Cs137 (pCi/kg)
|T200
1400
-------
to
-fc.
*
a?
2.
S*
»»
o*
B
n
e
3
2
o*
&
I
H
ST
e
9
¦8
o
o
er
a
D.
*0
a
©
o +
10
LH-42A
Total PCB (mg/kg)
100 200 300 400 500 600
700
c 20
t
5
30 -
40
50
• LH-42A Be-7
~ LH-42A Cesium
LH-42A lotalPCBs
500 1000 1500 2000
Be7orCsn7(pCi/kg)
LH-42C
Total PCB (mg/kg)
2500 ' 3000
10
10
c 20
I
a
30
40
50
IS 20
. . - - J-.
25
jX.
30 35
• LH-42C Be-7
~ LH-42C Cesium
I LH42C Total PCBs
loo
"355
600
w
1000
Be? or Cs"7 (pCi/kg)
t
&
10
20
30
40
50
50
¦ I ¦
LH-42B
Total PCB (mg/kg)
100 150 200
• UM2B Be-7
~ LH-42B Cesium
LH-42B Total PCBs
250
—Hr
300
0 200 400 600 800 1000 1200
Be7 or Cs137 (pCi/kg)
10
.£ 20
I
a
30
40
50
LH-42D
Total PCB (mg/kg)
5 10 15 20 25
~
LH-42D Bc-7
LH-42D Cesium
LH-42D Toul PCBl
100 200 300 400 500 600 700 800
Be7 or Cs"7 (pCi/kg)
-------
K>
<_/i
NO
v©
*
£
O
s
n
e
¦n
•-0
o
3
3
£
o"
*
BT
H
o
3
TS
M
e
a
»
B
a
*d
§
10
c 20
•5
Q.
-------
LH-44A
Total PCB (mg/kg)
10
15
20
10
5"
*
*>
(A
o
c 20
.g
H.
&
30
n
o
•1
!W
40
• LH-44A Be ?
~ LH-44A Cesium
I LH 44A Tnul PCBs
e
St
o*
»
o"
0+
tT
n
50
0 *-
200
400
600
0,5
Be7 or Csll7(pCi/kg)
LH-44C
Total PCB (mg/kg)
I 15
800
©
S
10
SB
B
a
o
o
c 20
S
D-
30
40
• LH-44C Be-7
~ LH-44C Cesium
I LH-44C Tolil PCBj
50
"-r_r-
50
100
150
200"
250
10
I 20
t
a
30
40
50
LH-44B
Total PCB (mg/kg)
0.5 I 15
• LH-44B Bc-7
~ LH-44B Cetium
I LH-44B ToUl PCBs
150
50 100
Be7 or Cs"7 (pCi/kg)
200 250
-------
1994 Low Resolution Core Profiles below the Thompson Island Pool
D-27
-------
C/3
32
APPENDIX E
-------
APPENDIX E
Analysis Of 1984 Sediment PCB Quantitation
-------
Analysis of 1984 Sediment PCB Quantitation
Jonathan B. Butcher
Tetra Tech, Inc.
June 19. 1998
Purpose
PCB concentrations reported by NYSDEC for the 1984 Thompson Island Pool sediment survey are
dependent on the Aroclor quantitation methods used and are not equivalent to results which would be
obtained using capillary column GC analysis for PCB congeners. A translation scheme is required to
make these data consistent with Phase 2 congener-based quantitations.
Summary
"Total PCBs" reported for the 1984 sediment data (calculated by NYSDEC as a sum of Aroclors)
provide a good representation of the sum of tri- and higher-chlorinated congeners. They do not
accurately reflect total of all congeners. A linear relationship can be used to correct these data to a
basis consistent with the sum of tri- and higher-chlorinated congeners (ETri + ) in the EPA Phase 2
data.
Introduction
Valid interpretation of historical trends in PCB concentrations cannot be made without consideration of
the changes in analytical methods which have occurred over time. That is, a comparison is valid only
when there is consistency in what is being measured. The most dramatic change in analytical methods
is that between the recent data, using state-of-the-art, capillary-column, PCB congener analyses, and
older analyses based on packed-column quantitation of Aroclor equivalents. Because an Aroclor is a
complex mixture of many individual congeners, interpretation of the older packed-column data raises
difficult technical issues. In addition, packed-column Aroclor quantitation methods have changed over
time, and these changes have significant implications for the interpretation of historical trends in the
data and the development of valid statistical relationships.
Because a commercial PCB mixture consists of many individual congeners, each with its own set of
chemical properties, introduction into the environment quickly changes the original mixture and the
relative proportions of the congeners. Processes such as weathering, dechlorination and biological
accumulation affect the individual congeners to varying degrees. Thus, analytical Aroclor quantitations
on environmental samples are not directly comparable to actual concentrations of PCB congeners.
Results of capillary column analyses do not have a direct interpretation as "Aroclors"; however, total
PCB concentration is readily estimated as the sum of individual congener concentrations.
The 1984 sediment survey (Brown et al.. 1988) represents the most comprehensive database on PCB
concentrations in Thompson Island Pool sediments. It is thus crucial to understand what is reported in
these data and estimate how well the NYSDEC reported total represents actual total PCBs that would
have been calculated by summing congener concentrations.
Analytical quantitations for the 1984 sediment survey were performed by Versar using packed-column
GC and Aroclor standards. Versar reported concentrations of Aroclors 1242, 1254, and 1260. The
chromatogram division flowchart described by Webb and McCall (1973) was used as a guideline to
determine which packed column peaks should be included in these calculations. They did not,
however, use the complete Webb and McCall method, nor did they report concentrations of lighter
E-l
-------
Aroclors.
Like the Webb and McCall (1973) approach, the method used by Versar is an apportionment method:
that is. the packed-column peaks are each assigned to an individual Aroclor. and the concentration of
that Aroclor is then simply the sum of the concentrations represented by those peaks. Versar used
"major'' peaks only, with the result that some degree of underestimation is inevitable for any peaks not
included in the quantitation. Indeed, NYSDEC determined that Versar's Aroclor 1242 estimates were
significantly underestimated, which "highlights the problem associated with omitting peaks from
calculations using the Webb and McCall analyses without correcting for the mass of PCB associated
with ignored peaks" (Brown et al.. 1988, p. 16). There was also concern that Versar had mis-
identified peaks. NYSDEC therefore recalculated Aroclor 1242 using a different method which
consisted of an average of the weighted responses of three packed column peaks. This recalculation is
a scaling, rather than apportionment, method, in which a response factor is used to scale up the peak
concentration to an Aroclor concentration. These re-calculated Aroclor 1242 estimates were summed
with the Aroclor 1254 and Aroclor 1260 Versar quantitations to yield the total PCB estimates reported
by NYSDEC and contained in the TAMS/Gradient database. (It should be noted that the database
reports the original Versar quantitation for Aroclor 1242, and does not directly give the NYSDEC
recalculated quantitation. The recalculated Aroclor 1242 estimate can. however, be retrieved by
subtracting the Aroclor 1254 plus 1260 concentrations from the reported Total PCB concentration.)
Because there is overlap between the congener composition of Aroclors 1242 and 1254, use of a
response factor scaling method for Aroclor 1242 can result in double-counting of congeners which
appear in both Aroclor 1242 and the packed-column quantitation peaks used for Aroclor 1254. The
original reapportionment method, which used major peaks only, is likely to underestimate PCB
concentrations. Finally, it is known that significant dechlorination has occurred in Thompson Island
Pool sediments, resulting in elevated concentrations of monochloro- and dichlorobiphenyls.
Methods
Performance of the 1984 quantitation scheme was investigated by performing "as if" numerical
experiments on congener quantitations from the Phase 2 High Resolution Core data. This consists of
interpreting the congener data "as if" they had been analyzed by the packed column methods used by
NYSDEC and comparing the results to the actual sum of congeners.
As noted above, Versar employed a Webb and McCall-type method for Aroclors 1254 and 1260. In
this approach, multiple packed-column peaks are used to estimate an Aroclor concentration. Each
packed-column peak is used to estimate the concentration of PCBs associated with that peak. The
concentrations of PCBs associated with m packed-column peaks are then summed to arrive at an
estimate of the total Aroclor concentration:
m
[Aroclor] - Area/ • RF (1)
i i
where RFp] is a response factor for the packed column peak. Versar did not use any factors to correct
for the fact that an Aroclor may not be completely represented by the selected peaks. In this approach,
an Aroclor concentration estimate is equal to the sum of concentrations of the n PCB congeners
associated with each of the m the packed column peaks:
-------
[Aroclor] EE [congener] (2)
, 1 . i ''
The NYSDEC Aroclor 1242 re-quantitations used an average of three quantitations based on responses
to single packed column peaks. Each individual estimate is obtained based on a response factor relating
the peak concentration to an Aroclor standard:
[Aroclor] = Area • RF s (3)
where Area! is the area associated with packed-column peak j and RF} is a response factor defined as
the concentration of standard Aroclor injected divided by the area of the selected packed-column peak.
Area of a packed column peak is equivalent to the concentration of individual congeners in that peak
divided by a packed-column response factor, defined as the concentration of standard Aroclor injected
multiplied by the weight percent of PCBs in the packed-column peak and divided by the area of the
selected packed-column peak. By definition, the ratio of the packed-column response factor to the
Aroclor response factor is equal to the weight percent of the PCBs in the packed column peak. An
equivalent estimate from congener data obtained from packed column peak j is approximately (Butcher.
1997):
n
i, [congener]^
[Aroclor] ~ —
1 wt % peakj
Note that this interpretation is not technically exact, as it does not take into account variability in
response factors among congeners within a packed-column peak. This does not. however, appear to
introduce significant bias (Butcher. 1997). The final estimate for Aroclor 1242 is then obtained as the
average over m peaks:
n
/ \conzener]
[Aroclor] ~ —
m ! i wt % peakj
To equate congener-specific analyses with packed-column data, information on the congeners
represented in packed-column peaks is required. Because the absolute retention time of a packed-
column peak may vary, many researchers adopted the convention of reporting retention times relative
to the retention time of a standard compound. For example. Webb and McCall (1973) reported
retention times relative to the retention time of p,p'-DDE. In this discussion, all packed-column peaks
are referred to by their retention time relative to p.p'-DDE, and individual PCB congeners are referred
to by their BZ numbers defined by Ballschmitter and Zell (1980). The packed-column peaks used for
quantitation and congeners associated with these peaks (Brown et al., 1984; Gauthier. 1994) are shown
in Table 1. Table 1 also shows the associated weight percents of congeners contained in a given RRT
peak in the April 1994 Aquatec analyses of Aroclor standards.
E-3
-------
Table 1. Quantitation Peaks and Congeners
Aroclor
RRT Peak
Associated Congeners (BZ #)
Weight Percent
1242
.28
15.17.18
13.9
.47
47.48.49.52.75
8.7
.58
41.64,72
3.5
1254
.98
85.87.97,119,136
8.6
1.04
77.110
10.4
1.25
82,107.118,135.144,149,151
14.3
1.46
105,/J2,146.153
7.6
1.60
130,137.141,165,176,179
12.7
1.74
129,138.158.175.178
8.6
1260
2.03
128.167.183,185.187
9.1
2.32
171.172,/7i,174,177,202
10.0
2.44
156.157.200
0.6
2.80
180.191.193
11.7
3.32
170.190
4.8
3.72
189.196.198.199.201.203
4.7
4.48
195.208
1.0
5.28
194.206
2.5
Note: congeners shown in italics do not have useable data in the Phase 2 Database.
Data
The analysis is based on the Phase II High Resolution Core data, using samples indicated as mainsteni
upper river and lower freshwater in the database (Release 3.7b). Both "P" samples and "A" samples
with PCB quantitations were included, yielding 241 sample points. Only the 126 "useable" (target and
nontarget) congeners were included. A total of eight congeners included within the packed-column
quantitation peaks are not available or not useable in the database; these are not. however, believed to
represent significant mass fractions. "Value 2" congener concentrations from the database were used,
which contain specific corrections for non-detects. All "R" rejected data were dropped.
Results
Using the congener data, estimates of reported Aroclor methods "as if" calculated by the 1984 packed
column methods were estimated. Total PCBs "as if" by the 1984 method were reconstituted as the sum
of Aroclors 1242. 1254 and 1260. Total PCBs "as if" calculated by the 1984 NYSDEC method are
E-4
-------
plotted against actual sums of PCB congeners for the High Resolution Core data in Figure 1. From this
plot, it is obvious that the NYSDEC sediment totals represent a consistent and significant underestimate
of the total concentration PCBs which wojU be calculated by summing congener concentrations. For
the higher concentration samples, the congener sums exceed the 1984-style Aroclor sums by a factor of
about 2.5, representing a serious discrepancy.
The reason for this discrepancy is simple: Most of the sediment samples contain a significant
proportion of dechlorination products, particularly BZ#1 (monochlorobiphenyl) and BZ#4
(dichlorobiphenyl). The lowest packed column peak used in the quantitation of NYSDEC totals (with
Aroclor 1242 recalculation) is RRT .28, which contains BZ015, BZ017 and BZ018. The latter two are
trichlorobiphenyls. while BZ#15 is a dichlorobiphenyl. Thus, the NYSDEC sediment quantitations
include only one of the dichlorobiphenyls and none of the monochlorobiphenyls. and will not reflect
any enhancement of concentrations in this range.
This suggests that the 1984 data should provide a better approximation to the sum of tri- through deca-
chlorobiphenyls, designated ETri+ (although a discrepancy may be present because Aroclor 1242 does
contain a small fraction of mono- and dichlorobiphenyls). In Figure 2. the sum of Aroclors estimated
from the High Resolution Core data "as if by the 1984 quantitation methods are plotted against
ETri+. It is obvious that the resulting numbers are in much closer agreement; further, the scatter in
the 1984-method results is substantially reduced, resulting in a nearly linear plot.
Because a linear relationship holds, a regression-based correction is attractive. This yields the
following relationship:
Tri' (ng''fcg) = 376.38 (ng/kg) ' 0.945 • 1984 Aroclor Sum (ng'kg)
with an R2 of 98.3 c/c and a standard error of 13.569 (uglkg). The intercept term is not significantly
different from zero, and a regression forced through zero yields the relationship
^ Tri*¦ (\ig!kg) - 0.944 • 1984 Aroclor Sum (|ig/kg)
The correction factor is expected to be less than 1 because Aroclor 1242 does contain about 14.6%
mono- and dichlorobiphenyls, which are not included in ETri + . The mono- and di-chlorobiphenyls
which do contribute to Aroclor 1242, but are not included in the NYSDEC quantitation scheme (i.e..
all but BZ #15) have a total weight percent contribution of 12.98 % in the April 1994 Aquatec analysis.
The correction factor to a tri- through deca-chlorinated homologue sum that would be expected based
on an accurate quantitation of Aroclor 1242 (but not dechlorination products) is 1/1.1298 = 0.885.
The actual correction factor is slightly higher, and likely reflects a small buildup of trichlorobiphenyl
intermediate degradation products.
E-5
-------
References
Brown. M P.. M B. Werner, C.R Carusone. and 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 Survey. New York State Department of Environmental Conservation. Albany.
NY
Ballschmitter, K. and M. Zell. 1980. Analysis of polychlorinated biphenyls (PCBs) by glass capillary
gas chromatography. Fresenius Z. Anal. Chem., 302: 20-31.
Butcher. J.B. 1997. Use of Historical PCB Aroclor Measurements: Hudson River Fish Data. Environ.
Tox. Chem., 16(8): 1618-1623.
Gauthier, T. 1994. Aroclor Translation Procedures. Memo to Ed Garvey (TAMS/NJ) from Tom
Gauthier, Gradient Corporation. Cambridge, MA.
Webb, R.G. and A C. McCall. 1973. Quantitative PCB standards for electron capture gas
chromatography. J. Chromatogr. Sci., 11: 366-373.
E-6
-------
Figure 1. PCBs in Sediment
Analysis of High Resolution Core Data
0 500000 1000000 1500000 2000000 250000(
Congener Total (ppb)
Figure 2. PCBs in Sediment
Analysis of Tri+ Congeners
0 200000 400000 600000 800000 10000C
Tri - Deca Congener Total (ppb)
E-7
-------
TAMS
Appendix F
-------
APPENDIX F
Statistical Summary Sheets
for
Chapter 4
-------
Log10(Length-Wt'd Avg) By Hotspot X
o>
>
<
•O
O)
c
0)
3.0
2.5
2.0
1 .5
c" 1 0
o 1 u
0.5
Mean
of all
samples
t
Ciroup
Mean
Data Point
Mean • I
Std. Dev!
Inner
Quartile /
DistaneeX,
Mean -1
Slil. Dev.
3*1
95%
Conlidence Interval
About Mean
NYSDI-C
1976-1978
Dala
l.im Resolution
Coring Dala
1994
Percentiles
• 90th
¦ 75th
Median
25th
I (1th
Tukey-kramer
Comparison on all .sample pairs'
a 0.05
I lotspot X
Note:
. Sec Key Diagram 2 lor an explanation of the I ukey-Kramer comparison.
Appendix F - Key Diagram 1
Statistical Summary for Hot Spots Below the TI Dam
-------
Tukey-Kramer Comparison *
angle greater than
90°
angle equal to
90°
angle less than
90°
o
not significantly
different
border line
significantly
different
a
significantly
different
Nolo:
1. In the Tukey-Kramer comparison. the center of each circle is alligned with the mean of the group it represents.
The circle diameter represents the 95% con faience interval about the mean. The outside angle of intersection
tells sou whether group means are significantly different. Circles for means that are significantly different
either do not intersect or intersect slightly so that the outside angle of intersection is less than 90 degrees.
If the circles intersect by an angle of more than 90 degrees or if they are nested, the means are not significantly
different.
Appendix F - Key Diagram 2
Statistical Summary of Hot Spots Below the TI Dam
-------
Statistical Analysis of Delta-M as a Function of 1984 Sediment ZTri+ Inventory
1
-------
Summary of Fit
RSquare
0.191714
RSquare Adj
0.1 77778
Root Mean Square Error
0.243499
Mean of Response
0.317782
Observations (or Sum Wgts)
60
Onewav A nova
t-Test
Estimate
Std Error
Lower 95%
. Upper 95%
^Assuming
Difference
0.247335
0.066685
0.1 13851
0.3808.9
t-Test
3.709
DF
58
Prob>ltl
0.0005
equal variances
Analvsis of Variance
Source
Model
Error
C Total
DF
1
58
59
Sum
of Squares
0.8156611
3.4389167
4.2545778
Mean Square
0.815661
0.059292
0.0721 1 1
F Ratio
13.7568
Prob>F
0.0005
Means for Onewav Anova
Level Number Mean Std Error
<10 g/mA2 20 0.482672 0.05445
>10 g/mA2 40 0.235337 0.03850
Std Error uses a pooled estimate of error variance
Means and Std Deviations
Level
<10 g/mA2
>10 g/mA2
Number
20
40
Mean
0.482672
0.235337
Std Dev
0.346325
0.1 72466
Std
Err Mean
0.07744
0.02727
Means Comparisons
Dif = Mean[i]-Mean[j]
<10 g/mA2
>10 g/mA2
<10 g/mA2
0.000000
-0.24733
>10 g/mA2
0.247335
0.000000
Alphas 0.05
Comparisons for all pairs using Tukey-Kramer HSD
q *
2.00177
Abs(Dif)-LSD <10 g/mA2 >10 g/mA2
<10 g/mA2 -0.15414 0.1 13847
>10 g/mA2 0.1 13847 -0.10899
Positive values show pairs of means that are significantly different.
2
-------
\
Wilcoxon / Kruskai-VVallis Tests (Rank Sums) J
Level Count Score Sum Score Mean
<10 g/mA2 20 794 39.7000
>10 g/mA2 40 1036 25.9000
(Mean-Mean0)/Std0
2.878
-2.878
2-Sample Test, Normal Approximation
S Z Prob>IZI
794 2.87751 0.0040
1-way Test, Chi-Square Approximation
ChiSquare DF Prob>ChiSq
8.3252 1 0.0039
Median Test (Number of Points Above Median)
—
Level Count Score Sum Score Mean
<10 g/mA2 20 15 0.750000
>10 g/mA2 40 15 0.375000
(Mean-Mean0)/Std0
2.716
-2.716
2-Sample Test, Normal Approximation
S Z Prob>IZI
15 2.71570 0.0066
1-way Test, Chi-Square Approximation
ChiSquare DF Prob>ChiSq
7.3750 1 0.0066
3
-------
Log Delta-mol»2 By 1984 Inventory+NVSDECSamp.Type J
1984 irvenory+NvSOECSamc Type
[Quantfles }
Level
minimum
19.0%
2 5 . 0 *4
median
75.0%
90.0%
maximum
Core <,0g/rnA2
0 105405
0 10803
0 246473
C
423296
3
827968
i
086288
i 367977
Core >'0g/mA2
C 0C277'
C 0'295
0 126361
c
2*05 79
0
393918
3
5 ' 6 1 51
0 614 151
Grab 1Q g/nv"2
0 0C2281
0 002511
0 085531
a
179616
0
27&C69
0
442029
0 454789
(Oneway Anova )
(Summary o< Fit
RSouare
RSouare Aoi
Rooi Mean Squa'e t'tot
Mean qt Response
Obse'vations tor Sum Wgts)
0 255173
0 2^5272
0 237882
0 3? 7782
60
(Analysis
Of
Variance }
Source
DF
Sum of Square*
Mean Square
F Ratio
Moaei
3
i 0856546
0 361835
6 3951
ErfO'
56
3 1689231
0 056588
Prob>F
C Total
59
4 2545778
0 0721 1 *
3 :008
(Means for Oneway Anovaj
Level
Core <"Cg'W2
Core >'
G'ab < 'Z^ rv 2
Gfac >*0 * ^ 2
Number
t7
29
Mean
C 527242
0 253761
2 C 284705
'3 C '86853
S!d Error
0 05769
0 04 496
: *682'
: 06598
Sid Erro? uses a ccoieo e$i>mate c4 e^or variance
(Means and Std Deviations )
L#v»l
Core «lGg.'rrA2
Core >i0g/m'2
Grab i0 g.'mA2
Number
• 7
28
2
'3
Mean
0 527242
0 253761
0 264705
0 186853
Std Oev
0 356701
0 *80885
C 0C9196
C «44236
Std Err Mean
: 0865"
0 03418
0 0C650
j04000
Means Comparisons j
OI» = M#«ri(ii-M#«n|j] Cor#,
Cor#. *1i)g/mAa
Grab. 10g?m* 2
Grab, >10 g/m*2
A pha= I Zi
Comparisons for
q *
2 64 794
Abs(Oif)-LSD
Core, t0g/mA2
G.ao. >10 g/m*2
Posi'ive values
i0g/m 2
0 273482
0 030944
0 OOOOCO
•0 06691
Grab. >10 g/m*2
0 340389
C 097852
3 0669C7
0 000000
all pairs using Tukey-Kramer HSD
Core.
t0g/mA2
::798C7
-: 430:9
0 *6835
¦Z 1445
Grab. >10
g/mA2
0831 1
38053
J '445
24707
s of f^ears *ra: a*e s
-------
[wilcoxon / Kruskal-Wallis
Tests
(Rank
Sums) J
Level Count Score
Sum Score Mean
(Mean-Mean0)/Std0
Co-e <*C5-m^2 -7
-2 do:-:
3 199
Co-'e >-C-3.-n*2 28
774
27 6429
• * 1 7S
GraD <'Cg/rrv-S 2
65
33
0 185
GraC >10 g tv*2 13
276
2' 23C9
2 ' £3
'•wav Tesi C"-Square Ascoximat-c
ChiSquare OF Prob>ChiSq
'1S238 3 0 3080
(Median Test (Number of Points Above Median))
Level
Count
Score Sum
Score Mean
(Mean-Mean0)/Std0
Core
i 7
'3
: 76471
2 657
Core
>'Cg.m'2
29
12
0 42957
-1 C26
G'aD
< 10^ TV* 2
2
2
' OCOCC
' 426
Grab.
>10 g.-mA2
~ 3
3
0 23C77
•2 175
1-way Test Cii-SQuare Approximation
ChiSquare OF Prob>ChlSq
10 9203 3 0 C122
Hudson Rease 3 5
Low Resoioton Ti Pool Cores
3/8/98
-------
Analysis of Fractional Change In Mole/mA2 as
for Tl Pool Cores Alpha = 0.05
Log(Delta-M)
vs Cohesive/Noncohesive Sediment Classification
Low Resolution Core Results
Log Delta-mol+2 By Cohesive/Noncohesive Class. J
[Quantiles ]
Level minimum
10.0%
25.0%
median
75.0%
90.0%
maximum
C 0 002771
0 04982
0
127276
0 218508
0 420494
0 592224
0 881222
N 0 002281
0 302682
0
153385
0 292994
C 698694
' 121 5
'¦ 367977
[Oneway Anova )
fsummary of Fit ]
RSquare
0 050B75
RSauare aoi
0 034511
Root Mean Square Error
0 263861
Mean of Response
0 317702
Observations for Sum Wgts)
60
(T?
est
Estimate
Sid E'tQ?
L...er 95*«
Upper 95*e
Difference
•0 13502
3 C77031
¦0 29002
0 C18372
t • T e S1
' 763
DF
58
Prob>ltl
0 0831
Assuming equai var ances
[Analysis of Variance )
Source DF Sum of Squares Mean Square F Ratio
Model i 0 216451 1 0 216451 3 1089
E-ror 58 4 038267 0 C69623 Prob>F
C Torai 59 4 2545778 C 072m 3 0831
[Means for Oneway Anova)
Laval Number Maan Sid Error
C 44 0 281563 *0 03978
N 16 0 417385 0 06597
Std Error uses a pooied estimate oi error variance
[Means and Std Deviations ]
Le' el Number Mean Std Dev Strl Err Mean
C 44 0 281 563 0 1 97533 3 02978
N '6 0 417385 0 396678 0 099*7
(Means Comparisons ]
Dlt = Mean[i]-Mean[|) N C
N 3 000000 3 1 35822
C 0 13582 0 030000
Aipha= I 35
Comparisons for all pairs using Tukey-Kramer HSD
q '
2 30*77
AbMDitl-LSD N C
N -3 '8674 •: 31838
C -3 01 838 • 3 M 2 61
Positive va'->es snow oa^s ot mea^s :^a! afe s.gn iicaniiy afferent
-------
(wilcoxon / Kruskal-Wallis Tests (Rank Sums) )
Level Count Score Sum Score Mean (Mean»MeanO)/StdO
C -4 '294 29 409' 794
S 16 536 33 5000 C '94
2-Samoie Test No'^ai Approximation
S Z Prob>IZI
536 : 79402 : 4272
t-wav Test Ci.-Sauare Approximation
ChiSquere OF Prob>ChiSq
0 6438 ' 0 4223
[Median Test (Number of Points Above Median) ]
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
C 44 20 0 454545 •' 158
S 16 1C 0 625000 ' 158
2Sampie Test Norma' Approximation
S Z Prob>IZI
10 1 15790 0 2469
1-way Test Cni-Square Approximation
ChiSquare OF Prob>ChiSq
1 3409 1 0 2469
Hudson Rrver Database Release 3 5
Low Reso'uton TI Pool Cores
3'8/98
-------
Analysis of Fractional Change in Mole/m/,2 as Log(Delta-M) vs Cohesive/Noncohesive Sediment Classification
for TJ Pool Cores Alpha = 0.10 Low Resolution Core Results
.Log De»a-mol»2 By Cohesive/Noncohesive Class.
.1- .... 1 .
Q
~
N AM PalfS
^ 7iikey-Kram«r
Cohesrve/Nonconestv» Class. 01
[Quantiles ]
Levet
minimum 10.0%
25 .0%
median
7 5.0% 90,0%
maximum
C
0.002771 004982
0.127275
0 218500
0.420494 0.592224
0.881222
N
0.002281 0.002682
0.153395
0.292994
0.698694 1.1215
1.367977
(Oneway A nova )
[summary of Fit"*)
R Square
RSquare Adj
Root Mean Square Error
Mean of Response
Observations (or Sum Wgts)
0.050875
0.034511
0.263661
C.317782
60
(t-Test)
Estimate
Std Error
Lower 05%
Uppef 95%
Difference
•0.13582
0.077031
•0.29002
0.018372
t-Teit
•1 703
DF
58
Prob>itl
0.0831
Assuming equal variances
(Analysis of Variance )
Sum
Source
Model
Error
C Total
OF
\
58
59
of Squares
0.2164S11
4.0381267
4.2545778
Mean Square
C2*6451
0.069623
C.072 * 11
F Ratio
3 1069
Prob>F
0.0831
(Means for Oneway Anova ]
Level Number Mean Std Error
C 44 0.2815 6 3 0.C3978
N 1Q 0.417385 006S97
Std Error uses a pooled estimate of error variance
(Means and Std Deviations j
Level Number Mean Std Oev Std err Mean
C 44 0.281S63 0 197533 0 02978
N 16 0.417385 0.396678 0 09917
(Means Comparisons ]
™ ¦
Oif«Meanti]-Mean(U N
c
N 0 000000
0.135622
C -0.13582
0.000000
A»pha» 0.10
Comparisons for all pairs
using Tukey-Kramer HSD
1 67155
Abs(Dif)*LSO N
c
N -0.15594 0 007060
C 0.007060 -0 094C3
Positive values snow pairs ot means tnat are significantly different
-------
(wilcoxon / Kruskal-Wallis Tests (Rank Sums) ]
Level Count Score Sum Score Mean (Mean-MeanO )'StdO
C 4a <294 23 4091 0 734
N -5 536 33 5C00 !¦ 794
. 2Sarr,pie Test Normal Approximation
S Z Prob>IZI
536 3 79402 C 42 72
'•way Test C^i-Souare Approximation
ChiSquare OF Prob>ChiSq
0 6438 ' 0 4223
(Median Test (Number of Points Above Median) ]
Level Count Score Sum Score Mean (Mean-MeanO)/StdO
C 44 20 0 454545 ¦¦ 158
N '6 '0 : 6250C0 ' *58
2-Sampie Test Ncrmai Approximation
S Z Prob>\Z\
10 1 '.5798 0 2469
t-way Test. Cni-Sauare Approximation
ChiSquare OF Prob>ChlSq
* 3409 ' 0 2469
Hudson River Database Release 3 5 Low Resolution Ti Pool Cces 3/8/98
-------
Analysis of Fractional Change in Mole/mA2 as Log(Delta-M) vs Be-7 Detection in Tl Pool Cores
Low Resolution Core Results
[Log Delta-mol»2 By Be-7 Detection )
[Quantiles J
Level
minimum
10-0%
25.0%
median
75.0%
9 0.0%
maximum
Be-7 Det
0 002281
: C39899
0 '77117
0 301799
0 467951
0 785365
: 367977
Be-7 Non
0 002771
0 C02796
0 090076
0 166507
0 205651
0 365193
0 428032
(Oneway Anova )
[Summary of Fit ]
RSouare
0 089425
RSquare Adi
0 073726
Root Mean Square Erro'
0 258447
Mean ot Response
0 317782
Observations ior Sun Wgts)
60
(t-Test )
Difference t-Test OF Prob>ltl
Estimate 0 199C79 2 307 53 3 C203
Std Error 0 0634 1 4
Lower 95J. :¦ C j 2 1 08
Upper 95% 0 366048
Assuming eauai variances
(Analysis of Variance )
Source OF Sum of S >s Mean Square F Ratio
Moaei • C 30^4676 0 380468 5 6960
Err0' 59 3 8741 102 0 066795 Prob>F
C Total 59 4 2545778 0 0721H 0 2203
(Means for Oneway Anova )
Level Number Mean Sid Error
Be-7 Del 48 0 357598 0 03730
Be-7 Non 12 0 ' 58520 0 07461
Std E"0t uses a oooieo estirrate ot error variance
(Means and Std Deviations ]
Level Number Mean Std Dev Std Err Mean
Be-7 Det 48 0 357596 0 281936 0 04069
Be-7 Ncn ^2 0 1 58520 OH 2076 0 03235
[Means Comparisons )
Dif = Mean(i)-Mean[j] Be-7 Oet Be-7 Non
8e-7 Det 0 C000C0 01 99078
Be-7 Non : 19908 C 000000
Aipha= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
q '
2 30177
Abs(Oif)-LSD Be-7 Det Be-7 Non
Be-7 Det : 1C55 j 032*03
Be-7 Non : 032103 ¦: 21 *2i
Positive values snow oairs ct nea^s trat are significantly ait-erent
-------
(Wilcoxon / Kruskal-Wallis Tests (Rank Sums) ] ]
Laval Count Scora Sum Scora Mean (Maan-MaanO)/Std0
Be-7 Det -i? 'SC8 33 5000 2 662
Be-7 Non •2 222 '3 500: 2 652
2-Saripie Test Ncrai Acprox manc^
S 1 Prob>lZI
222 ¦ 2 55*96 0 0000
1-way Test C~-SQua'e Approximator.
ChiSquara OF Prob>ChiSq
7 C820 ' 0 0078
[Median Test (Number of Points Above Median) ]
Laval Count Scora Sum Scora Maan (Maan-Maan0)/Std0
Be-7 Del 48 29 0 604*67 3 200
Be-7 Non ' 2 ' 0 083333 3 20C
2-Samcie Test Normal Approximation
S Z Prob>IZi
' -3 20048 0 0014
1-way Test Cni-Square Approximation
ChiSquara DF Prob>ChiSq
^0 2431 1 0 0014
Hudson River Dataoase Re-ease 3 5 Low Resolution Ti Pool Cores
3/8/98
-------
Analysis of Relative Change in Sediment Inventory as Mass/Area (MPA) as a
Function of the 1984 ^"Tri+
Quantiles
Level minimum
10.0%
2 5.0%
median
7 5.0%
9 0.0%
maximum
<10 g/mA2
0.10305
0.108832
0.221767
0.317801
0.713091
0.978971
1.333799
>10 g/mA2 0.002317
0.005341
0.081851
0.167332
0.336886
0.438039
0.548945
[Log(Delta mass) By 1984 ITri+ PCB Inventory
i/)
in
re
E
0)
Q
o>
o
<10 g/mA2 >10 g/mA2
1984 ITri+ PCB Inventory
All Pairs
T ukey-Kramer
0.05
(Oneway Anova
Summary of Fit
RSquare
RSquare Ad|
Root Mean Square Error
Mean of Response
Observations (or Sum Wgts)
0.217576
0.204085
0 228549
0.289599
60
t-Te st
Estimate
Std Error
Lower 95%
Upper 95%
Difference
0.251367
0.062591
0.126078
0.376656
t-Test
4 016
DF Prob>ltl
58 0.0002
Assuming equal variances
[Analysis of Variance
Source DF Sum of Squares Mean Square
Model 1 0.8424728 0.842473
Error 58 3.0296203 0.052235
C Total 59 3.8720931 0.065629
F Ratio
16.1286
Prob>F
0.0002
Means for Oneway Anova
Level Number Mean Std Error
<10 g/mA2 20 0.457177 0.051 1 1
>10 g/mA2 40 0.205810 0.03614
Std Error uses a pooled estimate of error variance
1
-------
Wilcoxon / Kruskal-Wallis Tests (Rank Sums) )
Level
<10 g/mA2
>10 g/mA2
Count Score Sum Score Mean (Mean-Mean0)/Std0
20
40
806
1024
40.3000
25.6000
3.066
-3.066
2-Sample Test, Normal Approximation
S Z Prob>IZI
806 3.06568 0.0022
1-way Test. Chi-Square Approximation
ChiSquare
9.4466
DF Prob>ChiSq
1 0.0021
Median Test (Number of Points Above Median) )
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
<10 g'mA2 20 15 0.750000 2.716
>10 gmA2 40 15 0.375000 -2.716
2-Sample Test. Normal Approximation
S Z Prob>IZI
15 2.71570 0.0066
1-way Test. Chi-Square Approximation
ChiSquare
7.3750
DF Prob>ChiSq
1 0.0066
Hudson River Database Release 3.5
Low Resolution Tl Pool Cores
3/9/98
-------
This Page Was Intentionally Left Blank For Pagination Purposes.
-------
Length-Weighted Average Comparison
log10(LWA mg/kg)
Hot Spot 25
1976-1978 vs 1994
Log10(Length-Wt'd Avg) By Hotspot
2 5-
HotSpOt
[Quantiles j
Level
25 76-70
25 94
minimum
1 380211
0.630713
10.0%
1 380211
0 630713
2 5.0%
1 *150403
0 738118
median
1 602819
1 692758
7 5.0%
2 538637
2 022499
9 0.0%
2 687529
2 611163
maximum
2 687529
2 611163
[Oneway Anova i
(Summary of Fit I
RSquare
0082278
RSquare Adj
0 016726
Root Mean Square Error
0 641831
Mean of Response
1 657556
Observations lor Sum Wgts)
16
it-Test
Estimate
Std Error
Lower 95°o
Upper 95°o
Difference
0.36238
0.32345
•0 33136
1 05611
t-Test
1 120
OF Prob>ltl ]
14 0 2814 i
Assuming equal variances
Analysis of Variance i
i Source
11 Modet
!| Error
'' C Total
"jF Sum of Squer*s
1 0 5170605
14 5 7672568
T5 6 2843173
Mean Square
0.617061
0 41 1947
0 418954
F Ratio
1 2552
Prob>F
0 2814
[(Means for Oneway Anovaj
Level Number Mean Std Error
25 76-78 7 1 86139 0 24259
25 94 9 1 49902 0 21394
Std Error uses a pooled estimate ot error variance
(Means and Std Deviations^)
Level Number Mean Std Dev Std Err Mean
25 76-78 7 186139 0 532695 0 20134 |
25 94 9 '49902 0 712800 0 23760!
-------
Means Comparisons i
Dif=Mean(i}-Mean( j]
25 76-78
25 94
25 76-78
0 000000
•0 36238
Alpha= 0.05
Comparisons for all
q *
! 2.14478
Ab»(Dif)-LSD 25 76-78
25 76-78 -0 73582
25 94 -033136
25 94
0 362377
0 000000
pairs using Tukey-Kramer HSD
25 94
-0 33136
•0 64893
Positive values show pairs of means that are significantly different.
(Wilcoxon
/ Kruskal-Wallis
Tests (Rank Sums) ]
Laval
Count Score Sum Score Mean (Mean-Mean0)/Std0
25 76-78
7 66
9 42857 0.635
25 94
9 70
7 77778 -0.635
2-Sample Test. Normal Approximation
S
Z Prob>IZI
66
0 63511 0.5254
1-way Test.
Chi-Square Approximation
ChlSquare
OF Prob>ChiSq
0.4734
1 0.4914
(Median Test (Number of Points Above Median))
Laval Count Scora Sum Scora Main (Maan-Maan0)/Std0
25 76-78 7 3 0.428571 0 488
25 94 9 5 0 555556 0 488
2-Sample Test. Normal Approximation
S Z Prob>IZI
3 -0 48795 0 6256
1-way Test. Chi-Square Approximation
ChiSquara DF Prob>ChiSq
0.2381 1 0 6256
(Van der Waerden Test (Normal Quantlles)j
Level Count Score Sum Score Mean (Mean«Mean0)/Std0
25 76-78 7 1 501040 0.214435 0.865
25 94 9-1 501048 -0.16678 -0.865
2-Sampie Test. Normal Approximation
S Z Prot»IZI
1 5010477 0 86536 0 3868
1-way Test. Chi-Square Approximation
ChiSquare DF Prob>ChiSq
0.7488 1 0 3868
Hudson River Database Release 3 5
Hot Spot 25
3/2/98
-------
Length-Weighted Average Comparison
log10(LWA mg/kg)
Hot Spot 28
1976-1978 vs 1994
(Log10(Length-Wt'd Avg) By Hotspot I
3.0"
All Pairs
Tukey-Kramer
0.05
Quantiles
Level
28 76-78
28 94
minimum
0 491362
1 190707
10.0%
1 042619
1 240007
2 S .0%
1 383277
2.03907
median
1 719663
2 365101
75.0%
1 948999
2 799789
90.0%
2.219556
3 049263
maximum
2.465383
3.073319
[Oneway Anova
[Summary of Fit
RSquare
RSquare Ad|
Root Mean Square Error
Mean of Response
Observations {or Sum Wgts)
0 286679
0.266298
0.498739
1 845207
37
(t-Test)
Estimate
Std Error
Lower 95%
Upper 95%
Assuming equal variances
Difference
-0.69244
0.18463
•1 06725
•0.31763
t-Teet
¦3.750
DF
35
Prot»ltl
0.0006
(Analysis of Variance )
Source
Model
Error
C Total
OF
1
35
36
Sum
of Squares
3 498839
8 705903
12.204742
Mean
Squere
3.49884
0.24874
0 33902
F Ratio
14 0662
Prob>F
0.0006
[Means for Oneway Anova J
Level Number Mean Std Error
28 76-78 27 1 65814 0.09598
28 94 10 2 35058 0 15771
S*d Error uses a pooled estimate of error variance
[Means and Std Deviations
Level
28 76-78
28 94
Number
27
10
Mean
1.65814
2.35058
Std Dev
0 469277
0 575438
Std
Err Mean i
0.09031
0.18197
-------
• ,! I
[Means Comparisons i
Dlf=Me»n(i]-Me»n( j)
28 94
28 76-78
28 94 28 76-78
0 000000
•0 69244
0 692430
0 000000
Alpha= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
q *
2 03012
Ab»(Dil)-LSD
28 94
28 76-78
28 94
-0 4528
0.317625
28 76-78
0.317625
•0.27557
Positive values show pairs of means that are significantly different.
VVilcoxon / Kruskal-Wallis Tests (Rank Sums) j
Lavtl Count Scora Sum Score Mean (Mean-MeanO)/StdO i
28 76-78 27 424 15 7037 -3.027
28 94 10 279 27.9000 3.027
2-Sample Test. Normal Approximation
S Z Prot»IZI
279 3 02682 0 0025
1-way Test. Chi-Square Approximation
ChiSquare OF Prot»ChiSq
92654 1 0 0023
(Median Test (Number of Points Above Median) )
Laval Count Scora Sum Scora Maan (Mean-MeanO)/StdO
28 76-78 27 10 0 370370 -2 290
28 94 10 8 0 800000 2 290
2-Sample Test. Normal Approximation
S Z Prob>IZI
8 229042 0 0220
1-way Test. Chi-Square Approximation
ChiSquara DF Prot»ChlSq
5.2460 1 0.0220
(Van der Waerden Test (Normal Quantiles) ]
Level Count Score Sum Scor* Mean (Mean-Mean0)/Std0
28 76-78 27 -7 058360 -0.29105 -3 133
28 94 10 7 050360 0 705036 3 133
2-Sample Test. Normal Approximation
S Z Prob>IZI
7 0T 83605 3.13349 0.0017
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
9.0107 1 0.0017
Hudson River Database Release 3 5
Hot Spot 28
3/2/90
-------
Length-Weighted Average Comparison
log10(LWA mg/kg)
Hot Spot 31
1976-1978 vs 1994
[Log10(Length-Wt'd Avg) By Hotspot j
3 0
31 76-78
31 94
HotSpOt
' All Paltl
0.0294
[Analysis of Variance )
Source
Model
Error
C Total
1
7
8
of Squares
2 2518461
2 1171810
4 3690271
Mesn
Square
.1 25185
030245
0 54613
F Ratio
7.4452
Prob>F
0.0294
[Means for Oneway Anova )
Level Number Mean Std Error
31 76-78 4 2 35103 0.27498
31 94 5 1 34439 0 24595
Std ^rror uses a pooled estimate of error variance
[Means and Std Deviations )
Laval
31 76-78
31 94
Number
4
5
Maan
2 35103
1 34439
Std Day
0.471285
0.602257
Std
Err Maan
0 23564
0 26934
-------
Means Comparisons i
Dif=Mean[i]-Mean( j]
31 76-78
31 94
31 76-70
0 00000
•V00664
31 94
1 00664
0 00000
Alpha= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
q •
2.36437
Abs(Dif)-lSD 31 76-78 31 94
31 76-78 -0.91945 0.134373
31 94 0.134373 0.82238
Positive values show pairs of means thai are significantly different
[wilcoxon / Kruskal-Wallis Tests (Rank Sums)
Level
31 76-78
31 94
Count
4
5
Score Sum
28
17
Score Meen
7,00000
3.40000
(Mean-Mean0)/Std0
1.837
•1.837
2-Sampie Test. Normal Approximation
S Z Prob>IZI
28 1 83712 0 0662
1-way Test. Chi-Square Approximation
ChiSquare DF Prob>ChiSq
3.8400 1 0.0500
[Median Test (Number of Points Above Median) )
Laval Count Score Sum Scora Maan (Maan-MaanO)/StdO
31 76-78 4 3 0.750000 1.556
31 94 5 1 0.200000 -1.556
2-Sample Test. Normal Approximation
S Z Prot»IZl
3 155563 0 1198
1-way Test. Chi-Squaie Approximation
ChiSquara DF Prot»ChlSq
2 4200 1 0 1198
[Van der Waerden Test (Normal Quantiles) j
Level Count Score Sum Score Meen (Mean-Mean0)/Std0
31 76-78 4 2 394226 0 598557 1.959
31 94 5 -2 394226 -0.47885 -1 959
2-Sample Test. Normal Approximation
S Z Prob>IZI
2.3942262 1 95855 0 0502
1-way Test. Chi-Square Approximation
ChiSquare DF Prob>ChiSq
3.8359 1 0.0502
Hudson River Database Release 3 5
Hot Spot 31
3/2/98
-------
Length-Weighted Average Comparison
log10(LWA mg/kg)
Hot Spot 34
1976-1978 vs 1994
ltl
0 0134
[Analysis of Variance
Source
Model
Error
C Total
DF
1
35
36
Sum
of Squeree
2274936
11 744567
14.019502
Mean
Square
2 27494
0.33556
038943
F Ratio
6.7795
Prob>F
0.0134
[Means for Oneway Anova )
Level Number Mean Std Error
34 76-70 28 1.63181 0.10947
34 94 9 1 05387 0.19309
Std Error uses a pooled estimate of error variance
[Means and Std Deviations]
Level
34 76-78
34 94
Number
28
9
Mean
1 63181
1.05387
Std Dev
0 542327
0 689507
Std
Err Mean
0.10249
0 22984
-------
I {, •
[Means Comparisons i
Oi( = Me«n(i]-Mean(|J 34 76-78 34 94
34 76-78 0 000000 0 577943
34 94 0 57794 0 000000
Alpha= 0 05 j
{Comparisons for all pairs jsing Tukey-Kramer HSD \
9 *
2.03012
Abs(Di*)-LSD 34 76-78 34 9-: j
34 76-78 0.3143 0 127326
34 94 0 127326 -0.55437
Positive values show pairs of means that are significantly different
(wilcoxon / Kruskal-Wallis Tests (Rank Sums) )
Lev«l Count Score Sum Scor* Main (M««n-MemO)/StdO
34 76-78 28 593 21 1786 2 142
34 94 9 110 12.2222 -2 142
2-Sample Tesl. Normal Approximation
S Z Prob>iZI
110 -2.14168 0.0322
1-way Test. Chi-Square Approximation
ChlSquir* DF Prob>ChlSq
4.6629 1 0 0308
(Median Test (Number of Points Above Median) J
Level Count Score Sum Scor* Mean (Mean»Mean0)/Std0
34 76-78 28 16 0.571429 1.799
34 94 9 2 0.222222 -1 799
2-Sample Test. Normal Approximation
S Z Prob>IZI
2 -1 79854 0.0721
1-way Test. Chi-Square Approximation
ChiSquare DF Prot»ChiSq
3.2348 1 0.0721
[Van der Waerden Test (Normal Quantiles) j
Laval Count Score Sum Score Mean (Mean-Mean0)/Std0
34 76-78 28 5 605596 0 203057 2 346
34 94 9 -5 685596 -0 63173 -2 346
2-Sampie Test. Normal Approximation
S Z Prob>l2l
-5 685596 -2.34619 0.01*
1-way Test. Cht-Square Approximation
ChiSquare DF Prob>ChiSq
5 5046 1 0.0190
Hudson River Database Release 3.5
Hot Spot 34
3/2/98
-------
Length-Weighted Average Comparison Hot Spot 35
log10(LWA mg/kg) 1976-1978 vs 1994
(Quantiles ]
Level minimum 10.0% 25.0%
35 76-78 1.048053 1 082673 1 489818
35 94 1 376084 1 376064 1 414618
median 75.0% 90.0% maximum
1812913 2 032337 2.132421 2.146066
1 721362 2.278457 2 400441 2 400441
(Oneway Anova )
(Summary of Fit )
•
R Square
0.01 1076
RSquare Adj
-0.06499
Root Mean Square Error
0 38444
Mean of Response
1.742001
Observations (or Sum Wgts)
15
(t - T e s t ]
Diffaranc* t-Tast DF Prob>ltl
Estimate 0.08565 0 382 13 0 7089
Stfl Error 0.224464
Lower 95% -0 57058
Upper 95% 0 399274
Assuming equal variances
[Analysis
of
Variance J
Source
DF
Sum of Squares
Mean Square
F Ratio
Model
1
0.0215194
0 021519
0 1456
Error
13
1 9213200
0.147794
Prob>F
C Total
14
1 9428394
0.138774
0 7089
[Means for Oneway Anova )
Level Numbar Maan Std Error
35 76-78 11 1 71916 0.11591
35 94 4 1 80481 0.19222
Std Error uses a pooled estimate of error variance
(Means and Std Deviations)
Laval Numbar Maan Std Dav Std Err Maan
35 76-78 11 1 71916 0.359964 0.10853
35 94 4 1 80481 0.456647 0 22832
-------
Means Comparisons
Olf = Mean[i]-Mean[j) 35 94 35 76-78
' 135 94 0 000000 0 085651
35 76-78 0 08565 0 000000
Alpna= 0 05
Comparisons for all pairs usiiig Tukey-Kramer HSO
q '
2 16040
Abs(Dit)-LSD 35 94 35 76-78
35 94 -0 58728 0 39928
35 76-78 -0 39928 0 35415
Positive values snow pairs of means that are significantly different.
(wilcoxon / Kruskal-Wallis Tests (Rank Sums) )
Laval
35 76-78
35 94
Count
11
4
Scora
Sum
87
33
Scora Maan
7 90909
8.25000
(Maan-Maan0)/Std0
•0.065
0 065
2-Sample Test. Normal Approximation
S Z Prob>IZI
33 0 06528 0 9480
1-way Test. Chi-Square Approximation
ChiSquara DF Prob>ChiSq
0.0170 1 0 8961
(Median Test (Number of Points Above Median) )
Laval Count Score Sum Scora Maan (Maan-Maan0)/Std0
35 76-78 11 5 0.454545 -0 151
35 94 4 2 0.500000 0.151
2-Sample Test. Normal Approximation
S Z Prot»IZI
2 0.15076 0.8802
1-way Test. Chi-Square Approximation
ChiSquara DF Prob>ChiSq
0.0227 1 0.8802
(Van der Waerden Test (Normal QuantilesT]
Laval Count Scora Sum Scora Maan (Maan-MaanO)ZStdO
35 76-78 11 -0.476837 -0.04335 -0 320
35 94 4 0.4768369 0.1 19209 0.320
2-Sample Test. Normal Approximation
S Z Prot»IZI
0.4768369 0 32045 0.7486
1-way Test. Chi-Square Approximation
ChiSquara DF Prob>ChiSq
0.1027 1 0.7486
Hudson River Database Release 3 5
Hot Spot 35
3/2/98
-------
Length-Weighted Average Comparison
log10(LWA mg/kg)
Hot Spot 37
1976-1978 vs 1994
[Log10(Length-Wt'd Avg) By Hotspot I
2 5
2 0
CT>
C
®
5" 0 5
o
0.0
Quantiles
Level
37 76-78
37 94
minimum
0.40654
-0.31995
10.0%
0 604725
-0.08059
25.0% median 75.0% 90.0% maximum
1563288 1661623 1921761 2.294222 2.303196
0 975853 1163684 1.326718 1.961995 2.115728
(Oneway Anova )
[Summary of Fit]
RSquare
RSquare Adj
Root Mean Square Error
Mean of Response
Observations (or Sum Wgts)
0.203731
0.167537
0.540768
1.392612
24
t-Test
Estimate
Sid Error
Lower 95%
Upper 95%
Difference
0.525604
0.221538
0.066166
0 985043
t-Test
2.373
OF
22
Prob>ltl
0.0268
Assuming equal variances
(Means and Std Deviations)
Level
37 76-78
37 94
Number
13
11
Mean
1.63351
1 10791
Std Oev
0 510717
0 574760
Std
Err Mean !
0.14165 |
0.17330 I
-------
¦Means Comparisons
Dif = Maan(i]-Mean(j] 37 76-76 37 94
37 76-78 0 000000 0 525604
! 37 94 -0.5256 n on^000
Aipha= 0.05
Comparisons for all pairs using Tukey-Kramer HSD
ChlSq
6.3516 1 0 01 17
(Median Test (Number of Points Above Median) j
Laval Count Scora Sum Scora Mean (Mean-Maan0)/Std0
37 76-78 13 11 0.846154 3.609
37 94 11 1 0.090909 -3 609
2-Sampfe Test. Normal Approximation
S Z Prob>IZI
1 -3.60943 0.0003
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
13.0280 1 0 0003
(van der Waerden Test (Normal Quantiles) ]
Laval Count Score Sum Scora Maan (Maan-Mean0)/Std0
37 76-78 13 5.162079 0.397083 2.340
37 94 11 -5.162079 -0.46928 -2.340
2-Sample Test. Normal Approximation
S Z Prob>i'
-5.162079 -2.34005 0.0 „
1-way Test. Chi-Square Approximation
ChlSquara OF Prob>ChiSq
5.4758 1 0.0193
Hudson River Database Release 3 5
Hot Spot 37
*2/98
-------
Length-Weighted Average Comparison
log10(LWA mg/kg)
Hot Spot 39
1976-1978 vs 1994
(Log10(Length-Wt'd Avg) By Hotspot
2 5
2 0-
¦*
<
"O
1.5-i
t
£
o>
c
0)
J
1.0-1
o
z
_J
0.5-
1
o.o- r
39 76-78 39 94
Motspot
All Pairs
Tukey-Kramer
0.05
(Quantiles
Level
39 76-78
39 94
minimum
0 946943
0.019359
10.0% 25.0% median 75.0% 90.0% maximum
104914 1490586 1631748 1668293 2 084753 2 459392
0 131136 0 663247 1 360283 1 745127 1 948978 2.066031
(Oneway Anova ]
(Summary of Fit ]
RSquare
0.172447
RSquare Adj
0.141797
Root Mean Square Error
0.497255
Mean of Response
1 387002
Observations (or Sum Wgis)
29
(t - T e s t )
Estimate
Std Error
Lower 95%
Upper 95%
Difference
0.438309
0.184706
0.059163
0.817455
•Teet
2 372
OF
27
Prot»ltl
0.0251
Assuming equal variances
(Analysis
of
Variance I
Source
OF
Sum of Square*
Mean Square
F Ratio
Model
1
1 3911740
1 39117
5.6263
Error
27
6.6760901
0.24726
Prob»F
C Total
28
8.0672641
0 28812
0.0251
[Means for Oneway Anova I
Level Number Mem Std Error
39 76-78 15 1 59860 0.12839
39 94 14 1 16029 0 13290
Sic Error uses a pooled estimate of error variance
(Means
and Std
Deviations ]
Laval
Number
Maan Std Dev
Std Err Maan
39 76-78
15
1 59860 0 332032
0 08573
39 94
14
1 16029 0 628347
0.16793
-------
[Means Comparisons j
Dif=Mean[i]-Mean[j]
39 76-78
39 94
39 76-78 39 94
0.000000 0.438309
•0.43831 0.000000
(Mean-Mean0)/Std0
1.768
-1.768
i Alpha= 0.05
I Comparisons for all pairs using Tukey-Kramer HSD
q *
2.05184
Abs(Dif)-LSD 39 76-78 39 94
39 76-78 -0.37256 0.059158
39 94 0.059158 -0.38563
Positive values show pairs of means that are significantly different.
(wilcoxon / Kruskal-Wallis Tests (Rank Sums)]
Level Count Score Sum Score Mean
39 76-78 15 266 17.7333
39 94 14 169 12.0714
2-Sample Test, Normal Approximation
S Z Prot»IZI
169 -1.76756 0.0771
1-way Test, Chi-Square Approximation
ChiSquare OF Prob>ChiSq
3.2019 1 0.0736
(Median Test (Number of Points Above Median) )
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
39 76-78 15 10 0.666667 2.016
39 94 14 4 0.285714 -2.016
2-Sample Test. Normal Approximation
S Z ProbMZI
4 -2.01581 0.0438
1-way Test, Chi-Square Approximation
ChiSquare OF Prob>ChiSq
4.0635 1 0.04 38
[Van der Waerden Test (Normal Quantiles) ]
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
39 76-78 15 4.401803 0.293454 1.787
39 94 14 -4.401803 -0.31441 -1.787
2-Sample Test. Normal Approximation
S Z Prob>IZI
-4.401803 -1.78698 0.0739
1-way Test, Chi-Square Approximation
ChiSquare DF Prob>ChiSq
3.1933 1 0.0739
Hudson River Database Release 3.5
Hot Spot 39
3/2/98
-------
Length-Weighted Average Comparison
log10(LWA mg/kg)
Dredge Location 182
1976-1978 vs 1994
[Log10(Length-Wt'd Avg) By Hotspot ;
(Quantiles ]
Level minimum
1 0.0%
25.0%
median
75.0%
9 0.0%
maximum
182 76-70 0 13077
-0.13077
0.786848
1 326644
1 .403721
1 403978
1 403978
102 94 1 255081
1 255081
1 255081
1 372084
1 489086
1 489086
1 489086
AM Pairs
Tukey-Kramer
0.05
(Oneway Anova
(Summary of Fit )
RSquare
0 069208
RSquare Adj
-0 08592
Root Mean Square Error
0.553044
Mean of Response
1.145877
Observations (or Sum Wgts)
8
(t - T e s t
Estimate
Sid Error
Lower 95%
Upper 95%
Difference
•0.30161
0 45156
•1.40653
0 00332
t-Teat
•0 668
OF
6
Prob>ltl
0.5290
Assuming equal variances
(Analysis of Variance )
Source DF Sum of Squares Mean Square
Model 1 0.1364512 0.136451
Error 6 1 8351472 0.305858
C Total 7 1.9715984 0.281657
F Ratio
0.4461
Prob>F
0 5290
[Means for Oneway Anova j
Level Number Mean Std Error
182 76-78 6 1 07048 0.22578
182 94 2 1.37208 0.39106
Sid Error uses a pooled estimate of error variance
Means and Std Deviations
Level
182 76-78
182 94
Number
6
2
Mean
1.07048
1 37208
Std Dev
0 601293
0 165467
Std
Err Mean
0.24548
0.1 1700
-------
(Means Comparisons
Dit = M«an{i]-Mean[j]
182 94 182 76-78
162
182
94
76-78
0.000000
•0.30161
0 301608
0 0OO000
Alpha= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
q '
2 44692
Abs(Dlf)-LSD 182 94 182 76-/8
182 94 -1 35325 0 80332
182 76-78 0 80332 0 78130
Positive values show pairs of means that are significantly different.
[Wllcoxon / Kruskal-Wallis Tests (Rank Sums) j
Laval Count Scora Sum Scora Maan (Maan-Maan0)/Std0
182 76-78 6 24 4 00000 0 033
182 94 2 12 6.00000 0 833
2-Sample Test. Normal Approximation
S Z Prob>iZI
12 0.83333 0.4047
1-way Tesl. Chi-Square Approximation
ChlSquara DF Prob>ChiSq
1 0000 1 0.3173
[Median Test (Number of Points Above Median) )
Laval Count Scora Sum Scora Maan (Maan-Maan0)/Std0
182 76-78 6 3 0 500000 0 000
182 94 2 1 0 500000 0.000
2-Sample Test. Normal Approximation
S Z Prob>IZI
1 0.00000 1.0000
1-way Test. Chi-Square Approximation
ChlSquara DF Prot»ChlSq
0.0000 1 1 0000
[Van der Waerden Test (Normal Quantiles)]
Laval Count Scora Sum Score Maan (Maan-Maan0)/Std0
182 76-78 6 -1080930 -0.18016 -1094
182 94 2 1.080930 0 540465 1.094
2-Sample Test. Normal Approximation
S Z Prob> IZI
1 08093 1 09355 0.2 7-'
1-way Test. Chi-Square Approximation
ChlSquara OF Prob>ChiSq
1.1959 1 0.2742
Hudson River Database Release 3.5
Dredge Location 182
3/2/98
-------
This Page Was Intentionally Left Blank For Pagination Purposes.
-------
Mass per Unit Area Comparison
log10(MPA g/m*2)
LoglO(MPA) By Hotspot
2 On —
' 81
' 6-1
/ - - \
14-1
25 76-78 25 94
HOlSpOl
Hot
1976-1978
Spot 25
vs 1994
All Pairs
Tukey-Kramer
0.05
¦ c
IQuantiles j
Level
25 76-78
25 94
minimum
0 8777
0.3374
10.0%
0 0777
0 3374
2 5.0%
0 9381
0 4305
median
0.9822
0 9997
7 5.0%
J 7302
1 55875
9 0.0%
i 8791
1 9457
maximum |
1 8791 j i
1 9457 |
(Oneway Anova
[Summary of Fit j
R Square
RSquare Adi
Root Mean Square Error
Mean of Response
Observations tor Sum Wgts)
i[t-Test )
| Estimate
(j Std Error
|! Lower 95°
! Upper 95°o
Difference
0.162427
0.262362
•0-40028
0.725137
0 026647
-0.04288
0 520609
1 129506
16
t-Test
0 619
DF Prob>ltl
14 0 5458
Assuming equal variances
'Analysis of Variance
Source
Mode'
Error
C Tolal
OF
14
15
Sum of Squares
0 1038812
3 7944738
3 8983550
Mean Square
0 10388 T
0 271034
0 259890
F Ratio
0 3833
Prob>F
0 5458
!i:l!
Means for Oneway Anova
1
i1 Level Number Mean Std Error
!' 25 76-78 7 1 22007 0 19677
j: 25 94 9 1 05044 0 17354
StO Error uses a pooled estimate ot error variance
\t ^
jlMeans and Std Deviations )
: Level Number Mean Std Oev
I25 76-78 7 1 22087 0 412873
! 25 94 9 1 05844 3 588610
Std
Err Mean i
0 15605 I
0 19620 I
-------
Means Comparisons
Dif = Mean{i]-Mean[j] 25 76-78 25 94
25 76-78 0 003000 0 162427
25 94 -0 16243 0 000000
1 Aipha= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
q •
2 14478
I Abs(Dif)-LSD 25 76-78 25 94
1125 76-78 0 59684 ¦ -.,u28
1125 94 -0.40028 C 5co. /
j J Positive values show pairs of means that are significantly different
iff
i
j (wilcoxon / Kruskal-Wallis Tests (Rank Sums)
I
Level Count Score Sum Score Mean (Mean-Mean0)/Std0 J
! 25 76-78 7 64 9 14286 0 423
25 94 9 72 8 00000 0 423
| 2-Sampie Test. Normal Approximation
S Z Prob>IZI
64 0.42340 0 6720
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
0 2269 1 0 6338
(Median Test (Number of Points Above Median) ;
Level Count Score Sum Score Mean (Mean-MeanO)'StdO
25 76-78 7 3 0.428571 0 488
25 94 9 5 0.555556 0 488
2-Sampie Tesl. Normal Approximation
S Z Prob>IZI
3 -0 48795 06256
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
0 2381 1 0 6256
(van der Waerden Test (Normal Quantiles) i
Level Count Score Sum Score Mean (Mean-Mean0)/Std0 !
25 76-78 7 0 9430254 0 134718 0 544 |
| 25 94 9 -0.943025 0 10470 -0 544
2Sampie Test. Normal Approximation j
S Z Prob>iZI |
0 9430254 0 54366 0 5867
1-way Test. C^i-Square Approximation
ChiSquare DF Prob>ChiSq
• : 0 2956 1 0 5867 I
i j
v ;
Hudson River Database Release 3 5
Hot Spot 25
3/2/98
-------
Mass per Unit Area Comparison
log10(MPA g/mA2)
Hot Spot 28
1976-1978 vs 1994
|Log10(MPA) By Hotspot i
2.5"
1 .5"
=> 1.0"
o.o-
28 76-78
HotSpOt
28 94
All Pairs
Tukey-Kramer
0.05
(Quantiles )
Level minimum 10.0% 25.0% median 75.0% 90.0% maximum
28 76-78 0.0364 0.55296 0.8807 1.0991 1.3284 1.54996 1.7958
28 94 0.5279 0.62384 1.727975 2.14065 2.302525 2.47444 2.4828
! (Oneway Anova ]
[Summary of Fit J
RSquare
0.444677
RSquare Adj
0.42881 1
Root Mean Square Error
0.455715
Mean of Response
1.307165
Observations (or Sum Wgts)
37
(t - T e s t I
Difference t-Test DF Prob>ltl i
Estimate -0.89309 -5.294 35 <0001 j
j Std Error 0.16870 i
| Lower 95% -1 .23557 j
; Upper 95% -0.55062 i
Assuming equal variances |
[Analysis of Variance
Source
Model
Error
C Total
DF
1
35
36
Sum
of Squares
5.820410
7.260656
13.089065
Mean
Square
5.82041
0.20768
0.36359
[Means for Oneway Anova j
Level Number Mean Std Error
28 76-78 27 1.06579 0.08770
28 94 10 1.95888 0.1441 1
Std Error uses a pooled estimate of error variance
P Ratio
28.0264
Prot»F
<.0001
( N !
[Means and Std Deviations i
Level Number Mean Std Dev Std Err Mean i
28 76-78 27 1.06579 0.403642 0.07768 j
28 94 10 1.95888 0.580475 0.18356 ;
-------
Means Comparisons
Dif = Meanfi]-Mean(j] 28 94 28 76-78
;:28 94 0 C03000 ? 393091
ii28 76-78 -0 89309 0 000000
ij
I j Alpha= 0 05
Comparisons for alt pairs using Tukey-Kramer HSD
¦i q'
:j 2 03012
|| Abs(Dif)-LSD 28 94 28 76-78
::28 94 -0.41 374 0 550612
: 28 76-78 0 550612 0 2518
1 Positive values snow pairs of means thai are significantly different
V ~
Wilcoxon / Kruskal-Wallis Tests (Rank Sums) j
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
28 76-78 27 406 15 0370 -3 642
28 94 10 297 29 7000 3 642
2-Sampie Test. Normal Approximation
S Z Prob>IZI
297 3.64244 0 0003
1-way Test. Chi-Sauare Approximation
ChiSquare OF Prob>ChiSq
13 3922 1 0 0003
Median Test (Number of Points Above Median)
Level Count Score Sum Score Mean (Meen-Mean0)/Std0
28 76-78 27 9 0 333333 -3 021
28 94 10 9 0 900000 3.021
2-Sample Test. Normal Approximation
S Z Prob>IZI
! 9 3 02098 0 0025
¦ 1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
1 9 1263 1 0 0025
r, .
-Van der Waerden Test (Normal Quantiles) I |
level Count Score Sum Score Mean (Mean-MeanO)/StdO
28 76-78 27 -9 115551 0 33761 -3 635
28 94 10 9 1 15551 0 91 1555 3 635
2 Sampie Test. Normal Approximation
S Z Prob>IZI
9 1 155506 3 63455 C 0003
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
13 2100 1 0 0003
Hudson River Database Release 3 5
Hot Spot 28
3 2-98
-------
Mass per Unit Area Comparison
log10(MPA g/mA2)
Hot Spot 31
1976-1978 vs 1994
LoglO(MPA) By Hotspot
2
=> 1
31 76-78
31 94
HotSpOt
All Pairs
Tukey-Kramer
0.05
[Quantilesj
Level
31 76-78
31 94
minimum
1.1019
0 0538
10.0%
1 1019
0 0538
2 5.0%
1 2292
0 3217
median
1 6991
1.1127
7 5.0%
1 9446
1 24795
9 0.0%
1.9971
1 3236
maximum
1 9971
1 3236
(oneway Anova :
[Summary of Fit !
RSquare
0 463996
RSquare Adj
0 387424
Root Mean Square Error
0 468659
Mean of Response
1 194356
Observations (or Sum Wgtsi
9
t-Test
Estimate
Std Error
Lower 95°
Upper 95®«
Difference
0 77390
0 31439
003049
1 51731
t-Test
2 462
DF Prob>111 I
7 0 0434 i
Assuming equal variances
[[Analysis of Variance i
Source
Model
Error
C Total
DF
1
7
Sum of Squares
1 3309360
1 5374870
2 8684230
Mean Square
1 33094
0 21964
0.35855
F Ratio !
6 0596 I j
Prob>F I I
0 0434 i!
(Means for Oneway Anovaj
Level Number Mean Std Error
31 76-78 4 1 62430 0 23433
31 94 5 0 85040 0 20959
Std Error uses a pooled estimate of error variance
I
(Means
and Std
Deviations
Level
Number
Mean Std Dev
Std Err Mean
31 76-78
4
1 62430 0 382344
0 19117
31 94
5
0 65040 0 524149
0 23441 I
-------
[Means Comparisons
I Dif = Mean(I ] ¦ Mea n( I ] 31 76-78 31 94
!31 76-78 D DOOOOO C 77390C
' i31 94 0 7739 G C0000C j
¦ Alpna= 0 35 |
Comparisons for all pairs using Tukey-Kramer HSO ;
q ' i
2 36437 J
i Abs(Dif)-LSD 31 76-78 31 94
¦31 76-78 -0.78353 C __j576
.31 94 0 030576 C "':81 |
Positive values show pairs of means that are significantly different j
Wilcoxon / Kruskal-Wallis Tests (Rank Sums) J
Level Count Score Sum Score Mean (Mean-MeanO)/StdO
31 76-78 4 27 6.75000 1 592
31 94 5 18 3.60000 -t.592
2-Sampie Test. Normal Approximation
S Z Prob>IZI
27 1 59217 0 1113
1-way Test. Chi-Square Approximation
ChiSquare OF Prot»ChiSq
2.9400 1 0 0864
[Median Test (Number of Points Above Median) j
Level Count Score Sum Score Mean IZ1
3 1 55563 0 1 198
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChlSq
2 4200 1 0.1198
I (van der Waerden Test (Normal Quantiles) J
Level Count Score Sum Score Mean
,j 31 76-78 4 2 123173 0 530793
:l 31 94 5 -2 1231 73 -0 42463
2-Sampie Test. Normal Approximation
;j S Z Prob>lZI
j 2 1231728 1 73682 C 0824
1-way Test Chi-Square Approximation
ChiSquare OF Prob>ChiSq
3 0165 1 0 0824
r
(Mean-Mean0)/Std0
1 737
•1 737
Hudson River Database Release 3 5
Hot Spot 3t
a 2/98
-------
I
Mass per Unit Area Comparison
log10(MPA g/mA2)
:Log10(MPA) By Hotspot
Hot Spot 34
1976-1978 vs 1994
_ 1
<
Q.
2
2 0 51
•0 5"
34 76-78
HotSDOt
34 94
All Pairs
Tukey Kramer
0 05
(Quantiles
Level
34 76-78
34 94
minimum
0 1809
•0 52
10.0% 25.0% median 75.0% 90.0% maximum
0 31707 0 731575 1 09195 : 373075 1 82125 1 8279
-0 52 0 0216 D 2849 '.02295 1 7019 1 7019
[Oneway Anova
[Summary of Fit
RSauare
RSquare Adj
Root Mean Square Error
Mean of Response
Observations (or Sum Wgts)
:t-Tes
L)
! Estimate
Sid Error
!
! 11 Lower 95°©
, j | Upper 95°o
i. ^Assuming equal
f
Difference
0 542802
0 201840
0 133047
0 952558
anances
0 171247
0.147569
0 526754
0 922503
37
t-Test
2 689
OF
35
Prob>Jtf
0 0109
Analysis of Variance i
Source
Model
Error
C Total
DF
35
36
Sum of Squares
2 006699
9 711435
1 1 718134
Mean Square
2 00670
0 27747
0 32550
F Ratio
7 2321
Prob>F
0 0109
[Means for Oneway Anova
Level Number Mean Std Error
34 76-78 28 1 35454 C 09955
34 94 9 0 51173 j 17558
Std Error uses a pooled estimate of error variance
.Means and Std Deviations
I Level
I 34 76-
' 34 94
Number
28
9
Mean
1 C5454
•: 51173
Std Oev
C 462196
0 702101
Std Err Mean '
0 08735 !
C 23403 i
-------
Means Comparisons ,
Dil = Mean[i)-Mean(j] 34 76-78 34 94
,34 76-78 :¦ jCCOOO 9 5428C2
'34 94 0 5-123 0 30C003
• Alpha= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
; q *
| 2 03012
. Abs(Dif)-LSD 34 76-78 34 94
i -34 76-78 0.2858 0 133041
' 134 94 0 133041 -0 5041 1
[positive values snow pairs of means that are significantly atferent. i
i (f ... "
1 Wilcoxon / Kruskal-Wallis Tests (Rank Sums)
' Level Count Score Sum Score Mean (Mean-MeanO)ZStdO
34 76-78 28 593 21 1786 2 142 j
34 94 9 110 12 2222 -2 142
2-Sample Test. Normal Approximation
S Z Prot»lZI
110 -2.14168 0 0322
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
4.6629 1 0 0308
^
Median Test (Number of Points Above Median) j
I
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
34 76-78 28 16 0 571429 1.799
34 94 9 2 0 222222 -1 799
2-Sample Test. Normal Approximation
S Z Prot»lZl
2 -179854 0 0721
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
3.2348 1 0.0721
[Van der Waerden Test (Normal Quantiles) J
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
j 34 76-78 28 5.669361 0 202477 2.339
34 94 9 -5 669361 0 62993 2 339
2-Sample Test. Normal Approximation
S Z Prob>lZI
-5 669361 -2 33949 C 0193
j 1-way Tesl Cni-Square Approximation
ChiSquare OF Prob>ChiSq
j 5 4732 1 0 0193
Hudson River DataBase Release 3 5
not Spot 34
3.2/98
-------
Mass per Unit Area Comparison
log10(MPA g/mA2)
Hot Spot 35
1976-1978 vs 1994
LoglO(MPA) By Hotspot
1 7 "1
1 6i
1 5-
1 .41
1.3"!
1 21
1 1-f
1
0 9 1
0 8~
0 7
0 6
0 5
35 76-78
Holspot
35 94
At) Pairs
Tukey-Kramer
0.05
fouantiles
Level minimum 10.0% 25.0% median 75.0% 90.0% maximum |
35 76-78 0 5455 0 58012 0 9357 11923 13698 146276 1 4764 | I
35 94 0 9186 C 9186 0 92745 107625 1502025 16032 i 6032 I !
(Oneway Anova
[Summary of Fit I
R Square
0.006357
RSquare Adj
-0 07008
Root Mean Square Error
0 303979
Mean ot Response
1 13104
Observations {or Sum Wgts)
15
(t - T e s t
Estimate
Std Error
Lower 95°o
Upper 95°o
Difference
¦0 051 18
0 177486
-0 43462
0 332250
t-Test
•0 288
OF
13
Prob>ltl
0.7776
I Assuming equal variances
[Analysis of Variance i
Source
Model
I Error
jc Total
OF
13
14
I
Sum of Squares
0 0076848
' 2012439
1 2089286
Mean Square
0 007685
0 092403
0 086352
F Ratio
0 0832
Prot»F
C 7776
(Means for Oneway Anova
Level Number Mean Std Error
35 76-78 11 1 11739 0 09165
35 94 4 1 16857 0 15199
Sid Error uses a pooled estimate of error variance
Means and Std Deviations I
Level
35 76-78
35 94
Number
11
4
Mean
1 11739
1 16857
Std Dev
0 300486
0 315343
Std
Err Mean
0 09060 i
0 15767 I
I
-------
Means Comparisons
Oif = Mean(i]-Mean[j]
,35 94
13 5 76-78
i; Alpna= D 05
Comparisons for all pairs using Tukey-Kramer HSD
q '
2 t 6040
,j Ab»(Dif)-LSD 35 94 35 76-78
!35 94 -0 46437 „ 33226
35 76-78 -0 33226 ^ .8003
[\ Positive values snow pairs ol means that are significantly different
•, f ^
Wilcoxon / Kruskal-Wallis Tests (Rank Sums) ;
11 Level Count Score Sum Score Metn (Mean-MeanOf/StdO
j, 35 76-78 11 88 8 00000 0 065
¦I 35 94 4 32 8.00000 0 065
'j
2-Sampie Test. Normal Approximation
S Z Prot»IZI
32 0 06528 0 9480
j 1-way Test. Chi-Square Approximation
ChiSquare DF Prob>ChiSq
0 0000 1 10000
^Median Test (Number of Points Above Median) j j
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
35 76-78 11 5 0 454545 -0 151 j
35 94 4 2 0 500000 0.151
! 2-Sampie Test. Normal Approximation
j| S Z Prob>IZI
2 0 15076 0 8802
I
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
0 0227 1 0 8802 |
35 94 35 76-78
: 00C000 0 051184
•0 05118 0 000000
f v J
IjVan der Waerden Test (Normal Quantiles) , |
| Level Count Score Sum Score Mean (Mean-Mean0)/Std0
j 35 76-78 11 -0 315508 0 02868 -0 212 !
i| 35 94 4 0 3155083 0 078877 0 212 !
2 Sampie Test. Normal Approximation
S Z Prob>IZI
0 3155083 0 21203 j 8321
! 1-way Test. Chi-Square Approximation i
I
1 ChiSquare OF Prob>ChiSq
0 0450 1 0 8321 j
Hudson River DataBase Release j 5
Mot Spot 35
3.2/98
-------
Mass per Unit Area Comparison
log10(MPA g/mA2)
Hot Spot 37
1976-1978 vs 1994
I LoglO(MPA) By Hotspot
! 2 0 5-
i 5
I 2
o>
'3001
•3 5"
•1 0"
37 76-78
37 94
HotSpOt
All Pairs
T ukey-Kramer
0.05
[[Quantiles j
Level minimum 10.0% 25.0% median 75.0% 90.0% maximum
37 76-78 0 0484 0.15648 0.94275 1041 127665 1 6246 1 6336
37 94 0 7894 0 5998 0 2005 0 5057 0 7705 12574 13456
[Oneway Anova i
[Summary of Fit i
R Square
RSquare Ad|
Root Mean Square Error
Mean of Response
Observations lor Sum Wgtsi
0 271216
0 23809
0 485648
0.767463
24
t-Test
Estimate
Std Error
j Lower 95°.
Upper 95°o
Difference
0 569285
0 198957
0 156676
j 981893
t-Test
2 861
DF
22
Prob>ltl i
0 0091 I
Assuming equal variances
(Analysis of Variance
Source
Model
Error
| C Total
OF
1
22
23
Sum ot Squares
i 9310063
5 1887958
7 1198021
Mean
Square
1 9310 t
0 23585
0 30956
F Ratio
3 1873
Prob>F
0 0091
[Means for Oneway Anova i
Level Number Mean Std Error
37 76-78 13 1 02838 0 13469
37 94 11 j 45910 0 14643
Std Error uses a pooled estimate of error variance
Means and Std Deviations
I Level
I j 37 76-78
1 37 94
Number
13
Mean
1 02838
C 45910
Std Dev
0 442819
0 532516
Std Err Mean i
0 12282 I
0 16056 i
-------
Means Comparisons
i Di I = M ea n( i ] - Me a n [ | ] 37 76-78 37 94
37 76-78 0 COOOOO I 5692B5
37 94 -0 56923 C 000000
AlpMa= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
i <5 '
!! 2 07387
:j Abs(Dif)-LSD 37 76-78 37 94
:!37 76-78 -0 39505 J "56672
i 37 94 0 156672 -0 42946
Positive values show pairs of means that are significantly different
j(wilcoxon / Kruskal-Wallis Tests (Rank Sums) ]
ii
j| Level Count Score Sum Score Mean (Mean-Mean0)/Std0
37 76-78 13 208 16 0000 2 607
; I 37 94 11 92 8.3636 -2.607
2-Sample Test. Normal Approximation
S Z Prob>lZI
92 -2 60714 0.0091
1-way Test. Chi-Square Approximation
ChiSquare DF Prob>ChiSq
6.9491 1 0 0084
(Median Test (Number of Points Above Median) :
Level Count Score Sum Score Mean (Mean»Mean0)/Std0
37 76-78 13 10 0 769231 2 807
37 94 11 2 0 181818 -2 807
2-Sampie Test. Normal Approximation
S Z Prot»IZI
2 -2.80733 0 0050
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
7 8811 1 0 0050
(van der Waerden Test (Normal Quantiles) j
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
37 76-78 !3 5 530456 0 425420 2 507
37 94 11 5 530456 -0 50277 -2 507
2-Sampfe Test, Normal Aoproxmat'on
S Z Prob>lZI
•5 530456 2 50704 0 C122
I
1-way Test. Chi-Square Approximation j
ChiSquare DF Prob>ChiSq
6 2853 1 0 0122 ;
Hudson River Database Release 3 5
Hot Spot 37
3 2/98
-------
Means Comparisons
Dif = Mean[tJ-Mean(|] 182 94
¦182 94 0 000000
182 76-78 -0 29362
Aipna= o 05
Comparisons for al
q *
2 44692
i Abs(Dif)-LSO 182 94 182 76-78
182 94 -1 23407 0 71465
182 76-78 0 71465 0 71295
182 76-78
:¦ 2936 1 7
¦: occooo
pairs using Tukey-Kramer HSO
, | Positive values snow pairs of means tnat are significantly different
I
j Wilcoxon / Kruskal-Wallis Tests (Rank Sums) J
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
182 76-78 6 24 4 00000 -0 833
182 94 2 12 6 00000 0 033
2-Sample Test. Normal Approximation
S
12
Z Prob>lZI
0.83333 0 4047
1-way Test. Chi-Square Approximation
ChlSquare
1.0000
OF Prob>ChiSq
1 0 31 73
[Median Test (Number of Points Above Median) j
Level
182 76-78
182 94
Count
6
2
Score Sum
3
Score Mean
0.500000
0.500000
(Mean-Mean0)/Std0
0 000
0.000
2-Sample Test. Normal Approximation
S Z Prob>IZI
1 0.00000 1 0000
!j 1-way Test. Chi-Square Approximation
|| ChiSquare OF Prob>ChiSq
i 0 0000 1 1 0000
(Van der Waerden Test (Normal Quantiles) j
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
182 76-78 6 -1080930 -0 18016 -1 094
182 94 2 1 080930 0 540465 1 094
2-Sample Test. Normal Approximation
S Z Prob>lZl
108093 109355 0 2742
1-way Test Chi-Square Approximation
ChiSquare OF Prob>ChiSq
1 1959 1 0 2742
Hudson River Database Release 3 5
Dredge Location 182
3.2/98
-------
Mass per Unit Area Comparison
log10(MPA g/mA2)
Dredge Location 182
1976-1978 vs 1994
|Log10(MPA) By Hotspot
•=*= Y
¦ 0
132 76-78
Hotspot
182 94
All Pairs
Tukey-Kramer
0 05
[Quantiles
11
Level minimum 10.0% 25.0% median 75.0% 90.0% maximum n
182 76-78 -0 5113 -0 5113 0 314825 0 82415 0 901 175 0 9014 0 9014 I |
182 94 0 8592 0 8592 0 8592 0 8819 0 9046 0 9046 0 9046 !|
Oneway Anova i
[Summary of Fit )
RSquare
RSquare Ad)
Root Mean Square Error
Mean of Response
Observations tor Sum Wgts)
0 078023 |
0 07564 |
0 504662 |
0 661687 j
8 I
(t-Test
! Estimate
j Std Error
Lower 95°
Upper 95°:
Difference
¦0 29362
0 41205
1 30188
0 71464
t-Test
0 713
DF Prob>ltl
6 0 5029 I
Assuming equal variances
(Analysis of Variance
Source
Model
Error
C Total
DF
Sum of Squares
0 1293161
1 5280994
1 6574156
Mean Square
0 129316
0 254683
0 236774
F Ratio
0 5078
Prob>F
j 5029
'Means for Oneway Anova i
Level Number Mean Std Error
182 76-78 6 0 588283 0 20603
182 94 2 0 881900 C 35685 ,
Std Error uses a oooied estimate of error variance i
(Means and Std Deviations
j Level
Number
182
162
76-'
94
Mean
0 588283
0 681900
Std Dev
0 552643
C 332103
Std
i
Err Mean i
0 22562 i
0 02270 i
-------
Means Comparisons
Dif = Mean(i)-Mean(j] 39 94 39 76-78
¦¦39 94 ; 0000C0 0 C40465
39 76-78 o 04047 0 0000C0
| Alpha= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
!i q '
2 05184
. AbS(Dif)-LSD 39 94 39 76-78
! 139 94 -0.42711 -0 37947
j:39 76-78 -0 37947 -0 41263
11
Positive values sDow pairs of means that are significantly different
V
Wilcoxon / Kruskal-Wallis Tests (Rank Sums)
!! '
j Level Count Score Sum Score Mean (Mean-Mean0)/Std0
!| 39 76-78 15 187 12 4667 -1637
39 94 14 248 17 7143 1 637
2 Sample Test, Norma' Approximation
S Z Prob>IZI
248 1 63663 0 1017
1-way Test. Chi-Square Approximation
ChiSquare DF Prob>ChiSq
2 7505 1 0 0972
[Median Test (Number of Points Above Median)
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
39 76-78 15 4 0.266667 -2 369
39 94 14 10 0 714286 2.369
2-Sample Test. Normal Approximation
S Z Prob>IZl
! 10 2 36858 0 0179
1-way Test. Chi-Square Approx>malion
ChiSquare OF Prot»ChiSq
56102 1 0 0179
([Van der Waerden Test (Normal QuantilesT]
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
: 39 76-78 15 -3 160932 -0 21073 -1283
j| 39 94 ' 4 3 160932 0 225781 1 283
;; 2 Sample Test Normal Approximation
h S Z Prob>IZI
!| 3 1609324 1 28323 j 1994
! i way Test. Chi-Square Approximation
i ChiSquare DF Prob>ChiSq
1 6467 1 0 1994
Hudson River Database Reiease 3 5
Hot Spot 39
3 Z'98
-------
Mass per Unit Area Comparison Hot Spot 39
log10(MPA g/mA2) 1976-1978 vs 1994
rf \ Si
[LoglO(MPA) By Hotspot j
2.0
1.5
1.0
<
a.
2 0.5
o
?
J 0.0
-0.5
[Quantiles
Level
39 76-70
39 94
minimum
0.5087
-0.7617
10.0%
0.5723
-0.5559
2 5.0%
0.9253
0.9223
median 7 5.0%
1.0112 1.0477
1.2435 1.553325
[Oneway Anova )
[Summary of Fit j
RSquare
0.001446
RSquare Adj
-0.03554
Root Mean Square Error
0.550744
Mean ot Response
1.030048
Observations (or Sum Wgts)
23
ft-Test )
Estimate
Std Error
Lower 95%
Upper 95%
Difference
-0.04047
0.204663
-0.4604
0.379465
»-Test
-0.198
DF
27
Prob>ltl
0.8447
Assuming equal variances
[Analysis
of
Variance ]
Source
OF
Sum of Squares
Mean Square
F Ratio
Model
1
0.0118573
0.011857
0.0391
Error
27
8.1896165
0.303319
Prob>F
C Total
28
8.2014738
0.292910
0.8447
;
Means for Oneway Anova j
Level Number Mean Std Error
39 76-78 15 1.01051 0.14220
39 94 14 1.05098 0 14719
Std Error uses a pooled estimate of error variance
(Means and Std Deviations]
Level Number Mean Std Dev Std Err Mean
39 76-78 15 1.01051 0.279450 0.07215
39 94 14 1.05098 0.738831 0.19746
90 0% maximum
1 ^4456 1.7898 |
17533 1.7738 j
-------
(Means Comparisons ]
DII>M*»n[i]-M*»n[j] 39 94 39 76-78
39 94
0.000000 0 040465
39 76-7«
•0.04047 0 000000
Alpha. 0.05
Comparisons
for aft pairs using Tukey-Kramer HSD
2 05184
Ab*(Dlf)-L3D
39 94 39 76-78
39 94
-0.42711 0 37947
39 7S-7S
-0 37947 -0 41263
Positive values jrow pairs of means mat are significantly different
(Wllcoxon / Kruskal-Wallis Tests (Rank Sums))
Laval Count Score Sum Scora Mean (Mean*MeanO)/SldO
39 76-78 15 187 12.4867 -1 637
39 94 14 248 t7.7143 1 637
2-Sample Test. Normal Approximation
S Z ProD>IZI
248 1.63663 0 1017
1-way Test. Cht-Square Approximation
ChlSquara DF Prob>CMSq
2.7505 1 0.0972
[Median Test (Number ot Points Above Median) j
Laval Count Scora Sum Score Mean IZI
10 2.36858 0.0179
1*way Teat. Chi-Square Approximation
ChlSquara OF Prob>ChlSq
5 6102 1 0.0179
(Van dar Waerden Test {Normal Quantiles) J
Laval Count Scora Sum Scora Maan
-------
Mass per Unit Area Comparison
log10(MPA g/mA2)
Oredge Location 182
1976-1978 vs 1994
[LoglO(MPA) By Hotspot
^1
<
a.
3
• o-
/ N
j
j
182 76-78
Hotspot
182 94
All Pairs
Tukey-Kramer
0.05
(Quantiles )
Laval
182 76-78
182 94
minimum
•0 5113
0.8592
1 0.0%
•0.5113
0.8592
25,0%
0.314825
0.8592
median
0.82415
0.8819
75.0%
0.901175
0.9046
90.0%
0 9014
0.9046
maximum
0.9014
0.9046
(Oneway Anova
(Summary of Fit)
R Square
0.078023
R Square Adj
-0.07564
Root Mean Square Error
0.504662
Mean e» Response
0 661687
Observations (or Sum Wgts)
8
(l - T e s t ]
Estimate
Sid Error
Lower 95%
Upper 95%
Assuming equal variances
Difference
0 29362
0.41205
1 30188
0 71464
l-Taat
0 713
OF Prob>lt!
6 0.5029
(Analysis of Variance J
Source
Model
Error
C Total
DF
1
6
7
Sum of Square*
0.1293161
1 5280994
1 6574156
Mean Square
0.129316
0.25*683
0.236774
F Ratio
0,5078
Prob>F
0.5029
(Means for Oneway Anova]
Laval Number Mian Std Error
182 76-78 6 0 588283 0 206Q3
182 94 2 0.881900 0.35685
Std Ettor uses a pooled estimate a! error variance
[Means and Std Deviations j
Laval Number Mean Std Dev Std Err Mean
182 76-78 6 0.588283 0 552643 0.22562
182 94 2 0.881900 0.032103 0.02270
-------
[Means Comparisons I
Dlf = Maan[i]-M*an(j] 182 94 182 76-78
182 94 0.000000 C 293617
182 76-78 -0.29362 0 000000
Alpha= 0 05
Comparisons for all pairs using Tukey-Kramer HSD
q *
2 44692
Abs(Dif)-LSD 182 94 182 76-78
182 94 -1 23487 0.71465
182 76-78 -0 71465 -0 71295
Positive values show pairs of means that are significantly different.
(Wilcoxon / Kruskal-Wallis Tests (Rank Sums) j
Laval Count Scora Sum
102 76-78 6 24
182 94 2 12
2-Sample Tesl. Normal Approximation
S Z Prob>IZI
12 0.83333 0 4047
1-way Tesl. Chi-Square Approximation
ChlSquara OF Prob>ChlSq
1.0000 1 0 3173
Scora Mean (Maan-Maan0)/Std0
400000 -0.833
6 00000 0.833
(Median Test (Number of Points Above Median))
Level Count Score Sum Score Mean (Meen-Mean0)/Std0
182 76-70 6 3 0.500000 0.000
182 94 2 1 0.500000 0.000
2-Sampie Test. Normal Approximation
S Z Prob>IZI
1 0.00000 1.0000
1-way Test. Chi-Square Approximation
ChiSquare DF Prob>ChlSq
0.0000 1 1 0000
(Van der Waerden Test (Normal Quantiles) )
Level Count Score Sum Score Mean (Mean-Mean0)/Std0
182 76-78 6 -1.080930 0 18016 -1.094
182 94 2 1 080930 0.540465 1.094
2-Sample Test. Normal Approximation
S Z Prob>IZI
V08093 109355 0 2742
1-way Test. Chi-Square Approximation
ChiSquare OF Prob>ChiSq
1.1959 1 0.2742
Hudson River DataBase Release 3.5
Dredge Location 162
3/2/98
------- |