HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
VOLUME 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
ADDENDUM TO THE DATA EVALUATION AND INTERPRETATION REPORT
FEBRUARY 1999
st4f
<3* %
• .
f mm 1
iSEZi
For
U.S. Environmental Protection Agency
Region II
and
U.S. Army Corps of Engineers
Kansas City District
Book 1 of2
TAMS Consultants, Inc.
TetraTech, Inc.
-------
PRO^
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
REGION 2
290 BROADWAY
NEW YORK, NY 10007-1866
FE8 1 7 l"9
To All Interested Parties:
The U.S. Environmental Protection Agency (EPA) is pleased to release this Responsiveness
Summary to the Low Resolution Sediment Coring Report (LRC) for the Hudson River PCBs
Superfund site. This document contains written comments from various reviewers of the LRC
and the Agency's responses to significant comments. In addition, the appendices to the
responsiveness summary include, a comparison of sediment PCB inventory between 1984 and
1994 on an area basis, rather than a point-to-point basis; a revised estimate of the Thompson
Island Pool sediment PCB inventory; and, a revised estimate of the loads measured in EPA's 1993
water-column sampling based on corrections to the site database described in the December 1998
Responsiveness Summary for the Database Report, Preliminary Model Calibration Report, and
Data Evaluation and Interpretation Report.
EPA's careful consideration of comments received on the LRC, and the additional analyses
contained in the appendices to the responsiveness summary, support the overall conclusions of the
LRC. The area-to-area analysis in Appendix A calculated a level of loss of PCB mass from highly
contaminated sediments in the Thompson Island Pool that is similar to the loss estimated by a
point-to-point comparison in the LRC. EPA acknowledges that there is considerable uncertainty
surrounding the loss values in these estimates, but stresses that there is statistically significant loss
of PCB mass despite this uncertainty. EPA therefore believes that it is appropriate to reaffirm the
following general conclusions from the executive summary to the LRC:
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.
In other words, the PCB contamination in the Upper Hudson River continues to release PCBs to
the water column, and it does not appear that burial by clean sediment is occurring significantly
enough to resolve the problem.
Internet Address (URL) • http://www.epa.gov
Recycled/Recyclable • Printed with Vegetable Oil Based Inks on Recycled Paper (Minimum 25% Postconsumer)
-------
2
The Data Evaluation and Interpretation Report (DEIR) and the LRC are currently being peer
reviewed by a panel of independent experts. In addition to those reports, EPA is providing the
peer reviewers copies of the responsiveness summaries to the DEIR and the LRC. The reviewers
will discuss their findings at a meeting to be held on March 16, 17 and 18, 1999 at the Marriott
Hotel in Albany, New York.
The technical concerns raised by both the peer review and the public comment processes are
valuable to EPA's evaluation of the Hudson River system. We are pleased to provide you this
response to public concerns on the LRC.
Sincerely yours,
William McCabe, Deputy Director
Emergency and Remedial Response Division
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
VOLUME 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
ADDENDUM TO THE DATA EVALUATION AND INTERPRETATION REPORT
FEBRUARY 1999
A :
UsEi
% PRO^
For
U.S. Environmental Protection Agency
Region II
and
U.S. Army Corps of Engineers
Kansas City District
Book 1 of 2
TAMS Consultants, Inc.
TetraTech, Inc.
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
FEBRUARY 1999
CONTENTS
BOOK 1 OF 2 Pages
TABLE OF CONTENTS
LIST OF CORRECTIONS v
LIST OF APPENDICES v
LIST OF TABLES v
LIST OF FIGURES vi
LIST OF ACRONYMS vii
I. INTRODUCTION AND COMMENT DIRECTORY
1. INTRODUCTION CD-I
1.1 Recent Developments CD-2
2. REPORT COMMENTING PROCESS CD-2
2.1 Report Distribution CD-2
2.2 Review Period and Informational Meetings CD-2
2.3 Receipt of Comments CD-3
2.3.1 Comments on the Low Resolution Sediment Coring Report CD-3
2.4 Distribution of the Responsiveness Summary CD-6
3. ORGANIZATION OF COMMENTS AND RESPONSES TO COMMENTS . . . CD-7
3.1 Identification of Comments CD-7
3.2 Location of Responses to Comments CD-7
3.3 Types of Responses CD-8
4. COMMENT DIRECTORY CD-10
4.1 Guide to Comment Directory Responsiveness Summary CD-10
4.2 Comment Directory for the LRC CD-13
II. RESPONSES TO COMMENTS
RESPONSES TO GENERAL COMMENTS LRC-1
RESPONSES TO SPECIFIC COMMENTS LRC-4
EXECUTIVE SUMMARY LRC-4
ACRONYMS LRC-4
GLOSSARY LRC-4
i TAMS/TetraTech
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
FEBRUARY 1999
CONTENTS
BOOK 1 OF 2 Pages
1. INTRODUCTION LRC-4
1.1 Purpose of Report LRC-4
1.2 Report Format and Organization LRC-4
1.3 Project Background LRC-4
1.4 Background for the Low Resolution Sediment Coring Program . . . LRC-4
1.5 Low Resolution Sediment Coring Program Objectives LRC-4
2. SAMPLING DESIGN AND iMETHODS LRC-5
2.1 Technical Approach for the Low Resolution Sediment Coring
Program LRC-5
2.2 Field Sampling LRC-8
2.2.1 Sample Locations LRC-8
2.2.2 Sample Preparation LRC-8
2.3 Sample Analyses LRC-9
2.3.1 PCB Congener Analysis LRC-9
2.3.2 Radionuclide Analysis LRC-11
2.3.3 Total Organic Carbon and Total Kjeldahl Nitrogen LRC-11
2.3.4 Physical Properties LRC-11
2.4 Summary of Analytical Results LRC-11
2.4.1 PCB Congener Analysis LRC-11
2.4.2 Radionuclide Analysis LRC-20
2.4.3 Total Organic Carbon and Total Kjeldahl Nitrogen LRC-21
2.4.4 Physical Properties LRC-21
3. INTERPRETATION OF LOW RESOLUTION SEDIMENT
CORING RESULTS LRC-25
3.1 Comparison between the PCB Results for the Low Resolution Cores
and the High Resolution Cores LRC-25
3.2 Interpretation of the Relationships Among the Low Resolution
Core Parameters LRC-29
3.3 Interpretation of the Low Resolution Core and the Side-Scan
Sonar Results LRC-31
3.4 Summary of Chapter 3 LRC-31
4. AN EXAMINATION OF HUDSON RIVER SEDIMENT PCB
INVENTORIES: PAST AND PRESENT LRC-32
4.1 Sediment Inventories of the Thompson Island Pool LRC-33
n
TAMS/TetraTech
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
FEBRUARY 1999
CONTENTS
BOOK 1 OF 2 Pages
4.1.1 A Comparison of 1984 and 1994 Conditions LRC-41
4.1.2 Assessment of Sediment Inventory Change Based on
the Original 1984 VTri+ Sediment Inventory LRC-44
4.1.3 Assessment of Other Potentially Important
Characteristics LRC-61
4.1.4 Implications of the Inventory Assessment LRC-62
4.2 Sediment Inventories of the Upper Hudson Below the Thompson
Island Dam LRC-62
4.2.1 Calculation of the Length-Weighted Average Concentration
(LWA) and Mass Per Unit Area (MPA) for Sediment Samples
Below the TI Dam LRC-65
4.2.2 Comparison of 1976-1978 Sediment Classifications
and the Side-Scan Sonar Interpretation LRC-66
4.2.3 Comparison of Sediment PCB Inventories: NYSDEC
1976-1978 Estimates versus 1994 Low Resolution Core
Estimates LRC-68
4.2.4 7Be in Surface Sediments LRC-78
4.2.5 Hot Spot Boundaries LRC-86
4.2.6 Comparison of the 1994 Hot Spot Inventories with
Other 1977 Estimates LRC-86
4.3 Sediment Contamination in the Near-Shore Environment LRC-86
4.4 Summary and Conclusions LRC-86
4.4.1 Sediment and PCB Inventories in the TI Pool LRC-86
4.4.2 Sediment and PCB Inventories Below the TI Dam LRC-86
4.4.3 Sediment Contamination in the Near-Shore
Environment LRC-88
4.4.4 Summary LRC-88
REFERENCES LRC-89
iii
TAMS/'TetraTech
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
FEBRUARY 1999
CONTENTS
BOOK 1 OF 2 Pages
APPENDICES
APPENDIX A Data Usability Report for PCB Congeners Low Resolution
Sediment Coring Study LRC-88
APPENDIX B Data Usability Report for Non-PCB Chemical and Physical
Data Low Resolution Sediment Coring Study LRC-88
APPENDIX C 1994 Low Resolution Core and 1984 NYSDEC Core
Profiles for the Thompson Island Pool LRC-88
APPENDIX D 1994 Low Resolution Core Profiles Below the Thompson
Island Pool LRC-88
APPENDIX E Memoranda from John Butcher of TetraTech Inc.
Concerning Historical PCB Quantitation LRC-88
APPENDIX F Statistical Summary Sheets for Chapter 4 LRC-88
BOOK 2 OF 2
III. COMMENTS ON THE LOW RESOLUTION SEDIMENT CORING REPORT
Federal (LF-1)
State (LS-1)
Local (LL-1)
Community Interaction Program (LC-1 through LC-4)
General Electric (LG-1)
iv
T AMS/T etraT ech
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
FEBRUARY 1999
CONTENTS
BOOK 1 OF 2 Pages
LIST OF CORRECTIONS
Section 2.4.1- PCB Congener Analysis LRC-11
Section 4.1.1- A Comparison of 1984 and 1994 Conditions LRC-41
Section 4.1.2- Assessment of Sediment Inventory change Based on the
Original 1984 STri+Sediment Inventory LRC-44
Section 4.2.2- Comparison of 1976-1978 Sediment Classifications and the Side-Scan
Sonar Interpretation LRC-66
Section 4.2.3- Comparison of Sediment PCB Inventories: NYSDEC 1976-1978
Estimates versus 1994 Low Resolution Core Estimates LRC-68
LIST OF APPENDICES
APPENDIX A A Comparison of PCB Sediment Inventories in the Thompson Island Pool,
1984 to 1994
APPENDIX B Revised Estimate of the 1984 Thompson Island Pool Sediment PCB
Inventory
APPENDIX C Revised Estimates of PCB and Suspended Solids Loads In the Upper
Hudson River
LIST OF TABLES
2.4.1 Statistic Summary of RPD (%) as a Function of Depth and 137Cs
Presence in the Core Bottom LRC-17
LG-1.2 Review of 1994 Low Resolution Sediment Core Completeness LRC-34
LG-1.9 Subreach Variogram Models for Natural Log of PCB Mass
Concentration, 1984 Thompson Island Pool Sediment Survey LRC-6
LG-1.19B Surface PCB Concentrations in NYSDEC 1984 Data Compared to
Texture Class LRC-57
LG-1.22 An Examination of GE Monitoring Results for Jan. 9, 1998
Ft. Edward Flow at 34,300 cfs LRC-59
TAMS/TetraTech
v
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
FEBRUARY 1999
CONTENTS
BOOK 1 OF 2 Pages
LL-1.1 Parameter Comparison for Brown et a!., 1988 and LRC
Inventory Analyses LRC-39
LIST OF FIGURES
2.4.1 A Relationship Between RPD and Total PCB Concentration for Low
Resolution Core Field Splits LRC-15
2.4.IB RPD Comparison for Low Resolution Core Split Samples based
on 137Cs Presence LRC-16
2.4.1C RPD Comparison for Low Resolution Core Split Samples based
on Sample Depth LRC-18
LF-1.3A Depth Difference of PCB Maximum and the First Appearance of Wood
Chips in All Low Resolution Core Samples LRC-22
LF-1.3B Depth Difference of PCB Maximum and the First Appearance of Wood
Chips in the Low Resolution Core Sampling Excluding Incomplete and
One-Segment Cores LRC-23
LF-1,3C Total PCBs and the C/N Ratio in High Resolution Cores from the Upper
Hudson LRC-24
LG-1.18A Distributions of the 7BeV137Cs Ratio in Low Resolution Cores Tops .... LRC-80
LG-1.18B Relationship Between 137Cs and 7Be' in Low Resolution Core Tops .... LRC-81
LG-1.18C Distribution of Decay-Corrected 7Be' in Surface Sediment from
Low Resolution Cores LRC-82
LG-1.19B Correlation of TOC Concentration and Porosity in TI Pool Surface
Sediments LRC-56
LG-1.21 The Number of Chlorine per Biphenyl vs. the GE/HydroQual
Dechlorination Ratios for the High Resolution Core Data LRC-10
LG-1.26A The Relationship Between the Number of Chlorides per Biphenyl
and the Moar Dechlorination Product Ratio for the High Resolution
Core Data LRC-26
LG-1.26B Relationship Among MDPR, MDPR* and Fractional Change in
Molecular Weight for All Post-1954 Freshwater High Resolution Core
Sediment LRC-27
TAMS/TetraTech
vi
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
FEBRUARY 1999
CONTENTS
BOOK 1 OF 2 Pages
CORRECTED LRC FIGURES
2-4 Distribution of Total PCB Concentrations in Low Resolution
Sediment Core Samples (corrected) LRC-12
2-6 Precision in Total PCB Concentration for Low Resolution
Core Field Splits (corrected) LRC-13
4-2 High Resolution Core 19 from the TI Pool (corrected) LRC-42
4-7 1984 Trichloro and Higher Homologues as MPA vs Mass Difference
Relative to 1994 - Log Scale (corrected) LRC-43
4-10 Distribution of the Percent Change in PCB Molar Inventory (DeltaM)
(corrected) LRC-45
Vll
TAMS/TetraTech
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY FOR
Volume 2C-A LOW RESOLUTION SEDIMENT CORING REPORT
Addendum to the Data Evaluation and Interpretation Report
HUDSON RIVER PCBs REASSESSMENT RI/FS
LIST OF ACRONYMS
ASTM
7Be
7Be
cm
137Cs
DEIR
DN
ECD
GC
GE
IQD
ITD
kg
kHz
LQ
LRC
LWA
MDPR
MDPR*
mg
MPA
MPI
MVUE
American Society for Testing and MW
Materials ng
Beryllium-7 NPDES
Be Concentration Decay-corrected to
September 1, 1994 NYSDEC
Centimeter
Cesium-137 PCB
Data Evaluation and Interpretation Report ppb
Digital Number ppm
Electron Capture Detector QA
Gas Chromatograph QAPjP
General Electric RPD
Interquartile Distance RRT
Ion Trap Detector RSD
Kilogram s
Kilohertz SAP
Lower Quartile SAS
Low Resolution Sediment Coring Report SOP
Length-Weighted Average SSW
Molar Dechlorination Product Ratio TC
Sum of BZ#1, 4, 8, 10, & 19 over Sum of TCL
All Congeners (molar basis) TI
Sum of BZ#1 ,4, 10 & 19 only over Sum TKN
of All Congeners (molar basis) TN
Milligram TOC
Mass Per Unit Area ng
Malcolm Pirnie, Inc. UQ
Minimum Variance Unbiased Estimator USGS
Molecular Weight
Nanogram
National Pollution Discharge Elimination
System
New York State Department of Environ-
mental Conservation
Polychlorinated Biphenyl
Parts per Billion
Parts per Million
Quality Assurance
Quality Assurance Project Plan
Relative Percent Difference
Relative Retention Time
Relative Standard Deviation
Standard Deviation (also as SD)
Sampling and Analysis Plan
Special Analytical Services
Standard Operating Procedure
Solid Specific Weight
Total Carbon
Target Compound List (Organics)
Thompson Island
Total Kjeldahl Nitrogen
Total Nitrogen
Total Organic Carbon
Microgram
Upper Quartile
United States Geological Survey
viii
TAMS/TetraTech
-------
Introduction
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY
VOLUME 2C-A: LOW RESOLUTION SEDIMENT CORING REPORT
ADDENDUM TO THE DATA EVALUATION AND INTERPRETATION REPORT
FEBRUARY 1999
I. INTRODUCTION AND COMMENT DIRECTORY
1. Introduction
The United States Environmental Protection Agency (USEPA) has prepared this
Responsiveness Summary for Volume 2C-A: Low Resolution Sediment Coring Report (LRC) for
the Hudson River PCBs Reassessment Remedial Investigation/Feasibility Study (Reassessment)
which is an addendum to the Data Evaluation and Interpretation Report. It addresses significant
comments received during the review of this Report.
For the Reassessment, USEPA has established a Community Interaction Program (CIP) to
elicit feedback from the public through regular meetings and discussion and to facilitate review of
and comment upon work plans and reports prepared during all phases.
The LRC is incorporated by reference and is not reproduced herein. No revised copy of the
LRC will be published as such. The comment responses and revisions noted herein are considered
to amend the Report. For complete coverage, the Report and this Responsiveness Summary must
be used together.
The first part of this Responsiveness Summary is entitled "Introduction and Comment
Directory." It describes the Report review and commenting process, explains the organization and
format of comments and responses, and contains a comment index or directory.
The second part, entitled "Responses", contains the USEPA responses to significant
comments. Responses are grouped according to the section number of the Report to which they refer.
e.g., responses to comments on Section 2.1 are found in the "Responses" Section 2.1 of the
Responsiveness Summary. Additional information about how to locate responses to comments is
contained in the Comment Directory. Tables and figures for the responses are found within the text
of the responses in Book 1 of the Responsiveness Summary . Book 1 also contains the appendices
prepared for the Responsiveness Summary. Note that each appendix begins with a table of contents,
listing the Figures and tables contained in the appendix. The respective tables and figures are
contained at the end of each of the appendices
CD-I
rAMS.Tciral'cch
-------
The third part, entitled "Comments on the Phase 2 Report", contains copies of the comments
submitted to the USEPA on the LRC. The comments are identified by commentor and comment
number, as further explained in the Comment Directory. These comments are found in Book 2.
1.1 Recent Developments
Since the issuance of the LRC, further review of the Report has revealed certain errors in the
text and figures of the document. Corrections to these errors are provided under the "Responses"
section of the Responsiveness Summary under section in which the error occurred.
In addition, an alternative analysis to the comparison of the Thompson Island Pool (TI Pool)
sediment inventory in 1984 and 1994 is presented in Appendix A. This approach addresses a number
of criticisms of the LRC analysis. The analysis is area-based, examines the change in trichloro and
higher homologue inventories only, and uses a more normally-distributed function to assess the bulk
change in inventory.
A revised estimate of the TI Pool sediment PCB inventory in 1984 is presented in Appendix
B. This revised estimate incorporates texture information and results in inventory estimates for fine-
grained and coarse-grained areas of the TI Pool.
Appendix C presents a revision of the water column PCB transport analysis originally
presented in the Data Evaluation and Interpretation Report (DEIR) (USEPA, 1997). The revisions
reflect the current understanding of flow and PCB transport conditions in the Upper Hudson. In
particular, revised flow estimates for the Stillwater and Waterford monitoring stations as well as a
potential bias in the TI Dam monitoring station precipitated these revisions. The need for these
revisions was originally noted in the Responsiveness Summary forVolumes 2A, B, and 2C, USEPA
(1998). Several figures from the DEIR were revised as a result of these developments. The revised
figures are included in Appendix C.
2. Report Commenting Process
This section documents and explains the commenting process and the organization of
comments and responses in this document. To find responses to particularly comments, the reader
should go to the Comment Directory on page CD-I3.
TAMS/TetraTech
-------
2.1 Report Distribution
The LRC was distributed to federal and state agencies and officials, participants in the
Community Interaction Program (CIP), and General Electric, as shown in Table 1. Distribution was
made to approximately 100 agencies, groups, and individuals. Copies of the Report were also made
available for public review in 17 information repositories, as shown in Table 2.
2.2 Review Period and Informational Meetings
USEPA held a formal 30-day comment period on the LRC. although USEPA has welcomed
comments on the Reassessment throughout the study. USEPA held a Joint Liaison Group meeting
that was open to the public to present the Report. The meeting was held on July 23. 1998 in Albany,
NY.
Minutes for the meeting will be contained in a binder entitled Project Documents Binder.
This binder is part of the project information available for public review at 11 of the 17 information
repositories (Table 2). Four of the six repositories that do not currently have a Project Documents
Binder (Marist Library, R.I. Library. SUN Albany Library , and USA Library ) are partial repositories
maintained primarily for their CD-ROM capability. The other two. Sojourner Truth Library at SUN
New Pails, and the Sea Grant office in Kingston, will have copies of Project Documents Binders in
the near future .
As stated in USEPA's letter transmitting the Report, citizens are encouraged to participate
in the Reassessment process and to join one of the Liaison Groups formed as part of the Community
Interaction Program. USEPA requested that all comments, including those of Liaison Groups, be
sent to USEPA.
2.3 Receipt of Comments
Comments on the Report were received in two ways: letters or other written submissions to
USEPA; and written statements submitted as follow-up to oral statements made during the meetings.
2.3.1 Comments on the Low Resolution Sediment Coring Report
A total of 8 comment sets were received, submitted by one federal agency, one state agency,
one local government; four Community Interaction Program participants; and General Electric.
Federal agency comments consisted of one set from the National Oceanic and Atmospheric
Administration CN'OAA) (LF-1. 8.'28'98).
CD-3
1AMS TelraTech
-------
One set of comments was received from the New York State Department of Environmental
Conservation (L.S.-l, 8/31/98).
Local government comments were submitted by the Saratoga Environmental Management
Council (LL-1,8/28/98).
Comments were submitted by four members of the Community Interaction Program
including J. Sanders (member, Science and Technical Committee: IX-1. 8/31/98), George Pitman
(member, Science and Technical Committee; LC-2, 8/29/98), T.Borden (chairperson, Agricultural
Liaison Group; LC-3, 8/30/98), and M. Pulver (co-chair, Agricultural Liaison Group and Fort
Edward Town Board; LC-4, 8/31/98).
General Electric (LG-1) comments constituted virtually a free-standing Report, with 54 pages
of text plus 46 pages of tables and figures, as well as two additional appendices.
2.4 Distribution of the Responsiveness Summary
This Responsiveness Summary, like all other documents prepared for the Reassessment, has
been distributed to the members of the Steering Committee, the Hudson River PCB Oversight
Committee, the Scientific and Technical Committee, NYSDEC and General Electric. In addition,
copies have been sent to the peer reviewers for the DEIR and LRC. This Responsiveness Summary-
has also been placed in the 17 Information Repositories and will be included in the Administrative
Record.
CD-4
TAMS/TetraTtfch
-------
TABLE 1
DISTRIBUTION OF REPORTS
HUDSON RIVER PCBs OVERSIGHT COMMITTEE MEMBERS
USEPA ERRD Deputy Division Director (Chair)
USEPA Project Manager
USEPA Community Relations Coordinator, Chair of the Steering Committee
NYSDEC Division of Hazardous Waste Management representative
NYSDEC Division of Construction Management representative
National Oceanic and Atmospheric Administration (NCAA) representative
Agency for Toxic Substances and Disease Registry (ATSDR) representative
US Army Corps of Engineers representative
New York State Thruway Authority (Department of Canals) representativ e
USDOI (USF&W) representative
NYSDOH representative
GE representative
Liaison Group Chairpeople
Scientific and Technical Committee representative
SCIENTIFIC AND TECHNICAL COMMITTEE MEMBERS
STEERING COMMITTEE MEMBERS
USEPA Community Relations Coordinator (Chair)
Governmental Liaison Group Chair and two Co-chairs
Citizen Liaison Group Chair and two Co-chairs
Agricultural Liaison Group Chair and two Co-chairs
Environmental Liaison Group Chair and two Co-chairs
USEPA Project Manager
NYSDEC Technical representative
NYSDEC Community Affairs representative
FEDERAL AND STATE REPRESENTATIVES
Copies of the Reports were sent to relevant federal and state representatives who have been inv olved w ith
this project. These include, in part, the following:
The Hon. Daniel P. Moynihan - The Hon. Michael McNulty
The Hon. Alfonse M. D'Amato - The Hon. Sue Kelly
The Hon. Gerald Solomon - The Hon. Benjamin Gilman
The Hon. Nita Lowey - The Hon. Richard Brodsky
The Hon. Maurice Hinchey - The Hon. Bobby D'Andrea
The Hon. Ronald B. Stafford
17 INFORMATION REPOSITORIES (.vtv Table 2).
CD-5
TA.MS.TetraTcch
-------
TABLE 2
INFORMATION REPOSITORIES
Adriance Memorial Library
93 Market Street
Poughkeepsie, NY 12601
Catskill Public Library
1 Franklin Street
Catskill, NY 12414
A Cornell Cooperative Extension
Sea Grant Office
74 John Street
Kingston, NY 12401
Crandall Library
City Park
Glens Falls, NY 12801
County Clerk's Office
Washington County Office Building
Upper Broadway-
Fort Edward, NY 12828
* A Marist College Library
Marist College
290 North Road
Poughkeepsie, NY 12601
* New York State Library
CEC Empire State Plaza
Albany, NY 12230
New York State Department
of Environmental Conservation
Division of Hazardous Waste Remediation
50 Wolf Road, Room 212
Albany, NY 12233
* A R. G. Folsom Library
Rensselaer Polytechnic Institute
Troy, NY 12180-3590
Saratoga County EMC
50 West High Street
Ballston Spa, NY 12020
* Saratoga Springs Public Library
49 Henry Street
Saratoga Springs, NY 12866
* A SUN at Albany Library
1400 Washington Avenue
Albany, NY 12222
* A Sojourner Truth Library
SUN at New Pails
New Pails, NY 12561
Troy Public Library
100 Second Street
Troy, NY 12180
United States Environmental Protection
Agency
290 Broadway
New York, NY 10007
* A United States Military Academy Library-
Building 757
West Point, NY 10996
White Plains Public Library
100 Martine Avenue
White Plains, NY 12601
Repositories with Database Report
CD-ROM (as of 10/98)
Repositories without Project
Documents Binder (as of 10/98)
CD-6
TAMS.TelraTech
-------
3.
ORGANIZATION OF COMMENTS AND RESPONSES TO COMMENTS
3.1 Identification of Comments
Each comment submitted for a Report was assigned a dual letter code. The first letter
references the Report (L for LRC) for which the comment was addressed and the second letter was
used to denote one of the follow ing:
F - Federal agencies and officials;
S - State agencies and officials;
L - Local agencies and officials;
C - Community Interaction Program Committees and Liaison Groups; and.
G - General Electric.
The letter codes were assigned for the convenience of readers and to assist in the organization of this
document; priority or special treatment was neither intended nor given in the responses to comments.
Once a letter code was assigned, each submission was then assigned a number, in the order
that it was received and processed, such as L.C-1, LC-2 and so on. Each different comment within
a submission was assigned its separate sub-number. Thus, if a federal agency submitted three
different comments under the same cover, these are designated LF-1.1. LF-1.2. LF-1.3.
The alphanumeric code associated with each reprinted written submission is marked at the
top right corner of the first page of the comment letter; the sub-numbers designating individual
comments are marked in the margin, as shown in the sample letter in Section 4 of this introduction.
Comment submissions are reprinted in numerical order by letter code in the following order: F. S,
L. C, and G.
3.2 Location of Responses to Comments
The Comment Directory, following this text, contains a complete listing of all commentors
and comments. This directory allows readers to find responses to comments and provides several
items of information. In several cases, the name of the agency or organization of the commentors
has been abbreviated, as follows:
- NOAA National Oceanic and Atmospheric Administration
- NYSDEC New York State Department of Environmental Conservation
- SCEMC/GLC Saratoga County Environmental Management Council Governmental
Liaison Committee
- GE General Electric Company
The comment directory table is organized as follows:
• The first column lists the names of commentors. Comments are grouped first by: F
(Federal). S (State). I. (Local). C (CIP) or G (General Electric) preceded by a L for the
LRC.
CD-"
I'AMS.TetraTcch
-------
• The second column identifies the alphanumeric comment code, e.g., LF. 1 -1. assigned to
each comment.
• The third column identifies the location of the response by Report section number. For
example, comments raised on Section 3.2 of the Report can be found in the
corresponding Section 3.2 of the Responses section, following the third tab of this
document.
• The fourth, fifth, and sixth columns list key words that describe the subject matter of
each comment. Readers will find these key words helpful as a means to identify subjects
of interest and related comments.
Responses are grouped and consolidated by section number in order that all responses to
related comments appear together to help achieve consistency among the responses and for the
convenience of the reader interested in responses to related or similar comments.
In a few instances, several commentors commented on the same or very similar items. These
comments are answered by one common response that addresses the common issue being raised.
Thus, a comment is not necessarily answered by an individualized response.
In other cases, different comments pertaining to the same Report section are made. Thus,
a section number may contain more than one response.
3.3 Types of Responses
Responses to comments include the types described below.
• General Responses
In some instances, comments were general and pertained to the Reassessment process
or the Report overall rather than to a specific section of it. Responses to these
comments are coded as General and appear at the very beginning of the Responses,
under the heading General.
• Specific Responses to Comments
These comments are answered in the Responses, grouped by the number of the
section of the Report to which they refer. A common response is provided when
commentors question the same or very similar items. In some cases, commentors
voiced opposite opinions about the same point, typically a controversial one, but both
comments took issue with the same part of the Report. The rationale for the Report's
findings or resolution of the issue may be contained in a common response
addressing the conflicting nature of the comments and the controversy surrounding
the issue.
CD-8
TAMS/TetraTech
-------
Additional References
Full citations are provided only for new references not identified previously in the
References section of the LRC.
• Corrections
Corrections to the text are noted in the appropriate Report section. No subsequent
action will be taken since the Report will not be reissued. A list of corrections to
the LRC is included in the Table of Contents. Revised figures for the DEIR can
be found in Appendix C.
¦ Acronyms
A table of acronyms originally provided with the LRC has been updated to reflect
discussions in this Responsiveness Summary. The table immediately follows the
Table of Contents.
CD-9
TAMSTetralech
-------
4. COMMENT DIRECTORY
A Comment Guide, a sample comment letter, a diagram illustrating how to find responses
to comments, and the Comment Directory follow.
As stated in the preface to this Responsiveness Summary, this document does not
reproduce the LRC. Readers are urged to utilize this Responsiveness Summary in conjunction
with the Report in order to fully understand the comments and responses.
4.1 GLIDE TO COMMENT DIRECTORY
RESPONSIVENESS SUMMARY
1 Step 1
Step 2
Step 3 ;
Find the commentor or the key
; words of interest in the
! Comment Directory. Comments
are separated by commentor
group.
Obtain Comment Codes and
Report Section. Find coded
comments following the tab in
Book 2 of this Responsiveness
Summary.
Find the responses follow ing the I
Responses tab in Book 1. See the !
Table of Contents to locate the !
page of the Responsiveness i
Summary for the Report Section. J
1
j Key to Comment Codes:
' Comment codes are in this format XY-a.b
: X=Report (L=LRC)
Y=Commentor Group
: (F=Federal, S=State. L=Local. C=Community Interaction Program, G=General Electric)
a=Letter or report containing comments
b=Numbered comment
Example:
Comment Response Assignment for the DEIR
| AGENCY
Comment
REPORT
KEY WORDS
Name
CODE
SECTION
I ¦ _ j
NOAA 'Rosman LF-1.1 4.4.3 Near Shore Sediment Concentration
Find comment under tab "Federal (LF)".
Find response under tab "Response" on page LRC-88 where comments relating to Section 4.4.3
arc discussed.
CD - 10
TAMS TetraTL-ch
-------
U.S. DEPARTMENT OF COMMERCE LF-1
National Oceanic and Atmospheric
Administration
National Ocean Service
Office of Ocean Resources Conservation and Assessment
SAMPLE LETTER Hazardous Materials Response and Assessment Division
Coastal Resources Coordination Branch
290 Broadway, Rm 1831
New York. New York 10007
August 28, 1998
Doug Tomchuk
U.S. EPA
Emergency and Remedial Response Division
Sediment Projects/Caribbean Team
290 Broadway
New York, NY 10007
Dear Doug:
Thank you for the opportunity to review the July 1998 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. The following comments are submitted by the National Oceanic and Atmospheric
Administration (NOAA).
Comments
The Executive Summary of the Low Resolution Sediment Coring Report highlighted four major
findings. The results of the nearshore sediment investigation were reported as additional findings
and the significance of these findings was downplayed. It was stated that the implications from the
two to three times increase in the estimate of the exposure point concentration would be addressed
in the Human Health Risk Assessment Implications to the Ecological Risk Assessment (ERA)
were not discussed. It was suggested that this increased estimate of PCB concentrations in
nearshore sediments should not substantially change the human health risk estimate from wading
and swimming (pg. 4-44); however, it may have serious implications for human health exposure
from consumption of fish and for ultimate remedial decisions. Furthermore, the ERA risk to
ecological receptors must consider the potential underestimate of PCBs in the nearshore
environment.
Four nearshore areas were sampled in approximately 4 feet of water. The water depth was chosen
since it posed a likely human exposure from wading and swimming. These shallow nearshore
areas are also of particular importance to biota because they provide refuge, feeding and spawning
habitat for fish and are an important source of contamination to prey species. In addition, PCBs in
these sediments may be most affected by daily changes in water level associated with hydropower
generation, as well as being vulnerable to scour from large debris (e.g., logs, root masses), ice
scour, and other disturbances.
The Low Resolution Coring Report attempted to quantify the potential underestimation of PCB
concentrations in nearshore sediments, but conceded that the results usefulness may be limited due
to the small sample size (n=l 1). Using data from all nearshore fine-grained low resolution TIP
cores within 50 feet of the shoreline yielded a somewhat larger dataset (n=19) and a higher 95
percent confidence limit (264 ppm PCBs compared to 151 ppm PCBs). Side-scan sonar nearshor;
samples that overlapped with the shoreline appeared to have been excluded from this analysis even
though these are important areas to ecological receptors. The limited characterization of nearshore
-------
THIS PAGE LEFT BLANK INTENTIONALLY
cd -1:
TAMS-TetraTech
-------
Comment Directory
-------
4.2 COMMENT DIRECTORY FOR THE LRC
AGENCY/
NAME
COMMENT
CODE
REPORT
SECTION
1
KEYWORDS
2
3
NOAA/Rosman
LF-1.1
4.4.3
Near Shore
NOAA/Rosman
LF-1.2
4.2.3
PCB Deposits
Fish
NOAA/Rosman
LF-1.3
2.4.3
C/N Ratio
NOAA/Rosman
LF-1.4
2.4.1
Homogeneity
NYSDEC/Pons
LS-1.1
4.2.4
Burial
NYSDEC/Ports
LS-1.2
4.1.2
PCB Mass Loss
Estimates
NYSDEC/Ports
LS-1.3
4.2.3
PCB Inventories
Loss
NYSDEC/Ports
LS-1.4
4.2.3
PCB Inventories
Underestimates
NYSDEC/Ports
LS-1.5
4.1.4
Within TIP
PCB Inventory
NYSDEC/Ports
LS-1.6
4.4.3
Near Shore
PCB Concentration
Sediment
NYSDEC/Ports
LS-1.7
general
Human Health
Risk Assesment
Scope of work
SCEMC/Balet
LL-1.1
4.1
PCB Inventories
1977-1984
1984-1994
SCEMC/Balct
LL-1.2
4.1
PCB Mass
Stat. Variance
Heteroaeneitv
SCEMC/Balet
LL-1.3
4.1
Sampling Design
Heterogeneity
SCEMC/Balet
LL-1.4
2.1
Below TIP
SCEMC/Balet
LL-1.5
2.3.1
Modeline
Redox
TC/TN
SCEMC/Balet
LL-1.6
2.4.1
Spatiality
PCB Inventories
SCEMC/Balet
LL-1.7
2.4.1
PCB Burial
SCEMC/Balet
LL-1.8
2.4.1
Fig 2.4 Ref Incorr.
Units Incorrect
SCEMC/Balet
LL-1.9
4.2.3
36% PCB Cone.
SCEMC/Balet
LL-1.10
2.4.1
Replicates
Heterogeneity
Analytical Uncert.
SCEMC/Balet
LL-1.11
3.2
Var. of Parameters
High Res. Cores
SCEMC/Balet
LL-1.12
4.1/App. E
Analytical Methods
1984 v. 1994
Corr. Factors
SCEMC/Balet
LL-1.13
4.1. l/Fig4-2
Core Profile
SCEMC/Balet
LL-1.14
4.1.1
Scour
Distant Between
Cores
SCEMC/Balet
LL-1.15
4.1.1/Fig 4-7
MPA3+
Tri+ Loss
SCEMC/Balet
LL-1.16
4.1.3
Scour
Cohesive Seds.
Non-coh. Scds
SCEMC/Balet
LL-1.17
4.2.1
Grab Samples
Extrapolation
SCEMC/Balet
LL-1.18
4.2.1
Two Data Sets
Difference
SCEMC/Balet
LL-1.19
4.2.1
SSW, '76-78
Correlation
Uncertainty
SCEMC/Balet
LL-1.20
4.2.2/Fig 4-18
Side Scan Sonar
Sed. Classification
SCEMC/Balet
LL-1.21
4.2.2/Fis 3-28
Sed. Classification
Correlation
SCEMC/Balet
LL-1.22
4.2.3
Hot Spot Bound.
SCEMC/Balet
LL-1.23
Fig 4-19, 4-20, 4.2.3
Fig. Ref. Incorrect
SCEMC/Balet
LL-1.24
4.2.3
PCB Loss
1994 v 1976-78
SCEMC/Balet
LL-1.25
4.2.3
Burial
PCB Profile
SCEMC/Balet
LL-1.26
4.2.3
Burial
Long-term Storage
Lone-term Loss
Sanders. J.
LC-1.1
2.1
Sample Depth
Putman. G.
LC-2.1
4.2
Consistency
Statistics
Putman, G.
LC-2.1 A
aeneral
Variance Estimates
Putman, G.
LC-2.1B
aeneral
Model Calibration
CD- l.i
TAMSHY'iraTech
-------
AGENCY/
• NAME
COMMENT
CODE
REPORT
SECTION
1
KEYWORDS
2
3
Putman, G.
LC-2.2
general
PCB Mass Flux
Putman, G.
LC-2.3
1.3
DNAPL
Putman. G.
LC-2.4
4.2.3
PCB Mass Flux
Putman, G.
LC-2.5
4.2.3
PCB Mass Loading
Putman, G.
LC-2.6
4.2.3
Diffusional Loss
Borden, T.
LC-3.1
general
Water Column
Fish
Pulver, M.
LC-4.1
general
40% Loss
Pulver, M.
LC-4.2
2.3.1
# of Samples
Heterogeneity
Areal Coveraae
Pulver, M.
LC-4.3
general
Analytical
Methodology
PCB3+
1994 Total PCBs
Pulver, M.
LC-4.4
4.1
84 PCB3+/'94 PCB3+
Mass Loss Estim.
Pulver, M.
LC-4.5
4.1.2
9" Sample Inter.
Highest Cone.
Pulver, M.
LC-4.6
4.1.2
73% PCB Loss
Hot Spot
Pulver, M.
LC-4.7
general
Highest PCB Level
Top 9 Inches
Pulver, M.
LC-4.8
general
Peer Review
Pulver, M.
LC-4.9
general
Dredging
GE
LG-1.1
Section 4
1984 PCB Inventory
Statistical
Uncertainty
GE
LG-1.2
Section 4.1
Matched pairs
PCB Mass Estimate
GE
LG-1.2A
4.1.1
Sample Numbers
GE
LG-1.3
Section 4.1
Consistency
GE
LG-1.4A
4.1.2
Tri+ Inventory
PCB Mass Estimate
40% loss
GE
LG-1.4B
4.1.2
40 % Loss
GE
LC-1.4C
4.1.2
Other PCB Meas.
GE
LG-1.4D
4.1
Mechanisms
GE
LG-1.4E
4.1.2
PCB Mass Est.
40% Loss
GE
LG-1.5A
4.2.4
7Be
Burial
GE
LG-1.5B
4.2.3
Burial
GE
LG-1.5C
3.1
Dechlorination
GE
LG-1.5D
4.2.3
PCB Inventories
GE
LG-1.5E
4.1.2
Implausible
Mass Loss
GE
LG-1.5F
4.2.3
PCB Inventories
GE
LG-1.6
4.2.3
PCB inventories
GE
LG-1.7
4.2.3
l37Cs
PCB inventory
GE
LG-1.8
4.1
1984 Grab samples
PCB depth profile
GE
LG-1.9
Section 2.1
Sample Location
Heterogeneity
H-7 site
GE
LG-1.10
4.1,4.2
% mass change
PCB mass
GE
LG-1.11
4.1.2, Appendix E
PCB 3+(1984)
Total congeners
GE
LG-1.12
4.2,4.3
Geometric mean
Delta PCB
Arithmetic Mean
GE
LG-1.13
4.1.2
PCB mass loss
GE
LG-1.14
4.1.2
Implausible
Mass Loss
GE
LG-1.15
4.1.2
Fate and Transport
Mass Loss
CD-14
TAMS/TetraTech
-------
AGENCY/
NAME
COMMENT
CODE
REPORT
SECTION
1
KEYWORDS
2
3
GE
LG-1.16
4.1.2
Mass Loss
Water Column Data
GE
LG-1.17
No relevant section
(GE's 1998 data)
4.1.2
GE-1998 data
GE
LG-1.17A
4.1.2
PCB?.
GE-1998 Data
GE
LG-1.17B
4.1.2
PCBV
Fish
GE
LG-1.18
4.2.4
7Bc
Deposition
GE
LG-1.18A
4.2.4
7Be
Burial
GE
LG-1.18B
4.2.4
7Be
GE
LG-1.18C
4.2.4
7Be
GE
LG-1.19A
4.1.2
Water column PCB
Concentration
Areal Flux
GE
LG-1.19B
4.1.2
Water Column PCB
Concentration
Organic Carbon
GE
LG-1.20
4.1
Surface Sediment
Sediment-Water
Exchange Processes
GE
LG-1.21
2.3.1
Fish Flesh Studies
(NOAA)
PCB Congeners
Exposure ratio
GE
LG-1.22
4.1.2
January 1998 Flood
Water Column
GE
LG-1.23
Sect. 3.2
137Cs
Scour
GE
LG-1.24
4.4.2
PCB mass
Below TIP
Parametric Statistics
GE
LG-1.25
3.1
Dechlorination
GE
LG-1.26
3.1
Arochlor 1242
Mass loss
Dechlorination
GE
LG-1.27
3.1
Dechlorination
Log of the PCB
Concentration
GE
LG-1.28
4.1.2
Fate and Transport
Mass Balance
Model
GF.
LG-1.29
Section 2.1
Sample Selection
Bias
"Purposive"
Sampling
GE
LG-1.30
2 2 1
7 locations below TIP
GE
LG-1.31
3.1
Cross-Contamination
Rejected Data
GE
LG-1.32
2.4.1
Low Concentration
Data
Log-normal
Distribution
GE
LG-1.33
3.1
Purposive Data
Selection
GE
LG-1.34
4.2.1. Figure 4-17,
Table 4-3
SSW
GE
LG-1.35
4.2.1
Measurement
Methods
GE
LG-1.36
4.2.1
Grab Samples
Extrapolation
GE
LG-1.37
3.2
Regression
GE
LG-1.38A
4
MVUE method
Statistics
GE
LG-1 38B
2.4.1
Lognormal bias
GE
LG-1.38C
4.2.3
PCB inventory
Statistics
GE
LG-1.38D
4.2
Statistical methods
CD-15
TAMS/TetraTech
-------
AGENCY/
COMMENT
REPORT
KEY WORDS
NAME
CODE
SECTION
1
2
3
GE
LG-1.38E
4.1.2
Hot spots
Delta ratio
GE
LG-1.38F
4.2
Statistics
GE
LG-1.38G
4.2
Statistics
GE
LG-1.38H
4.1.2
Statistics
PCB inventorv
GE
LG-1.381
4.2.3
PCB inventorv
Zonal areas
GE
LG-1.38J
4.1.2
Uncertainty
PCB inventorv
GE
LG-1.39A
4.2.4
Be
GE
LG-1.39B
4.2.3
PCB inventory
PCB change
GE
LG-1.39 C
3.1
Dechlorination
GE
LG-1.40A
4.2.3
Burial
GE
LG-1.40B
4.1.2
Loss vs gain
GE
LG-1.40C
4.2
Statistics
GE
LG-1.41
general
Modelling
Sediment Transport
PCBs
CD-16
TAMS/TeiraTech
-------
Responses
-------
HUDSON RIVER PCBs REASSESSMENT RI/FS
RESPONSIVENESS SUMMARY
VOLUME 2C-A: LOW RESOLUTION SEDIMENT CORING REPORT
ADDENDUM TO THE DATA EVALUATION AND INTERPRETATION REPORT
FEBRUARY 1999
Responses to General Comments
Response to LC-2.1A
The issue of variance was addressed in response to comment DC-4.6 in the Responsiveness
Summary for Volumes 2A. 2B and 2C (USEPA, 1998b), which states as follows:
Each of the three sources of variance mentioned (variance shown in analyses from the site;
variance caused by sampling methods; and known physical sources of variance including cross
channel and vertical inhomogeneity in the PCB distribution in the river) are examined separately
below. The Phase 2 data were generated in such a way as to minimize unwanted sources of
variability so that the actual trends in the data would not be obscured.
For variability shown in analyses from Rogers Island, Phase 2 data are compared to General
Electric data for the same period. Figure 3-105 of the DEIR shows monthly PCB loads from above
Bakers Falls. Bakers Falls to Rogers Island, and Rogers Island to the TI Dam. The Phase 2 flow-
averaged estimates agree well with the General Electric estimates in both total load and distribution
for July through September 1993 (post construction at the Bakers Falls area).
The precision in the Phase 2 sampling methods is determined by comparing the split sample
data. Figure DC-4.6 shows the relative fractional differences of the ten split samples analyzed for
Total PCBs taken during the flow-averaged and transect events. Although the distributions are right
skewed, the median values are low at 0.10, 0.13, and 0.13 for the dissolved, suspended and whole
water samples, respectively. Each sample required a large volume of water in order to achieve the
low quantitation limits for PCBs, which in turn necessitated a long sampling period. This may be the
cause of the occasional high relative fractional difference.
The impact of physical sources of variance due to cross channel and vertical inhomogeneity
in PCB distribution in the Hudson River is shown in Figure 4-22 of QEA's March 1998 report,
Thompson Island Pool Sediment PCB Sources. General Electric's routine composite sample of east
and west channels at Rogers Island is compared to shallow and deep samples taken at six stations
in a river cross section just upstream of Rogers Island. This was performed on 2 separate occasions.
While there are differences among the samples, it is clear that the routine sample provides a
reasonable estimate of the Hudson River PCB concentration at this station. The routine sample is
comparable to the Phase 2 sample at Rogers Island which was stationed before the river splits into
east and west channels.
Response to LC-2.1B
Discharges from the Snook Hill and Moses Kill add only about five percent to the total flow
at the TI Dam. Contrary to the writer's assertion, the discharges from the Snook Kill and Moses Kill
are included in all flow and PCB load analyses presented in the DEIR. As far as their contributions
LRC • 1
TAMSTetrarech
-------
to suspended solid loads, this issue, including the HydroQual study, will be examined as part of the
Baseline Modeling Report. These tributaries do not contribute significantly to the Upper Hudson's
total PCB load. See also Appendix C of this Responsiveness Summary (Interpretation of the Revised
Estimates in the Upper Hudson, Spring Flood): and response to comment DC-4.7 in the
Responsiveness Summary for Volumes 2A, 2B and 2C.
Response to LC-2.2
See response to comment DC-4.6 in the Responsiveness Summary for Volumes 2A, 2B and
2C (USEPA, 1998b), which is reprinted in the Response to Comment LC-2.1A, above.
There has also been a careful examination of the loads at Rogers Island, as reported in
Section 3.4.2 of the DEIR. This section examines the biweekly data collected by GE and finds a
relatively consistent basis on which to estimate loads at Rogers Island. Note that the Phase 2 water
column sampling was conducted in 1993.
Response to LC-3.1 and LC-4.1
EPA never described the loss of approximately 40% of the PCB inventory from areas of high
concentration in the Thompson Island Pool as either a "crisis" or an "emergency." Nevertheless,
the conclusions reached in the LRC, particularly the loss of PCBs from the Thompson Island Pool
sediments coupled with the lack of widespread burial of contaminated sediments, are serious because
they imply that PCB contamination in sediments is spreading from areas of high concentration into
the rest of the river.
Although it is true, as suggested in Comment LC-3.1, that the Low Resolution Sediment
Coring Report relies on previously collected data, the complex and time-consuming analysis of that
data was completed only shortly before the LRC was issued.
Response to LC-4.3
After learning the results of the LRC analyses indicating the loss of approximately 40% of
the PCB inventory from areas of high concentration in the Thompson Island Pool, and of the lack
of widespread burial of contaminated sediments, it was entirely appropriate for the Agency to
evaluate whether some action could be taken to prevent the spread of contamination before an
overall remedy for the site is selected. In fact, it is a standard practice in the Superfund Program
to examine site information as it is generated and analyzed to determine if some expedited action is
necessary to contain or prevent the spread of contamination from a known and identified source,
even before an overall remedy is chosen for the entire site. After carefully reviewing the available
options for an early action, however, on December 17, 1998, EPA notified the public that the
Agency was not able to identify a feasible and appropriate interim action that would have a
significant impact on the loss of PCBs from Upper Hudson River sediments.
LRC-2
TAMS/TetraTech
-------
Response to I.C-4.7
The depth in the sediment to which fish and other organisms are potentially exposed to PCBs
will be investigated further in the Baseline Modeling and Ecological Risk Assessment reports.
Response to LC-4.8
The LRC will undergo peer review between January' and March, 1999.
Response to LC-4.9
Remedial action alternatives for the Hudson River PCBs site will be addressed in the
Feasibility Study.
Response to 1,0-1.41
This comment is simply a status report of GE's sediment modeling efforts. The USEPA
appreciates these contributions and will consider them under the comments to the Baseline Modeling
Report. Note that the USEPA has constructed its own sediment transport model.
Response to LS-1.7
Comments on the Human Health Risk Assessment Scope of Work will be addressed in the
responsiveness summary for that document.
LRC • 3
TAMS/TctraTech
-------
Responses to Specific Comments
EXECUTIVE SUMMARY
ACRONYMS
GLOSSARY
Ao significant comments were received on the Executive Summary, Acronyms or Glossary.
1. INTRODUCTION
1.1 Purpose of Report
1.2 Report Format and Organization
No significant comments were received on Sections 1 through 1.2.
1.3 Project Background
Response to I.C-2.3
The USEPA agrees that the loads originating above Rogers Island have decreased markedly
since the early 1990s. However, loads originating within the TI Pool have not. Additionally, while
DNAPL was shown to be present at the release areas near the GE Hudson Falls facility, its presence
has not been demonstrated at Rogers Island or at points downstream. See also the response to
comment DG-1.3. third paragraph, in the DEIR Responsiveness Summary (USEPA, 1998b). Note
that the USEPA Phase 2 water column sampling was conducted in 1993.
1.4 Background for the Low Resolution Sediment Coring Program
So significant comments were received on Section 1.4.
1.5 Low Resolution Sediment Coring Program Objectives
No significant comments were received on Section 1.5.
LRC-4
TAMS TctraTcch
-------
2.
SAMPLING DESIGN AND METHODS
2.1 Technical Approach for the Low Resolution Sediment Coring Program
Response to LC-1.1
The USEPA acknowledges the suggestions made by the writer concerning sampling
techniques and data interpretation. It should be noted that the primary purpose of the low resolution
coring program was to examine inventory changes and not to establish release mechanisms, howev er.
Response to Lfi-1.9
In this commenLthe writer critiques the degree of certainty associated with each of the paired
TI Pool sampling locations. However, the writer incorrectly assesses the degree of uncertainty in
location. In both the NYSDEC and the USEPA studies, the uncertainty in location was estimated to
be ± 3 ft. These uncertainties are properly combined as follows assuming the stated error to represent
2 standard deviations about an individual point:
Total uncertainty -2* \jj>!22 - 3/22 = 4.2ft
Thus, the writer exaggerates the degree of uncertainty in the location differences. In addition, since
this uncertainty is presumably random, some locations are undoubtedly closer and some further than
estimated. Thus, the best estimate of the average distance between 1984 and 1994 sampling locations
remains the mean distance of separation or 3.7 ± 4.2 feet, and not the 10 feet stated by the writer. It
should be noted as well that 56 of the 60 paired locations (93 percent of the locations) were separated
by less than 8 feet, which is within the mean plus its uncertainty.
The writer also contends that net changes with the clusters of samples are simply random.
As shown in Plate 4-20 of the Report as well as by the area-based analysis included as Appendix A
of this Responsiveness Summary, this is not true. The majority of the clusters show loss whether
considered on an individual location basis or as whole clusters. This is also evident in Figure 4-5
of the Report which shows both the Total PCB as well as the Tri-1- inventory to be lower in 1994
relative to 1984. Thus, the losses are not random within the clusters although there is some
variability. The data as presented by the writer shows this quite clearly, with many of the plots
showing mass loss for the entire cluster. That the absolute amount of mass loss is not constant is to
be expected since mass loss is expected to be dependent upon the amount of mass originally present
(as well as other factors). Thus, higher values have greater mass loss.
Further, the writer also contends that the H-7 area represents the only data set useful for
estimating close range variability. This is not true. The H7 site is located in the northern part of the
Thompson Island Pool, where a high degree of heterogeneity and limited spatial correlation is
expected based on the variogram analysis presented in the DEIR. Hot spots reoccupied for the Low
Resolution Sediment Coring Study, however, were primarily from the more southerly portions of
the Thompson Island Pool, where spatial correlation is much stronger (see Table LG-1.9. which is
based on Table 4-4 of the DEIR). Specifically, only 4 of the 60 paired locations are found in the
region near H-7.
l.RC • 5
TAMSTetraTech
-------
Table LG-1.9
Subreach Variogram Models3 for Natural Log of PCB Mass Concentration,
1984 Thompson Island Pool Sediment Survey
Subreach 5
1163000 -
1170100 N
Subreach 4
1170100 -
1177000 N
Subreach 3
1177000 -
1181900N
Subreaches
1 and 2
1181900 -
1191700 N
Observations
235
320
238
321
Nugget
0.750 (.284)
0.484 (.154)
0.0 (-).
1.54 (.108)
Sill-Nugget
1.520 (.282)
1.092 (.153)
1.733 (.060)
0.203 (.106)
Practical Range (ft)b
340 (75)
280 (68)
286(49)
582 (521)
Anisotropy Ratioc
1.0
1.5
2.5
1.0
Major Axisd
N 10° W
N 35° W
-
Note:
a. Variograms are exponential models, showing fit along the major axis and anisotropy ratio. Standard
errors of the coefficients from the least squares estimation are shown in parentheses.
b. A value of 2 times the practical range was used as the length of the major axis of the polygon
associated with each 1994 location. This distance represents the distance of separation at which
variance between point pairs approaches that of the population as a whole.
c. This ratio represents the ratio of the major axis over the minor axis of the ellipse associated with each
sampling point.
d. This represents the orientation of the major axis. Essentially this orientation causes the ellipse to be
oriented in the direction of river flow. This angle is not defined when the anisotropy ratio is unity (1).
Source: USEPA, 1997
T AMS/T etraT ech
-------
The variogram analysis in the DEIR addresses primarily long-range spatial correlation, with
practical ranges ofvariograms extending up to 521 feet. This scale primarily addresses the coherency
of hot spot areas, i.e., it confirms that high MPA values tend to be found together within discrete hot
spots, as is appropriate to estimate total PCB mass. Small scale spatial correlation was not addressed
in the DEIR, but evidence as to shorter scales is available in the 1984 sediment survey. While most
samples in the 1984 survey were collected on a 125-foot sampling grid, NYSDEC also investigated
finer-scale variability. This included 19 clusters of 5 samples each on 10 foot separations (four
cardinal directions around a central point), and a 10-core east-west transect at 10 foot intervals. A
total of 321 samples pairs are available at a separation distance of less than 30 feet within the 1984
Thompson Island Pool sediment data set.
Analysis of the variogram of closely spaced log-transformed MPA data suggests that the
short range structure has a sill near 1.0 at a separation distance of 20 feet. Note that this sill
(equivalent to the local variance estimate) is considerably less than the sill of larger scale structures
on a subreach basis, which ranges from 1.73 to 2.27 (DEIR, Table 4-4). A Gaussian model was fit
to the short-range variogram structure of 1984 log-transformed MPA (all samples), with the
following characteristics:
Practical Range (a) 25 feet
Sill -Nugget (C,) 0.778
Nugget (C0) 0.254
The unstandardized Gaussian variogram model may be defined on separation distance h as
y(h) -- C. - (C0 - c.y
exp
/ 3 h2
in which the practical range, a, is the distance at which the variogram value is 95% of the sill. The
correlogram (cross-correlation) function between points for h greater than zero is
C -exp
/ -3h2^
P(A)
V
C0 + C,
Sixty five percent (39 out of 60) of the 1984-1994 pairs were collocated within 4 feet, at which
distance the correlation coefficient is 0.70. Ninety five percent (57 out of 60) are collocated within
10 feet, at which distance the correlation coefficient is 0.47. This contrasts markedly with the
comment that "The semi-variogram showed that PCB levels m ere correlated only within a distance
of 5 feet... Beyond 5 feet, the variance was near that of the population. . even within the distance of
5 feet, the correlation was weak... " The latter conclusions apply only to the very heterogeneous H7
site, and are clearly not applicable to other hot spots within the Thompson Island Pool.
The remaining contentions of the writer concerning the rejection of various sample pairs to
yield a smaller set of paired coring locations are thus refuted by the discussion above. Nonetheless.
l.RC • ¦
TAMSTetraTech
-------
the suggestion by the writer that an area-based comparison be made for the various clusters of 1994
coring locations has merit. Appendix A of this Responsiveness Summary presents such an analysis.
• This analysis yields a similar degree of statistically significant mass loss as that presented in the
original Low Resolution Sediment Coring Report.
Response to LG-1.29
In selecting its coring sites, the USEPA attempted to minimize sediment variability and focus
on those areas best quantified by both USEPA and NYSDEC. In this regard, areas which were
screened but otherwise unanalyzed were avoided. Additionally, since spatial correlation was clearly
evident in the 1984 data, it was most prudent to take advantage of this fact by placing the samples
relatively close together while matching the previous sampling locations. Since the chief goal of the
sampling program was to establish the degree of change and not to re-establish sediment inventories,
it was most important that the 1994 effort match the locations of the 1984 program as much as
possible to reduce variability due to factors other than true inventory change. As stated in the Report,
the main conclusions apply to fine-grained sediment and not to the entire Pool. More importantly,
the document states that it is the occurrence of inventory loss itself and not the absolute magnitude
of this loss which is the most important conclusion of the Report, thus refuting any scenario wherein
sediment burial and PCB inventory gain in Upper Hudson sediments are assumed to represent the
major fate-and-transport processes. The samples as collected and analyzed in the LRC and this
Responsiveness Summary provide a sufficient basis for this conclusion.
Response to LL-1.4
The area represented by the cores below the Thompson Island Dam is approximately 526,000
m2. A breakdown of these areas by hot spot is given in Table 4-8 of the LRC. The near-shore area
of the TI Pool can be approximated as two six-mile-long strips of river bottom extending about 50
feet from shore. This area is approximately 300,000 m2. These area can be compared to the hot spot
areas of the TI Pool itself as estimated by Tofflemire and Quinn, 1979), which total approximately
520.000 m2.
2.2 Field Sampling
2.2.1 Sample Locations
Response to LG-1.30
For the areas below the TI Dam, the hot spot boundaries formed the basis for the sampling
design, not the dredge locations. The dredge boundaries were made available to USEPA after the
sampling effort was completed and simply served as an alternate basis for comparison.
2.2.2 Sample Preparation
No significant comments were received on Sections 2.2 through 2.2.2.
LRC-8
TAMS/TetraTech
-------
2.3 Sample Analyses
2.3.1 PCB Congener Analysis
Response to LC-4.2.
The USEPA believes that some, if not all, of the fish body burdens are derived from PCBs
associated with the sediments. The pathways for fish exposure may be direct, as in a benthic food
pathway, or indirect via sediment release to the water column, with subsequent direct absorption
across the gills, or via a water column (pelagic) food pathway. In either case, the PCB contamination
in fish is at least partially derived from the sediments. The fact that the sediments are losing
inventory over time does not require that fish levels increase over time. In fact, it is expected that
since fish body burdens are related to the sediments, fish body burdens would decrease over time,
paralleling the decline in the sediments. Fish data collected to date indicate some decline over time
despite the recent releases from the GE facilities. It now appears that fish levels have returned to the
conditions seen in the years just prior to the 1991 GE release event, a period where water column
and fish levels were most likely governed by the sediment PCB inventory. Thus there is no
inconsistency in the low resolution coring results and the historical trend in fish body burdens. The
issue with the sediments is the length of time required for the remaining inventories to be either re-
released and transported downstream or truly buried. To date, there is little evidence to suggest that
the sediment releases are declining to any substantive degree despite the major reduction in the GE
discharges from the Hudson Falls facility. The issue of the PCB congener fingerprint in fish will be
discussed in the Ecological Risk Assessment. However, the ratios used by GE and referenced here
by the writer do not provide a sufficient key to link the fish to recently released PCBs. See response
LG-1.21 concerning the applicability of the fish ratios.
Response to LG-1.21
The writer suggests that proposed congener ratios can be used to directly link fish body
burdens to their PCB source. GE's proposed use of several, specific congener ratios to determine
the degree of dechlorination was addressed in detail in responses to comments DG-1.19 and DG-1.20
in the Responsiveness Summary for Volumes 2A, 2B and 2C. As explained in those responses. EPA
does not believe that the suggested congener ratios are good predictors of the degree of
dechlorination. Thus, EPA does not believe that these congener ratios are useful in determining the
source of PCBs.
While the figure provided by the writer does show the trend of the mean ratios with chlorines
per biphenyl, the graphs oversimplify the wide range of variability. This is illustrated in Figure
LG-1.21 ( a copy of Figure DG-19A from the DEIR responses). The very wide range in values
suggests that these ratios may not have the predictive power assigned to them by the writer. Note for
example that the 56:49 ratio can vary more than an order of magnitude at 3 Cl/BP. Additionally, the
fish samples collected in 1993 in the TI Pool may have been directly influenced by the large,
water-borne loads present from 1992-1993 relative to other food web-related pathways.
Nonetheless, there may be some use for these and other congener ratios in the ecological
investigation. These issues will be further explored later in the Ecological Risk Assessment.
l.RC - 9
FAMSTetraTcch
-------
Upper Hudson Samples
Upper Hudson Samples
0
«
01
c
o
a
n
3
o
9— 74:49
¦0 592 » 0.367x R'= 0.309
o
o
3.5
3 2.5 2
Cl/BP
1.5 1
66:49
y = -0.822 + 0.447x R = 0.408
Ck°S-.Q5
O
Oft
: o
3 2.5 2
Cl/BP
Upper
>-
X
Hudson Samples
>•=-0.137 ~0.0836X R:= 0.13
60:49
CD Q
Cl/BP
Upper Hudson Samples
>• = -0 489 + 0.262x R:= 0.298
X
c
o
JS
zj
-------
Response to LL-1-5
The reduction/oxidation potential data and the total carbon/total nitrogen data are not used
directly in the model. Much of the data used for modeling comes from long term monitoring
programs of the Hudson River. The reduction/oxidation potential data and the total carbon/total
nitrogen data, like all of the data gathered from the low resolution cores, were to be used to gain an
understanding of the processes affecting PCBs in the river. See response to LG-1.28.
2.3.2 Radionuclide Analysis
2.3.3 Total Organic Carbon and Total Kjeldahl Nitrogen
2.3.4 Physical Properties
No significant comments were received on Sections 2.3.2 through 2.3.4.
2.4 Summary of Analytical Results
2.4.1 PCB Congener Analysis
Correction to Section 2.4.1 - PCB Congener Analysis
Figure 2-6 is incorrectly referenced as Figure 2-4 on page 2-18. The text of the second full
paragraph on page 2-18 should read:
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 2-6 shows the level of
precision attained for field replicates. The average RPD was 36 percent, and the
median RPD was 27 percent.
The units on Figure 2-4 were corrected to read mg/kg in the revised figure. The values of the abscisa
of the upper figure in Figure 2-6 are corrected in the revised Figure 2-6, included in this Report.
Based on the discussions at the peer review meeting in January of 1999 as well as those with
the NYSDEC, the issue of the low resolution coring RPD and its potential correlation with other
parameters was explored. The RPD can be considered an integrating measure of uncertainty and so
the parameter is useful in estimating the uncertainty of the entire sampling and analysis process.
Because the RPD is a relative measure of uncertainty, it tends to increase near the "edges" of the
measurement range, that is, near the analytical detection limit and the analytical maximum
quantitation level. In particular, the RPD tends to increase when the absolute measurement error
becomes large relative to the amount being quantified. In any given data set, it is anticipated that the
sample pairs with the highest RPD values represent samples with concentrations very close to the
LRC-ll
TAMS/TetraTech
-------
c
3
O
U
c
3
O
U
All Core Segments
200"
200 400 . 600 800 1000 1200
Total PCB Concentration (mg/kg)
1400
-3-2-10123
Log (Total PCB Concentration (mg/kg))
Shallow Sediment
140
120
100
80'
I 60
U
40
20"
1 f 1
0 200 400 600 800 1000 1200 1400
Total PCB Concentration (mg/kg)
0 12 3 4
Log (Total PCB Concentration (mg/kg))
Source: TAMS/Gradienl Database, Release 3.5
TAMS
Figure 2-4 (corrected)
Distribution of Total PCB Concentrations in Low Resolution Sediment Core Samples
-------
o
U
o
U
12
10-
8 -
6"
4 "
2"
20 40 60 80 100
Relative Percent Difference (RPD)
t—i—i—i—|—i—r
l—|—i—i—i—r
0.5 1 1.5 2
Regression Slope of Field Split Pairs
BZ#52 Normalized Congener Results
2.5
Source: TAMS/Gradient Database, Release 3.5
TAMS
Figure 2-6 (corrected)
Precision in Total PCB Concentration for Low Resolution Core Field Splits
-------
detection limit. Figure 2.4.1 A shows the relationship between RPD and Total PCB concentration
for the low resolution sediment coring program. The figure demonstrates the tendency of lower RPD
levels corresponding to higher Total PCB concentrations. This trend, while statistically significant,
is not a very strong one and much of the variation in the RPD is unrelated to concentration. The
results indicate that PCB concentrations greater than 20,000 jag/kg (20 ppm) have a mean RPD of
28 percent while those higher than 50,000 ng/kg (50 ppm) have a mean RPD of 17 percent.
Given the issue of potential cross-contamination discussed in the LRC and its relationship
to 137Cs, the l37Cs levels of the split pairs were examined. This is illustrated in Figure 2.4.1 A as black
and white symbols, with black symbols indicating the bottom core layers from cores lacking :37Cs
in the core bottom, much as was done in Figure 3-3 of the LRC. White symbols represent all top and
middle segment split pairs plus the bottom core segments where ,37Cs was present in the core bottom.
This figure and all subsequent figures in this discussion represent 22 of the 23 low resolution
sediment core split pairs. One sample, LR-016-2436, was excluded from this analysis due to its lack
of detected congeners. Only two congeners were detected in both of the split samples for this core
segment. All of the remaining core split pairs had at least 23 congeners detected in both split
samples. Sample LR-016-2436 apparently represented PCB-free sediments underlying the
contaminated, recent deposition. Its Total PCB concentration was less than 1 |ig/kg. The other split
pair samples had concentrations in the range of 135 to 83,000 ng/kg.
In the LRC, the data analysis showed that bottom core segments from cores lacking 137Cs
in the core bottom frequently had higher dechlorination ratios than would be predicted from the
sediment PCB concentration. This was attributed to the occurrence of cross-contamination by
overlying core segments or to the sample stratigraphy wherein a small layer of PCB-bearing
sediments would have to be blended with a larger, uncontaminated region of the core segment. Both
cross-contamination and incomplete mixing of a small quantity of PCB-bearing sediment would
yield the divergence from the expected trend (see Section 3.1 of the LRC). Similar to the finding in
the LRC for the entire low resolution coring data set, the bottom core segments in the low resolution
coring split sample pairs which lack 157Cs in the core bottom are the segments most likely to have
a high RPD. This is consistent with the interpretation of these core segments as described above. The
distinctly higher RPD levels are illustrated more clearly in Figure 2.4. IB which shows a set of "box
and whisker" diagrams for the coring data grouped on n7Cs absence or presence. Despite the small
size of the group lacking "7Cs, it is still significantly higher in RPD than the remaining samples. This
difference is statistically significant and is summarized in Table 2.4.1.
The absence of l37Cs marks the bottom core segment as a likely candidate for this concern.
This is made clearer when the RPD is examined as a function of depth. In Figure 2.4.1C, the split
pair data are grouped by position within the core. The diagram shows a clearly lower RPD value for
the top core segments relative to the underlying ones. This difference, while not statistically
significant, indicates that the top core segments, where the majority of PCB inventory was shown
to occur, have the lowest RPD levels. The mean and median RPD levels for these samples are 22 and
20 percent, respectively. This represents a substantial improvement over the low resolution sediment
core data set as a whole (37 percent mean RPD).
This consideration plus the general trend toward lower RPD at higher sediment
concentrations indicate that the individual sediment inventory estimates presented in the LRC (which
integrate the top and underlying core segments) have a much lower degree of uncertainty than was
LRC - 14
T' MSTetiaTech
-------
i i i i i i
' > ¦ ' ' : ' ' -• 'III
120
100
80
g,
Q
CL
aC
60-
40-
20-
0
0.1
RPD (%) = 72.451 - 23.901 log(Total PCB (mg/kg)) R"= 0.40
Legend
o
o
o
* 1?7Cs Nondetect"
° I37Cs Detect2
o
c
I O | I 1 I I I I
10
100
1000
Split Mean Total PCB (mg/kg)
Notes:
1. Mean Total PCB concentration for field split pair.
2. Open symbols represent top and middle core segments as well as bottom core segments
when 137Cs was detected at the bottom of the core. Closed symbols represent bottom core
137
segments when Cs was not detected in the core bottom.
Hudson River Database Release 4.1 TAMS/TetraTech
Figure 2.4.1A
Relationship Between RPD and Total PCB Concentration for Low
Resolution Core Field Splits
-------
o
(L
DC
125-
100"
75-
50-
25-
I
Bottom Layer witl
,37Cs Absent
Top and Middle Layers
plus Bottom Layer When
137Cs is Present
All Pairs
Tukey-Kramer
0.05
Legend
Percentile
90 _
Median
. . Inner Quartile Distance
Mean +1 Std. Error
Mean
Mean -1 Std. Error
Data Points
10
Note: 1. Separation of circles indicates that the two groups are statistically different
Hudson River Database Release 4. J
Figure 2.4.1B
TAMS/TetraTech
RPD Comparison for Low Resoution Core Split Samples
Based on 137Cs Presence
-------
Table 2.4.1
Statistic Summary of RPD (%) as a Function of Depth and 137Cs Presence in
the Core Bottom
Events
No. of Samples
Mean
RPD1 (%)
Median
Std. Error
Depth
Top Layer
9
22.1
20
6.1
Middle and Bottom
Layers
14
45.8
34.7
10
137Cs
Present2
Top and Middle
Layers plus Bottom
Layer when,37Cs is
Present
17
26.9
20
6
Bottom Layer,
137Cs Absent
5
71.3
62.3
17.8
Note:
1. RPD is based on the Total PCB concentration in low resolution sediment core field split pairs.
2. Samples are grouped according to the presence or absence of l37Cs in the core bottom. The
first group includes all top and middle core segments plus the bottom core segments when
l3'Cs was detected in the bottom core slice. The second group represents the the bottom core
segments where "'Cs was not detected in the bottom core segment.
-------
125
100 —
75"
50-
25-
All Pairs
Tukey-Kramer
Middle and Bottom Layers
Top Layer
0.05
Legend
Percentile
90 _
75
Median
25
10
Inner Quartile Distance
Mean +1 Std. Error
Mean
Mean -1 Std. Error
Data Points
Note: Lack of separation of circles indicates that the two groups are not statistically different
Hudson River Database Release 4.1 t-.. <% « ¦, TAMS/ TetraTech
Figure 2.4.1C
RPD Comparison for Low Resoution Core Split Samples
Based on Sample Depth
-------
suggested by the original RPD analysis presented in the LRC. Based on all above analysis, it can be
concluded that poor homogenization or cross-contamination of deeper core segments characterized
by lower PCB concentrations are the principal contributors to the relatively high RPD levels reported
for the low concentration core samples. It is likely that the majority of the PCB mass estimates from
the low resolution cores have uncertainties closer to 20 percent rather than the 37 percent originally
presented in the LRC.
Response to LF-1.4
No response required.
Response to LG-1.32
The exclusion of low-level samples in order to find quantitative relationships among Total
PCBs and other parameters is an attempt to exclude samples wherein the expected relationships are
unlikely to apply. In these instances, issues of accurate quantitation, sample cross-contamination and
related factors are likely to mask the true relationships of PCBs with other parameters. It may have
been appropriate for USEPA to have used different criteria to select and exclude these data, but these
exclusions have little impact since only one of the many relationships examined was converted to
a quantitative basis. Specifically, only solid-specific-weight was evaluated quantitatively for use in
estimating 1977-78 inventories. As such, it had a very limited impact since the density factor only
varies about ±30 percent about its average value.
Response to LG-1.38B
The writer correctly identifies an inconsistency between the formulas on page 4-28 and Table
4-13. Table 4-13 was mislabeled, and should have noted that the formula represents a simpler,
approximate estimate of the mean as given by Gilbert (1987) and not the MVUE. It should be noted
here that the estimator is applied consistently here and therefore sufficient for the comparisons made.
Nonetheless, the USEPA acknowledges that the MVUE would have provided a more rigorous
comparison.
Response to LL-1.6
The complete sentence to which the comment refers is, "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." These cores were selected to
understand the change in PCB inventory in the hotter areas of the fine-grained sediments, i.e. local
conditions. This study was not designed to or used to generate a 1994 Thompson Island Pool PCB
inventory nor are the data used in that manner in the LRC. It is strictly used to surmise the direction
and degree of change in a limited number of representative areas. In this context, the data set is
sufficient.
LRC -19
TAMSTetraTech
-------
Response to LL-1.7
The 28 Phase 2 high resolution cores were collected from areas of relatively continuous
sedimentation of fine-grained material along the length of the Hudson River. Twelve historical
sample collection sites which had previously produced high quality cores with readily interpretable
analytical results were reoccupied. To select the remaining 16 locations, 55 preliminary cores were
subjected to screening for radionuclide abundance to ascertain the capability of the sediment to
produce an interpretable profile. This rigorous selection process was designed to collect datable
cores; thus, the majority of the cores show a characteristic total PCB profile with a peak at depth.
This does not mean that all cores collected from the fine-grained areas of the Hudson River have this
profile.
The majority of the low resolution cores have the PCB maxima in the top-most core layer.
Since these samples are homogenized from zero to nine inches below the surface, the exact location
of the peak cannot be ascertained. The peak concentrations could occur anywhere within the ten
inches, even at the surface. It cannot be inferred by the profiles of the selected high resolution cores
that the peak concentrations are always buried by several inches of cleaner sediment. More
importantly, the occurrence of the PCB maxima in the 0-9 inch interval directly refutes the GE
contention that PCBs are being "deeply" buried since at most, the peak is only a few inches below
the surface. Only occasionally was the peak overlain by nine or more inches of less contaminated
but still not clean sediments.
In addition, PCBs within the top few inches of the surface in shallow near-shore areas are
subject to disturbance by watercraft. Although scour by high-flow events may be an unlikely
transport mechanism in these areas, this does not preclude the possibility of PCB transport to the
water column through other currently less-well-defined mechanisms. For instance, since it is
relatively clear that the current TI Pool load is not produced by a flow/scour process, other processes
must work to create the water-column load in the TI Pool.
Response to LL-1.8
USEPA acknowledges the error in the LRC figure. This correction has been noted in the
Correction to Section 2.4.1 of the Low Resolution Sediment Coring Report.
Response to LL-1.10
See response to comment LG-1.6 for discussion of analytical uncertainty. See response to
comment LG-1.1 for discussion of the validity of the 1984 versus 1994 inventory analysis given the
number of sampling points.
2.4.2 Radionuclide Analysis
No significant comments were received on Section 2.4.2.
LRC - 20
TAMS/TetraTech
-------
2.4.3 Total Organic Carbon and Total Kjeldahl Nitrogen
Response to LF-1.3
In response to this comment USEPA has prepared a further analysis of the wood chip and
PCB data available for the low resolution cores. The field geologist's descriptions concerning the
absence or presence of wood chips were retrieved from the database and correlated with the
occurrence of the PCB maximum. This was accomplished by subtracting the upper depth of the core
segment containing the PCB maximum from the upper depth of the first core segment found to
contain wood chips. Thus core segments with coincident wood chips and PCB maxima were
assigned a value of zero. If the PCB maximum overlay the wood chip occurrence, a negative value
was obtained. If the converse was true, a positive value, was obtained. The results for all low
resolution cores are presented in Figure LF-1.3A. It is clear from this figure that the PCB maximum
and the appearance of wood chips are coincident (difference equal to 0) in the majority of the low
resolution cores, given their coarse resolution. The next largest group is those cores where wood
chips were not reported. The next most important groups are those occurring at about 10 inches
separation, close to the typical slice thickness. However, these non-zero difference cores scatter
relatively equally about the value of 0, indicating no definitive trend. These results are consistent
with the conclusion based on the C/N ratio in the LRC, that is, woody materials are present
throughout Upper Hudson sediments and do not reliably predict the C/N ratio.
This issue was explored further by removing those cores which might affect the distribution
due to their special conditions, i.e.,one segment cores (one PCB analysis) and incomplete core (cores
where the PCB maximum depth is uncertain). These results are presented in Figure LF-1.3B. The
relationships are the same as noted in Figure LF-1.3A.
As a last analysis, the C/N ratio was examined for the high resolution cores along with the
total PCB levels. These results are shown for six cores in Figure LF-1.3C. These profiles show that
the C/N maximum predates the PCB maximum in every case, indicating that the ratio is not a good
predictor for Total PCBs. The results also show that major wood chip releases occurred prior to the
onset of GE operations, probably commensurate with wood processing operations in the early part
of this century.
2.4.4 Physical Properties
No significant comments were received on Section 2.4.4.
LRC-21
TAMS/TetraTech
-------
-30 -20 -10 0 10 20 No Wood
Depth difference (inches) Chips Present
Note:
1. Depth difference is defined as the top of the core segment with the PCB maximum
minus the top of the first segment found to contain wood chips. Thus a core with a PCB maximum
in the top layer (0-9 inches) and a first reported occurence of wood chip in the next layer
(9-18 inches) would have a depth difference of (0-9) or -9.
Hudson River Database Release 4.1 TAMS/Tetra Tech
Figure LF-1.3 A
Depth Difference of PCB Maximum and the First Appearance
of Wood Chips in All Low Resolution Core Samples
-------
2
O
U
0 ^
*
-30 -20 -10 0 10 20
Depth Difference (inches)
No Wood
Chips Present
Note:
1. Depth difference is defined as the top of the core segment with the PCB maximum
minus the top of the first segment found to contain wood chips. Thus a core with a PCB maximum
in the top layer (0-9 inches) and a first reported occurence of wood chip in the next layer
(9-18 inches) would have a depth difference of (0-9) or -9.
Hudson River Database Release 4.1 TAMS/Tetra tech
Figure LF-1.3B
Depth Difference of PCB Maximum and the First Appearance
of Wood Chips in the Low Resolution Core Samples
Excluding Incomplete and One-Segment Cores.
-------
Log (Total PCBs (ppb))
1 2 3 4 J 6 7
20 '
40 "
?
o
£ 60
J
80
100
—r-; ¦ i i i ' 1 , 1 I r-r-n
10 15 20 25 30 35 40
C/N Ratio (Molar)
RM 159.0
Log (Total PCBs (ppb))
1 2 3 4 5 6 7
ii>. i i i I .
20
-i
_40 -
I J
£60^1
- J
80
100-h—
10 15 20 25 30 35 40
C/N Ratio (Molar)
RM 185.8
Log(Total PCBs (ppb))
12 3 4 5 6 7
10 15 20 25 30 35 40
C/N Ratio (Molar)
RM 166.3
LogfTotal PCBs (ppb))
2 3 4 5 6
o -
1
1
r
C
£
20-
-
r
tit
h
X
-
i
40-
-
|
B
~
-
E 60-
80-
100 I ' r • • | . 1 I pr-i—, •¦>!<<
10 15 20 25 30 35 40
C/N Ratio (Molar)
RM 188.5
Log (Total PCBs, (ppb))
1 2 3 4 5 6 7
11 i . i . I .
20-
. 40-
* 60-
80-
100-
§
X
t-
r
T»
40
ON Ratio (Molar)
RM 177.8
Log (Total PCBs, (ppb))
2 3 4 5 6
' ¦ ¦ i . i 1 i , i l .
20-h
40-
§ :
£60-
80-
100-
10 15 20 25 30 35 40
C/N Ratio (Molar)
RM 1893
Hudson River Database Release 4.1
TAMS/TetraTech'
Figure LF-1.3C
Total PCBs and the C/N ratio in High Resolution Cores from the Upper Hudson
-------
3.
INTERPRETATION OF LOW RESOLUTION SEDIMENT CORING RESULTS
3.1 Comparison between the PCB Results for the Low Resolution Cores and the High
Resolution Cores
Response to LG-1.5C
The USEPA does not dispute that recent release events from the Hudson Falls facility have
added PCBs to the Pool inventory. If the GE data show that the April-September 1991 measurements
at the TI Dam were more dechlorinated, this may be due to a number of factors including the
addition of new material to the Pool sediments. More importantly however, as shown in the DEIR,
the congener pattern of the net gain of congeners to the water column across the Pool has not
changed in a consistent manner over the seven year period, suggesting that there has been no
substantive change in the nature of the PCBs responsible for the Pool load gain.
Response to LG-1.25
This Report was intended to examine changes in the PCB inventory of the sediments of the
Upper Hudson. The issues raised by the writer will be examined and addressed in later Phase 2
Reports, in particular, the Ecological and Human Health Risk Assessment Reports. It should be
noted, however, that the extent of dechlorination is found to be greatest in highly-contaminated
sediment, while it is relatively minor in sediments with low levels of PCB contamination typical of
most of the Hudson.
Response to LG-1.26
It has been demonstrated that the change in molecular weight (AMW) is algebraically related
to the number of chlorines per biphenyl (Cl/BP) (see response DG-1.19 of the Responsiveness
Summary for Volumes 2A, 2B and 2C). It has also been shown in both the DEIR and the LRC that
the relationship between the AMW and the molar dechlorination product ratio (MDPR) is linear (R:
= 0.90 or higher), meaning that the dechlorination products are produced in direct proportion to the
degree of chlorine mass loss from a sediment mixture. Therefore, since these measures track each
other and are directly related to Cl/BP, they directly reflect the degree of dechlorination in a sample.
The quality of these measures can be contrasted with the relationship between Cl/BP and the ratios
proposed by GE. This is illustrated by comparing Figure LG-1.21 and Figure LG-1.26A. The ratios
proposed by the writer yield R: values of 0.4 or less, indicating the substantially greater scatter and
therefore reduced sensitivity in these relationships. In particular note the broad ratio ranges seen
around 2.5 to 3 Cl/BP, levels commonly found in the Hudson.
The discussion concerning the insensitivity of the MDPR to the conversion of BZ#8
(2,4'-dichlorobiphenyl) to BZ#1 is misleading. This conversion is only important at extremely high
degrees of dechlorination. At levels more typical of Hudson sediments, BZ#8 forms an important
intermediate whose inclusion in the MDPR enhances the usefulness of this ratio. This is illustrated
in Figure LG-1.26B which compares the MDPR and MDPR*. MDPR* is defined as the sum of
BZ#1,4,10 and 19 (excluding BZ#8) over the sum of all congeners on a molar basis. Figure LG-
1.26B shows that the MDPR* also tracks the change in molecular weight although not as well as the
orginal MDPR. However, both ratios are clearly superior to the ratios proposed by the writer.
LRC-25
XMS/TetraTech
-------
y = 1.62 - 0.453x R = 0.94
0.8 -
.2 g
1 Q
S 0.6
o w
!S °
j—< •-*
o
0)
cd
Q *
H O 0.4
*2 3
£ o
0.2 -
0
<§>
© cp
'g/i ' i i i i i i i i i
3.5 3 2.5 2 1.5 1
Cl/BP
Note:
Data from Phase 2 Upper Hudson samples shown.
Hudson River Database Release 4.1 TAMS/TetraTech
Figure LG-1.26A
The Relationship Between the Number of Chlorines per Biphenyl
and the Molar Dechlorination Product Ratio for the High Resolution Core Data
-------
0 2 —
Reproduced from Figure 4-21 of the DEIR
WD
S
£
U
o.i
£ 3
SJD ^
'3 I*
£ -
^ ti
O
L*
<
o
.2
u
*©
~ -5-0 1 —
Theoretical Relationship
A1232 •
Leeend
A 0 6
Regression Fit
+ Kqunaleni to number of chlonnes per biphenyl (sec response LG-1.19).
T
02
04
06
0 8
IBZ#1.4.10. & 19
Z AH Congeners
(MDPR*)
10.0795
Hudson River Database Release 4.1
TAMS/TetraTech
Figure LG-1.26B
Relationship Among MDPR, MDPR* and Fractional Change in Molecular
Weight for All Post-1954 Freshwater High Resoution Core Sediment
-------
Ultimately, it is the close correlation between the AMW and the MDPR which validates
USEPA's methodology. The AMW expression, just as Cl/BP. integrates the entire sample for the
degree of dechlorination. The strong linear relationship between the MDPR and AMW indicates that
the MDPR also represents the degree of dechlorination in the sample and that the conversion of
BZ#8 to BZ#1 does not substantively affect the usefulness and sensitivity of the MDPR to the
dechlorination process.
Response to LG-1.27
The correlation between extent of dechlorination and PCB mass is well established by the
analyses presented in the DEIR. In the discussion presented there, it was shown that below 30
mg/kg, the degree of dechlorination was not predictable and that many samples found below this
level did not show substantive levels of dechlorination. These samples had low values for AMW, and
high values for Cl/BP, indicating that the dechlorination process was unimportant for reducing
sample mass or toxicity in these instances. This included nearly all of the Lower Hudson as well as
locations from the Upper Hudson. The possibility that a few select congeners may be rapidly
dechlorinated in the sediment at low concentrations does not change the fact that the majority of the
heavier congeners remain intact in the sediments at low concentrations.
The removal of the "cross-contaminated samples" was predicated on the strong relationship
developed from the high resolution cores. The arguments presented in the LRC are sufficient to
justify- the exclusion of these samples in the examination of the dechlorination ratios. Note, however,
that these samples were nol excluded from the sediment inventory calculations performed in Chapter
4 of the Report.
Response to LG-1.31
The criteria for selection of data to be rejected were presented in detail in the Report and
provides a sufficient basis for their exclusion. The know ledge of the geochemical processes affecting
PCBs and their relationships in the sediment is an essential foundation necessary before any
statistical tests can be applied. This is an important precept, since relationships between PCBs and
other parameters cannot be discerned if the data set contains many sediments deposited prior to
appearance of PCBs in the Hudson.
Nonetheless, the most important conclusions of the Report stem from the analyses described
in Chapter 4 of the LRC. In these analyses, no samples were excluded based on the rejection criteria,
i.e.. all core sections were included in creating the mass-per-unit-area (MPA) values subsequently
used in estimating the degree of change over time. Thus the comment does not apply to these
analyses.
Response to LG-1.33
The data analysis presented in Figures 3-2 and 3-8 of the LRC does not represent data
censoring. In fact the difference in the data sets represented in these figures is the result of a model
used to predict those points which were likely candidates for cross-contamination. PCB
concentration was not used as a criterion in selecting the points excluded from Figure 3-8. This is
extensively discussed in the text and intervening figures. This analysis was done to confirm an
LRC • 28
TAMSTetraTech
-------
already proven relationship {i.e., the relationship between total PCB mass and the degree of
dechlorination). This was developed as a part of a prior study completed on Hudson River sediments
for the US EPA. As such the data presented in Figure 3-8 were intended to check for consistency with
the previous relationship and not prove the original premise.
The US EI PA disagrees with the writer's contention that the low resolution coring results do
not confirm the findings of the DEIR that dechlorination is proportional to Total-PCB concentration.
Specifically, both the 1994 low resolution coring and 1992 high resolution coring data sets provide
nearly identical regressions between the change in molecular weight (AMW) and the dechlorination
product content of the sample (MDPR) to that predicted by the simple dechlorination model. This
relationship will only hold if the dechlorination process does not involve the inner chlorine atoms
on the PCB molecule. The minor differences between the two regressions for the two data sets is
probably attributable to minor changes in the analytical techniques and are not indicative of real
differences between the two data sets. Thus the high resolution core finding in this regard is
confirmed by the low resolution results.
Similarly, the low resolution cores results, when corrected for what is arguably a likely
cross-contamination issue, yield a relationship between the degree of dechlorination and the total
PCB mass of the sample which parallels that of the high resolution cores. This confirms the general
high resolution core result finding, that dechlorination increases with total PCB mass. Subsequently,
when the impact of the different sampling techniques are considered, it is clear that the parallel
relationships converge to a single relationship. While this step does not constitute a confirmation,
it does represent an interpretation of the data which is consistent with the original premise. Thus the
low resolution coring data confirm both general findings and are consistent with the specific finding
in the latter case.
3.2 Interpretation of the Relationships Among the Low Resolution Core Parameters
Response to LG-1.23
b7Cs cannot be directly applied in the manner suggested for the same reason that surface sediment
PCB levels cannot be used to infer scour. Both ,3,Cs and PCBs are particle-reactive agents whose
distribution in the sediments is based upon deposition rates, deposition history, history of release and
bioturbation. For this reason, the high resolution cores cannot be used to establish the l37Cs level of
sediments of comparable age throughout the Pool. This would be the same as using these cores to
establish the sediment PCB levels throughout the Pool. As is well known, PCB levels (and
presumably "Cs levels) vary widely in the Pool. Thus the use of surface l?"Cs levels as a measure
of the depth of scour in a theoretical high resolution core does not establish the approximate depth
(or time horizon) of scour since the initial level of :::Cs in the coring location is not known. Based
on the difficulty of finding suitable coring locations, it is unlikely that most areas experience the
same levels of ,;,Cs as seen in the high resolution cores. Indeed, even among the high resolution
cores sediment :3:Cs levels can vary by nearly an order of magnitude. See the surface n€s in the
region between RM 185 and 195 on Figure 3-63 of the DEIR. An important compounding factor is
the degree of bioturbation at a site. This process serves to homogenize i:,'Cs levels and reduce and
broaden the peak levels.
IRC - 29
TAMSTetraTech
-------
The net result is that surficial ;37Cs cannot be used a priori as a measure to determine the
depth scour, or as suggested by the writer, to exclude the possibility of scour without knowing the
detailed variation of 137Cs in the core. The premise proposed by the writer would be useful in an ideal
setting but must be rejected here as a basis to eliminate scour as an important mass loss process.
Response to LG-1.37
The commentor addresses the fact that, for any pair of measurements subject to uncertainty,
initial high values will tend to decrease, and initial low values will tend to increase, on resampling.
This "regression toward the mean" effect results because uncertainty, whether due to uncertainty in
resampling location or to analytical uncertainty, results in a sample being more like the local mean
value than the other member of the pair under comparison.
The USEPA acknowledges that the "regression toward the mean" effect exists, and should
have been noted within the Report. This issue would be more significant if the split of the samples
at an MPA of 10 g/m2 was arbitrary and unsupported by geophysical evidence. In fact, the split point
is logical, as "the greater-than-10-g/m2 group corresponds to sediments typically found in hot spot
areas (LRC, page ES-4)." Thus, the analysis which was performed is consistent with a geophysical
hypothesis that mass loss has occurred from concentrated hot spots, with some of the mass being
locally redistributed into less contaminated areas. If this mechanism is accepted as reasonable, it is
not appropriate to perform statistical analysis of the data without stratifying on the basis of whether
1984 sediment samples were representative of hot spot or non-hot spot conditions.
The fact is clear that higher concentration sediment samples show a statistically significant
decline in PCB inventory from 1984 to 1994 (Figures 4-12 through 4-14 jof the LRC). A part of this
sample change is due to actual mass loss, and a part may be due to the "regression toward the mean"
effect. For the "regression toward the mean" effect to constitute a major portion of the difference,
however, would require the presence of significant uncertainty due to either analytical uncertainty
or location errors.
The effect of analytical uncertainty has first been minimized by examining multiple measures
(MPA, DeltaM, DeltaPCB), where the latter two ratios are designed to diminish the importance of
analytical variability, as described in the response to comment LG-1.39B. More importantly, it
should be noted that the estimates of sediment inventory' change in the Thompson Island Pool are
based on comparison of 1994 Total PCB mass per unit area to 1984 estimates of Tri+ mass. Clearly,
monochlorbiphenyl and dichlorobiphenyl mass was present in 1984, although not measured. The
1984 MPA estimates are thus known to be biased low, which will minimize resulting estimates of
mass loss. Indeed, among the higher-concentration 1992 high resolution coring samples, Tri-
represented less than 50 percent of the total PCB mass present. The mass loss estimates presented
in the Low Resolution Sediment Coring Report are conservative. In this manner, the USEPA has
presented a strong test for mass loss from the more-contaminated sediment areas which should not
be affected by the analytical uncertainty.
The commentor appears to attribute significance to the "regression toward the mean" effect
primarily based on locational uncertainty, stating that "the geostatistical evidence from the 1984
survey shows that very short-scale spatial variability is often comparable to total variability." This
statement is untrue within the areas sampled in 1994. Indeed, short-scale spatial variability was not
LRC - 30
TAM S/T etraT ech
-------
examined in the 1997 DEIR, and cannot be evaluated from the figures presented in that Report. The
response to comment LG-1.9 shows that samples are highly correlated within the short range of
locational uncertainty applicable to reoccupying the 1984 sample locations.
In sum, the commentor has correctly pointed out an additional source of uncertainty which
could result in a slight high bias in the estimated amount of mass loss from highly contaminated
sediments between 1984 and 1994. This uncertainty does not, however, invalidate the general
finding of mass loss from higher concentration hot spot sediments.
Response to LL-1.11
The Phase 2 high resolution cores were selected from areas of relatively continuous
sedimentation of fine-grained material. As a result, the bulk sediment properties, grain-size
distribution and some chemical parameters, such as total organic carbon, show little variation. The
correlation with Total PCBs is poor for these samples, because the material is all of one type. It is
known that l37Cs is correlated with Total PCBs due to the similar deposition histories of these
parameters. This relationship is displayed in Figures 3-53 through 3-55 of the DEIR (USEPA, 1997)
which show the concentration Total PCBs and 137Cs plotted versus depth.
In contrast, the Phase 2 low resolution cores were selected to obtain new estimates of the
sediment PCB inventories at a number of locations in the TI Pool and to refine the PCB mass
estimates for a limited number of historic hot spot locations below the TI Pool. These objectives did
not require the cores to be of only fine-grained sediment. The low resolution sediment core samples
show a wider range of values for the bulk sediment and chemical properties. These value can be
correlated with Total PCBs and measures of dechlorination.
3.3 Interpretation of the Low Resolution Core and the Side-Scan Sonar Results
No significant comments were received on Section 3.3.
3.4 Summary of Chapter 3
A'o significant comments were received on Section 3.4.
LRC • 31
TAMSTctraTcch
-------
4. AN EXAMINATION OF HUDSON RIVER SEDIMENT PCB INVENTORIES: PAST
AND PRESENT
Response to LG-1.1
The writer asserts that the various estimates created in the USEPA Reports for the inventor)'
of PCBs in the TI Pool sediments are largely the result of the sediment heterogeneity. While the
sediments are certainly heterogeneous in their PCB content, this is not the reason for the large
differences between the estimates given in Brown et al., 1988 and the Data Evaluation and
Interpretation Report (USEPA, 1997). Rather, the wide differences are the result of the assumptions
used to estimate the sediment inventory. The estimate by Brown et al. is largely based on subjective
evaluation of the data, which while useful, may yield an overestimate of the actual inventory. As part
of the DEIR, the sediment inventories were estimated at 14.5 and 19.3 metric tons, based on two
separate statistical techniques, kriging analysis and Theissen polygonal declustering, respectively.
Of these techniques, the kriging analysis is the more rigorous since it examines how contamination
varies as a function of distance. These techniques are described in detail in the DEIR. Neither
technique as applied in the DEIR was able to directly account for the sediment textures mapped by
the side scan sonar. When this information is used in conjunction with the Theissen polygonal
declustering, the inventory estimate obtained is 14.7 metric tons, essentially the same as that
obtained via the kriging analysis. Appendix B of this Responsiveness Summary describes the new
analysis.
The writer also asserts that the number of samples collected was too small to discern a
difference between the 1984 and 1994 collection events. This again is incorrect. In the analysis of
the low resolution coring data, various statistical techniques were applied to test for the statistical
significance of the difference between the 1984 and 1994 data. These tests take into account the
number of samples available. The Report only presents those differences which were shown to be
statistically significant at the 95 percent confidence level. The 95 percent confidence level is the
gLxierally accepted significance level for most statistical tests.
Response to LG-1.38A
The writer is correct in noting that the Minimum Variance Unbiased Estimator (MVUE)
method is based on an underlying log-normal distribution. Minor deviations from a true log-normal
distribution introduce minor errors to the MVUE. However, the writer's concern over data censoring
is unwarranted. As discussed in the text as well as in response LG-1.33, the data sets used in the
mass inventory estimates did not exclude any sample data. Additionally, no inventor)' estimates were
nondetect. It is important to note that underlying data distribution is not known except through the
data collected. The fact that the sample distribution is not perfectly log-normal does not disprove the
log-normality of the underlying population. Thus, the use of the MVUE is justified in light of the
greater probability that the underlying population is log-normal. This is evident in LRC Table 4-6.
which illustrates the high probability that the sediment PCB data are log-normally distributed.
The writer's contention that the simple arithmetic mean is preferable to the MVUE due to the
uncertainties in the population's shape is untrue. Specifically, if the underlying population is
log-normal, as has been suggested by many different studies of PCB contamination in Hudson River
sediments, then the use of the MVUE provides a minimum variance estimate (which the arithmetic
LRC - 32
TAMSTetralech
-------
mean is not, (Gilbert, 1987)) and is preferred, particularly when the sample set is small and likely
to be log-normal. Small data sets typically do not sufficiently capture the complete population
distribution and tend to under represent the values far from the mean. As a result, the simple
arithmetic mean may not provide the best estimate of the true mean. Further evidence for an
underlying log-normal distribution is found in LRC Table 4-6 which shows all of the hot spots to
have a rather high probability of a log-normal distribution. In each case, the test for normal
distribution yielded lower probabilities. Lastly, as is evident in LRC Table 4-7. there is little
difference between the two estimators and thus, the writer's concern makes little difference in the
ultimate result. The statistical analysis only yielded significant differences when the mass values
varied more than a factor of 2 so that the small (roughly 5 percent) difference between the estimators
is unimportant.
4.1 Sediment Inventories of the Thompson Island Pool
Response to LG-1.2
The writer asserts that a list of conditions must be met in order for the USEPA Report to
provide "credible and persuasive results." While the USEPA agrees that meeting the conditions
posed would be useful in assessing differences between the 1984 and 1994 surveys, the list is nol
an essential list of criteria for the evaluation. Each of the conditions adds to the uncertainty of the
analysis. However, the statistical tests used in the USEPA's evaluation account for this uncertainty
by examining the variability of the data collected. In reply to each of the conditions listed, the
USEPA offers the following:
1. This condition is incorrect. Since the USEPA's evaluation is based on the average difference
between the sixty 1984-1994 sample pairs, the true requirement is that the two data sets represent
unbiased estimates of the sediment PCB inventory at the sampling location at the time of collection.
Therefore, the differences represent unbiased estimates of the actual difference. By examining the
average difference and testing for its statistical difference from zero, the analysis is not dependent
upon the absolute accuracy of any individual pair. Rather, by taking the average, the analysis
accounts for the fact that some pairs may overestimate or underestimate the actual difference, but
on average, since the data are unbiased, the mean difference will indicate the direction and
magnitude of change. The uncertainty calculations performed as part of this analysis indicated that
the reported degree of change was statistically different from zero and therefore that the direction
of change (i.e., inventory loss) was discernable from the data. The relatively large uncertainty in the
actual magnitude of the change was accounted for in part by the creation of a lower bound
(minimum) estimate of the mass loss. In this manner, since the minimum estimate of mass loss was
shown to be statistically different from zero, it is likely that the magnitude of mass loss was at least
comparable or perhaps greater than that estimated. This analysis is described in detail in Section 4.1
of the Report.
2. This condition was only truly necessary for the 1994 data set. To this end, the cores were
analyzed for :37Cs in the bottom-most slice. As discussed in the text, most (46 / 60) of the sediment
cores had no lj7Cs in this slice. The remaining 14 of the 60 cores did have detectable levels in the
bottom-most slice but 5 of these were shown to have falling l37Cs, indicating that the vast majority
of the PCB inventory had been captured. Thus the 1994 data set represents the entire PCB inventory
in nearly all TI Pool cores. This is summarized in Table LG-1.2. If the 1984 coring work did not
l.RC - 33
TAMS.Tetra'Iech
-------
Table LG-1.2
Review of 1994 Low Resolution Sediment Core Completeness
Core
Characterization
Total Thompson
Island Pool
Locations
Matched 1984
to 1994
Locations
Change in Sum of Tri+ and
Dechlorination Products
(BZ# 1,4,8.10,19)4
Change in Sum of Tri+
Homologues Only4
Original 1984 Inventory'
Number of
cores with loss
Number of
cores with gain
Number of
cores with loss
Number of
cores with gain
Cores less
than 10 g/mJ
Cores greater
than 10 g/mJ
Complete1
61
46
27
19
42
4
15
31
Nearly
Complete2
5
5
3
2
5
0
0
5
Incomplete3
10
9
5
4
8
1
4
5
Total
76
60
35
25
55
5
19
41
% Nearly
Complete or
Better
87
85
86
84
85
80
79
88
TAMS/Tcira Tech
1. No cesium-137 present in bottom slice of core,
2. Cesium-137 levels decline from surface to bottom core slice.
PCB maximum evident, i.e. maximum concentration occurs above deepest slice in core.
3. Rising or steady cesium-137 with no PCB maximum evident.
4. Based on 60 matched 1984 to 1994 locations only.
-------
capture the entire sediment inventor)', then these estimates represent lower bounds on the actual
PCB inventory present in 1984. If, in this case, the 1994 inventory is lower than the 1984 based on
the measured values, the difference represents the minimum mass loss. Again the data analysis has
been performed in a manner to conservatively estimate the actual mass loss, thus assuring the
absolute direction of change in the inventory.
3. The comparability of the 1984 grab samples to core samples was closely examined by
NYSDEC (Brown et a!., 1988). While the grabs have greater uncertainty associated with the
sediment depth they represent relative to the cores, this does not preclude their use in the analysis,
since, again, the statistical tests used account for uncertainty. The similarity of the '84 grab - '94 core
pairs to that of the core-core pairs is shown in Figure 4-15 of the Report. For both the >10 g/m: and
the <10 g/m: groupings, the '84 grab - '94 core pairs were not found to be statistically different from
the core-core pairs. Indeed, the match between the core-core pairs and the '84 grab - '94 core pairs
was quite close for the >10 g/m2 group (the more contaminated sediments).
4. Strictly speaking, the same sediment cannot be sampled in both 1984 and 1994 since the
1984 sampling effort removed and did not replace the sediments that were collected. However for
the purposes of the low resolution coring analyses, it is only necessary that the 1994 samples be
collected from the same area as that of 1984 samples. In this manner, on average, the difference
between the 1984 and 1994 results will represent the average change in the sediment inventory. The
semivariogram analysis performed on area H-7 by GE demonstrates the heterogeneity of the
sediments in the upper reach of the TI Pool, an area characterized by coarse-grained sediments. This
analysis does not apply to most of the 1994 sample collection sites which are located further
downstream. Only four 1994 locations were placed in this area of the Pool.
As shown in Figure 4-9 of the DEIR, semivariogram analysis of 1984 samples collected in
the upper portions of the Pool yield the same level of variability seen by GE at H-7. However,
semivariogram analysis of the lower two-thirds of the Pool shows significant spatial correlation, also
shown in Figures 4-10 to 4-12 of the DEIR. Thus it is clear that the remaining 56 paired sampling
locations occupied in 1994. with a median separation of 3 feet from the 1984 location, can expect
to represent the same sediment as collected in 1984. (A discussion of the semivariogram analysis
is presented in Section 4.2.4 of the DEIR.)
5. The objective of the 1994 effort was not to completely reestablish a new inventory for the
TI Pool. Rather it was to assess changes in the inventory documented in 1984. The writer
misconstrues the information in his calculation of sampling density for the 1994 survey. In fact, 16
areas of the TI Pool were examined at exactly the same sampling density as originally performed by
NYSDEC, since they are collected from precisely the same locations as those occupied by NYSDEC
in the areas studied. These 16 areas were selected from throughout the TI Pool and represent a range
of contamination and sediment textures, although they focus principally on fine-grained sediments.
As such, these areas are subject to the same processes affecting all sediments of the Pool and can be
considered representative of 1994 sediment conditions in the Pool. The analysis of the 60 coring
locations in the TI Pool as well as the 80-1- locations below the TI Dam all support the conclusion of
substantive sediment PCB loss from the sediments. The data set is sufficient for the purpose for
which it was intended, i.e., the assessment of the direction and approximate magnitude of the change
in sediment PCB inventory between 1984 and 1994.
LRC • 35
TAMS TctraTcch
-------
Response to LG-1.3
The USEPA disagrees with the premise posed by the writer that either the 1984 or 1994 data
sets are insufficient to characterize the PCB inventories in the areas examined. Both data sets were
subject to extensive quality assurance and quality control procedures and represent valid estimates
of the sediment concentrations at the time of sampling. While both data sets (as well as those
collected by GE) contain some degree of uncertainty, the data were compared by employing
statistical tests which account for the uncertainties in the data.
With regard to quantitation, the USEPA asserts that more reliable quantitation for the 1984
and 1994 data sets has not been demonstrated. In fact, both the USEPA and NYSDEC analytical
programs employed near state-of-the-art techniques at the time of implementation. In both cases,
sufficient data were obtained to assess uncertainties in the sampling and analytical procedures.
Ultimately, it is the construction of an average degree of change in the sediment PCB inventories
which yields the most powerful statistical tests to confirm the direction and magnitude of change.
The USEPA acknowledges that a better understanding of the mechanisms responsible for the
mass losses estimated in the Low Resolution Sediment Coring Report would be useful. However,
it is not necessary that these mechanisms be understood in order for the measured difference to be
considered valid. More importantly, it is essential to integrate the net change in sediment inventory
caused by the assortment of mechanisms so that the nature and scale of these mechanisms can be
constrained. The mechanisms discussed by the writer are not known in sufficient detail or magnitude
to provide their own constraints in the absence of data on sediment losses or gains.
Response to LG-1.4D
The purpose of the LRC was not to study PCB release mechanisms but rather to identify
changes in PCB inventories in the TIP and in hot spots below the Thompson Island Dam, essentially
integrating the net effect of these mechanisms. Comments on GE mechanism estimates have been
provided in the DEIR Responsiveness Summary. USEPA does not believe that these estimates
themselves provide sufficient independent constraints on the actual amount of mass loss.
Additionally, the USEPA provided a critique in Book 3 of the Responsiveness Summary for
Volumes 2A, 2B and 2C of the GE/QEA model which forms the baasis for the writer's assertions
in this comment. In this critique, several of the underlying assumptions made by GE/QEA are
shown to be of questionable validity (see Sections 2 and 3 of the critique in particular), thereby
rendering the flux estimates attributed to the various mechanisms very uncertain. As such the
GE/QEA model cannot be used to refute the findings of the LRC with regard to mass loss.
Response to LG-1.8
Inclusion of grab samples from the 1984 data set was necessary since about two-thirds of the
1984 NYSDEC locations were represented by grab samples. Thus exclusion of these from the low
resolution coring analysis data set is unjustified. NYSDEC assessed the data from paired cores and
grabs to determine the depth to assign to the grab samples (Brown et al., 1988).While the USEPA
acknowledges that the use of grab-to-core comparisons introduces some uncertainty into the analyses
presented in the LRC, this uncertainty is again incorporated in the statistical tests which still yielded
LRC - 36
T AM S/T etraTech
-------
statistical significance for the differences found despite all the uncertainties present. As further
evidence in support of the use of the grab-core pairs, the LRC presented an analysis of the Delta
values derived for core-core and grab-core pairs in Figure 4-15 of the Report. This analysis as
illustrated by the figure demonstrates that the grab-core pairs are not statistically different from the
core-core pairs and in fact yield very similar results for the sediments with PCB inventories greater
than 10g/m2.
Response to LG-1.10
The writer's contention that spatial heterogeneity is so great that the TI Pool inventory cannot
be determined precisely is inaccurate as discussed in detail in the response to LG-1.1. The writer
also contends that the 1994 sampling density was too low for the comparisons made. This is
incorrect since, within the clusters, the sampling density matched that of the NYSDEC study.
Finally, with regard to the number of samples, the writer contends that the USEPA acknowledges
that its sample set is too small but then uses it anyway. The USEPA acknowledges that the data set
is too small to constrict an overall sediment inventory estimate, not that it cannot be used to study
the changes in inventory within the areas studied.
In implementing the 1994 low resolution sediment coring program, the USEPA sampling
program focused on several concerns dealing with sediment homogeneity. Specifically, the low
resolution core clusters were selected so as to minimize local sediment heterogeneity by selecting
areas where sediment PCB inventories did not vary greatly (roughly a factor of 2% range).
Additionally, the USEPA sampled at the high end of the sediment distribution to characterize the
changes in the most contaminated sediments. These sediments represent the sediments most likely
to release PCBs to the water column since their high concentrations yield the strongest gradients to
drive this release. Lastly, due to their higher concentrations, accurate quantitation of PCB levels by
both 1984 and 1994 could be expected since detection limit thresholds were avoided. Figure 5
presented by the writer confirms that the 1994 locations represent the more contaminated sediments
oftheTIPool.
Figure 6 presented by the writer is used to contend that the selected 1994 locations are not
representative of the hot spots in general. Flowever, this comparison should be made on the basis of
the hot spots studied only. In addition, Brown et al, 1988 noted that many of the 1977 delineations
did not appear valid based on the 1984 sampling locations. It is the USEPA's intention to apply the
loss calculations to the fine-grained sediment PCB inventories defined by the 1984 sampling and the
1992 side-scan sonar survey. It is clearly stated that the mass loss calculations apply to sediments
of 10 g/m2 or higher, which are typical of hot spots. The writer has incorrectly inferred this
statement to mean that all hot spot sediments have greater than 10 g/m2. By focusing on the actual
inventory, the low resolution coring results can be used to estimate changes in the more
contaminated, fine-grained sediments {i.e., sediments greater than 10 g/m2). Figure 6 confirms that
the 1994 survey sampled among the more contaminated sediments and thus can be considered
representative of these sediments.
The writer incorrectly states that the mass loss calculations of the LRC apply to all sediments
of the TI Pool. There is in fact no inference that the mass loss seen for sediments greater than 10
g/m2 applies to all sediments. Indeed, the Report suggests that there may be PCB gains in the
LRC - 37
TAMS/TetraTech
-------
coarse-grained sediments of the Pool, perhaps in part due to the redistribution of PCBs from the
higher inventory areas.
See also LG-1.2, part 5 for additional discussion of these issues.
Response to LG-1.20
The issues raised here are addressed at length in USEPA's critique of the GE Model Report
prepared by QEA. This is included as Book 3 of the Responsiveness Summary for Volumes 2A, 2B
and 2C. As noted in the executive summary of the critique, the USEPA finds the following:
The congener signature of the TIP load is consistent with a weathered,
partially-deehlorinated PGB source although not as fully dechlorinated as some
buried hot spot sediments. The assumption that pore water flux is the only summer
loading pathway appears to be incorrect. Instead, new analyses conducted for this
review suggest that the summer TIP load is a mixture of pore water flux and bulk
loading of fine sediment, perhaps driven by bioturbation.
Additionally, the USEPA notes that the match shown on the writer's Figure 19 is a
comparison of measured data with the model construct, essentially a regression exercise and does
not constitute proof of the model assumptions. The partition coefficients for the dissolved organic
carbon-bound PCB congeners are not well known and those estimated by applying a simple, constant
correction factor do not agree with those reported by USEPA in the DEIR. The concentration of
dissolved organic carbon (DOC) is also not a well-known parameter since the values used by GE are
based on frozen samples. Freezing will likely cause precipitation of DOC, thereby potentially
introducing both variability and bias to the reported values. The issue of the compositing process is
also important here since it may mask relationships between concentration and congener ratios in
the sediments.
Response LG-1.19B also discusses some issues related to this comment.
Response to LL-1.1
The 1984 estimate of the TI Pool (Brown, et al. 1988) of 23 metric tons of PCBs is
substantially lower than the 1978 estimate of 61 metric tons (MPI, 1978). As discussed in Brown,
et al (1988) there are several differences in calculation methods, assumptions and quantitation that
result in this apparent 62 percent loss. These differences are listed on Table LL-1.1. From this listing
it is clear that the 1984 analysis has more precision than the 1978 analysis, because:
¦ The 1984 study of the TI Pool included more than three times the number of samples in the
TI Pool than in 1978,
¦ 40% of the 1984 locations are cores versus 33% in 1978,
¦ The 1984 grab depth is assigned by sediment texture and these depths were estimated
through statistical analysis of sediment cores of similar texture, and
¦ Specific weight was analyzed for most of the 1984 samples, but none of the 1978 samples
LRC - 38
TAMS'TetraTech
-------
Table LL-1.1
Parameter Comparison for Brown et al., 1988, and LRC Inventory Analyses
No. Sample Locations Used in Analysis
1978
Inventory t
1984
o Inventory
1984
Point
1994
o Point
1984
Sampl
1994
e Area
Total Number of Locations
No. Grabs
No. Cores
313
209
104
1014
607
407
60
15
45
60
0
60
197
85
112
70
0
70
Quantitation
1221,
1016. 1254
1242,
1254, 1260
Trichloro and Higher
Homologues
Trichloro and Higher
Homologues
Specific Weight Analysis
None
1.042 g/cc
Yes, except
for gravel
1.3 g/cc
for gravel
Yes,
except for
gravel
1.3 g/cc
for eravel
Yes
Yes. except
for gravel
1.3 g/cc for
gravel
Yes
Grab Depth Estimate
Coarse-Grained Sediments
Fine-Grained Sediments
24"
24"
12"
17"
12"
17"
Not
Required
12"
17"
Not
Required
Sources: Hudson River Database Release 4.1 and Brown. 1988 TAMS/TetraTech
-------
The conservative grab depth of contamination and constant specific weight serve to bias the
1978 inventory estimate high. The quantitation in 1978 included Aroclors 1221 and 1016 and
captured the monochloro- and dichlorohomologue fractions. As discussed in LRC Appendix E, the
1984 quantitation did not capture these fractions. By 1984, a significant amount of dechlorination
could have occurred, but this mass is not accounted for in the 1984 inventor}' estimate. The 1984
inventory estimate is underestimated, because this portion of the inventory is not included.
The LRC point-to-point analysis used the same number of sample locations in 1984
inventory estimates and 1994. Only 25 percent of the 1984 sample locations utilized for the point-to-
point analysis were grabs (see response to comment LG-1.8 for the justification for using grab
samples). Specific weight was analyzed in the laboratory for both the 1984 and 1994 samples. The
comparison is made on a constant quantitation basis using trichloro and higher homologues only.
A second sample area comparison is provided in Appendix A of this Responsiveness Summary
which yields a 43% molar loss of trichloro and higher homologues (with upper and lower 95 percent
confidence limits of 2 percent and 59 percent loss) from local areas of fine-grained sediment in the
TI Pool.
The LRC analysis recognized and accounted for the differences between the 1984 and 1994
sampling events, whereas no attempt was made to reconcile the differences between the 1978 and
1984 estimates. Because the 1978 to 1984 comparison was without a common basis, but the 1984
to 1994 has a common basis, it is reasonable to reject the 1978 to 1984 mass difference of 62%, but
accept the 43% (on average) molar loss from local areas of fine-grained sediment in the TI Pool
betw een 1984 and 1994.
The decision by Brown et al. not to assess the difference between the 1978 and 1984
inventory estimates inventory estimates was most likely limited by the issues discussed above.
Nonetheless, it is clear that the river was not working in a different manner during the period 1978
to 1984 than in the period 1984 to 1994. This is evident in the water column transport analysis
performed in the DEIR (see DEIR Figures 3-100 and 3-101). In both figures the mass loss from the
TI Pool from 1978 to 1985 is so great that it is readily discerned from the sporadic USGS monitoring
data, rhe inability of the USGS to track the lighter congeners limits the usefulness of this data as the
trichloro- and higher homologue decreases over time after 1985. This limitation is subsequently
superceded by the GE data which begin in 1991 and continue to demonstrate the mass loss from the
TI Pool. The writer's contention that Brown et al. *s perspective was somehow better than the current
understanding and should therefore be used as a benchmark for deciding the impact of sediment
PCBs is not accepted by the USEPA.
Response to LL-1.2
The comparison in the LRC between the 1984 and 1994 inventories was designed to
characterize local areas of finer sediment. The study was not designed to create a 1994 inventory,
but to determine the direction of change and an estimate of the degree of change in these local areas.
A subsequent analysis based on area-to-area averages is presented in Appendix A of this Report.
This analysis confirms the results presented in the LRC. The statistical techniques utilized take into
LRC - 40
I AMSTeiraTech
-------
account the number of samples available and account for the uncertainty as discussed in the response
to comment LG-1.1.
Response to I.I.-1.3
The analysis presented in the LRC for the TI Pool inventory is a point-to-point comparison
between sixty 1984 cores or grabs and sixty 1994 cores, not a Thompson Island Pool inventory
estimate comparison. Not all 1984 cores were used in the analysis, only the sixty locations
reoccupied in 1994. These sixty locations can be considered representative of the most contaminated
areas of the TI Pool. The analysis shows that these locations have lost inventory. By inference, it is
likely that similar sediments throughout the TI Pool have also lost inventory. A similar conclusion
was found based on area comparisons. (See Appendix A of this Responsiveness Summary.)
The issue of spatial heterogeneity is addressed in the responses to comments LG-1.9 and LG-
1.10.
Response to M.-1.12
The analysis of 1984 sediment PCB Aroclor-based quantitation in Appendix E of the LRC
provides a translation scheme to make the 1984 data consistent with the Phase 2 congener-based
quantitation. The method is quite successful, yielding a linear relationship between the trichloro- and
higher homologues and the 1984 method sum of Aroclors with a correlation coefficient (r) of 0.983.
It should be noted that this relationship is based solely on the 1994 data. Nonetheless, it is still the
best approach currently available to establish a consistent analytical basis between the 1984 and
1994 sediment data sets. By applying the correction factor developed in Appendix E. a reasonable
basis of comparison between 1984 and 1994 sediment samples is achieved.
At the time of the preparation of this Report, a more direct analysis of the 1984 and 1994
methods was being prepared by Dr. R. Bopp of Rensselaer Polytechnic Institute for NYSDEC. This
analysis will be reviewed by the USEPA in the near future.
4.1.1 A Comparison of 1984 and 1994 Conditions
Corrections to Section 4.1.1 - A Comparison of 1984 and 1994 Conditions
Figure 4-2 has been revised to show the correct units of depth.
Figure 4-7 has been revised. The graphs are reversed to match the discussion in the text and
the units have been corrected.
Response to LC-4.4
See response to comment LG-1.2, part 5 and LG-1.10.
I.RC-41
"ANlSTetraTech
-------
Total 137Cs (pCi/kg)
0 2000 4000 6000 8000 1 104
1991
0.
10.
1963
a,
Core 19 (RM 188.5)
2.5 10
0
Total PCBs (mg/ke)
Legend:
Source: TAMS/Gradient Database, Release 3.5
? Cs137
I
o PCBs
TA
Figure 4-2 (corrected)
High Resolution Core 19 from the TI Pool
-------
50'
C/5
V
°-
3
DO
-
/-¦s
O
-
o
£
^So
E
o
-50-
V
u
c
X
L.
O
-C
Tr-
ee
o
•
a
£
X
i
rr
o
-100"
S
C
ON
¦
VJ
55
2
C
-150"
r3
s
u
JZ
o
•u
-200-
. —- - :
c
~5b
CJ
5 T:r
C 30
.JJ OS
^ tj-
~~1 o
iS 2
« - .
2
2
o
ioo-
50-
o-
-50"
-100 -
-150"
-200"
-250"
Note:
1. See text for definition.
••
-250 —I 1—i—i i i 1111 1—i—i i i 1111 1—i—i i i i 111 1—i—i i i i n
0.1 1 10 100 1000
1984 MPA (g/m2) for Trichloro and Higher Homologues1
•
r
A
1—I—I II 1111 1—I—I I I 11 n 1—I—I I I 1111 1 1—I I II11
0.1 1 10 100 1000
1984 MPA (g/m") for Trichloro and Higher Homologues
Source: TAMS/Gradient Database, Release 3.5
Figure 4-7 (corrected)
1984 Trichloro and Higher Homologues as MPA vs
Mass Difference Relative to 1994 - Log Scale
TAMS
-------
Response to LG-1.2A
As discussed in the Response to LG-1.2, the USEPA investigation focused on areas of local
homogeneity in PCB contamination to establish the degree of change during the period 1984 to
1994. It is clear from the semivariogram analysis presented in the DE1R (see Section 4.2.4 of the
DEIR and part 4 of response LG-1.2) that GE selected an unsuitable area (area H7) of the TI Pool
to study the spatial relationship in sediment PCB contamination, as documented by both their study
and the NYSDEC 1984 survey. In addition to largely avoiding this area of the Pool in the 1994, the
USEPA low resolution coring program also took care to select clusters of coring sites with relatively-
minor variation in total PCB mass per unit area and sediment texture based on the NYSDEC 1984
survey. Lastly, the statistical tests employed confirm the significance of the changes measured. If
the data set was insufficient or its uncertainty too great, the statistical tests would have yielded no
statistical significance to the differences calculated.
Response to LL-1.13
USEPA acknowledges the error in the LRC figure. This correction has been noted in the
Corrections to Section 4.1.1 of the Low Resolution Sediment Coring Report. See Response LL-1.6
and LL-1.7.
Response to LL-1.14
See responses to comments LG-1.9 and LG-1.10. Since only four sampling locations pairs
were separated by more than eight feet, the table requested is not necessary, and will not provide a
useful comparison and is not included here.
Response to LL-I.15
USEPA acknowledges the error in this LRC figure. This correction has been noted in the
Corrections to Section 4.1.1 of the Low Resolution Sediment Coring Report.
4.1.2 Assessment of Sediment Inventory Change Based on the Original 1984 £Tri+ Sediment
Inventor)
Correction to Section 4.1.2
The x axis on the lower diagram in Figure 4-10 was corrected to show the correct values and
units. It is included in this report.
Response to LC-4.5
The 1984 to 1994 TI Pool inventor)' comparison used a common quantitation basis of
trichloro and higher homologues. See Appendix E of the LRC for a discussion of the 1984
quantitation and translation scheme implemented in the LRC analysis. See also Response LG-1.11.
LRC • 44
TAMSTctraTcch
-------
u
c
3
U
40'
35"
30 ~
25 ~
20"j
15 ~i
10-i
5 ~i
1
1
m
Delta
i 1 1 1 1 r
10 15
i
1 1 1 i
20
M
14"
-1
Notes:
1. See text for definition of Delta
-0.4
Loss'
0.5
2.0
4.3
8.0
13.8
23.1
jam j
Delta (log scale)
M
2. The value of 2 is added to Delta^ prior to taking the logarithm of the value in order to translate the
distribution away from zero and negative numbers which do not have defined logarithm values.
The scale is the actual Delta value and not the log.
Source: TAMS/Gradient Database. Relea.se 3.5
Figure 4-10 (corrected)
Distribution of the Percent Change in
PCB Molar Inventory (Delta^)
TAMS/Tetra Tech
-------
Response to LC-4.6
The value of 73 percent loss of Tri+ PCBs is based on GE's calculation and not the
USEPA's. It is based on the assumption of a log normal distribution of the parameter DeltaM. The
use of the DeltaM function to estimate the mean degree of inventory change in this manner is not
appropriate since the function is highly skewed and neither normally nor log-normally distributed.
Its use in the LRC is appropriate since the analysis there was focused on detecting real change.
However, the average degree of change predicted by the DeltaM function is not considered accurate
and was not reported in the LRC. To better address this issue, a ratio estimator whose statistical
properties are much better defined was used to estimate the average mass loss for the fine-grained
sediments of the Pool. This is presented in Appendix A of this Responsiveness Summary.
Response to LG-1.4A
The Executive Summary of the LRC should have noted that the loss estimate represents the
median mass loss from the sediments. The intention of the program was to test whether the sediment
loss or gain was occurring and whether this change was statistically significant. The writer should
note that there has been much made of the scenario wherein "dirty," contaminated sediments are
being overlain and "buried" by "clean" sediments. In light of the revelations concerning the leaking
GE facilities at Hudson Falls, it is unlikely that any "clean" sediments have been deposited in the
Upper Hudson in the last 20 years. Indeed this is verified by the high resolution cores which show
the continued contamination of recently deposited sediments. Nonetheless it is still possible that
"cleaner" sediments still serve to sequester more contaminated ones if the more contaminated areas
are being overlain with these materials. However, in this instance, sediment inventories (i.e., PCB
mass-per-unit-area) should increase since additional sediment, even if it is less contaminated than
sediments that were deposited earlier, will add additional PCB mass. At the same time, previously
deposited sediments would be isolated from the water column, thus isolating and securing their PCB
inventor>'. In the low resolution coring analysis, the loss of PCB inventory was found to be
statistically different from zero, thereby rejecting the premise that the sediments had either gained
PCBs or had simply stayed the same. The results showed a statistically significant loss, effectively
discrediting the burial scenario and demonstrating the absence of TI Pool-wide sediment burial and
sequestering.
Response to LG-1,4B
In the discussions contained in the LRC, comparisons were made among several different
interpretations of the analytical results. Total PCBs estimates, the sum of trichlorinated and higher
homologues, and the sum of trichlorinated and higher homologues plus the five dechlorination
product congeners (BZ#1, 4, 8, 10 and 19), were compared in various ways. The purpose of the
various comparisons was to simply demonstrate that regardless of which assumption was used
concerning the quantitation of the 1984 data set, the same major conclusion was obtained, i.e., there
has been a significant net loss of PCB inventory from the sediments despite the evidence for
continual input from GE sources upstream. Presumably this loss has released PCBs to the overlying
water column and sediments elsewhere in the river. In fact, the Report focuses on the sum of
trichlorinated and higher homologues plus the five dechlorination product congeners (BZ#1, 4, 8,
10 and 19), which provides a minimum PCB mass loss estimate. The writer is referred to pages 4-10
to 4-12 of the LRC which discuss in detail the construction of the PCB inventory estimates. As
LRC - 46
TAMS/TetraTech
-------
stated in the Report, the comparison of this sum for the 1994 samples to the sum of trichlorinated
and higher homologues in 1984 conservatively assumes that all dechlorination occurred post-1984.
A recent study performed by at Rensselaer Polytechnic Institute (McNulty, 1997) developed an in
situ dechlorination rate for sediments deposited prior to 1984. A revised calculation is included in
Appendix A of this Responsiveness Summary, which includes a dechlorination rate derived from
McNulty's work.
With regard to the estimation of the actual PCB mass represented by the 1984 measurements,
the USEPA believes that the discussion presented in Appendix E of the LRC is the best current basis
for these data. Nonetheless, the analysis in LRC Appendix E is based on GE's attempt to reproduce
the NYSDEC 1984 technique and as such does not represent proof that this interpretation is correct.
Thus, it was important to examine other possible interpretations to show that regardless of the basis
used, the 1984 inventory has substantively declined.
Response to LG-1.4C
The writer's contention that 10.8 metric tons have left the TI Pool is incorrect.
Specifically, the mass loss calculated for the TI Pool is attributed to the fine-grained sediments,
which represent about 8.7 metric tons of PCBs, or about 60 percent of the TI Pool inventory. This
estimate is discussed in detail in Appendix A of this Responsiveness Summary. More importantly,
the USEPA does not contend that all PCBs lost from the fine-grained sediments have left the TI Pool
but simply that they have left the fine-grained sediments. As noted in the LRC, the area of the Pool
characterized by coarser sediments appears to have seen an apparent gain in PCB inventory,
probably in part due to the redistribution of PCBs from the more contaminated fine-grained
sediments.
The writer's contention also assumes that no dechlorination occurs between 1984 and 1994.
This is the opposite assumption from that used in the LRC (i.e., the LRC assumes that no
dechlorination occurs prior to 1984 and that all dechlorinated congeners present in 1994 were
produced between 1984 and 1994). The writer's assumption yields an upper bound on the amount
of PCBs lost from the sediment (rather than the lower bound estimated by the USEPA's calculation).
While the USEPA believes that dechlorination largely ceases a few years after deposition, there is
some evidence to suggest that it does continue at a slow rate after PCB-contaminated sediments have
been in place for about one year (McNulty, 1997). See also the discussion in Appendix A of this
Responsiveness Summary.
Response to LG-1.4E
The 80 percent mass loss is a number constructed by the writer and does not represent the
USEPA's estimate of the true mass loss from either the Pool or even from the fine-grained
sediments. The data set referenced by the writer (consisting of only 12 points) is probably too small
to make the comparison concerning the rate of mass loss over time between 1994 and 1998.
I lowever. as is evident from Figure 13 in the writer's commentary, both the 1994 and 1998 inventory
estimates are substantially lower than the 1984 inventory, again confirming the finding that the
fine-grained sediments have lost substantial mass since 1984. The writer is also referred to response
LG-1.14 for further discussion.
LRC - 4?
T AMSTetraTech
-------
Response to LG-1.5E
The USEPA acknowledges that a better understanding of the mechanisms responsible for the
mass losses estimated in the Low Resolution Sediment Coring Report would be useful. The writer's
contention that modeled mechanisms can only yield an 18 percent mass loss does not provide a
constraint on the actual mass loss since it is not known whether all mechanisms are represented or
correctly modeled. However it is not necessary that these mechanisms be understood in order for
the measured difference to be considered valid. More importantly, it is essential to integrate the net
change in sediment inventory caused by the assortment of mechanisms, so that the nature and scale
of these mechanisms can be constrained. The mechanisms discussed by the writer are not known in
sufficient detail or magnitude to provide their own constraints in the absence of data on sediment
losses or gains. The suggestion of erosion as the sole source of these mass losses is strictly the
writer's conclusion and not the position of the USEP-A. It is the USEPA's contention that there are
probably several mechanisms, working separately or in conjunction with each other, which are
responsible for the measured mass loss. Finally, the limited number of cores presented by GE is too
small a set to provide useful constraints on the USEPA data set. Additionally, as noted in Response
LG-1.7, both the GE and USEPA data show a marked decline in the PCB inventory relative to the
1984 study.
Response to LG-1.11
Several comparisons were made between the 1984 and 1994 measurements based on different
assumptions about the reported values in 1984 and the likely processes affecting the sediment PCB
inventory. These include the following:
1. Total PCBs in 1994 vs. Total PCBs in 1984
2. Total PCBs in 1994 vs. sum of trichlorinated and higher homologues (Tri+) in 1984
3. Sum of trichlorinated and higher homologues (Tri+) in 1994 v.v. sum of trichlorinated and
higher homologues (Tri+) in 1984
4. Sum of trichlorinated and higher homologues (Tri+) plus 5 specific dechlorination congeners
(BZ~1.4. 8. 10 and 19) in 1994 vs. sum of trichlorinated and higher homologues (Tri-1-) in
1984
The detailed discussion of these choices and how they w:ere developed is discussed in Section 4.1
of the LRC. Only comparisons 2 and 4 were used to quantitate the change in sediment inventory.
With regard to comparison 2 (the writer's issue) this comparison demonstrates that a mass loss is
evident even when all PCBs present in 1994 are considered. Ultimately, these choices were made
to demonstrate that no matter what assumptions are made about the 1984 data set. the direction of
change in the TI Pool sediment PCB inventory is found to be the same. i.e.. loss. By focusing on the
fourth comparison listed above, the analysis presents a minimum estimate of PCB loss from the
sediments. By showing that this comparison yields a statistically significant loss, it can be assured
that an actual, substantive loss has taken place and that the hypothesized sediment burial scenario
proposed by GE can be rejected. This loss combined with that obtained from examining the degree
of dechlorination yielded the 40 percent mass loss discussed in the Report.
The USEPA agrees with the writer that the Tri+ estimate is probably the best representation
of the 1984 data, which is why it was used for the most rigorous comparisons in the Report.
l.RC-48
TAMS "TetraTech
-------
Nonetheless, the analyses presented in the LRC demonstrate that the direction of change in the TI
Pool inventory is the same regardless of the interpretation of the 1984 results as Tri+ or Total PCBs.
See also responses LG-1.4B and LG-1.12 as well as Appendix A of this Responsiveness
Summary for related discussions of this issue.
Response to LG-1.13
The USEPA uses all points to evaluate change. The data were simply separated into two
groups based on the original 1984 inventory. The group less than 10 g/m2 yielded mass gain while
the groups greater than 10 g/m2 yielded mass loss. The mass loss estimated by the point-to-point
comparison was confirmed by an area-based analysis. This is presented in Appendix A of this
Responsiveness Summary. This analysis yielded a net mass loss of 43 percent excluding
dechlorination losses. The purpose of the comparisons made in the LRC was to examine 1994
conditions relative to 1984 conditions and not the validity of the 1978 hot spot designations.
Therefore the probability plots developed by the writer are not appropriate to define the sample
groups for comparison, i.e. the type of sediments characterized.
The USEPA indicated that locations greater than 10 g/m2 were typical of hot spots.
However, our classification was not based on this designation and yielded, more directly, that more
contaminated sediments lost mass while less contaminated sedimentary have gained mass. The issue
of hot spot boundaries is immaterial to this calculation since hot spot boundaries were not used.
Nonetheless, if the criteria is decreased to 5 g/m2. a statistically significant loss for sediments greater
than 5 g/m2 is still obtained. As discussed elsewhere in the responses, the USEPA rejects the premise
that some of the sample pairs should be rejected because of separation distance (See LG-1.9). The
calculation of the Tri-1- difference alone as suggested by the writer assumes no dechlorination loss
and so represents a kind of upper bound on the degree of mass loss. This mass loss calculation is
presented in LG-1.12.
Response to LG-1.14
The comparison of the Tri4- values for 1984 and 1994 as promoted by the writer assumes no
dechlorination loss occurs over the 10 year period, resulting in an overestimate of the Tri+ mass loss.
Dechlorination rates for PCB-contaminated sediments have been documented by McNulty (1997)
and shown to be low though not negligible. In addition, as calculated in response LG-1.12. the mass
loss of Tri- based on the molar change would be 70 percent, not 80 percent as suggested by the
writer. However, as recommended by the writer, an area-based mass loss estimate was completed
as well. The analysis is summarized in Appendix A of this Responsiveness Summary. The result of
the analysis yielded a mass loss of 43 percent, excluding dechlorination losses, and is again
applicable only to the fine-grained sediment areas. The mass loss estimate would not apply to the
entire TI Pool inventor)' as incorrectly inferred by the writer. USEPA integration of the fine-grained
sediments indicates that about 8.7 metric tons are found in these areas assuming the 1984 values to
represent Total PCBs (8.2 metric tons are obtained if the Tri+ assumption is applied; see Appendix
B). The writer has incorrectly applied the mass loss to the 14.5 metric ton estimate given in the
DEIR, representing the entire Pool inventory. The mass loss rate of 43 percent would yield a PCB
inventory loss of 3.5 tons, but as stated elsewhere in these responses, not all of these PCBs would
LRC ¦ 4<3
1AMS TetraTech
-------
necessarily leave the Pool. Some would be deposited elsewhere, presumably among the less
contaminated sediments of the Pool. Thus the calculation as presented by the writer does not present
an independent constraint on the degree of PCB loss and cannot be used to dismiss the USEPA mass
loss estimates.
Response to LG-1.15
Fate and transport mechanisms do not provide sufficient independent constraints to disprove
the mass loss estimates developed in the LRC. The need for an independent mass loss estimate for
the purpose of fate and transport modeling was one of the main reasons for the low resolution coring
program.
The estimation of dechlorination losses based on McNulty (1997) is presented in Appendix
A of this Report. These results are summarized here. Essentially, McNulty shows that changes in the
congener patterns of matched sediment layers between cores collected 8 years apart suggest
dechlorination loss continues at a slow rate, after PCB-contaminated sediments have been in place
for about one year. Dechlorination loss as indicated by the shift from trichloro and higher
homologues to mono and dichlorinated homologues ranges from - 8.4 to -19.3 percent. The fact that
some layers actually show positive shifts toward higher homologues is probably the result of
analytical uncertainty. Nonetheless, the results can be used to estimate a net rate of 4.7 percent mass
loss for the 8 year period. This is much lower than the estimate put forth by the writer. The writer
has chosen a single layer (representing -1968). the layer exhibiting the highest rate of dechlorination
and clearly an outlying estimate for dechlorination throughout the core as a whole. Thus the writer's
contention that dechlorination loss from 1984 to 1994 amounts to 12 percent is clearly an
overestimate. This value is similar to that used in the LRC, in which the 11 percent dechlorination
value was clearly identified as an overestimate but was used to construct the lower bound estimate
of the actual molar loss from the sediments. In light the data presented by McNulty (1997), it is clear
that this value is much too high and that the conclusions of the DEIR concerning dechlorination are
correct, i.e., the vast majority of dechlorination occurs soon after deposition with little modification
after the first year. The writer's assertion that a 10 percent mass loss occurred over the ten-year
period is also inconsistent with their assumptions regarding a Tri+ to Tri+ comparison. It should
also be noted that the mass loss calculated by the writer is incorrectly applied to the entire Pool
inventory instead of to the inventory of the fine-gained sediments only.
The writer also presents estimates of other mechanisms which are proposed to provide further
constraint on the degree of mass loss. These other mechanisms are much more poorly constrained
than dechlorination and are contingent upon knowing surficial conditions in all areas of the TI Pool,
conditions which are not well defined. In addition, the processes of diffusion and groundwater
transport are not well documented in the Upper Hudson. Modeling results suggest that flow-induced
shear may not be sufficient to yield the PCB loss but little in situ data on the vertical mixing of
sediments is available to constrain these models. Lastly, other potentially important mechanisms
probably exist which are not addressed in the GE model. In particular, bioturbation, a process well
documented in other systems, is not addressed and is capable of enhancing both resuspension and
porewater exchange. Comments on the March, 1998 GE Report discuss these issues in greater detail.
LRC - 50
TAMS/TetraTech
-------
These comments are provided as Book 3 of the Responsiveness Summary for Volumes A, B and C
of the Phase 2 Report.
The writer also cites the water-column monitoring data collected by the USGS and GE as an
additional constraint on the mass loss from the fine-grained sediments. It is emphasized here that the
loss from these sediments is not inherently loss from the TI Pool, Re-deposition of some portion of
the fine-grained sediment PCB losses are likely elsewhere in the Pool. Evidence to support this
comes from the results of the less contaminated sediments (less than 10 g.'nr) as discussed in the
I RC which were indicated to have seen an inventory gain. Secondly, as acknowledged in Chapter
4 of the LRC, the magnitude of the change in inventory is relatively uncertain although definitely
different from zero and so cannot be used as a criterion for dismissal of the conclusion of mass loss
from Pool sediments.
Finally, model mechanisms as currently understood do not provide a basis for the dismissal
of the measured estimate of PCB mass loss because of the uncertainties involved. Failure of the
model to match the estimated loss may be because the GE model simply does not accurately depict
the system.
The IJSEPA did not estimate a 10.8 metric ton PCB loss from the TI Pool, as asserted by the
writer. Even assuming that the entire mass loss from the fine-grained sediments left the Pool (an
unlikely prospect as discussed in LG-1.15) at the rate calculated in the LRC (30 percent mass loss)
or as revised in LRC Appendix A and in response LG-1.12 (43 percent mass loss), the mass
transport rate would be much smaller. Using the 8,2 metric ton estimate for the trichlorinated and
higher homologue inventory of the fine-grained sediments as provided in Appendix B of this
Responsiveness Summary, the mass loss from these sediments would be 3.5 metric tons as
trichlorinated and higher homologues. Even if this entire loss were to leave the Pool (unlikely as
described above), the average transport rate would be 2.1 lb/day (1 kg/day), which is well within the
range of values obtained by the USGS and GE.
It should also be noted that the US EPA does not accept the GE estimates for PCB mass
transport, which are developed on the basis of a rating curve. The calculation techniques and load
estimates developed in the DEIR (USEPA, 1997) should be applied instead.
The comment also makes several other statements or inferences which require correction.
Specifically, the USGS used a packed column procedure to measure PCB concentrations only until
1987 after which the procedure was switched to a capillary column procedure. Also, the USGS
record at Schuylerville (sometimes used as a surrogate for the TI Dam concentration) only extends
to 1989 and not 1991 as indicated by the writer. USGS stations further downstream are quite distant
(at least 13 miles) and cannot be directly used in place of this monitoring station as a measure of the
TI Pool load due to the potential occurrence of PCB loss or gain in the intervening river section.
Response to LG-1.17
The comparison made between the USEPA and GE sediment inventories is based on an
assumption of a constant (linear) rate of mass loss between 1984 and 1994, This is an unlikely
LRC-51
TAMSTetraTech
-------
prospect since mass loss is typically driven by concentration gradients, which would have been much
higher in the earlier portion of the 10 year period. An exponentially declining release rate is much
more likely. It should also be noted that the GE data set described in this comment was not available
for USEPA review at the time of the preparation of these responses, thereby limiting the ability of
the USEPA to respond to these comments.
Response to LG-1.17A
The data set used by GE for this comparison was clearly quite small (12 locations) and
insufficient to estimate differences between 1994 and 1998. No data are provided to assess GE
sample reproducibility nor is any other uncertainty analysis presented. It should be noted, however,
that both data sets show a substantial loss relative to the 1984 inventory, as shown in Figure 13 of
the GE comments. The contention that mass loss should occur continuously between 1984 to the
present from all areas of the Pool is simply incorrect and represents a significant oversimplification
of the issue. The LRC demonstrated that some areas gain while others lose but that the overall trend
was downward.
Response to LG-1.17R
The contention of low fish body burdens in 1997 is based on a limited sample set collected
by GE and not by NYSDEC. This data set may not be directly comparable to earlier NYSDEC data.
The USEPA will evaluate the 1997 NYSDEC fish data when it becomes available.
With regard to the other issues raised by the writer, the PCB loss rate is unlikely to be linear,
as discussed in LG-1.17A, therefore extrapolating a linear decline scenario to present conditions is
largely a useless exercise. The declines in fish body burdens are consistent with a sediment loss
scenario which is non-linear, such as an exponential decline. Both the measured fish body burden
decline seen from the 1984 to 1996 and the measured sediment inventory decline are consistent with
the major losses from the sediments occurring in the early 1980's and a subsequent decline in the
sediment PCB loss rate, yielding proportionately lower levels in fish. Fish body burdens, although
they have responded to the recent release events from the Hudson Falls facility, still suggest an
underlying source since they have declined only slightly relative to the late 1980's. These results
indicate that food chain derived PCBs as well as PCBs from on-going sediment release will serve
to sustain fish body burdens for the foreseeable future.
The limited GE data also represent a different time of year than most of the previous NYSDEC
sampling and therefore may not be directly comparable. Thus it is inappropriate to speculate on the
nature of the 1997 fish body burdens until they are released by the State. The GE data do not
represent an extensive survey nor have they been shown to match the NYSDEC data on a
quantitative basis. With regard to the recent decline in fish body burdens, it is important to note that
as of 1996, fish body burdens were at or just slightly below the fish levels measured in 1989, two
years prior to the major release event. Thus the fish body burden ''recovery" alleged by GE largely
represents a return to the river conditions which prompted the initiation of the Hudson River
Reassessment in 1990.
LRC • 52
TAMS/TetraTech
-------
The USEPA also notes in the few diagrams provided by GE concerning their 1998 sediment
collection (the data were not received in sufficient time for review prior to the preparation of these
responses) that the surface sediment layers show no substantive decline in the top 5 cm despite the
controls put in place at the Hudson Falls facility during the last few years. These preliminary results
indicate either the absence of recent deposition at these sites or else the deposition of re-released
PCBs originating from other sediments.
Response to LG-1.19A
The figure presented does not suggest a linear increase in water column load across the TI
Pool and in fact suggests several points where PCB load increases markedly (e.g., RM 192.5 and
190). Nonetheless, the distribution of hot spots and fine-grained sediments is such that if these areas
were the main sources, then the river PCB load might increase somewhat linearly as well. In fact,
float survey data not presented by the writer shows substantially higher concentrations (some greater
than 200 ng/L) in the near-shore environments relative to the main channel, suggesting enhanced
PCB transfer in these areas. These areas have substantial levels of biological activity which may
enhance PCB transfer and help to create the strong seasonal variation seen in the more recent
monitoring data. It is entirely possible that water column loads from the TI Pool are produced from
near-shore fine-grained sediments and simply mixed into the main channel flow. This would
potentially explain the discontinuities in the main channel loads as the River's passage through the
Pool is directed by river bends and narrows which force horizontal mixing. Even under low flow
conditions, water within the Pool is in motion, serv ing to homogenize concentrations. Thus the
contention that the float survey data indicate that the TI Pool load is produced in a uniform areal
manner is not inherently supported by the data. Additional evidence for enhanced release in the
near-shore environment comes from the TI Dam monitoring station maintained by GE which
sometimes shows substantially higher PCB concentrations relative to the main channel, indicative
of incomplete mixing of loads produced in the near-shore area with the main channel. These loads
also show enhanced concentrations of lighter congeners as might be expected from more
concentrated and subsequently more dechlorinated PCB inventories. Much of the writer's contention
in this comment is based on assumptions made in the construction of GE's PCB transport model
which likely has significant flaws. (See Book 3 of the Responsiveness Summary for Volumes 2A.
2B and 2C).
Response to LG-1.19B
This comment is addressed as part of the critique of the GE Model Report contained in Book
3 of the Responsiveness Summary for Volumes 2A, 2B and 2C of the Phase 2 Report. A portion of
the I'SEPA critique (Section 5) is repeated here.
. ,.(GE)/QEA implicitly sets up the hypothesis that known mechanisms of
flux from "old" hot spot sediments in the TIP (considered to be hydrodynamic
erosion, diffusion, and pore water advection) are not sufficient to account for the
"anomalous" TIP load. Therefore, additional mechanisms are needed to provide a
newer, enhanced PCB load to surficial sediments in the TIP. Three additional
mechanisms are postulated:
I.RC-53
TAMS.'TeiraTech
-------
1. PCB DNAPL loading in bedload along the sediment-water interface
2. Pulse loading of PCBs due to periodic flooding of the Baker Falls plunge
pool
3. Transport of oil-soaked sediment into the TIP at the time of the Allen Mill
collapse.
As an implied result of these '"additional mechanisms'", GF./QEA claims that
organic-carbon normalized PCB surface sediment concentrations are similar across
the TIP. and that these active sediment concentrations are disconnected from buried
hot spots...
...QEA (1998, Table 4-6) presents information showing that mean PCB
concentration in surface sediments, when normalized to organic carbon
concentration, is similar in the hot spot and non-hot spot areas, and is similar for fine
and coarse sediment. They then state (p. 48): "The flux of PCBs from surface
sediments to the water column depends on the organic carbon normalized PCB
concentration... Regions of the river with equal surface sediment organic carbon
normalized PCB concentrations and composition contribute equally to the water
column PCB load."
This argument is flawed. Suppose PCB concentrations on organic carbon are
everywhere the same, but location A has a high weight percent of organic carbon,
while location B has almost no organic carbon. Obviously, location A has a much
greater mass of PCBs per volume of sediment and is likely to contribute more PCB
load to the w ater column, even if similar pore water concentrations are calculated for
each location under equilibrium conditions. What QEA's argument primarily reflects
is that hot spot areas are "hot'" because they have more fine-grained sediment with
high organic carbon concentrations.
QEA's argument is invalid for any source mechanisms that involve bulk
sediment movement (scour, bioturbation, etc.), and only partly valid for
consideration of a purely pore water source from sediments. It is true that equilibrium
partitioning assumptions imply that the observed apparent pore water concentration.
CPWa. (including both dissolved and colloidally-sorbed PCBs) should be proportional
to the organic-carbon normalized PCB concentration, but this is not the only factor.
Rearranging Equation (3-29; USEPA, 1997) yields:
r
PW.a
t \
Cp "0 + m DOC^DOC^
\focj ^oc
where CP is the particulate concentration.
9 is the saturated porosity,
is the mass of DOC per volume of pore water,
A.'dcx- is the partition coefficient to dissolved organic carbon, and
A'or is the partition coefficient to sediment organic carbon.
IRC • 54
TAMS TetraTech
-------
Inspection of this equation shows that the apparent pore water concentration depends
not just on the organic-carbon normalized sediment concentration but also on 9 and
wD0C. As both porosity and the concentration of dissolved organic carbon tend to
increase in fine-grained, organic sediments, the pore water concentration should also
be higher in hot spot areas.
Analysis of the 1991 GE data from the 0-5 cm layer in the TIP reveals wide
ranges in TOC concentration (from 4,961 to 69.474 ppm) and in porosity (from 16
to 70 percent). With a few exceptions, TOC concentration increases with porosity
{see Figure LG-1.19B). This correlation indicates that inferences of pore water source
strength cannot be based on organic carbon normalized PCB concentrations alone.
In Phase 2 results (USEPA, 1997, p. 4-20) it was noted that 'locations
with...finer-grained sediments have consistently higher median and mean PCB
levels."' The 1984 NYSDEC data also show a strong relationship between sediment
texture class and total PCB concentration, with the highest concentrations in the
finest grained sediments. Table [LG-1.19B] shows the averages of NYSDEC top core
section and grab sample results for the near-surface layer. These results show a clear
increase in average PCB concentration for sediments with finer texture and higher
organic content. Results are similar for sample medians, except in the case of
sediments classified as clay. A portion of these samples are believed to include intact,
uncontaminated glacial clays. In any case, it appears clear that it is inappropriate to
compare sediment concentrations as a source of pore water flux unless both organic
carbon fraction and porosity are taken into account.
It should be noted that it is reasonable to expect a smoothing out of surface
concentrations relative to buried hot spot concentrations. However, such a general
smoothing of surface sediment concentrations does not indicate that the surface PCB
inventory is unconnected to buried hot spots. PCBs introduced into the water column
by erosion or other disturbance of bulk sediment would be subject to local-scale
settling, spreading concentrations. Some settling may also occur of PCBs loaded to
the water column via pore water advection. following partitioning to solids in the
water column, while lateral interflow could also "'smear" the pore water signal.
Response to LG-1.22
USEPA cannot comment fully on these data, having only just received them prior to the
completion of this Responsiveness Summary. However, the approach used by the wTiter is inherently
flawed. Specifically, the approach assumes that the sediments responsible for the PCB load across
the TI Pool during this event have a single congener pattern. This is highly unlikely given the broad
range of mixtures present in the Pool. Instead, the PCB load generated by the river's passage through
the Pool represents the integration of what is undoubtedly a broad range of sediment congener
patterns whose net result is to produce the patterns seen at the TI Dam. It is likely that this mixture
represents both recently deposited, fresh Aroclor-1242 like mixtures as well as recently exposed,
older and more altered mixtures and recently re-released and redeposited older PCB mixtures.
Various combinations of these mixtures are capable of yielding the mixture seen at the TI Dam.
Nonetheless, the ratios selected by the writer do demonstrate that the passage through the Pool does
LRC • 55
TAMSTeitaTech
-------
100,000
80,000
•60,000
Q.
CL
O
240,000
20,000
"r.i Ei__
II
a u
* "fcp 1 fe '
|3 Ml
20
—I 1 H
40 60
Porosity (%)
80
100
Hudson River Database Release 4.1
Figure LG-1.19B
Correlation of TOC Concentration and Porosity in TI Pool Surface Sediments
TAMS/TeiraTcch
-------
Table LG-1.19B
Surface PCB Concentrations in NYSDEC 1984 Data Compared to Texture Class
Texture
Class
Interpretation
Average
Total PCBs
(mg/'kg)
Median Total
PCBs
(mg/kg)
Median
Specific
Weight (g/cc)
Sample
Count
FS-GRV
Fine sand
and gravel
14.7
9.1
0.9
7
CS-WC
Coarse sand,
wood chips
16.9
10.7
1.1
9
GRAVEL
Gravel
19.8
14.1
—
127
CS-SND
Coarse sand
25.0
13.8
1.25
22
GR-WC
Gravel, wood
chips
29.9
29.3
—
19
FS-WC
Fine sand,
wood chips
47.3
25.7
0.9
79
CLAY
Clay-
54.9
6.7
1.0
10
FN-SND
Fine Sand
80.8
31.1
0.8
290
MUCK
Muck
121.1
103.8
0.5
14
-------
markedly shift the congener spectrum to a more altered mixture as illustrated by a brief examination
of the GE samples collected on Jan. 9, 1998.
Three samples collected on Jan. 9, 1998 were examined in terms of their molecular weight
and Cl/BP to note the impact of the passage through the Pool as well as to Schuylerville. This
information is summarized in Table LG-1.22. The data show a marked decline in the molecular
weight from Rogers Island to points downstream.
The load at Rogers Island is particularly interesting since its AMW is negative (-0.06)
relative to Aroclor 1242. This sample is notable as well in the near complete absence of monochloro-
and dichloro-homologues. As EPA learned, GE was performing remedial work at the GE Hudson
Falls Plant Site in the river near the pumphouse in January 1998. This work involved the removal
of debris and sediment which contained high concentrations of PCBs. At the time GE performed
sampling in the river, there was a high flow event occurring which could have mobilized PCBs from
this area. Therefore it is highly likely that GE has captured a truly recent sediment deposit
undergoing resuspension. In this scenario, nearly all of the lighter congeners have been lost in the
deposition process, presumably via partitioning to the water column. It is likely that this sample
represents the unaltered surface sediment GE has been attempting to find. It is interesting to note that
this evidence has occurred upstream and no! within the TI Pool.
The total load at the TI Dam has a AMW value of 0.08, representing about a 9 % mass loss
by dechlorination. However, this represents the molecular weight of the total load. The change in
this value from Rogers Island to the TI Dam is quite large, at 0.14. This indicates that the molecular
weight of the net load is substantially lower than that of the total load. This in turn suggests that the
net additional PCB load from the Pool to the water column during this high flow event was produced
from sediments with an average AMW in the range of 0.1 to 0.2, since the AMW value rose so
much during the passage through the Pool. Only if the entire PCB load entering the TI Pool from
above Rogers Island is deposited within the TI Pool (an unlikely scenario based on the 1993 spring
transect event), does the AMW of the sediment source approach that of the TI Dam load.
The transit from TI Dam to Schuylerville serves to raise the molecular weight somewhat and
decrease the apparent level of dechlorination to a 5.8 percent mass loss. This may be accomplished
in several ways, such as loss of the lightest congeners by gas exchange or resuspension of sediments
less dechlorinated than those typical of the TI Pool. However, when examined with respect to the
Rogers Island load, both the TI Dam and Schuylerville stations show major reductions in the
molecular weight of the water column load, presumably by the scouring of sediments at 10 percent
or higher dechlorination mass loss.
Based on this initial analysis, the data suggest that sediments dechlorinated to at least 9
percent mass loss (AMW greater than 0.08) are responsible for the changes seen. In reality the level
of dechlorination may be much higher.
Response to LG-1.28
The L'SEPA is currently undertaking development of PCB fate-and-transport models to better
understand the processes affecting PCBs. Nonetheless, it is data sets such as the one obtained for the
LRC which provide the necessary constraints on the model and not the other way around. Simply
LRC • 58
TAMS "letraTcch
-------
Table KG-1.22
An Kxamination of (>R Monitoring Results for Jan. 9, 1998
Kt. Edward Flow at 34,300 cfs
Station
PCB
Concentration
ng/L
I'lux at
34.3(H) el's
lli/day
Molar
Concentration
nmol/L
Mol. Weight
g/mole
AMW
Kelaiive id A1242
Cl/BP
Theoretical
Mass Loss by
Dcchlorinalion
(%)
Rt. 197 Bridge
(Rogers Island)
71
13
0.252
281.7
-0.06
.3.70
TI Dam West
142
26
0.581
244.6
0.08
2.63
9.2%
Seluiylervillc
253
1.001
252.4
0.05
2.85
5.8%
Aroclor 1242
265.7
0.00
3.24
TAMS/Tetra Tech
-------
being unable to explain a phenomenon does not inherently make the measurement of the
phenomenon wrong or inaccurate. The LRC presents analyses which show statistically lower median
inventories in the sediments of the Upper Hudson. Further analyses provided as part of this
Responsiveness Summary improve on the calculations provided in the LRC and yield similar
degrees of change for both median and mean mass loss from TI Pool sediments (56 and 43 percent
loss, respectively). As stated elsewhere, the writer has incorrectly applied the original mass loss
estimates to the entire Pool, as well as assumed that all lost PCBs have left the TI Pool instead of
being at least partially redistributed elsewhere in the Pool. Ultimately, the major conclusion from
the LRC is not the degree of absolute mass loss but rather that the fine-grained sediments of the
Upper Hudson have only served as temporary storage for PCB contamination and that they have
re-released much of their burden to the river.
See also LRC Appendix A which presents the statistical analysis described above.
Response to LG-1.38R
The statement referenced by the writer could have been more well written. This statement
was not intended to represent an ad hoc basis for assigning significant differences but rather
represents the end result of the statistical analyses performed earlier in the LRC. Specifically, the
statistical analysis used in the hot spot comparisons completed earlier in the Report did not yield
statistically significant differences, unless the 1976-1978 and the 1994 inventory estimates differed
by a factor of 2. This is simply an observation based on the fact that the when the 1984 and 1994
geometric means were found to be statistically different, the arithmetic means differed by a factor
of 2 or more. That is, the inventory had to be at least halved or doubled in order for the statistics to
confirm the difference as significant. This can be seen in LRC Table 4-8. Hot spot areas whose
arithmetic means differed less than a factor of 2 relative to the 1976-1978 inventory did not yield
statistically significant differences. The minimum statistically significant difference relative to the
1976-1978 data was seen for Hot Spot 34. where the difference was almost exactly a factor of 2
(1976/950 = 2.1). Since all of the subsequent estimates for these hot spots were based on the same
1976-1978 data set, this observation provided a useful basis for evaluating the important differences
between estimates.
Response to 1.G-1.38H
The value of 2 added to each of the DeltaM values represents the minimum value required to
permit the calculation of log values for all points. This addition does not modify the distribution but
only shifts its center. Specifically, the addition creates a three-parameter log-normal distribution
function, wherein the distribution is shifted but no change is made in its actual shape, i.e.. there is
no effect on the lower end characteristics nor on the distribution parameters. Only the log-transform
affects the shape of the distribution, but this is a standard statistical practice when an underlying
log-normal distribution is suspected. This is further discussed in Chapter 12 of Gilbert (1987).
Although the USEPA believes that the DeltaM function to be a useful one, there has been sufficient
concern over its use that the USEPA has prepared an alternate approach. This approach is described
Appendix A and utilizes a ratio estimate rather than a Delta estimate. See Appendix A for the
ratio-based estimates of the sediment inventor)' changes.
I.RC • 60
TAMSTe:raTech
-------
Response to LG-1.38J
The US EPA agrees that it is useful to provide measures of uncertainty when presenting
information, but it is not possible to display this information everywhere. Table 4-8 is already quite
complicated as it is presented, and inclusion of further data would only worsen this problem. Instead
the uncertainties are presented prior to the introduction of this table in Table 4-7. Previous estimates
of sediment inventory did not provide rigorous measures of uncertainty which is why this
information was reconstructed for the 1976-1978 data set in this Report.
Response to LG-1.40B
The sign test applied by the writer was performed on the most conservative estimates of mass
loss. When such a comparison is made solely on the basis of the Tri • inventory, such a test is likely
to prove a statistical significance. More importantly, however, is the additional area-based analysis
presented in Appendix A. This comparison is based on area-based means for the cluster areas of the
1994 sampling event, wherein the ratio of 1994 to 1984 inventories in mole/area is used as the
regression variable. This provides a parameter which is well described by parametric statistics and
is found to have statistically significant differences with respect to zero (no change) for both the
median and mean estimates.
Response toLS-1.2
See response to comment LG-1.4A. See also Appendix A which provides a mean mass loss
estimate and its associated uncertainty.
4.1.3 Assessment of Other Potentially Important Characteristics
Response to LL-1.16
From the last paragraph of Section 4.1.3 on p. 4-18:
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.
This states that the 1994 cores show PCB loss relative to the 1984 samples which is greater
in the fine-grained sampling areas than the coarse-grained sampling areas. No mechanism is
proposed to account for this loss, because the USEPA program was not designed to unequivocally
determine the means of PCB transport. The mechanisms need not be defined in order for the
measured difference to be considered valid.
LRC-61
TAMSTciraTech
-------
4.1.4 Implications of the Inventory Assessment
Response to LS-1.5
No response required.
4.2 Sediment Inventories of the Upper Hudson Below the Thompson Island Dam
Response to LC-2.I
See response to comment LG-1.38G.
Response to LG-1.12
The USEPA agrees with the writer that the arithmetic mean should be used to calculate the
net changes when integrating over the entire Pool. Nonetheless, the 40 percent mass loss presented
in the LCR does represent a useful value for comparison since it represents the median change in the
sediments, i.e., any individual location is likely to see a mass loss comparable to this level. However,
the estimation of the mean mass loss from the function DeltaM is not straight forward since the
function is neither normally nor log-normally distributed. While the use of this function is
appropriate for the testing of the direction and statistical significance of change, it is not the best
function for the estimate of the scale of the mean loss. For this reason, the USEPA has prepared a
separate analysis to estimate the magnitude of the mean change in inventory. This analysis is
described in detail in Appendix A of this Responsiveness Summary. The approach and results are
summarized below.
Based on the suggestions of several of the reviewers (including the writer), the USEPA has
prepared a revised mass loss estimate, implementing four important changes. First, the estimate of
mass loss is now based on an area-based comparison. This comparison reduced the number of data
pairs available but also reduced some of the variability since the comparisons are now based on area
averages and not point estimates. Second, the USEPA used an estimate of the dechlorination rate
based on the work of McNulty (1997) rather than the upper bound (maximum possible) rate
originally applied in the LRC. Third, no distinction was made based on the original 1984 inventory.
i.e.. all sampling areas were considered in the examination without regard to the greater-than 10
g/m:. less-than 10 g/m: classification previously used. Nonetheless, the results are still considered
indicative of fine-grained sediments and not the entire Pool due to their locations along the sides of
the river channel.
Lastly, the USEPA used a ratio rather than a Delta statistic to estimate the change in mass.
Specifically, the ratio of the 1994 to 1984 sediment inventories is used as the variable in the statistics
rather than the Delta function. This yields a statistically better "behaved"' log-normal function as
shown in Figure A-8 in Appendix A. In this instance, the ratio can be converted to a Delta value after
the statistical calculations are applied by simply subtracting one from the ratio as shown below:
LRC ¦ 62
TAMS TetraTech
-------
1994 MPA^ - 1984 MPAJ,
1984 MPA1f
1994 MP A.
L. -1
1984 MPA,
In this manner, the ratio is tested for statistical significance relative to 1 (i.e.. log (0)) since this
represents the absence of change (i.e., 1984 =1994). The result is then converted to the Delta
function afterwards. This avoids the creation of negative values prior to the log conversion. In the
original analysis presented in the LRC. the creation of negative Delta values necessitated the addition
of 2 prior to taking the log of the delta values in order that a log value could be defined for all delta
values. This approach also avoids the creation of the asymmetric distribution characteristic of the
delta function.
The end result of the revised calculation was to yield a mean mass loss of -45 percent
including dechlorination. The 95 percent confidence interval about this value was -4 to -59 percent,
thus excluding 0 and indicating that the change was statistically significant. The median mass loss
as estimated by the geometric mean was -57 percent with a 95 percent confidence interval of -33 to
-72 percent. These values represent the total mass losses from the sediment. Correcting for
dechlorination loss yields only a minor decrease in the mean mass loss estimate, to -43 percent. The
range about this estimate is +1 to -58 percent. (The fact that this uncertainty now includes zero is
not considered important since the median mass loss is still statistically different from zero at -56
percent (range of -31 to -72 percent).) Thus the mean mass loss of -43 percent represents the mass
loss from the sediment which is not the result of in situ dechlorination but rather represents the mass
loss from the sediment to the water column of the TI Pool. Presumably some portion of this loss
passes over the TI Dam while another portion is deposited at lower concentrations elsewhere in the
Pool. The lower concentrations result from the mixing with less contaminated sediments in the water
column and river bottom.
The results obtained in the analysis presented in Appendix A are comparable to the results
originally reported in the LRC, Specifically, the median value of -56 percent (range of -31 to -72
percent) is within error of the value obtained for the molar loss of the trichlorinated and higher
homologues of -28 percent (range of -2.9 to -50 percent). These values both represent the median
losses from the sediment to the overlying water column. More importantly, both estimates show the
statistical significance of the net sediment loss. On this basis, the hypothesis of simple burial and
sequestering of contaminated sediments must be rejected, as discussed in the LRC.
Response to LG-1.38D
The USEPA agrees that a multivariate model would be interesting to complete, but it is
largely peripheral to the main topic of mass loss. It has been demonstrated elsewhere (USEPA. 1997)
that dechlorination is proportional to sediment PCB concentration. Therefore, the statement listed
by the writer is not based on the analysis of the low resolution coring results and does not constitute
a misinterpretation of the data. Other statements concerning the correlations of PCBs and '"Cs are
I RC - 6 i
TAMSTetiaTech
-------
based on the LRC analysis but have been demonstrated elsewhere as well. (Bopp and Simpson.
1989: USEPA. 1997).
Response to LG-1.38F
The purpose of the discussion in Section 4.3 of the LRC was simply to show that the estimate
based on 1984 data could represent a significant underestimate if applied in a risk assessment. This
95 percent confidence limit value was all that was reported in the 1991 USEPA Phase 1 Report and
so establishes the basis for comparison. The writer is correct in noting that the 95 percent confidence
limit value is subject to a number of influences related to sampling. Nonetheless, the 95 percent
confidence limit value provided in the Phase 1 Report still appears low, even with respect to the
mean estimates reported on the line above in LRC Table 4-13. This table summarizes the results for
shallow, near-shore sediments based on the 1984 and 1994 surveys. Based on this, the comparison
is still appropriate and not misleading.
Response to LG-1.38G
A more detailed analysis of variance, while probably interesting, was not necessary to
support the basic conclusions of the Report. Additionally, it is unlikely that sufficient information
is available to accurately and completely represent all components of variance in each of the surveys
utilized. This deficiency would then return the analysis to the original approach used here. That is.
by the use of mean and standard error estimates, the analysis presented simply assumes that the
various sources of variance are unbiased and that the total variance is reflected in the standard error
estimates.
It should be noted, however, that the discussion of the semivariogram analysis in this
comment is inaccurate. The upper portion of the TI Pool, i.e., the area studied extensively by GE (H-
7), is subject to very short-scale spatial variability. This is confirmed by analysis of both the
NYSDEC 1984 and GE 1991 data sets from this region. However, in the areas of the Tl Pool
principally sampled during the low resolution sediment coring program, the short-scale spatial
variability is substantially lower than the total variability observed between widely-separated
locations. Thus, the writer over-states the actual uncertainty of the data. This is further discussed in
response LG-1.9.
Response to LG-1.40C
The analyses presented in the LRC were never intended to be exclusively statistical analyses.
Indeed, statistics are simply tools with which to test hypotheses and do not represent an end in
themselves. Knowledge of the geochemical processes which can affect PCB inventories, as well as
the history of PCB release and transport, is essential before undertaking any statistical tests. These
tests simply provide numerical support for the apparent geochemical changes which have occurred
over time. The examples listed by the writer represent the geochemical processes which form the
hypotheses for subsequent statistical testing and analysis. Without the prior geochemical knowledge
to propose these hypotheses, there would be little purpose to perform any statistical analyses.
Ultimately, it is the knowledge of PCB geochemical fate and transport, supported by the data
collected and subsequently tested with appropriate statistics, which provide the basis for the
LRC • 64
TAMS'IctraTcch
-------
conclusions drawn in the Report, It is the combination of these components which make the
conclusions of the LRC most defensible.
4.2,1 Calculation of the Length-Weighted Average Concentration (LVVA) and Mass Per Unit
Area (MPA) for Sediment Samples Below the TI Dam
Response to LG-1.34
The criticism raised by the writer is a valid one but given the data available, there was no
other basis to establish the sediment density. The USEPA was aware of this issue at the time of the
preparation of the Report. As a result, the length-weighted average concentrations were also
compared between the 1976-1978 and the 1994 surveys. These values do not require solid-specific
weight and provide an alternate basis for comparison. The comparisons of the 1976-1978 to 1994
length-weighted averages yielded similar results to those obtained for the mass-per-unit-area
comparisons.
Response to I G-1.35
The issue of various analytical techniques was discussed at length in Appendix E of the LRC.
It certainly would have been preferable to be able to run the techniques on identical samples, but the
descriptions of the historical techniques are less than complete so that reconstruction of the original
techniques is difficult. Reconciliation was addressed to some degree by a series of samples collected
by GE in 1991 and 1992 which form part of the discussion in Appendix E.
Response to LG-1.36
This issue was addressed in the previous Reports dealing with the 1976-1978 data sets. The
analysis in the Report relies on the relationships developed in Malcolm-Pirnie, 1992. Figure 4-23
of the LRC is derived from the Malcolm-Pirnie Report and shows the relationship to be unbiased
although the extrapolation of grab samples to depth adds additional uncertainty relative to the core-
based inventory estimates. This issue, as well as that discussed in LG-1.35. add uncertainty to the
individual hot spot PCB inventory estimates. Nonetheless, given the magnitude of change found for
several of the hot spots, it is unlikely that main conclusion for Section 4.2 will be directly affected.
That is. the sediment PCB inventories of the Upper Hudson below the TI Dam are not static zones
simply undergoing burial but are instead subject to various processes which serve to re-release the
PCB contamination originally stored there.
Response to LL-1.17
As stated on p. 4-21 of the LRC. the 1994 length-weighted average concentrations for the 0-
12" interval were calculated, "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 12-inches, the reported concentration for the segment was used without
modification." A last condition, which was not mentioned in the text, is that if the concentration of
the second layer dropped below 10% of the top layer, the LWA was set equal to the top layer
concentration. A review of the 1977-76 LWA concentrations indicates that the second layer was
LRC • 65
TAMS TetraTech
-------
nearly always used if the top-most layer was less than 12-inches thick. Both data sets are subject to
the same affect of dilution if the inventory is unchanged. It is unlikely that this approach badly
underestimates the LWA since the PCB maxima were typically found in the uppermost core
segment.
See response to comment LG-1.8 for a discussion of the inclusion of the 1984 grab samples.
Response to LL-1.18
The data for the NYSDEC 1976-78 Sediment Survey was taken from a Malcolm Pirnie (MPI)
draft report written in 1994 for NYSDEC (MPI, 1994). Upon comparison between the data in the
MPI report and the data in the Hudson River Database Release 3.5, it was evident that numerous
cores shown in the report drawings were absent from the database. There were also instances where
core or grab concentrations did not match between the report and the database. Because there was
no means of checking the data included in the database which had been provided by NYSDEC. the
data in the MPI report was used. The hard copy of data in the report was manually converted into
electronic files and then checked. Coordinates were digitized from the report drawings. The
conversion to an electronic media was both time consuming and painstaking, but performed to
provide the highest quality analysis possible. In this manner, the USEPA chose the better, not the
more convenient, data set.
Response to LL-1.19
The statement made on page 3-18 of the LRC regarding the strength of the correlation
between solid-specific weight and Total PCBs is too strongly worded and essentially incorrect. The
trend in the data was obscured by binning the data finely in the box and whisker plot (LRC Figure
3-15). In fact, as shown on LRC Table 3-7, solid-specific weight has a regression coefficient of -54
percent, second only to percent solids for the bulk sediment properties. Solid-specific weight is one
of the better predictors for Total PCBs.
In addition to a mass basis, the hot spots below the TI Dam were compared on a length-
weighted basis which is independent of the solid-specific weight of the samples. From the graphs
shown in LRC Figure 4-22, both the length-weighted and mass bases give similar results. Only Hot
Spot 39 is different, with a statistically significant different loss between 1976-78 and 1994 for the
length-weighted average and no change on a mass basis. But as discussed in the text, because the
PCB mass at Hot Spot 39 appears to have been poorly captured in both sampling events, conclusions
drawn for this area are uncertain.
4.2.2 Comparison of 1976-1978 Sediment Classifications and the Side-Scan Sonar
Interpretation
Correction to Section 4.2.2 -Comparison of 1976-1978 Sediment Classifications and the Side-Scan
Sonar Interpretation
Figure 3-27 incorrectly referenced as Figure 3-28 on page 4-25. The text should read.
LRC - 66
TAMS-TetraTech
-------
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 3-27. Note that
Figure 3-27 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 3-27). 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.
Response to LL-1.20
The side scan sonar can distinguish between fine-grained and coarse-grained sediments,
although fine sand which is on the boundary of fine and coarse-grained material cannot be
distinguished. This issue is addressed in the quoted passage from the LRC referenced in the
peceding correction to Section 4.2.2. The agreement is good between the NYSDEC visual sediment
classification and the side scan sonar, with the knowledge that the side scan sonar cannot resolve the
fine sands.
Response to LL-1.21
USEPA acknowledges the error in the LRC text. This correction has been noted in the
Correction to Section 4.2.2 of the Low Resolution Sediment Coring Report.
As discussed in the response to LL-1.20, the side scan sonar analysis is not sensitive in the
range of fine sands. The result is that fine sands can be characterized as either fine-grained or coarse-
grained sediments with equal probability. For the USE? A data approximately two-thirds of the fine
sands samples were located in fine-grained areas, but nearly half of the NYDEC fine sand samples
were located in fine-grained areas. Both data sets show a significant split for the samples
characterized as fine sand, but the higher percentage of fine sands located in fine-grained areas for
the USEPA is most likely due to the selection of sample location. The USEPA samples were
intentionally placed in fine-grained areas while the NYSDEC sampling locations were selected by
overlaying the TI Pool with a sampling grid. NYSDEC sample locations are roughly evenly split
between fine and coarse-grained sediments, while the USEPA samples are predominantly fine-
grained.
LRC ¦ 67
TAMS TetraTcth
-------
4.2.3 Comparison of Sediment PCB Inventories: NYSDEC 1976-1978 Estimates versus 1994
Low Resolution Core Estimates
Correction to Section 4.2.3 - Comparison of Sediment PCB Inventories: NYSDEC 1976-1978
Estimates versus 1994 Low Resolution Core Estimates
Figures 4-19 and 4-20 are incorrectly referenced as Figure 4.2-3 and 4.2-4, respectively, on page 4-
27. The text should read.
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 4-19 and 4-20. 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 4-19 and 4-20 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).
Response to LC-2.4
This comment will be taken under consideration during the preparation of the Baseline
Modeling Report.
Response to LC-2.5
The USEPA disagrees with the writer's contention that water column loads provide a critical
constraint on PCB mass loss estimates derived from the low resolution coring analysis. The mass
loss estimates apply only to the fine-grained sediments. The net change in the coarse-grained
sediment inventories is unclear. Therefore, the load gain across the TI Pool is not directly linked to
the mass loss estimates derived in the LRC and Appendix A of this Responsiveness Summary. While
discerning the exact nature of the source mechanisms is useful, the fact remains that the fine-grained
sediments of the Upper Hudson River have lost a substantive portion of their prior inventories
despite the continued released from the GE facilities. With regard to high-flow releases from the TI
Pool, GE has recently (January 1998) obtained data from a one-in-15-year flow event which should
provide input on TI Pool resuspension loads driven by flow. As discussed in Book 3 of the
Responsiveness Summary for Volumes 2A, 2B and 2C, (USEPA, 1998b), however, it appears that
much of the TI Pool load is not derived by flow-driven shear stress. This issue will be more closely
examined in the Baseline Modeling Report.
LRC • 68
TAMS TetraTech
-------
Response to LC-2.6
This comment will be taken under consideration during the preparation of the Baseline
Modeling Report.
The USEPA agrees that recent trends more clearly exhibit the load originating from the TI
Pool. However, the earlier years of data (1991-1996) also clearly illustrate the TI Pool load in both
magnitude and congener pattern. The USEPA disagrees with the contention that TI Dam loads for
1991-1995 are not partially derived from TI Pool sediment inventories which were primarily
deposited prior to 1977 and does not see any inconsistency in its perspective in this regard. The
water column loads leaving the TI Pool are quite distinct from those entering the Pool and represent
what is clearly a spatially altered source. The USEPA does not state that the TI Pool was responsible
for the majority of the total load of PCBs during the period 1991 to 1995, but rather that the TI Pool
is responsible for the majority of the load during low flow conditions. In particular, conditions prior
to June 1993 were dominated by loads originating upstream of Rogers Island, as noted in the DEIR.
Nonetheless, the importance of the TI Pool load should not be dismissed, since low flow conditions
are particularly important to biological uptake because they represent the water column exposure
concentrations during the period of maximum biological activity. Additionally, the conditions post-
1995 are probably similar to those which existed prior to September 1991 (the Allen Mills structure
failure), thus returning the Upper Hudson River to the condition wherein sediments probably played
a more important role in governing water column and fish PCB concentrations. These were the
conditions which existed prior to the USEPA Reassessment study.
Response to I.F-1.2
.An independent assessment of the PCBs lost to the water column from the sediments of the
TI Pool can be made by comparison with the estimated load at the TI Dam. Comparison with the fish
body burdens is complicated by fish exposure to sources other than the TI Pool sediments. In
particular, the load from above Rogers Island was a significant source to the water column between
1984 and 1994.
Note that the mass loss fraction given in the LRC (28 percent) represents a median and n£>l
a mean loss, as incorrectly stated in the Report. The loss of trichloro and higher homologues from
the predominantly finer areas of the TI Pool is estimated at -43 percent with upper and lower 95
percent confidence limits of 1 and -58 percent, respectively (See the revised calculations in
Appendix A). The amount of trichloro- and higher PCBs estimated to be present in the finer-grained
areas of the TI Pool in 1984 is 8,200 kg (See Appendix B). Using these values, the trichloro- and
higher homologue loss to the water column between 1984 and 1994 (assuming 10 years and 365
days/year) is 0.97 kg/day on average with bounds of 0 kg/day and 1.3 kg/dav. A net gain is not
calculated here since the estimated median mass loss is clearly negative, precluding any net mass
gain by the fine grained sediments of the Pool.
The load at the TI Dam generated by the sediments of the TI Pool can be estimated using the
data in Figures 3-86 and 3-87 of the DEIR. Figures 3-86 and 3-87 show the cumulative load of
trichloro- and tetrachloro homologues at Rogers Island and the TI Dam over nearly five years (1991-
1995) to be 422 kg or approximately 0.23 kg/day. Tri- and tetrachloro homologues are the greatest
contributors to the trichloro- and higher homologue load, making this value a slight underestimate.
LRC - 69
"" ' MSTctraTech
-------
This trichloro- and higher homologue load at the TI Poo! is four times smaller than the
expected value of 0.97, but falls well within the 95 percent confidence limits.. Also note that the 43%
loss of trichloro- and higher homologues represents all forms of loss excluding dechlorination. Note
that the 43 percent mass loss also incorporates redistribution within the TI Pool, which may represent
a significant fraction of the total. The point-to-point comparison of 1984 to 1994 cores suggests that
the noncohesive sediments may be gaining inventory while the cohesive sediments have lost PCB
mass.
Response to LG-1.5B
See Response to Comment LG-1.7.
Response to LG-.l .5D
The writer's premise in this comment is that a single pattern can be developed which
represents the sole source of PCBs to the water column. Any "match" achieved is strictly dependent
on the mechanisms assumed to produce it. In this instance, the writer is almost certainly incorrect
in assuming a single mechanism when almost certainly more than one is involved. This issue is
discussed in detail in Book 3 of the Responsiveness Summary for the Database Report, PMCR and
DEIR. Specifically, it is unlikely that the partition coefficient data is sufficiently precise to uniquely
constrain any congener pattern and preclude a specific mechanism.
The use of PCB fish tissue data alone is inappropriate to isolate the nature of the sediment
source since fish do not bioaccumulate less chlorinated congeners as efficiently as heavier ones and,
therefore, fractionate the congener distribution. This process serves to partially obscure the
fingerprint of the source material. Greater depth of analysis is required before any single source can
be removed from further consideration in this regard.
A discussion of the January 1998 high flow sampling data collected by GE is found under
response LG-1.22.
Response to LG-1.5F
In the hot spots below the TI Dam, the samples were distributed throughout the hot spot areas
and were comparable to the sampling densities applied by NYSDEC in 1977-1978. As noted in the
LRC, the premise for loss is simply dependent on complete recovery of the contaminated sediment
interval in 1994 and not in the earlier survey. This is because an incomplete core in 1977-1978
versus a complete core in 1994 will show a mass gain, as long as there has been no true PCB loss.
Thus, for those hot spots below the TI Dam exhibiting statistically significant mass loss, there is no
doubt that these areas must have truly lost PCB mass. The measured gain at Hot Spot 28 appears to
be, in fact, the result of the hypothesized example given above, i.e., the collection of incomplete
cores in the earlier study followed by complete cores in the latter study. Only in this case, the earlier
study missed so much of the inventory that it is unclear whether any actual mass loss had occurred
as well.
LRC - 70
TAMS/T c traTec'n
-------
Response to I.G-1.6
The use of averages and confidence limits to compare the 1984 and 1994 conditions
intrinsically incorporates the uncertainties in the data. The USEPA does not assume no error; but.
in fact, explicitly tests to see if the uncertainties in the estimates of the means render the differences
in the means statistically meaningless (Note: the estimates were found to be statistically different).
The data were not analyzed assuming no error in the measurements but simply that the data represent
unbiased estimates of the true values.
The USEPA does not suggest or state that the Delta functions reduce analytical uncertainty.
The L'SEPA agrees with the writer's statement that no mathematical transformation can reduce
analytical uncertainty. However, the use of a Delta function or other similar ratio is to convert the
analytical uncertainty to an approximately constant value. That is, analytical uncertainty is typically
a small percentage of the absolute value reported, regardless of it magnitude (i.e., analytical error is
often given as a percentage of the value reported). Thus, by dividing by the 1984 concentration, the
differences between 1994 and 1984 are normalized to account for analytical uncertainty as an
approximately constant percentage of the reported value.
For example, assume that an analytical precision of 20 percent is attained for a set of samples
and two measurement sites are examined, one at 300 g/m2 and one at 5 g/m2 based on 1984
measurements. Subsequent sampling in 1994 yields 250 and 12 g/m2 for these locations. It is clear
that the absolute change at the 300 g'm2 site is much greater than that for the 5 g/m' site (-50 vs. +7).
However, as a percentage of the original value, the relative change at the 300 g/m2 site is much
smaller ((250-300)/300 or -17 percent) vs. that at the 5 g/m2 site ((12-5 )/5 or 140 percent). In this
context, it is clear that the inventory change at the 5 g/m2 site is relatively more important and
exceeds its analytical precision of 20%. The change at this 300 g/m2 site, although large in absolute
value, does not represent a large fraction of the site inventory and, in fact, falls below the level of
analytical precision, suggesting the change is not significant. Thus by the use of the Delta function,
the importance of any 1984 to 1994 change in PCB inventor)' can be assessed relative to the likely
analytical uncertainty.
Analytical uncertainty was estimated to average 36 percent for individual 1994 samples. (The
median uncertainty was 27 percent, meaning half of all replicate pairs agreed to within 27 percent
or less.) However, it should be noted that the uncertainty of the average concentration difference
from 1984 to 1994 will be substantially less since the error on the average will decrease with the
number of samples utilized in the average. Although the tests used in the LRC. are more rigorous and
incorporate the uncertainty of both the 1984 and 1994 data sets, an example of how the uncertainty
of the average decreases relative to the individual uncertainty is given as follows:
Error on Average
Individual Uncertainty
/ No. Of Samples In Average
Error on Average
36%
•f 45
1£%
6.71
5.4%
LRC ¦ "!
IAMS TetraTech
-------
Forty-five represents the number of samples in the locations greater than 10 g/m:. Based on this
calculation, the uncertainty on the average is about 7 times less than the individual uncertainty.
Lastly, it should be noted that the use of averages and confidence limits to test for statistical
significance intrinsically incorporate the analytical as well as all other uncertainties inherent in the
data.
Response to LG-1.7
See Response LG-1.2 and Table LG-1.2, with regard to the number of complete and
incomplete cores. USEPA acknowledges that the truly incomplete cores of the TI Pool add some
additional uncertainty since they are potentially biased low with respect to the mass estimate.
However, this would have excluded only 9 of the 60 1994 matched core sites or 15 percent. Only
10 of the 76 cores placed in the TI Pool were considered incomplete, representing only 13 percent
of the total number of cores. Of the 9 paired cores, 5 represent sites with greater than 10 g/m2. The
remainder represent sites with less than 10 g/m2. Thus, the effect of excluding these cores would be
spread across both 1984 inventory core groups (i.e.. <10 g/m2 and <10 g/m2). Only 5 of the 9 show-
mass loss when considering the dechlorination products while 8 of the 9 show loss based on the Tri-
sum. Based on the roughly equal distribution of these cores among the main groups in two of the
three cases just mentioned and the relatively small number of incomplete cores out of the total, it is
unlikely that these cores serve to greatly bias the statistical outcome.
This assertion is verified in the calculations presented in Appendix A which compare the
area-based mean inventories for 1984 and 1994 both with and without the incomplete cores. In all
comparisons, the results show that the estimate of the mean mass loss changes less than 5 percent
based on the exclusion or inclusion of the incomplete cores. Thus the inclusion of these cores, while
potentially adding some uncertainty, does not affect the major conclusions of the LRC. .
The direct comparison between the USEPA 1994 data and the GE 1998 data is restricted by
the small sample size of the GE data set, as it is probably insufficient to conduct the proper statistical
analyses to support their conclusions. GE did not occupy any incomplete coring sites as defined by
the USEPA criteria (described in Chapter 2 and pages 4-32 to 4-33 of the LRC) so it is not possible
to test these sites to see the extent of contamination missed. Nonetheless, as discussed previously,
the few new GE data confirm the findings of the LRC, showing major loss of PCB inventory from
the sediments between 1984 and 1998. as depicted in GF/s Figure 13 of their comments.
Response to LG-1.38C
The analysis of PCBs and related ancillary parameters was presented for interest and
completeness. The USEPA does not believe that the data set is sufficiently detailed to develop a
covariate model for PCB mass loss, since these ancillary parameters are not consistently measured
between the older and newer data sets and some potentially important parameters are not available
for the portions of the various data sets.
i.rc -
TAMSTiiraTech
-------
The 1984 survey does not provide specific information on the spatial distribution of sediment
contamination in the areas of Hot Spots 25 through 39, since these areas are located downstream of
the areas studied in 1984. As shown in the DEIR, the spatial relationships vary from reach to reach
and are not readily extrapolated. It should be noted that the 1984 and 1994 means are estimated in
the same fashion, thereby providing a consistent basis for comparison. To a large extent, the spatial
variation is already accounted for, in that the areas were previously surv eyed and selected as areas
of high contamination. Side-scan sonar images showed Hot Spots 25, 28, 34 and 35 to consist
primarily, though not exclusively, of fine-grained sediments (see LRC Plates 4-21 to 4-24). thus
much of the spatial relationships are effectively included by the averaging process. It should be noted
as well that all of the hot spots studied, except Hot Spot 34, represent areas similar in shape and
design to the semivariogram results obtained for the 1984 sunev of the TI Pool. Specifically, each
hot spot area represents an elongated or elliptic zone whose major axis parallels the direction of river
flow and whose minor axis is perpendicular to this flow. Essentially, the original layout of the hot
spots inherently incorporated the spatial relationships anticipated in each area.
Lastly, it should be noted that samples collected in 1994 demonstrate that areas outside hot
spots are typically much less contaminated than those within the hot spots, supporting the original
area designations.
Response to LG-1.39B
The text notes that there is a need to differentiate analytical variability from real change. The
writer is correct in noting that the text is technically wrong in suggesting that analytical uncertainty
can be diminished. However, while there is no means to reduce the degree of uncertainty attributable
to analytical variability, it is possible to isolate it in a fashion so that real change can be more readily
discerned. Analytical variability is such that it typically represents a percentage of the reported value,
and it occurs such that measured values are equally likely to overpredict and underpredict the true
value, (i.e., the errors are unbiased). Analytical techniques are generally applied so that this
variability is a small percentage of the expected range of concentrations to be measured. Otherwise
little information can be obtained from the analysis. For example, little confidence is placed in a
reported value if its associated uncertainty is close to its absolute value (i.e., 100 percent
uncertainty). On the other hand, a value with a 10 percent level of uncertainty is frequently
considered "good" when estimating sediment contamination. In either case, the absolute value of the
uncertainty of the reported value is dependent upon the magnitude of the value itself. These
percentages essentially represent the absolute uncertainty divided by the reported value. By dividing
by the measured value, the uncertainty is converted to a constant (e.g., 10 percent). Thus, examining
the differences between 1984 and 1994 on a delta or ratio basis, it is possible to effectively isolate
the analytical uncertainty by assuming it to represent a constant, unbiased percentage of the reported
value. True analytical uncertainty will express itself as both positive and negative changes within
a small range of zero, yielding a delta function that is not statistically different from zero. On the
other hand, if real change has occurred, the deltas will be typically one-sided, with a mean
statistically different from zero.
It is useful to illustrate this point in an example. Two coring sites are originally occupied and
found to contain 15 and 300 g/nr of PCB contamination, respectively. Analytical uncertainty is
I.RC - ""3
TAMSTetraTech
-------
estimated at 10 percent. Upon a subsequent resampling, the values obtained for these sites are 5 and
270 g/m2, respectively. This yields absolute differences of -10 and -30 g/m2. Based on absolute
values, it appears that the sediment has lost more mass from the 300 g/m2 site and that this is the
more important mass loss. However, when these differences are viewed in the context of the absolute
inventory of the site, the relative importance of these losses relative to the ability to measure them
becomes evident. Specifically, the 10 g/m: represents a loss of 66 percent of the original inventory
from the 15 g/m2 site, while the 30 g/m2 loss represents only a 10 percent loss from the 300 g/m2 site
and is probably within the measurement error. Thus, the smaller 10 g/m2 loss would be expected to
represent a true decline in the sediment inventor} while the greater 30 g/m2 loss is within
measurement error and may in fact represent no real change. Because of the wide range of
concentrations obtained from both the 1984 and 1994 sediment samples (over several orders of
magnitude), it is necessary to put the measured differences in context of the original inventory so
that real change can be assessed relative to the likely analytical uncertainty.
In the case of no real change, delta values such as that for 300 g/m2 site would be expected
to occur as both positive and negative differences with a mean value close to zero. The range of
values obtained for delta would reflect the total variability in the measurements, including the
analytical variability. It is for this reason that the delta function was tested for its statistical difference
with respect to zero. The results obtained for the LRC yielded a delta value for sediments greater
than 10 g/m2 which was negative and statistically different from zero, indicating that real change had
occurred.
The finding of net PCB loss from the sediment is not a result of regression toward the mean,
as contended by the writer. As noted elsewhere in these responses and in the LRC, the net loss was
found for the minimum loss estimate, (that is, loss of Tri+ homologues plus the five dechlorination
product congeners). It is even more evident when the Tri+ sums alone are compared or when a minor
correction for dechlorination is included. In this case, most sites show loss, regardless of inventory.
This is further borne out by the analysis presented in Appendix A of the Responsiveness Summary.
In this presentation, area-based averages are compared for 1994 and 1984. Again, statistically
significant losses are found, comparable to those estimated from the delta function analysis presented
in the LRC.
Response to LG-1.40A
Statistical evidence for the change in PCB inventory was extensively developed in the
Report. Inherent in this finding is the fact that if the PCB inventory is declining and there is no
evidence for its in situ destruction, then the inventory cannot be undergoing burial. Supporting
evidence for the lack of wide-spread burial is multifold. Thus, this conclusion does not hinge on any
single result or analysis. The differences in both the depth and concentration of PCB contamination
between 1984 and 1994 are direct evidence of the lack of burial since the sediment inventory has
decreased and the depth of contamination has largely decreased or remained the same. The premise
put forth by GE that the most contaminated areas of the river were being rapidly sequestered by
burial is clearly inconsistent with this evidence. This is because the inventories are principally still
within the top 9 inches of sediment, just as they were in 1984. Hence, rapid, "deep" burial (i.e.. on
the scale of 9 to 12 inches) is not occurring. Thus, if burial has occurred, in most instances it is
LRC - "¦>
TAMSTetraTech
-------
limited to a few inches, certainly not deep and probably not beyond the reach of sediment
resuspension or biological activity in many cases.
If sediments were experiencing burial, sediment PCB inventories would be expected to
increase, since, as has been well documented by GE as well as other monitoring efforts, the area
above Rogers Island continues to contribute fresh PCBs to the Upper Hudson (at least up to 1996;
the cores were collected in 1994). Thus, additional sediment would also add PCB inventory. Yet all
evidence suggests PCB loss from the sediment.
The fact that the very unique environments of the high resolution cores show burial of the
PCB maximum does not, in fact, mean that this phenomenon is wide-spread, or even common. These
cores are not considered representative of sedimentation patterns throughout the Pool since in fact,
such environments are difficult to find. Note that several of the high resolution cores collected from
the Pool could not be dated due to their apparent variations in sediment deposition. In fact, potential
application of the cores as representative of Pool-wide conditions was strongly assailed by GE on
several occasions. The high resolution cores, in reality, provide information on the nature of annual
transport and on the nature of the type of material which is actually deposited in the river, but they
do not provide any information on the pervasiveness of this phenomenon. The only direct evidence
for real change in the sediment inventories comes from the low resolution cores themselves.
Evidence for the absence of extensive burial is consistently seen in both the TI Pool and in
the areas downstream, since many sediment areas show loss over time. There is also the direct water
column evidence exhibited in Figures 3-100 and 101 of the DEIR. which show extensive loading
of the water column with PCBs between Ft Edward and Schuylerville throughout much of the early
1980s, despite the USGS' inability to track the lighter congeners. The loss from this region of the
river can be inferred, despite the somewhat random timing of the USGS sampling events at the
various stations. Presumably, the same mechanisms responsible for these losses are in part
responsible for the continued losses since 1984. Indeed, it is difficult to understand how the areas
originally responsible for the early 1980s loads could convert from source areas to storage areas. It
is far more likely that these areas continue to lose inventory in the same manner as before but simply
at a lower rate, since much of the inventory has already been depleted.
Lastly, additional evidence is provided by the "Be data which show that burial is not in
evidence in many places, based on its absence. As noted in response LG-1.18, the :Be results do not
prove the absence of burial as originally asserted in the LRC. However, neither does its presence
prove long term burial since episodic depositional events followed by periodic resuspension events
will yield 'Be bearing sediment, if sampled at the right time. At a minimum, in those areas lacking
"Be. the results suggest the deposition rate to be quite slow in these areas, amounting to no more than
a few tenths of a centimeter per year. Over a period of 20 years, a deposition rate of 0.5 cm per year
would accumulate roughly 10 cm (4 inches), This would leave the sediment PCB maximum
associated with the early 1970's well within the biologically active region of the sediments, despite
the passage of two decades.
The USEPA agrees that the lack of change in the sediment inventory depth does not prove
the absence of burial, but it does preclude the "deep" burial inferred by GE. Indeed, there was at least
one major depositional event in the Upper Hudson, since PCBs are found throughout the Pool. The
Ft. Edward dam removal in 1973 and the subsequent 100 year flood in 1976 are. of course, the most
LRC ¦
TAMSTetraTcch
-------
likely culprits. However, as is evidenced by the USGS records of the early 1980s, the GE records
of the 1990s and the USEPA studies of 1992-1994, the sediments continue to release PCBs to the
water column, as seen in the water column gains across the Pool as well as in the sediment losses
reported in the LRC.
Response to LL-1.9
See response to comment LG-1.6.
Response to LL-1.22
Because the hot spot locations are uncertain, these boundaries were not used in the analysis.
Rather the dredge location areas (which approximate the hot spot boundaries) were digitized from
the MP1, 1994 Report and used. In addition, the sample locations for the 1976-78 had been surveyed
and the coordinates known at the time of the 1994 sampling. The 1994 USEPA cores were located
to assess the same areas delineated in 1976-78. It is evident from LRC Plates 4-21 through 4-28 that
the areas delineated in 1976-78 by NYSDEC were resampled in 1994 with a similar sample density.
Response to LL-1.23
USEPA acknowledges the error in the LRC text. This correction has been noted in the
Correction to Section 4.2.3 of the Low Resolution Sediment Coring Report.
Response to LL-1.24
See the responses to comments pertaining to the 1976-78 versus 1994 PCB inventory
comparison below the Thompson Island Dam (LL-1.17. LL-1.18 .LL-1.19, LL-1.22, and I.L-1.26).
The USEPA disagrees with the writer's contention. The areas sampled in both surveys were covered
at similar sampling densities and yielded statistically significant differences while directly or
indirectly accounting for the uncertainties. For example both the LWA and the MPA results show
statistically significant differences but only MPA incorporates sediment density and its uncertainty.
Response to LI.-1.25
See response to comment LL-1.7. The USEPA disagrees with the writer's contention. As
discussed in response LG-1.40A, high resolution cores cannot be used to infer burial on a wide
spatial scale. They are not typical of the rate of sediment deposition throughout the river.
Response to LI.-1.26
Of the 13 low resolution cores taken from Hot Spot 28, six have the PCB concentration
maximum in the top core segment and four have the PCB concentration maximum in the second
segment. The remaining three cores contain only one segment for PCB analysis (see the core profiles
in Appendix D of the LRC). Long-term storage of PCBs is not assured, because six of the cores or
46 percent have profiles indicative of scour or at a minimum no change in inventors'. This is based
on the occurrence of the PCB maximum in the top core segment as well as the presence of a smaller
but still substantial PCB inventory at depth.
LRC • ~6
TAMSTciraTech
-------
The inventory estimate for Hot Spot 28 was 1.850 kg in 1976-78 and 20.382 kg in 1994. At
face value this suggests that between 1977 and 1994. 18.532 kg of PCBs were deposited at Hot Spot
28. This is more than the 1984 Thompson Island Pool inventory estimate of 14,900 kg (see
Appendix B of this Responsiveness Summary). But. as discussed on page 4-35. only a small fraction
of the PCBs downstream was transported after 1977:
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 DE1R (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 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.
See response to comment LG-1.24 for more discussion.
The writer inappropriately separates the results into "good"' and "bad" data. The 1976-1978
data collected for Hot Spot 28 were not "bad" but simply did not capture the entire inventor}'. The
values themselves were probably accurate for the sediments measured. The reason that losses can
be assured is related to this. Since in most instances, the 1994 cores captured all of the recent
deposition as documented by the ,37Cs analysis, they represent all of the PCB contamination at the
location. Conversely, the earlier data do not have this assurance and so there is the possibility that
further contamination lay below the core or grab. As a result, the earlier estimate can be thought of
as a minimum inventor) estimate. Any differences between the 1976-1978 and 1994 surveys must
then represent a minimum difference since the 1976-1978 inventory estimate may be low. Thus any
comparison indicating PCB loss can be assured since in reality the actual loss may be bigger than
estimated. Additionally there are water column data to suggest that sediment losses are occurring,
thus substantiating these results as well. In the case of gains, the opposite is true. Since the 1976-
1978 inventory estimate is a minimum, any gain estimate represents the maximum gain since the
actual difference between the current measurments and the 1976-1978 inventory is probably less
than estimated. In the case of Hot Spot 28. this is precisely the issue, i.e., the earlier study failed to
capture the deeper, more contaminated sediments and thus underestimated the 1976-1978 inventory.
Hence the apparent gain is not real.
Response to LS-1.3
No response required.
LRC • T
TAMS'IctraTcch
-------
Response to LS-1.4
No response required.
4.2.4 7Be in Surface Sediments
Response to LG-1.5A
See Response to Comment LG-1.18
Response to LO-1.18
The writer proposes several issues in this comment with regard to the use of the 7Be results.
In particular, the writer attempts to use the 7Be data to determine a deposition rate for the sediments
and then state that the USEPA has potentially misclassified some low deposition rate sites as
nondepositional. While the USEPA agrees that there may be some low level of deposition which is
not detected by the sampling technique, it is probably lower than that suggested by the writer. The
entire approach proposed by the writer presupposes a knowledge of the geochemistry of 7Be which
is currently not available. Specifically, the temporal variability of 7Be deposition is not well known,
nor are there much data on the 7Be levels in depositing sediment. The approach used by the writer
to estimate the maximum undetectable deposition rate is not supported by the USEPA since it
requires this currently unavailable information. The writer's premise is also based on the assumption
that 7Be levels in depositing sediment are the same everywhere. This is also probably not true. These
issues are discussed in greater detail below. Essentially, the USEPA believes that it is important to
be cognizant of the limitations of the current understanding of 7Be geochemistry and to avoid "over
interpreting" the 7Be data.
As stated in the LRC, the presence of 7Be was used to discern those areas where recent
deposition had occurred. Due to its short half life, 7Be presence in the surfical layer specifically
indicates the presence of sediments deposited within the last 6 months to a year. However, 7Be
presence only proves recent deposition, not long term deposition. Sporadic events of deposition and
resuspension will yield 7Be-bearing sediments even though there is no long term burial.
Alternatively, mixing of sediment layers by biological activity ("bioturbation") can serve to mask
the presence of recent deposition by diluting the surface material with underlying, 7Be-free
sediments. Ultimately, it is only the presence and depth of more persistent tracers, such as 137Cs or
PCBs themselves which can provide a true measure of the deposition rate at a given location.
Part of the writer's premise is predicated on the delivery of 7Be during a single event in the
spring of 1994. This argument would also imply that 7Be levels are homogeneous throughout the
study area. Alternatively, this hypothesis would hold that the 7Be-to-;37Cs ratio would be constant,
since both constituents are to be delivered by the same event. Evidence for the delivery of "Cs
suggests that at least this radioisotope is principally delivered during the spring runoff event since
it currently has no atmospheric deposition component, unlike 7Be. 7Be input, while potentially
dominated by spring deposition, is not exclusively tied to this event since atmospheric production
and fallout are relatively continuous throughout the year. Thus surface sediment 7Be levels can be
partially replenished after the major depositional event of the year. This additional input of 7Be
LRC - 78
TAMS/TetraTech
-------
undermines the assumption of a single 'Be depositional event occurring around April 17, 1994 as
proposed by the writer. The occurrence of additional 7Be input, also noted by the writer but not
discussed, would serve to increase 7Be levels and yield an overestimate of the actual deposition rate
using the model promoted by the writer.
The likelihood that the sediment deposition of 7Be and l37Cs are not linked is supported by
the "Be/'"Cs ratio results for the 0-1 inch layer (surficial sediment). This ratio provides additional
evidence that the deposition rate of "Be is not a well-known phenomenon and cannot be used in the
manner suggested by the writer. The data for the ratio of 7Be/l37Cs are shown in Figure LG-1.18A.
This figure represents all of the low resolution coring results. Note that the "Be results are all decay-
corrected to a single date, Sept. 1, 1994 as follows:
-1og{2)4 (Date of Interest • Count Date). 53 28 days)
Date of Interest 'Count Date ^
where Cdllt =7Be concentration on date specified in pCi/kg.
This date is approximately midpoint in the sample counting period. Correcting to this date eliminates
decay concerns when examining the results while also avoiding the uncertainties associated with
decay correction over a long period of time (such as to April 17). This diagram shows that this ratio
varies over an order of magnitude (less than 0.25 to 6). Given that l37Cs is principally delivered
during the spring high flow event, these results suggest that 7Be and l37Cs are not linked to the same
pathways and, in particular, that "Be input is not simply governed by a spring high flow depositional
event. Figure LG-1.18B illustrates the absence of correlation between the two isotopes as well as
shown by the poor regression line drawn in the figure. As described above, irCs is derived almost
exclusively from soil erosion since there is no direct atmospheric input. Thus if 7Be were simply-
related to spring deposition, its ratio to 157Cs would remain relatively constant since both would be
delivered in essentially the same manner from the same source materials.
The detection limit for "Be is an important component of the writer's analysis. 7Be
measurements decay-corrected to September 1. 1994 suggest a detection limit of 400 pCi/kg decay
(.see Figure LG-1.18C). The value of 400 pCi/kg is a lower value than that obtained by examining
the detection limits but is probably more accurate since it is based on the levels detected and not an
estimation of the detection limit. This result suggests a lower threshold for "Be detection than
suggested by the writer.
The purpose of the analyses presented above, is not to present an alternate estimate of the
actual deposition rate "detection limit" achieved by the sampling but rather to simply show the
uncertainties in the writer's approach. It is USEPA's opinion that "Be cannot be used for the purpose
suggested by the writer since its geochemical input is too poorly known.
Ultimately, the "Be data provide some information of the occurrence of very recent deposition
at the sampling site. These data cannot be used to infer a deposition rate at the resolution suggested
by the writer since the initial conditions as well as the input function are not well known.
Additionally, these data cannot be used to infer long-term or continuous deposition since episodic
deposition and scour will also yield measurable levels of 7Be if sampled at the appropriate time.
LRC - n9
JAMS TetraTech
-------
Be'/ Cs (unitless)
Notes:
7 7
1. Be' represents sample Be activity decay-corrected to September 1st, 1994
(see text for discussion).
2. ^0 samples were not detected for Be. Cs was found in all low
resolution core tops.
7 137
3. Be' and Cs are reported for 0-1 inch of sediment from each core.
Hudson River Database Release 4.1 TAMS/TetraT>
Figure LG-1.18A
7 137
Distributions of the Be'/ Cs Ratio in Low Resolution Cores Tops
-------
3000
60
U
£3.
ca
2500
2000
1500
1000
500
OO
o
o o
„ o „
R= 0.13733
500
137
1,000
Cs (pCi/kg)z
1.500
2.000
Notes:
1. Be' is decay-corrected concentration to September 1st, 1994 (see text for discussion).
7 1T 7
2. Be' and " Cs are reported for 0-1 inch of sediment for each core.
3. Plot represents results for 112 low resolution cores. Six values were excluded as
outliers based on Mahalanobis analysis. Plot also excluded the 50 samples which were
non-detected for ?Be.
Hudson River Database Release 4 1
TAMS/TctraTech
Figure LG-1.18B
137 7
Relationship Between " Cs and Be' in Low Resolution Core Tops
-------
Notes:
7
1. Be' is decay-corrected concentration to September 1st, 1994 (see text for discussion).
7
2. Be' reported for 0-1 inch of sediment cores.
Hudson River Database Release 4.1 TAMS/Tetra
Figure LG-1.18C
Distributions of Decay-Corrected Be' in Surface
Sediments from Low Resolution Cores
-------
USEPA does acknowledge that there are inherent limitations in the "Be data but it is unlikely that
deposition rates as high as 0.5 cm/yr have been missed. However, as noted in Response LG-1.40A.
a deposition rate of 0.5 cm/yr will only yield 4 inches of sediment in 20 years, leaving the peak
sediment PCB concentrations near the surface and within the biological active zone. Lower
deposition rates would of course leave the layers even closer to the sediment surface. The writer's
model is too simple to explain the "Be results, especially given the lack of knowledge concerning
the temporal "Be input to the river. The USEPA still considers the interpretation of non-detect "Be
levels as indicative of sites with little or no very recent deposition and as sites potentially undergoing
scour.
Response to LG-1.18A
This comment has a large number of issues which are discussed separately below. Much of
the comment is based on a data set obtained by GE during June through August of 1998. This data
set had not been submitted to USEPA in time for complete review prior to the preparation of this
Responsiveness Summary and as such the USEPA comments on the diagrams provided by GF.
should be considered preliminary.
The ability to discern statistically significant trends with the limited data set that GE
collected in 1998 is highly unlikely based on the results presented by GE as of the time this report
was being prepared.
GE presents the results of only 12 cores and attempts to use this much smaller data set as a
basis to discredit the much larger USEPA effort (76 cores from the TI Pool and 94 cores from
TIDam to Lock 2). Nonetheless, some useful information can be obtained from the GE cores as
presented. According to GE's contentions, deposition rates in the Tl Pool are about 1 cm per year
so that these layers would represent 5 years of deposition. USEPA does not accept this deposition
rate and believes that it is too high and certainly not applicable throughout the Pool. In nearly all
cores presented, there was no apparent PCB decline in the upper 5 cm despite the major load
reductions in PCBs entering the TI Pool over the past 5 years. These trends suggest that
sediment-derived PCBs may be contaminating any recently deposited sediment as it is deposited.
This would be consistent with sediment PCB loss as documented in the LRC.
This contention is also supported bv the variability in the surface ratios of the sediments. For
both the MDPR and Peak 46/32, surface (0 - 1 cm) sediments exhibit a wide range (MDPR of 0.3
to 0.8; Peak 46/32 ratio of 0.25 to 0.6, based on the graphs provided since USEPA does not have the
actual data) which is inconsistent with a single source such as that of the Hudson Falls facility. These
ratios are also substantially displaced toward greater degrees of dechlorination relative to Aroclor
1242. For the MDPR, initial ratios should be less than 0.14 (the ratio in Aroclor 1242) since
water-column transport yields a suspended matter mixture that is fractionated toward the heavier
congeners. Thus these ratios suggest that the material being deposited is not recently released
Hudson Falls contamination but rather represents material which has been re-released from the
sediments or perhaps a mixture of both recent and re-released contamination. Variations in these
ratios may result from the degree of PCB dechlorination in nearby sediments, the ratio of recently
released to re-released PCBs. and the period of time spent in the water column prior to redeposition.
The level of dechlorination w ould presumably be related to the local concentrations since, as show n
in the DEIR (USEPA. 1997 ). the degree of dechlorination varies with PCB mass. As to the ratios
LRC ¦ 83
TAMSTctraTcch
-------
found in deeper sediments (1-5 cm), it is interesting to note the reversals present in several of the
profiles from more dechlorinated ratios to less dechlorinated to more dechlorinated in this very short
distance. Perhaps in these instances, the 1991 event is evident as the ratio reversal. In these instances
this would place the 1991-1992 horizon around 3 cm deep, yielding a deposition rate of 3 cm/6 years
or a 0.5 cm/year deposition rate, placing at maximum 11 cm or 4.5 inches of sediment over the peak
concentrations associated with the 1970-1975 period, hardly an example of "'deep'' burial. Profiles
lacking this reversal suggest even slower rates of deposition since they lack sufficient sediment
thicknesses to resolve this event. The presence of a few centimeters of sediment also does not
necessarily represent long term deposition and sequestering of sediment PCBs since deposition may
be transient, present for a few years only to be removed by a one-in-three or one-in-five year flow
event.
In addition to the issues raised above, there are also analytical differences to be considered.
Although USEPA and GE analytical data have been shown to indicate generally similar trends in
water column loads, there has not been a direct reconciliation of the two analytical techniques for
sediment. Of particular note is the use of only one analytical column by the GE investigators while
the USEPA technique is based on a two column technique with 10 percent of samples confirmed by
a third column. The GE data are also based on Aroclor standards and not the congener-specific
standards utilized by USEPA. The USEPA technique is designed to be more conservative in its
approach with more internal checks as well as a formal data validation program to certify data
quality. These analytical differences may serve to create systematic differences between the two data
sets.
Regardless of the most recent depositional trends and potential analytical differences, the
most useful comparison is provided in Figure 13 of the GE comments. Specifically, this diagram
shows that both the USEPA and GE data yield substantially lower sediment inventories relative to
those obtained in 1984. The limited GE data set is probably too small to be shown statistically
different from either prior set of efforts but the trend to lower inventories relative to 1984 is clearly
suggested by the 1998 data.
The clear trend presented in Figure 13 should be contrasted with the Figures 11 and 12
presented by the writer. Specifically these figures attempt to suggest that the large difference
between the 1984 and 1994 data sets is comparable to that between the 1994 and 1998 GE data sets.
This is simply untrue and misconstrues the measurement uncertainty. As presented in the diagrams,
the average difference between 1984 and 1994 is -80 percent. This is larger than the estimate
obtained by USEPA (Appendix A) but is used here for the purposes of this discussion. Applying this
difference to a 100 g/m: 1984 Tri+ inventory would leave 20 g/m: in 1994. The writer contends that
a similar scale "gain" occurs from 1994 to 1998. This is not true. If the 89 percent rise as Delta is
applied to the 1994 inventory of 20 g/m\ this yields a 1998 inventory of 38 g'm ; representing a
Delta of -62 percent relative to 1984. This is well within the uncertainties associated with the
estimates of the 1984 to 1994 differences but nonetheless indicates a major loss of inventory from
1984 to the present. Thus the differences between the 1994 and 1998 data sets are substantially
smaller than the 1984 to 1994 differences. This is shown quite clearly in Figure 13 which show s the
1984 Tri-i- inventory levels relative to the 1994 and 1998 data. Apparently the error bars on the
diagram represent the individual points and not the uncertainties about the mean values. It should
be noted as well that the data have been ''filtered'' to include only those pairs separated by 5 feet.
LRC • 8-1
JAMS TeiraTech
-------
USEPA does not believe it is appropriate to "filter" the data as suggested by the writer, as discussed
in response LG-1.9.
Response to LG-1.18B
The USEPA agrees that erosion cannot be inferred for specific areas under specific flow-
events, however, neither can long term deposition. The distribution of sediment resuspension and
settling asserted by the writer is principally based on modeling assumptions and not on
measurements. Other processes may affect the sediment transport rate besides simple resuspension
and settling within the normal river boundaries. Among some important processes which occur only
during high flow events is the deposition of sediments in backwater areas and near-shore areas which
are normally found above water. These areas are subject to both deposition and erosion during the
high flow events while subject to surface runoff erosion during the period between major flood
events. These special processes may affect both sediment and PCB transport during these events,
making it difficult to characterize the system as a whole.
The model mentioned by the writer is described in an attachment to their comments as
Appendix B. As attached, this information is not sufficient for a thorough review. It is unclear at this
time as to whether the assumptions made in assembling this model are appropriate or well
constrained by the available data. The USEPA has not received sufficient information so as to review
the GE models. The USEPA intends to rely on its own modeling efforts to examine deposition
phenomena.
Response to LG-1.18C
As mentioned in response LG-1.18B, the USEPA has not yet received sufficient information
so as to review GE's sediment transport model. While the contentions put forth by the writer sound
interesting, it is unclear whether the model is sufficiently constrained by available data to make its
output meaningful. With regard to Hot Spot 14, it should be noted that the absence of berylium-7 in
several sites is not the only evidence for lack of burial and possibly scour in this area. A large area
of fine-grained sediments was found along the western side of Hot Spot 14 in which lineated
sediment structures were found indicative of sediment scouring by flow. See response LG-1.18 for
a discussion of berylium-7 and its relationship to long-term deposition. Lastly, it should be noted
that the degree of variability in sediment PCB inventories is in part an indication of the degree of
heterogeneity in sediment deposition and resuspension. The writer is reminded that assigning an
average deposition rate to an area largely ignores this fact and may mask important local rates of
sediment resuspension.
Response to LG-1.19A
The USEPA acknowledges that the text could have been written more clearly to explain the
relationship between "Be and PCB inventory. In both the TI Pool and the areas below the TI Dam,
the absence of "Be was shown to coincide with lower PCB inventories (median inventory at 4 g/nr).
Areas with 'Be present had higher PCB inventories (median inventory of 10 g/nr). In the case of
the TI Pool, the coincidence was shown to be statistically significant. For the areas below the TI
Dam, the data set was considered too small to provide a useful statistical test but still yielded the
expected relationship. Individual comparisons did not prove as useful, since not all sites with lower
LRC - 85
TAMSTetraTcch
-------
PCB levels relative to previous studies were nondetect for 'Be. As discussed in the Report, however,
this was considered to be evidence of temporary, very recent deposition in an environment which
had clearly undergone PCB loss, possibly via scour. Nonetheless, when considered as a whole, the
'Be results when examined on an absence/presence basis were consistent with the anticipated trend.
Thus the text of the Report is not inconsistent.
The issue of the 7Be data is discussed in the response to LG-1.18 as well and the writer is
referred to that section.
Response to I.S-1.1
No response required.
4.2.5 Hot Spot Boundaries
4.2.6 Comparison of the 1994 Hot Spot Inventories with Other 1977 Estimates
4.3 Sediment Contamination in the Near-Shore Environment
4.4 Summary and Conclusions
4.4.1 Sediment and PCB Inventories in the TI Pool
Xo significant comments were received on Sections 4.2.5 through 4.4.1.
4.4.2 Sediment and PCB Inventories Below the TI Dam
Response to LG-1.24
GE implies that the available evidence has too high a level of uncertainty to draw any
conclusions regarding change of PCB mass over time in hot spots located below the Thompson
Island Dam. While there is considerable uncertainty in the LRC estimates, as described in detail in
the Low Resolution Sediment Coring Report, the USEPA contends that the data are sufficiently
precise to draw firm conclusions regarding loss of PCB mass from several of the downstream hot
spots.
This comment first presents an argument based on the sampling distribution of differences,
and notes that the 95% confidence interval on the difference between 1976-1978 and 1994 arithmetic
mean MPA includes zero for each hot spot analyzed below Thompson Island Dam. The standard
error of the mean difference is calculated from the standard error of the mean for the individual
samples as
This standard error is then used to create a confidence interval about the difference in
arithmetic means based on a spread of two standard errors, as presented in GE's Table 11.
LRC - 86
TAMSTetraTech
-------
In fact, the Low Resolution Sediment Coring Report {see p. 4-30) does not contend that the
arithmetic differences are significantly different from zero or that the 1994 arithmetic means are
significantly different from 1976-1978 means at the 95% confidence level (although the differences
in MPA for Hot Spots 28,31, and 37 are significantly different at the 90% confidence level). Rather,
the Report demonstrates that the geometric means (means of the log-transformed data) are
significantly different for four of the downstream hot spots in terms of MPA and five of eight in
terms of LWA at the 95% confidence level. The geometric mean is an estimate of the median (50th
percentile) of an arithmetic distribution. A significant decrease in median MPA can reasonably be
concluded to represent a decrease in total mass. Differences in arithmetic means are not statistically
significant at the 95% confidence level because the uncertainty in estimating arithmetic means from
small samples of skewed distributions is taken into account in the MVUE estimator of the standard
error. Essentially, the MVUE-based confidence limit addresses the small but finite statistical
possibility that a small number of unobserved very high values might be present but not sampled,
thus lending uncertainty to the arithmetic mean estimate.
Accounting for geophysical evidence in addition to statistics leads to the conclusion that
inventories have indeed declined. Consider the alternative hypothesis that the median has indeed
decreased, but the arithmetic mean has not. For this to occur, the decline in the median would need
to be compensated for in the average by an increase in the high MPA values in the right-hand tail
of the distribution. In other words, PCB mass would either need to be concentrated in a smaller
volume, or new high-concentration PCB mass would need to be implaced in a few isolated pockets.
There is no evidence for physical mechanisms which would account for either possibility in the
upper Hudson below Thompson Island Dam. Therefore, a statistically significant decrease in the
geometric mean can indeed be interpreted as representing a significant decrease in PCB mass.
USEPA agrees with the comment that uncertainty in mass change calculations is, in part, due
to the small number of samples. The 1994 effort was not designed to be an exhaustive resurvey of
PCB mass in downstream hot spots, and, in any case, the density of samples available from 1976-78
is also low. As noted, uncertainty is also introduced by the necessity of extrapolating grab samples
and interpreting sediment density in the 1976-78 samples. The presence of these sources of
uncertainty does not invalidate the basic findings of the discussion, that mean PCB mass appears to
have declined since 1976-78 in most of the hot spots below Thompson Island Dam; rather it effects
the degree of statistical certainty which can be applied to conclusions regarding the magnitude of
change.
Finally, GE states that "the best indication of the overall uncertainty of the approach TAMS
used is the result obtained for Hot Spot 28. The implausibly large increase of mass in this hot spot
is dismissed." This argument addresses the statistics, but once again fails to consider the physical
evidence present. In fact, there is good evidence to indicate that the 1976-78 sampling did not core
below 12 inches in this hot spot, whereas later sampling suggests a significant amount of the PCB
mass is present at depth. As stated on page 4-35 of the Report, "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."
LRC-87
T AM S'Te traTcc h
-------
4.4.3 Sediment Contamination in the Near-Shore Environment
Response to LF-1.1
No response required.
Response toLS-1,6
No" response required.
4.4.4 Summary
No significant comments were received on the Summary.
Appendices A, B, C, D, E, F
No significant comments were received on the Appendices,
LRC-88
TAMS/TetraTech
-------
REFERENCES
Flood, R.D. 1993. Analysis of Side-Scan Sonar, Bathymetric, Subbottom, and Sediment Data from
the Upper Hudson River between Bakers Falls and Lock 5. State University of New York at Stony
Brook, Marine Science Research Center. September, 1993. Report to TAMS Consultants, Inc. for
the Hudson River PCB Reassessment RI/FS.
McNulty, Anne, K. 1997. In-Situ Anaerobic Dechlorination of Poly chlorinated Biphenvls in Hudson
River Sediments. Master of Science Thesis Submitted to Rensselaer Polytechnic Institute, Troy,
NY.
Tofflemire, T.J., and S. 0. Quinn. 1979. PCB in the Upper Hudson River: Mapping and Sediment
Relationships. NYSDEC Technical Paper No. 56. March 1979. NYSDEC, Albany, New York.
USEPA, 1998a. Low Resolution Sediment Coring Report. Phase 2 Report, Volume 2C-A.
Prepared by TAMS Consultants, Inc. and Tetra Tech, Inc., July.
USEPA, 1998b. Hudson River PCBs Reassessment RI/FS Responsiveness Summary for Volumes
2A: Database Report, 2B: Preliminary Model Calibration Report, and 2C: Data Evaluation and
Interpretation Report. Prepared by TAMS Consultants, Inc., Limno-Tech, Inc., Tetra Tech, Inc., and
Menzie-Cura & Associates, Inc., December.
LRC-89
T AM S 'TetraTech
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
Appendix A
-------
APPENDIX A
A Comparison of PCB Sediment Inventories
in the Thompson Island Pool, 1984 to 1994
Low Resolution Sediment Coring Study
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
APPENDIX A
A Comparison of PCB Sediment Inventories
in the Thompson Island Pool, 1984 to 1994
Low Resolution Sf.dimf.nt Coring Study
TABLE OF CONTENTS
Page
Discussion A-l
List of Tables
A-l Subreach Variogram Models for Natural Log of PCB Mass Concentration. 1984
Thompson Island Pool Sediment Survey
A-2 Comparison of Tri+ MPA Arithmetic Means for All NYSDEC 1984 Sample Points
in the Sample Areas and Co-Located 1984 to 1994 Sample Points
A-3 Number of Locations in Sample Areas for 1984 and 1994
A-4 Shapiro-Wilks Test and Ratio of Arithmetic Mean to Standard Deviation for Sample
Points in Sample Areas for 1984 and 1994
A-5 Selection of Cluster Area Best-Estimate-of-Mean for 1984 and 1994 Each Sample
Area
A-6 Estimate of the Average Molar Change in the Sediment PCB Inventory (MPA)
Trichloro- and Higher Homologues
A-7 Shift in Homologue Group Distributions (Mole Percent) for Matched Cores in the
Thompson Island Pool
A-8 Selection of Best-Estimate-of-Mean for 1984 Results After Correcting for
Dechlorination Loss
A-9 Estimate of the Average Molar Change in the Sediment PCB Inventory Excluding
Dechlorination (Trichloro- and Higher Homologues)
List of Figures
A-l Construction of the Cluster Areas Using the Subreach Semivariogram Models
A-2 1994 and 1984 Sample Locations in Cluster Areas with Side Scan Sonar Sediment
Classification
A-3 1984 vs. 1994 Sediment Tri+ MPA Best-Estimate-of-Mean Basis
A-4 1984 vs. 1994 Sediment Tri4- MPA Based on Arithmetic Mean and MVUE
A-5 1984 vs. 1994 Sediment Tri+ MPA Best-Estimate-of-Mean Basis (1994 Complete
Cores Only)
A-6 Best Estimate of Tri+ MPA for Cluster Areas: 1984 Cores Only vs. 1994 Cores
A-7 Best Estimate of Tri+ MPA for Cluster Areas: 1984 Grabs Only vs. 1994 Cores
A-8 Distribution of Tri+ MPA Best-Estimate-of-Mean Ratios (1994/1984) for Cluster
Areas
i TAMS TctraTech
-------
APPENDIX A
A Comparison of PCB Sediment Inventories
in the Thompson Island Pool, 1984 to 1994
Low Resolution Sediment Coring Stl dv
TABLE OF CONTENTS
A-9 Cesium-137 and Total PCB Profiles for Cores Collected in 1983 and 1991 from the
Thompson Island Pool
A-10 Distribution of Tri- MPA Best-Estimate-of-Mean Ratios (1994/1984) for Cluster
Areas with 1984 MPA Corrected for Dechlorination Loss Between 1984 and 1994
n
TAMS TctraTech
-------
Appendix A
A Comparison of PCB Sediment Inventories in the Thompson Island
Pool, 1984 to 1994
This Appendix describes an alternate statistical analysis for the estimation of the
direction and degree of change in the PCB inventory in the sediments of the TI Pool. The original
statistical analysis presented in the Low Resolution Sediment Coring Report was based on a
comparison of the results of the extensive 1984 NYSDEC survey of the PCB inventory at the TI
Pool and a series of matched sediment cores collected by the USEPA in 1994. Specifically, the
PCB inventories from a set of sixty sampling locations in the TI Pool were compared on a point-
to-point basis to provide a quantitative understanding in the direction and extent of change of the
PCB inventory of the Thompson Island Pool. The LRC concluded that the sediment PCB
inventory has substantially declined, presumably by re-release to the river. Additionally, the
report concluded that there is no evidence of extensive burial.
The statistical analysis presented in this Appendix examines the 1984 and 1994 data from
an area-based perspective, as opposed to the point-to-point comparison used in the LRC. This
analysis simply presents an alternate basis to examine the change in the PCB sediment inventory
between 1984 and 1994. In part, this analysis is to address the concerns expressed by some
reviewers that the point-to-point comparisons presented in the LRC may be biased due to a
"regression toward the mean'' effect or by the distance of separation between the 1984 and 1994
sampling locations. Although the USEPA does not accept these criticisms as wholly valid, the
statistical analysis presented here was designed to avoid these issues.
Outline of Analysis
The area-based examination required the construction of area-based estimates for the areas
to be compared. The examination presented in this Appendix is based on the procedure outlined
below. A detailed discussion of the individual steps follows this outline:
1. As described in the LRC, the 1994 sampling locations were arranged in clusters
and placed in areas of apparent local homogeneity in the PCB inventory and texture, based on the
1984 sampling results. These groups form the basis for the area-based comparison of PCB
inventory between 1984 and 1994.
2. The semivariogram analysis presented in the DEIR was used to establish the "area
of influence" around each of the 1994 sampling locations. Essentially, either a circle or an ellipse
of "influence" was defined for each 1994 sampling location. The shape, size and orientation of
each "area of influence" was dependent on the section of the TI Pool in which it was located. All
seventy-six 1994 TI Pool locations were considered in this manner and not simply the ones
specifically matched to the 1984 locations.
3. The clusters of 1994 sampling locations were grouped into larger areas based on
the overlap of the individual "areas of influence.'" These areas essentially corresponded to the
A-1
TAMSTetraTech
-------
original clusters developed in the sampling plan for the low resolution coring program. The shape
and orientation of the larger areas was again defined by the section of the TI Pool in which it was
located, following the proportions defined from the semivariogram analysis. The size was defined
as the minimum area which would encompass the individual ''areas of influence" for the cluster of
1994 sampling points. Clusters with extensive overlap were combined into a single large area.
4. All 1994 cores in a given cluster area were used to establish a mean PCB inventory
for the cluster for 1994. Similarly, all 1984 cores and grabs in the cluster area were used to
establish the mean PCB inventory for the cluster for 1984. Additionally, the 1984 sampling
points were separated based on sampling method (i.e., core or grab), and used to establish cluster
area means based on the specific sampling technique. Similarly, the incomplete cores from the
1994 sampling program were excluded and an alternate inventory estimate for the cluster areas
was obtained and contrasted with the 1984 cluster means. The calculations were based on the
trichloro- and higher homologue sums for 1984 and 1994. The 1984 trichloro- and higher sum was
based on the calculation technique described in Appendix E of the LRC. As part of this
construction of means for each cluster area for each of the sampling programs, the best basis for
estimating the mean was examined. For each cluster mean estimate, either the arithmetic (i.e.,
simple) mean or the minimum-variance-unbiased-estimator of the mean (MVUE) was selected as
the cluster mean. Like the point-to-point comparison presented in the LRC, the area-based
comparisons are considered to be representative of the fine-grained sediments of the TI Pool.
5. The cluster means for the 1984 and 1994 sample data were then compared and
used to estimate the net change for the entire set of clusters. The estimate of net change was
based on two separate statistical approaches. In the first approach, a linear regression on the
1984 and 1994 cluster means was used to estimate the mean inventory change. In the second
approach, the ratio of the mean 1994 inventory over the mean 1984 inventory for each cluster
was the variable used in the statistical analysis since this ratio was found to be statistically better
"behaved" than the Delta function used in the LRC. Specifically, when the 1994/1984 ratios were
examined, the distribution of ratios was found to be log-normally distributed. By comparison, the
Delta functions of the LRC were neither normally nor log-normally distributed, although they
were closer to log-normal than to normal distributions. All of the individual cluster ratios were
then examined as a whole to establish the direction and degree of change for the fine-grained
sediments of the entire TI Pool.
6. The comparisons were based on the entire cluster set for both 1984 and 1994 as
well as subsets of the 1984 and 1994 data sets, based on sampling technique (i.e., 1984 core or
grab) and on 1994 core completeness. These comparisons demonstrated the robustness of the
mass loss regardless of the assumptions concerning the underlying data sets.
7. Lastly, a correction for in situ dechlorination from 1984 to 1994 was obtained
from the data collected by McNulty, 1997 and applied to the 1984 PCB data. This permitted the
calculation of the trichloro- and higher homologue mass loss from the sediments exclusive of any
dechlorination loss.
Establishing a Basis for Comparison
In this comparison, the samples are grouped by the cluster locations developed for the
low resolution sediment coring program instead of comparing the 1984 samples to the 1994 cores
A-2
T AMS/T etraT ech
-------
on a point-to-point basis. The shape, orientation and dimension of the area associated with each
of the 1994 sampling points are determined by the results of the geostatistical analysis of the
Total PCB mass in the TI Pool using the 1984 data, as discussed in Section 4.2.4 of the DE1R.
Results of this analysis are listed in Table A-l, which is based on Table 4-4 of the DEIR.
In order to establish the "areas of influence" which were spatially correlated with the
1994 cores, circles or ellipses were drawn centered on each 1994 sampling point. The dimensions
of these circles or ellipses were based on the appropriate practical range, anisotropic ratio and
orientation of the major axis given in Table A-l. Figure A-l is a map illustrating the "areas of
influence" defined for each 1994 sampling location in the TI Pool. For subreaches 1, 2. and 5 as
defined in the DEIR. the '"area of influence" was defined as a circle about the sampling point
whose radius was given by the practical range developed from the semivariogram analysis. The
practical range can be thought of as the distance from a sampling point where the sediment
inventory would be expected to correlate with the original sampling point. Beyond this distance,
no correlation between the original sampling point and other sampling points is evident. In these
three subreaches, no directional component (i.e., downstream or cross-stream) to the spatial
correlation was evident. Hence, the spatial correlation is considered isotropic (equal in all
directions) and a "circle of influence" was defined.
For subreaches 3 and 4. a directional component to the spatial correlation was evident. In
these subreaches, the "areas of influence" were defined as ellipses whose major axes were
oriented parallel to river flow. The degree of anisotropy was used to determine the ratio of the
major axis to the minor (cross-stream) axis for the ellipses. The degree of anisotropy is
essentially a measure of the ability to estimate river PCB inventor}' conditions upstream and
downstream of a sampling point relative to conditions cross-stream.
After establishing the "areas of influence" about each sampling point, overlapping areas
were combined to form a single, larger area for the entire cluster. The larger area was defined with
the same proportions as the smaller areas it encompassed. The area was defined as the smallest
area sufficient to encompass all of the smaller polygons (see Figure A-l). Note that these cluster
areas were defined solely on the basis of the 1994 sampling locations. This procedure yielded 14
cluster areas for comparison.
Utilizing these cluster areas as overlays in a geographical information system, the 1984
and 1994 sampling locations contained within these areas were identified. This approach
expanded the sample basis for estimating the 1984 and 1994 sediment inventories, yielding 243
locations for 1984 and 70 locations for 1994. This should be compared to the 60 paired locations
used for both 1984 and 1994 in the point-to-point comparison presented in the LRC. Because
overlapping areas of influence were combined into larger cluster areas, none of the 1994 sample
locations and only one out of 243 of the 1984 sample locations were contained in more than one
cluster area. Figure A-2 presents a map of the TI Pool, illustrating the locations of the 1984 and
1994 sampling locations considered in this analysis. The bounds of each cluster area along w ith
associated 1984 and 1994 locations contained within each cluster area are shown.
Processing of the Data
On the previous section, a set of samples for 1984 and 1994 was established as a basis for
the creation of area-based mean sediment PCB inventories. In this section, the basis for
A-3
TAMS TetraTech
-------
calculating these means is presented. The original calculations presented in the LRC examined the
sum of trichloro- and higher homologues in 1984 against the sum of tri-chloro- and higher
homologues plus the congeners BZ#1, 4, 8, 10 and 19 for 1994 on a molar basis. This calculation
was presented as a minimum estimate of the loss of trichloro- and higher homologues to the w ater
column since it assumed the absence of dechlorination products in 1984. Subsequent discussions
and review suggest this may have been too conservative an approach since evidence for the
occurrence of dechlorination in 1984 was reported by several authors (e.g., Bopp et al., 1985).
In light of this, the PCB inventory as moles of trichloro- and higher homologues per unit
area (abbreviated as Tri~ MPA) was selected as the initial basis for comparison in this
calculation. A correction based on an estimate of the actual rate of dechlorination developed from
McNulty. 1997 will be applied later in the analysis. The individual Tri+ MPA estimates for each
sampling location (i.e., core or grab) were calculated using the equations provided in Chapter 4 of
the DE1R. with the exception that the corrected factor of 0.944 was applied to the 1984 data as
described in Appendix E of the LRC.
The data treatment applied to the grab samples was the same as that used for the kriging
analysis as discussed in the DEIR. Specifically, sample depth for grab samples was assigned
based on sediment texture: 12" for coarse-grained samples and 17" for fine-grained samples, as
originally defined by N'YSDEC.
Co-located 1984 sample pairs (i.e., field duplicates) were treated in the manner described
in the DEIR for the kriging analysis: for core pairs or grab pairs the values are averaged, for core-
grab pairs the core value is used, and for pairs in which one sample was analyzed with GC/ECD
and the other screened with mass spectrometry, the GC/ECD value is used. Treatment of field
duplicates in this manner reduced the number of sampling locations for 1984 from 243 to 197.
Representativeness of the Data
Several comments on the LRC raised the issue of the representativeness of the low
resolution core sampling locations and their associated 1984 sample result. The following
discussion addresses this concern in the context of the cluster area means. While the USEPA does
not accept all the commentors' claims as correct in this regard, the following approach
demonstrates that issues do not pertain to the area-based cluster estimates used in this Appendix.
Table A-2 presents several sets of arithmetic means for the cluster areas. The first column
of data lists the cluster area Tri+ MPA arithmetic means for all 1984 samples (197 in total,
excluding duplicates) contained within the clusters. The second column represents the arithmetic
means of the 1984 sample locations reoccupied in the Phase 2 sampling event, grouped by cluster
area. Notably, the cluster area mean values for all 1984 points are less than the mean values of the
reoccupied sample location results in 10 of the 13 cluster areas containing matched 1984 to 1994
sample locations. (One cluster area, LR-13, had no matched 1984-1994 sampling locations as
presented in the LRC.)
When all of the 1984 samples contained in the cluster areas are considered together, the
mean for all 197 points is less than half of the value for the 59 re-occupied locations (0.061
moles/m- for all points in the sample areas and 0.134 moles/m2 for the re-occupied locations).
These means are considered significantly different, assuming an uncertainty of two standard
A-4
TAMS. I etra l ech
-------
errors around the mean. This is indicative of the fact that the low resolution coring program
attempted to sample among the more contaminated sediments of the TI Pool, anticipating that
any sediment mass loss would be most easily discernible in these areas.
For all fine-grained areas in the Thompson Island Pool, the average Tri-t- MPA is 0.050
moles/m2 (based on the revised 1984 inventory analysis presented in Appendix B), which is
within the uncertainty of the mean for the one hundred and ninety-seven 1984 samples. This
comparison demonstrates that the 14 cluster areas studied as part of the low resolution coring
program and defined based on the semivariogram analysis can be considered representative of all
fine-grained sediments in the Tl Pool when estimating change in the PCB inventory.
The last column in Table A-2 presents the arithmetic means for the 14 cluster areas based
on the 1994 coring locations. The 1984 inventory estimates are higher than the 1994 estimates in
nearly every case, regardless of whether the matched points or the entire cluster areas are
considered. These differences will be shown to be statistically significant later in this Appendix.
In addition to examining the representativeness of the PCB concentrations themselves, the
physical nature of the cluster areas was also reviewed. Specifically, both the reported sediment
sample textures and the side-scan sonar results were examined in this context. The number of
sampling locations assigned to each cluster area is listed in Table A-3. The sampling locations are
further divided by sediment classifications based on side-scan sonar for both 1984 and 1994
locations, visual texture classification for 1984 samples and principal fraction based on laser-
grain-size distribution for 1994 samples. For the majority of cluster areas (11 of 14). the fine-
grained samples represent the majority of sampling points in the cluster, based on all four
measures, thus supporting their classification as fine-grained areas. In two instances, the areas
were dominated by coarse-grained samples and locations and were classified accordingly.
The general classification for the remaining cluster area (LR-06&07) was not as easily
resolved. Figure A-2 shows the cluster areas and sampling locations superimposed on the side-
scan sonar sediment classifications. In most instances, the cluster areas captured a majority of
fine-grained sediment with some coarse-grained samples. This pattern is also evident in cluster
area LR-06&07, in which a band of fine-grain sediments runs down the center. According to the
side-scan sonar analysis, 17 of the 24 1984 sample locations fall into the coarse-grained region,
but 14 of the 24 samples are fine-grained by visual texture classification. Similarly, the 1994
samples were exclusively classified as fine-grained based on laser-grain-size distribution analysis
but 4 of the 7 were classified as coarse-grained based on the side-scan sonar classifications. On
the basis of the specific samples, cluster area LR-06&07 was classified as fine-grained.
The original low resolution sampling program was principally designed to focus on fine-
grained sediments so as to examine the change in PCB inventory in the more contaminated regions
of the TI Pool. As shown above, the selected 1994 sampling locations and the associated 1984
locations were principally classified as fine-grained and as such can be considered representative
of fine-grained sediment regions. This assessment is also supported by the agreement of the mean
Tri+ inventory of the cluster areas with the mean Tri+ inventor) for the entire domain of fine-
grained sediments from the TI Pool.
A-5
TAMS/TetraTech
-------
Calculation of the Cluster Area Mean Inventories
The assessment of the mean mass loss from the sediments of the TI Pool is derived from
the mass estimates for each cluster to be compared. Thus these estimates need to accurately
represent the data available for each cluster. To this end, the data distributions for each of the
cluster areas for each of the sampling programs was examined to assess the degree of normality
and log-normality. The arithmetic mean for truly normal data distributions can be estimated by
calculating the arithmetic average of the sample population. Alternatively, when an underlying
distribution is log-normal, the MVL'E may represent the best estimate of the arithmetic mean.
Guidance on the basis for selecting the arithmetic mean or the MVUE was obtained from Gilbert.
1987.
The general log-normal nature of the groups of samples which comprise the individual
cluster areas is illustrated by the statistics provided in Table A-4. This table lists the results of
the Shapiro-Wilks W Test for normality for cluster area by year. The W statistic plus the
probability that the underlying distribution is normal is given for the original Tri- MPA values
plus the log-transform of these values. The closer the W statistic is to a value of one. the more
normal the underlying distribution. Most cluster areas have W values closer to one for the log
transform. The majority of cluster areas have probability values greater than 0.05 (5 percent)
indicating that the distributions could be log-normal. Since 71 percent of the 1984 cluster areas
and 64 percent of 1994 cluster areas appeared more log-normal than normal for 1984 and 1994,
respectively, all areas were assumed to be log-normal. The W test results are also listed for the
1994 cores that completely capture the PCB inventory at the sample location (i.e., complete or
nearly complete cores only). Nine cores were excluded from the 1994 cluster area analysis
because they were considered possibly incomplete representations of the sediment PCB
inventory at their respective locations as discussed in the LRC and response LG-1.2. The
majority of sample area distributions for the 1994 complete cores also have greater W values and
probabilities for the log-transformed data relative to the untransformed values. It should be noted
as well that the entire set of 1994 and 1984 values also indicate underlying log-normal
distributions, consistent with the individual cluster areas.
Because the W test suggests that the underlying distribution of the data may be log-
normal, the minimum variance unbiased estimator of the mean (MVUE) may be a better predictor
of the true population mean than the sample arithmetic mean. Guidance obtained from Gilbert.
1987 states that if the coefficient of variation is greater than 1.2, the MVUE is a better predictor
of the mean, otherwise the arithmetic mean can be used to estimate the population mean. The
coefficient of variation is defined as the ratio of the standard deviation to the arithmetic mean for
the sample group {i.e.. the individual cluster areas). This ratio is listed for each sample area on
Table A-4. The 5 of the 13 individual cluster areas in 1984 plus the entire 1984 data set (i.e., 197
points) have values greater than 1.2 indicating that the MVUE should be used to estimate the
mean in these instances. Only two of the cluster areas in 1994 utilizing all cores as well as
complete cores only have values greater than 1.2. The MVUE will be used in these instances as
well. Calculation of the MVUE is based on Gilbert, 1987.
Table A-5 summarizes both the arithmetic mean and MVUE estimates for all cluster areas
for the Tri+ MPA in mole/m2. The last column in the table represents the best estimate of the
mean based on the coefficient of variation criterion described above. The best estimate values are
used throughout the remaining discussion in this Appendix.
A-6
TAMS-TctraTcch
-------
Regression-Based Estimates of Mass Loss
Two approaches were used to estimate sediment PCB mass loss based on the Tri+ MPA
data. The first of these uses a regression analysis between the 1984 and 1994 best-estimate-of-
the-mean values from each of the cluster areas. Essentially, a regression of the form:
1994 MPA = a * 1984 MPA ^ b
was examined for the set of cluster area means. An initial examination of these regressions
indicated that the intercept term "b" was not statistically different from zero. As a result, the
regression was forced through zero, yielding the form:
1994 MPA = a * 1984 MPA
The advantage of this form is that the slope term "a" can be directly interpreted as a mass loss
estimate. The nature of this form of estimate is such that the cluster areas farthest from the
overall average 1984 and 1994 MPA values {i.e.. the average of all cluster areas) weigh more
heavily in the determination of the slope "a." The other approach to be used to estimate the mass
loss will weigh all clusters equally, regardless of their MPA value.
The results of the single coefficient linear regression directly comparing the Tri4- sediment
inventories (MPA) for 1984 and 1994 are shown in Figure A-3. A diagram of the regression as
well as some summary statistics are provided. The diagram shows the regression line along with
the estimated uncertainty. The fact that the uncertainty about the regression does not include the
line with a slope of unity (i.e., 1994 - 1984) indicates that the slope is statistically different from
unity and therefore indicates a statistically significant mass loss. The slope of the regression can
be converted to the Deltas expression used throughout the LRC as follows:
DeltaM = 1994 MPA - 1984 MPA = 1994 MPA - 1
1984 MPA 1984 MPA
Slope - 1
Thus the DeltaM estimated from Figure A-3 is (0.41 - 1) or -59 percent ± 19 percent. This
represents an estimate of the mean mass loss of trichloro- and higher homologues from the
sediments, including any dechlorination loss.
It should be noted that although care was taken to select the best estimate of the mean
cluster inventories, similar results are obtained if the arithmetic means or the MVUE values are
used exclusively. This results are presented in Figure A-4. The slopes obtained from these values
(0.39 for the arithmetic means and 0.37 for the MVUE values) agree quite well and deviate less
than one standard error from with the value of 0.41 obtained from the set of best estimates.
Figure A-5 represents a similar analysis utilizing only the complete 1994 cores for the
1994 cluster area MPA estimates. The values presented are based on the best estimates of the
mean, as was done in Figure A-3. This analysis addresses the concern that the incomplete cores
may substantively underestimate the 1994 sediment inventory and thereby overestimate the
A-7
TAMS TetraTech
-------
1984 to 1994 mass loss. The results of this analysis show that this concern is unwarranted since
the regression on the 1994 complete cores yields a Delta\i value quite close and well within error
of the value obtained for the entire 1994 data set.
Similar concerns were raised over the comparability of the 1994 cores to the set of 1984
cores and grabs. Two additional analyses were completed to address this concern. Figures A-6
and A-7 summarize these analyses. In Figure A-6. the 1994 cluster area MPA values are matched
to the 1984 cluster area MPA values based on 1984 cores only. This analysis yields a slope of
0.31 or a mass loss estimate of -69 percent - 16 percent. This agrees well with the mass loss
estimate obtained using all 197 of the 1984 sampling points. In Figure A-7. the 1984 mean
estimates were constructed based solely on the grab samples. This analysis also yielded a mass
loss estimate within error of the original estimate presented in Figure A-3. On the basis of these
sample subsets, it is clear that a statistically significant mass loss between 1984 and 1994 has
occurred for fine-grained sediments, regardless of the basis used to estimate the loss.
In preparing the regression-based mass loss estimates, it is noted that the regressions do
not consider the uncertainties in the individual cluster area means. An initial investigation of these
uncertainties suggests that they will not impact the conclusion of significant PCB mass loss from
the Upper Hudson sediments.
Ratio-Based Estimates of the Mean Mass Loss
An alternate basis of mass loss was developed from the 1984 and 1994 cluster area best-
estimate-of-the-mean values. In this instance the ratio of the 1994 MPA to 1984 MPA was
examined rather than the absolute values as was done in the regression analysis. This parameter is
related to the Deltaivi function used in the LRC as follows:
Thus the value for Deltas can be obtained after the mean ratio is obtained. The Deltas function
is a measure of the percent change in the sampling areas between 1984 and 1994. This parameter
was originally used directly to characterize the degree of change between 1984 and 1994. The
parameter was also shown to be skewed even under a log transform, making estimation of a mean
value for Deltas difficult. After further analysis, it was found that the ratio of the 1994 and
1984 inventories was a statistically "better behaved" function whose central tendency was easily
defined and whose distribution appeared normal under a log transform. This is clearly evident in
Figure A-8.
The distribution of 1994 MPA/1984 MPA is shown in Figure A-8 and can be better
described as log-normal (vs normal). Table A-8 provides the summary calculation for this
estimate of the mass change. On a ratio basis, the cluster area median mass loss. DeltaM^ was
estimated to be -57 percent with a range of -33 to -72 percent, larger but still within the
uncertainty of the original LRC median estimate of -40 percent. This is summarized on Figure A-
8. Note that the median Deltas estimate is based on the mean of the log-transformed data (i.e..
the geometric mean) which is presented in Figure A-8.
Delta\i
1994 MPA - 1984 MPA
1984 MPA
1994 MPA
1984 MPA
A-8
TAMSTetra'I ech
-------
Before the mean mass loss could be calculated, the coefficient of variation had to be
examined. The coefficient of variation for the cluster area mean ratios based on all 1994 points as
well as for the ratios using complete 1994 cores only are provided in the table. In both cases, the
coefficient is less than 1.2. indicating the arithmetic mean as the best estimate of the men ratio.
Nonetheless, the underlying log-normal distributions as shown by the W statistic also provided
on Table A-6, indicate that the uncertainty estimates should be derived based on this
consideration. The calculation of the uncertainty of the arithmetic mean given an underlying log-
normal distribution is described in Gilbert. 1987 and was utilized in providing the estimates the
table. These confidence limits for the ratio were converted to delta by subtracting one from the
values. Based on the ratio, the mean Deltas is -45 percent, with a 95 percent confidence range
between -59 percent and -4 percent. On this basis, the loss of trichloro- and higher homologues
between 1984 and 1994 is estimated to be -45 percent including any dechlorination loss. This
value is similar to the median mass loss of -40 percent originally estimated in the LRC . Both loss
estimates are found to be statistically significant since the 95 percent confidence limits do not
contain zero. Note that the arithmetic confidence limits are also provided in Table A-6. These
limits also exclude zero and are provided simply for comparison.
The DeltaM value was recalculated using only the 1994 cores which were complete or
nearly complete by the cesium-137 and total PCB profiles. Based on the ratio, the mean DeltaM
is -50 percent, with a 95 percent confidence range of -63 percent to -13 percent. This result is
quite similar to the Deltas values calculated using all 1994 cores. Because the incomplete cores
should under represent the true amount of PCBs in the location, the DeltaM excluding the
incomplete cores was expected to increase. This is not the case because incomplete cores
frequently yielded higher molar inventories relative to other 1994 cores from the cluster areas.
Based on this analysis, the mean mass loss including any in situ dechlorination losses is
estimated to be -45 percent with a range of -59 to -4 percent. This agrees well with the
regression-based estimate of -59 percent with a range of -78 to -40 percent.
Correction for Dechlorination
The degree to which the loss of tri- and higher homologues is due to dechlorination cannot
be directly assessed using the 1984 data, because the amount of mono- and dihomologues present
was not measured. A rough approximation of the degree of dechlorination between 1984 and
1994 can be calculated using the data found in McNulty, 1997. In this thesis, preserved cores
from 1983 and 1991 were analyzed for PCBs on a congener-specific basis and dated using
cesium-137. As shown in Figure A-9, the cores represent areas of near-continuous deposition in
the Thompson Island Pool (RM 188.5 and 188.6) and exhibit the profiles typical of high-
resolution cores. The homologue distributions and the DeltaM"s for trichloro- and higher
homologues are shown in Table A-7. The average percent change in the trichloro- and higher
homologue fraction is -4.7 percent. If the 1984 Tri+ MPA is multiplied by 0.953 (1-0.047). this
approximates the amount of trichloro- and higher homologues that would have been present in
1984. but were lost to dechlorination. The factor of 0.953 was applied to all of the 1984 samples
as an estimate of the dechlorination loss. The recalculated cluster mean values for the 1984 results
are listed in Table A-8.
Using the revised cluster mean estimates, the 1984-1994 mass loss was recalculated to
estimate loss exclusive of dechlorination (i.e., loss from the sediments). The ratio calculation
A-9
VMS TetraTech
-------
followed the same process as described for the total Tri- mass loss and is summarized in Table
A-9 and Figure A-10. Based on this calculation the median mass loss as estimated from the
geometric mean is -56 percent with an uncertainty range of -72 to -31 percent. This is within the
uncertainty of the median (as compared to the mean) mass loss estimate of -28 percent (range of
-48 to -5 percent) originally estimated in the LRC. The mean change in Tri— inventory expressed
as Deltas is -43 percent (i.e.. 43 percent molar loss) with an uncertainty range (or 95 percent
confidence interval) of -58 percent to -1 percent (see Figure A-10). The inclusion of a positive
upper limit in this instance is not taken as significant since the median mass loss is so large, the
mass loss was statistically different from zero for the 1994 complete cores (discussed below) and
the total Tri-;- mass loss was shown to be statistically different from zero. The mean mass loss
excluding dechlorination is statistically different from zero at the 90 percent confidence level.
Table A-9 also presents the results for the mean mass loss excluding dechlorination based
on the 1994 complete cores and the 1984 cluster means. These results are statistically different
from zero as well, and support the conclusion of mass loss for the sediments of the TI Pool.
Conclusions
Loss of trichloro- and higher homologues has been demonstrated with statistical certainty
in the TI Pool between 1984 and 1994. This degree of loss cannot be explained by dechlorination
alone and is estimated to be -43 percent, excluding dechlorination losses. This estimate is
interpreted as a loss of 43 percent of the sediment inventory to the overlying water column.
Following its release, some of this PCB mass would be transported downstream while some
would be redeposited in other areas of the TI Pool.
Direct evidence for dechlorination loss suggests a loss of about five percent over the
period 1984 to 1994. Combining loss from the sediments with dechlorination mass loss yields a
mean mass loss of -45 percent with an uncertainty of -4 to -59 percent. (Note that the five
percent dechlorination loss does not add directly to the 43 percent sediment loss since the 1984
inventory appears in both the numerator and denominator of the Deltas function.)
The median mass loss estimate excluding dechlorination was -56 percent or -57 percent
when dechlorination was included. These values are based on a statistically well-behaved ratio
function and local area-based averages.
Regression analysis of the sediment PCB inventory data (i.e.. Tri+ MPA) yielded slightly
higher estimates of mass loss with an average mass loss as Deltas of -59 percent and an
uncertainty range -40 to -78 percent. The mean mass loss estimates obtained by regression were
shown to be rigorous, regardless of the exclusion of various data types due to concerns over their
representativeness or comparability. The mean mass loss by regression agrees well with the mass
loss estimated by the ratio of the 1994 to 1984 Tri-" MPA.
These results compare favorably with the original 40 percent median mass loss estimated
in the LRC. This original value included a maximum dechlorination mass loss of 12 percent and a
median loss from the sediments of 28 percent. The original results are based on point-to-point
comparisons using a less-well-defined statistic (Delta^). In view of this, the revised estimates
presented in this Appendix are to be used in subsequent analysis of PCB contamination in the
Hudson River.
A-10
TAMS'I etra Tech
-------
APPENDIX A
TABLES
-------
Table A-l
Subreach Variogram Models* for Natural Log of PCB Mass Concentration,
1984 Thompson Island Pool Sediment Survey
Subreach 5
1163000-
1170100 N
Subreach 4
1170100 -
1177000 N
Subreach 3
1177000-
1181900N
Subreaches
1 and 2
1181900-
1191700 N
Observations
235
320
238
321
Nugget
0.750 (.284)
0.484 (.154)
0.0 (--)
1.54 (.108)
Sill-Nugget
1.520 (.282)
1.092 (.153)
1.733 (.060)
0.203 (.106)
Practical Range (ft)b
340 (75)
280 (68)
286 (49)
582 (521)
Anisotropv Ratioc
1.0
1.5
2.5
1.0
Major Axisd
_
N 10° W
N 35° W
.
Note:
a. Variograms are exponential models, showing fit along the major axis and anisotropv ratio. Standard
errors of the coefficients from the least squares estimation are shown in parentheses.
b. A value of 2 times the practical range was used as the length of the major axis of the polygon
associated w ith each 1994 location. This distance represents the distance of separation at which
variance between point pairs approaches that of the population as a whole.
c. This ratio represents the ratio of the major axis over the minor axis of the ellipse associated with
each sampling point.
d. This represents the orientation of the major axis. Essentially this orientation causes the ellipse to be
oriented in the direction of river flow. This angle is not defined when the anisotropv ratio is unity
(1).
Source: USEPA, 1997
TAMSTetraTech
-------
Table A-2
Comparison of Tri+ MPA Arithmetic Means for All NYSDEC 1984 Sample Points in the
Sample Areas and Co-Located 1984 to 1994 Sample Points
1984 Results
1994 Results
1984 Sample Locations with
All 1984 Sample Locations
matching 1994 Locations.
All 1994 Sample Locations
in Cluster Areas
grouped by Cluster area
in the Cluster Areas
No. of
No. of
Mean Tn+ MPA
Location
Mean Tri+ MPA
No Of
Mean Tri+ MPA
Cluster Area1
Locations
fmoles/m i
s
i moles/m")
Locations
(moles/m")
LR-01
14
0053
4
0.086
4
0 013
LR-02&03
12
0.166
6
0.281
6
0 011
LR-04&18
24
0.093
5
0.336
9
0074
LR-05
14
0.070
4
0.143
5
0 067
LR-06&07
24
0058
7
0 099
"7
0.029
LR 08
9
0.021
5
0025
5
0011
LR 09
14
0.057
6
0.106
6
0 013
LR-10
8
0.139
4
0 244
4
0036
LR-11
9
0 117
3
0 259
3
0.103
LR-I2
8
0018
5
0 019
5
0 018
LR 13:
6
0033
3
0015
LR 14
23
0.028
4
0026
4
0.007
LR-15
30
0.013
4
0 006
4
0016
LR-17
2
0.091
0 091
?
0.021
Total Locations
197
59
70
Arithmetic Mean
0.061
0 134
Standard Error
0 007
0 020
All Fine Grained Area
of the TI Pool5
0050
Hudson River Database Release 4 1 TAMS/TetraTech
Notes'
1 The LR-16 duster is not included in this analysis, because there is only one reoccupied
1984 sample location in the cluster This is the sixtieth matched 1984 to 1994 sample location.
2. Three sample points in Sampling Area LR-1 3 were reoccupied in 1994. but were excluded
from the pairwise analysis because the samples were only screened by mass .spectrometry
3 Tri-t- MPA in the fine-grained areas of the TI Pool is calculated from the Total PCB
Inventory Estimate described in Appendix B
-------
Table A-3
Number of Locations in Sample Areas for 1984 and 1994
Sampling
Visual Texture Principle
Laser Principle Fraction
Principal
Area
No. of Locations'
Sediment Classification by Side Scan Sonar
Fraction Classification
Analysis
Side Scan
Clusters
in Sampling Area
(No of locations)
(No. of locations)
(No. of locations)
Sonar
1984
1994
1984
1994
Clay, Silt or
Coarse Sand
Silt or Fine Medium Sand
Texture of
1984
1994
Pine
Coarse Rocky
Fine
Coarse
Fine Sand
or Gravel
Sand
or Gravel
Region
LR-OI
14
4
1 1
3
4
11
3
4
Fine
LR-02&03
12
6
9
2 1
5
1
10
2
5
1
Fine
LR-04&I8
24
9
18
6
9
22
2
9
Fine
LR-05
14
5
14
5
12
2
5
Fine
I.R-06&07
24
7
7
17
4
3
14
10
7
Fine2
LR-08
9
5
7
2
4
1
8
1
5
Fine
I.R-09
14
6
7
7
4
2
10
4
6
Fine
LR-IO
8
4
4
4
1
3
8
4
Fine
I.R-1 1
9
3
2
7
3
6
3
2
1
Fine
LR-12
8
5
2
6
5
5
3
5
Fine
I.R-13
6
3
6
3
1
5
1
2
Coarse
LR-14
23
4
14
9
3
1
14
9
4
Fine
LR-15
30
4
8
22
1
3
6
24
1
3
Coarse
I.R-17
2
5
2
5
2
5
Fine
Total
197
70
105
91 1
45
25
129
68
63
7
Hudson River Database Release 4.1 I AMS/TetraTech
Notes:
1 1984 field co-locates count as one point and were handled as described in IJSP.I'A, 1997
a Core-core or grab-grab pairs were averaged
b l or core-grab , the core value was used
c If one sample in a pair was screened with mass spectroscopy and the other sample analysed with (jC-HX'D, the GC'-FCD value was used
2 This area was considered fine-grained based on the 19X4 and 1994 sampling data only
-------
Table A-4
Shapiro Wilks Tosi and Ratio of Arithmetic Mean to Standard Deviation for Sample Points in Cluster Areas for 1984 and J994
1994
Complete
19X4
1994
1994 (
oinplele Cores Only
1984
1994
Cores
I 'nltansfoi ined
1 .og 10 Transt
>i j n
Untransforined
1 ,og 10 Transform
Untransfonned
Log 10 Transform
Coefficient of
Cluster Are
N
W
Piob
-------
Table A-5
Selection of Cluster Area Best-Estimate-of-Mean for 1984 and 1994
Each Sample Area
Number of
Arithmetic
Coefficient of
Sample
Best Estimate
1984
Mean
MVUE
Variation
Locations
of the Mean
l.R-01
0.053
0.080
1.1)
14
0.053
LR-02&03
0.166
0.155
1.40
12
0.155
LR-04& 18
0.093
0.098
1.75
24
0.098
LR-05
0.070
0.107
0.98
14
0.070
LR-06&07
0.058
0.058
1.21
24
0.058
LR-08
0.021
0.021
0.47
9
0.021
LR-09
0.057
0.057
1.29
14
0.057
LR-10
0.139
0.177
0.91
8
0.139
LR-II
0.117
0.119
1.03
9
0.117
LR-12
0.018
0.018
0.45
8
0.018
LR-13
0.033
0.037
0.51
6
0.033
LR-I4
0.028
0.030
1.27
23
0.030
LR-15
0.013
0.012
1.08
30
0.013
LR-17
0.091
0.091
-
0.091
Number of
Arithmetic
Coefficient of
Sample
Best Estimate
1994
Mean
MVUE
Variation
Locations
of the Mean
LR-01
0.013
0.013
0.335
4
0.013
LR-02&03
0.011
0.014
1.313
6
0.014
LR-04&18
0.074
0.075
0.71 1
9
0.074
LR-05
0.067
0.067
0.767
5
0.067
[.R-06&07
0.029
0.031
1.353
n
0.031
LR-08
0.011
0.011
0.628
5
0.011
LR-09
0.013
0.021
0.631
6
0.013
LR-10
0.036
0.036
0.212
4
0.036
I.R-11
0.103
0.099
1.003
3
0.103
LR-12
0.018
0.017
0.913
5
0.018
LR-13
0.015
0.015
0.643
3
0.015
LR-14
0.007
0.007
0.315
4
0.007
LR-15
0.016
0.018
0.621
4
0.016
LR-17
0.021
0.027
0.686
5
0.021
Number of
1994 Complete
Arithmetic
Coefficient of
Sample
Best Estimate
Cores
Mean
MVUE
Variation
Locations
of the Mean
LR-01
0.013
0.013
0.337
4
0.013
LR-02&03
0.01 1
0.012
1.464
5
0.012
LR-04&18
0.074
0.075
0.71 1
9
0.074
LR-05
0.067
0.067
0.767
s
0.067
LR-06&07
0.031
0.031
1.503
>
0.031
LR-08
0.011
0.011
0.710
3
0.011
LR-09
0.011
0.017
0.715
5
0.011
LR-10
0.036
0.036
0.212
4
0.036
l.R-11
0.043
0.043
0.158
0.043
LR-12
0.019
0.018
0.969
4
0.019
LR-13
0.015
0.015
0.645
3
0.015
LR-14
0.00''
0.007
0.314
4
0.007
LR-15
0.013
0.014
0.749
3
0.013
LR-17
0.021
0.027
0.685
5
0.021
Hudson River Database Release 4.1 TAMS 'I etraTech
Notes:
1. The best predictor of the mean is the arithmetic mean if the coefficient of variation
is less than 1.2. otheru ise it is the MVl.'K.
-------
Table A-6
Estimate of the Average Molar Change in the Sediment PCB Inventory (MPA)
Trichloro- and Higher Homologues
Ratio of 1994 to 1984 Sediment PCB
Inventory (Tri— MPA)
( Best Estimate of the Mean )
unitless
Complete 1994 Cores
All 1994 and 1984
Only and All 1984
Cluster Area
Samples
Samples'
LR-OI
0.247
0.247
LR-02&03
0.091
0.075
LR-04&18
0.754
0.754
LR-05
0.956
0.956
1.R-06&07
0.527
0.538
LR-08
0.522
0.524
LR-09
0.226
0.202
LR-10
0.261
0.261
LR-11
0.879
0.370
LR-12
1.011
1.084
LR-13
0.450
0.449
LR-14
0.231
0.231
I.R-15
1.267
1.029
l.R-17
0.228
0.228
Coefficient of V ariation2
0.6?
0.67
Estimate of the Mean Ratio
0.55
0.50
(Arithmetic)
DeltaM
-45%
-50%
Shapiro-Wilks Test
Untransformed
W
0.91
0.89
Prob<\V
0.14
0.08
Loe-Transformed Data
W
0.94
0.94
Prob
-------
Table A-7
Shift in llomologue Group Distributions (Mole Percent) lor Matched Cores in the Thompson Island Pool
I960 Sediment
1963 Sediment
1968 Sediment
1973 Sediment
1975 Sediment
1 lomolog Group
1983
1991
1983
1991
1983
1991
1983
1991
1983
1991
Mono
1 1.29
15.52
10 13
23.80
13.24
30.26
13.14
23.72
15.54
24.71
l)i
37.78
36.34
32 64
29.1 I
37.24
29 51
39 65
32 01
38.31
29.30
Tri
34 82
33.71
35.93
29 67
31 14
24.41
3 1.43
28 98
30.34
29.56
letra
1 1.52
10.07
14.31
10.80
12.27
9 84
1 1 56
9 55
10.36
10.56
I'enta
2-49
2.30
4.28
3.60
3 68
3.22
2 00
3.22
3 12
3.29
1 lexa-Deca
1 91
1 95
2.28
2.39
1.70
1.91
1.89
1.53
1.34
1.63
Tri'
50.74
48.03
56.80
46.46
48 79
39 38
46 88
43.28
45 16
45.04
Delta Tri *
-5.3%
-18.2%
-19.3%
-7.7%
-0 3%
1 lomolog Group
1976 Sediment
1983 1991
1979 Sediment
1983 1991
1980 Sediment
1983 1991
1982 Sediment
1983 1991
Mono
l)i
In
letia
I'enta
1 lexa-Deca
14.94
37 81
30 66
10.9 I
3.30
1.39
17.77
3 1 94
32.53
12 43
3 61
1 59
14.17
35.95
32.15
1 1.58
3.84
141
20.03
31.70
3 1 04
1 1 82
3.58
1.68
17.26
36 64
32.01
9.37
2.95
114
25.86
31.15
3 1.09
9.33
1 69
0 83
14 00
37.20
34 84
10.00
2.86
1 02
17 68
29.98
36.34
10.91
2 88
2.07
Tri ~
Delta Tri -
46 26
50 16
8 4%
48 98
48.12
-1.8%
45.47
42 94
-5.6%
48.72
52 20
7.1%
Average Delta Tri'
-4.7%
Source: McNulty, 1097 (Table 8) TAMS/TetraTech
-------
Table A-8
Selection of Best-Estimate-of-Mean for 1984 Results After
Correcting For Dechlorination Loss
Cluster Area
Arithmetic
Mean
MVUE
Coefficient
of Variation
Number of
Sample
Locations
Best Predictor
of the Mean
LR-01
0050
0.080
1.11
14
0.050
LR-02&03
0 158
0.155
1.40
12
0.155
LR-04&18
0.089
0 098
1.75
24
0.098
LR-05
0.067
0.107
0.98
14
0.067
LR-06&07
0.056
0.058
1.21
24
0.058
LR-08
0.020
0.021
0.47
9
0.020
LR-09
0054
0.057
1.29
14
0.057
LR-10
0.133
0.177
091
8
0.133
LR-1I
0 112
0.018
1.03
9
0.112
LR-12
0017
0.119
0.45
8
0.017
LR-13
0.031
0.037
0.51
6
0.031
LR-14
0.027
0.030
1.27
23
0.030
LR-15
0.012
0 012
1.08
30
0.012
LR-17
0.087
0.091
--
2
0.087 :
Hudson River Database Release 4 1 TAMS/TetraTech
Notes:
1. The best predictor of the mean is the arithmetic mean if the coefficient of variation
is less than 12. otherwise it is the MVL'E.
2 The arithmetic mean was selected for LR-17 due to the
small number of samples available.
-------
Table A-9
Estimate of the Average Molar Change in the Sediment PCB Inventory Excluding Dechlonnation
( Trichloro- and Higher Homologues)
Ratio of 1994 to 1984 Sediment PCB
Inventory (Tn+ \1PA)
( Best Predictor of the Mean )
unitless
Complete 1994
All 1994 and 1984
Cores Only and All
Cluster Area
Samples
1984 Samples'
LR-01
0.261
0.261
LR-02&03
0.090
0.077
LR-04&18
0 752
0 752
LR-05
1.002
1.002
LR-06&07
0.534
0.533
LR-08
0.550
0.538
LR-09
0.225
0.201
LR-10
0.274
0.273
LRU
0.921
0.388
LR-12
1.050
1.125
LR-13
0.478
0.477
LR-14
0.234
0.234
LR-15
1.321
1.072
LR-17
0.241
0.241
CoefTicienl of Variation2
0.67
0 67
Estimate of the Mean Ratio
0.57
0.51
(Arithmetic)
DeltaM
-43%
-49%
Shapiro-Wilks Test
Untransformed
W
0.91
0.89
Prob-Transformed Data
W
0.94
0 95
Prob
-------
APPENDIX A
FIGURES
-------
LR-14
LR-04&18
SUBREACH 4
SUBREACH2
USEPA 1994 Low Resolution Cores
x NYSDEC 1984 Grabs
o NYSDEC 1984 Cores
Cluster Area
Small Polygons with Dimensions from Variogram Models
—- — Subreach Boundaries
Shoreline
LR- 3
O X
SUBREACH 3
LR-15
LR-17
SUBREACH 5
-02&03
LR-11
LR-10
LR-01
LR-09
400 0 400 800 Feet
LR-08
SUBREACH4
LR-06&07
LR-05
Thompson
Island
Dam
Hudson River Database Release 4.1
Figure A-l
Construction of the Cluster Areas Using the Subreach Semivariogram Models
TAMS
-------
m lAas
UWW [ W
-------
o:
[jnc of No Change
/ i
? |
2 S
+ —¦
'Z v
H 3
t r
9< —
i 2
2
s
0 15
0 I
1994= 0.41* 1984
Regression
Uncertaintv
0.05
-j
0 05
0 1
0 15
o:
1984 Tri+ MPA
Best Estimate imoles/m*l
(Summary of Fit
¦>
RSquare
•
RSquare Adj
•
Root Mean Square Error
0.02671 1
Mean of Response
0.03129
Observations (or Sum Wgts) 14
t
Parameter
-
Estimates
X
Term
Estimate
Std Error
t Ratio
Prob>ltl
Lower 95%
Upper 95%
Intercept
Zeroed 0
0
•
•
0
0
Best Predictor
0.4095612
0.087919
4.66
0.0004
0.2196236
0.5994989
Effect Test]
Source Nparm DF Sum of Squares F Ratio Prob>F
Best Predictor 1 1 0 01548296 21.7005 0 0004
Notes:
Numbers on diagram represent cluster numbers
Dash ed lines represents regression through 0 and us uncertaintv
Regression has a slope statistically different t'rom unity (11.
Hudson River Database Release 4 I TAMS/TetraTech
Figure A-3
1984 vs. 1994 Sediment Tri+ MPA
Best-Estimate-of-Mean Basis
-------
0 15
«y l
<~ 5
9- e
0 05 -
Ijne of No Change
/
< >
a- ± /
ry 1994 = 0J9* 1984
_Rcgressioti
L'ncen.iinjv
o -- - - -
0 0 05 0 1 0 15
1984 PCS Tri+MPA
Arithmetic Mean (moleV'rn'i
0 15
< i
5 1
Lmc of No Change
/
1994 = 0.3? • 1984
0 05
Reiresiion
I'nccruinu
0 05
o
1984 Trt+ MPA
.MWE fmoles/m"^
Nines:
Numbers on diagram represent cluster numbers.
Dashed lines represents regression through 0 and us uncertainty
Regressions ha\e slopes statistically different from unity 111.
Hudson R;%cr Database Release 4 I TAMS/TetraTech
Figure A-4
1984 vs. 1994 Sediment Tri+ MPA
Based on Arithmetic Mean and MVUE
-------
U
0.15
Line of No Change
< E
a. ^
5 5
+
0.05
1994 = 0.33 » 1984
Regression
L'ncertaintv
i?
* . -'3
0
D05 0.) 0.15
1984 Tri+ \1PA
Best Estimate (moles/m")
0.2
(Summary of Fit )
RSquare
0
RSquare Adj
?
Root Mean Square Error
0.021344
Mean of Response
0.026672
Observations (or Sum Wgts)
14
(Parameter Estimates )
Term Estimate Std Error t Ratio Prob>ltl Lower 95°° Upper 95co
Intercept Zeroed 0 0 9 9 0 0
Best Predictor 0.3289175 0.070252 4.68 0.0004 0.177147 0.480688
¦ —
/ N
(Effect Test )
Source Nparm DF Sum of Squares F Ratio Prob>F
Best Predictor 1 1 0.00998597 21.9207 0.0004
Notes:
Numbers on diagram represent cluster numbers
Dashed lines represents regression through 0 and its uncertainty.
Regression has a slope statistically different from unity (l).
Hudson River Database Release 4 1 TAMS-TeiraTecl-
Figure A-5
1984 vs. 1994 Sediment Tri+ MPA
Best-Estimate-of-Mean Basis ( 1994 Complete Cores Only)
-------
u:
o 15
ljne of No Change
0 1
0 05
r
I
Regression
L'ncertaintv
1994 = 0.31 • 1984
)05 0 1 0 15
1984 Best Estimate Tri +
MPA Cores Onlv imoles/m')
_ — .. «-'
(Summary of Fit")
RSquare
9
RSquare Adj
9
Root Mean Square Error
0.029858
Mean of Response
0.032102
Observations (or Sum Wgts)
13
(Parameter Estimates )
Term
Estimate
Std Error
t Ratio
Prob>ltl
Lower 95% Upper 95%
Intercept Zeroed
0
0
n
7
0 0
1984 Cores only
0.3134883
0.080171
3 91
0.0021
0.1388105 0.488166
(Effect Test )
Source Nparm DF Sum of Squares F Ratio Prob>F
1984 Cores only 1 1 0 01363086 15 2900 0.0021
N OIL'S.
Numbers on diagram represent cluster numbers.
Dashed lines represents regression through 0 and its uncertainty
Regression has a slope statistically different from unity (1).
Hudson River Database Release 4 1 TAMS/Tetra Tech
Figure A-6
Best Estimate of Tri+ MPA for Cluster Areas:
1984 Cores Onlv vs. 1994 Cores
^
-------
o:
0 15
Lne of No Chance
C <
x a.
T 2
0 1
0 05
Regression
Ur.cert runts
1994 = 0.57 • 1984
&
0 05 0.1 0 15
1984 Best Estimate Tn*
MPA Cirabs Onlv tmcles/m* i
(Summary of Fit )
RSquare
?
RSquare Adj
?
Root Mean Square Error
0.029225
Mean of Response
0.03129
Observations (or Sum Wgts)
14
(Parameter Estimates )
Term
Estimate
Std Error
t Ratio
Pro b>itl
Lower 95% Upper 95%
Intercept Zeroed
0
0
9
?
0 0
1984 Grabs only
0.5747982
0.143759
4.00
0.0015
0.2642253 0.885371
(Effect Test )
Source Nparm DF Sum of Squares F Ratio Prob>F
1984 Grabs only 1 1 0.01365464 15.9867 0.0015
Notes:
Numbers on diagram represent cluster numbers
Dashed lines represents regression through 0 jnd its uncertainty
Regression has a slope statistically different from unity 111
Hudson River Database Release 4 I TAMS/TetraTech
Figure A-7
Best Estimate of Tri+ MPA for Cluster Areas:
1984 Grabs Onlv vs. 1994 Cores
-------
1994 Hesl Lstmiate Mean/1984 Best Kstimale Mean m Cluster Areas
• • ::: •
r 20 o
1 5 0
- to o
"5 0
00
25
50
75 1 00 1 25 1 50
Quantiles
maximum
quartile
median
quartile
minimum
100 0%
99 5%
97 5%
90 0%
7 5 0%
50 0%
25 0%
10 0%
2 b%
0 5%
0 0%
1 2675
1 2675
1 2675
1.1394
0 8983
0 4857
0 2306
0 1586
0 0910
0 0910
0 0910
(Moments )
Mean
Std Dev
Std Error Mean
Upper 95% Mean
Lower 95% Mean
N
Sum Weights
Natural Log of 1994 Best Hstimate Mean/19X4 Hesl Lslimale Mean in ( luster Areas
[quantiles j
r 15
t o
2 5 -2 0
15 -10 -0 5
quartile
median
quartile
minimum
100.0%
99.5%
97.5%
90 0%
7 5 0%
50 0%
25.0%
10 0%
2 5%
0.5%
0 0%
0 2370
0 2370
0 2370
0 1241
0 1079
7249
4673
941 5
3972
3972
3972
(Moments
Mean
Std Dev
Std Error Mean
Upper 95% Mean
Lower 95% Mean
N
Sum Weights
0 54653
0 36685
0 09804
0 75834
0 33472
14 00000
14 00000
(jest for Normality^)
Shapiro-Wilk W Test
W Prob< W
0 906724 0 1400
Units:
0 84742
0 76525
0.20452
-0 40557
-1 28926
14 00000
14 00000
Test for Normality^)
Shapiro-Wilk W Test
W ProtxW
0 935325 0 3486
1994 moles/m
1984 moles/m~
DeltaM
Best Estimate of Mean
-45%
Upper Confidence Limit'
-4%
Lower Confidence Limit'
-59%
Geometric Mean
-57%
Upper Confidence Limit2
-33%
Lower Confidence Limit2
-72%
Nutes:
1 Confidence limit derived from hesl estimate of confidence limits given underlying log-normal nature ol the data
Note (lie higher probability of the W statistic lor the log-transformed data, supporting this approach Sec lexl lor discussion
2 Confidence limits based on log transformed data .is given above under Natural Log statistics.
Hudson River Database Release 4 I Figure A-8 I AMSAI etra letli
Distribution of the Tri+ Ml*A Best-Kstimate-of-Mean Ratios (1994/1984) for Cluster Areas
-------
1990
Cesium-137 Profiles
Upper Hudson - Thompson Island Pool
Activity (pCi/Kg)
0 5000 10001) 15000 2000(1
, * \ I I "I
1980
1970
Year
1960
V
i
Core Collected
m 1991
b
"I
H
KX
KM
101
7
Core Collected
in 198.1
1950
1940
Total PCB Profiles
Upper Muson - Thompson Island Pool
1990
1980
1970
Year
1960
1950
19-10
Concentration (ppm)
500 1000
Core Collected
in 1991
1500
• I
b ™
P
IO)
)0\ H*
a'm"
¦J
Core Collected
in I9K.<
I mm McNulty, 1997 I[?ipure 2) I AMS/'Ielu IclIi
Figure A-9
Cesium-137 and Total PCB Profiles for Cores Collected in 1983 and 1991 from the Thompson Island Pool
-------
1994 Best F-'slnnale Mean/1984 Rest Hstimate Mean in C'lusler Areas
• -
1 5
1 0
5
25 50 75 1 00 1 25 1 50
(Quarttiles ]
(Moments "}
maximum
100 0%
1 3210
Mean
0 56643
99 5%
1 3210
Std Dev
0 38260
97 5%
1 3210
Std Errot Mean
0 10225
90 0%
1 1855
Upper 95% Mean
0 Z8733
qua rtile
7 5 0%
0 9413
Lower 95% Mean
0 34552
median
50 0%
0 5060
N
14 00000
quartile
25 0%
0 2370
Sum Weights
14 00000
10.0%
0 1575
i-—
2 5%
0 0900
0 5%
0 0900
minimum
0 0%
0 0900
Nalural Log of 1994 Best Hslnnate Mean/19X4 Best Hstimale Mean in Cluster Ar
I:—
(jest for Normality ]
Shapiro-Wilk W Test
W Prob
-------
Appendix B
-------
APPENDIX B
Revised Estimate of the 1984 Thompson Island Pool
Sediment PCB Inventory
Estimation of the 1984 Thompson Island Pool Sediment PCB Inventory Using Thiessen
Polygons and the Side Scan Sonar Results
TAMS/TetraTech
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
APPENDIX B
Revised Estimate of the 1984 Thompson Island Pool
Sediment PCB Inventory
Estimation of the 1984 Thompson Island Pool Sediment PCB Inventory Using Thiessen Polygons
and the Side Scan Sonar Results
TABLE OF CONTENTS
Page
Discussion B-l
List of Tables
B-l NYSDEC Sediment Survey Visual Texture Classifications and Assigned Sediment
Type
B-2 Previous and Revised Thompson Island Pool Sediment Total PCB Inventory
Estimates
List of Figures
B-l Cohesive Sediment Mass per Unit Area
B-2 Noncohesive Sediment Mass per Unit Area
l
TAMS'TetraTech
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
Appendix B
Revised Estimate of the 1984 Thompson Island Pool Sediment PCB Inventory
Estimation of the 1984 Thompson Island Pool Sediment PCB Inventory Using
Thiessen Polygons and the Side Scan Sonar Results
An estimate of the 1984 sediment total PCB inventory in the Thompson Island Pool using
geostatistical analysis is presented in Chapter 4 of the Data Evaluation and Interpretation Report
(DEIR). USEPA, 1997. This estimate used data from the 1984 NYSDEC sediment samples but did
not consider sediment texture. Sediment texture is relevant because PCB concentrations are strongly
correlated with shallow sediment texture, in that higher concentrations of PCBs are found in areas
of finer-grained, shallow sediments. A similar degree of correlation was noted between Total PCB
concentration and the side-scan sonar signal itself. (As discussed below, the side-scan sonar results
form the basis for the assignment of sediment texture.) LRC figures 3-19 and 3-30 demonstrate the
strength of the relationships between Total PCBs. sediment texture and side-scan sonar signal. The
mean PCB concentration varies nearly an order of magnitude in correlation with these properties.
A revised estimate of the Thompson Island Pool PCB inventory in 1984 is presented in this
Appendix which takes into account the relationship between PCB mass and sediment texture. The
purpose of this analysis is to provide an alternate estimate of the sediment PCB inventory while also
providing separate estimates for areas of fine-grained and coarse-grained sediments. The latter
estimates could not be obtained from the previous analysis and were needed for modeling purposes.
Sediment texture information is available in two forms: visual texture classification for the
sample points collected in 1984 and side-scan sonar sediment classification for the entire river
bottom of the TI Pool, obtained in 1992 (see Section 4.1.1 of the DEER for a complete discussion
of the side scan sonar analysis). In this revision, the NYSDEC core and grab samples are separated
into cohesive and noncohesive groups based on the 1984 visual texture classification. Noncohesive
sediments typically are coarse-grained, such as medium to coarse sand or gravel. Fine-grained
sediments, such as fine sands, silts and clays, are generally considered cohesive sediments.
The primary distinction between cohesive and noncohesive sediments is that cohesive
sediments exhibit interparticle attractions whereas noncohesive sediments do not. Cohesive
sediments exhibit very different flow-driven resuspension behavior than noncohensive sediments
as a result of the inter-particle bonds. The mechanistically different resuspension processes for these
two sediment types will be accounted for with different mechanisms in the Hudson River PCB
transport models. The models require determination of the cohesive and noncohesive sediment areas
and associated PCB mass.
In general, samples classified as predominantly clay, silt or fine sand were classified as
cohesive sediment. The remaining samples which are predominantly sand, coarse sand or gravel are
assigned to the noncohesive group. One sample had an ambiguous classification (FC) and was
grouped with the noncohesive sediments. There are 503 cohesive sample locations (221 grabs, 282
cores) and 591 noncohesive sample locations (470 grabs. 121 cores). A list of the visual texture
B-!
T A\1 S'TeiraTech
-------
classes and the assigned sediment types is provided in Table B-l.
A brief description of the geospatial technique used in this analysis is transcribed from page
4-33 of the DEER (USEPA. 1997):
A simple method for addressing the problem of irregular sample spacing (or
coverage) and clustering of data is a graphical technique known as polygonal
declustering (Isaaks and Srivastava, 1989). As with other approaches to estimating
total mass from spatial data, this relies on a weighted linear combination of the
sample values. Weighting is formed graphically, however, without any assumptions
regarding the statistical distribution of the data, and spatial correlation is not
explicitly modeled. In this method, the total area of interest is simply tiled into
polygons, one for each sample, with the area of the polygon representing the relative
weighting of that sample. The polygons, called Thiessen polygons or polygons of
influence, are drawn such that a polygon contains all the area that is closer to a given
sample point than to any other sample point. Polygonal declustering often
successfully corrects for irregular sample coverage. Because no complicated
numerical methods need be applied, polygonal declustering provides a useful rough
estimate of total mass to which the estimates obtained by other methods can be
compared.
In the analysis presented here, Thiessen polygons are formed around all 1984 cohesive
sample points. This procedure was repeated for the noncohesive sample points. Using the side scan
sonar sediment classifications, the Thiessen polygons are clipped so that the mass per unit area for
the cohesive sample points (based on visual texture classification) is applied only to cohesive areas
of the river (defined by side-scan sonar) and, similarly, the mass per unit area for the noncohesive
sample points is applied only to the noncohesive areas. For the side scan sonar sediment
classification, cohesive areas are defined as fine- or finer-grained and noncohesive areas are coarse-
or coarser-grained based on the original interpretation of the side-scan sonar images (Flood, 1993).
The means of calculating the mass per unit area is the same as described in the DEIR (USEPA,
1997).
Figure B-l shows cohesive sediment sample points and the associated Thiessen polygons.
The areas which are cohesive by the side scan sonar analysis are shaded to indicate the PCB mass
per unit area derived from the corresponding cohesive sediment samples. The noncohesive data are
shown in Figure B-2. This diagram is constructed in a fashion similar to Figure B-l, only based on
non-cohesive sediment areas and noncohesive sediment samples.
The revised sediment Total PCB mass estimate for the entire Pool (14.9 metric tons) is in
close agreement with the previous estimates presented in the DEIR (14.5 metric tons, see Table B-2).
The estimated trichloro- and higher homologue inventory present in 1984 can be calculated by
multiplying the mass of Total PCBs by 0.944, as discussed in Chapter 4 and Appendix E of the LRC.
As discussed in the text, it is likely that the 1984 measurements most accurately represent the sum
of the trichlorinated to decachlorinated homologues (Tri+). This correction yields the values given
in the last column of Table B-2. The estimate for the Tri+ inventory of the entire Thompson Island
Pool is 14.1 metric tons. Based on the discussion in Appendix E of the LRC, it is clear that while the
inventory of trichlorinated and higher homologues is relatively well known for 1984, the total PCB
inventory is less well known and, in fact, may be underestimated by a large percentage.
B-2
TAMS/TetraTech
-------
The issue of the estimation of sediment mass has been extensively addressed in Chapter 4
of the DEER. As noted in this discussion, the spatial correlation of the individual sediment mass
estimates obtained via sediment cores and grabs varies from subreach to subreach of the TI Pool.
This is evident in the semivariogram analysis presented in the DEER. Thus in subreaches 3, 4 and
5. where spatial correlation is high, the estimates for the sediment PCB inventory are relatively well
known and local inventories can be considered well-described . In subreaches 1 and 2. where spatial
correlation is poorer, the ability to infer local estimates for sediment inventory will be more limited
and will have a greater dependence on the local sampling density rather than inference from other
locations. As an overall estimate of the TI Pool, or as a basis for estimating the PCB inventories of
large segments of the Pool such as the regions of fine-grained sediments, these uncertainties
represent only minor concerns. While the analysis presented here does not permit the calculation of
a statistically-based uncertainty, the fact that the Thiessen polygons, when corrected for sediment
type, yield a sediment PCB mass estimate for the TI Pool (14.9 metric tons) which is within 3
percent of the mass estimate based on kriging (14.5 metric tons). This suggests that the uncertainty
in these estimates is small and will have minimal impact on the Reassessment findings.
lAMS-TctuTcch
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
APPENDIX B
TABLES
-------
Table B-1
NYSDEC Sediment Survey Visual Texture Classifications and Assigned Sediment Type
No. of
Texture
Sediment Type
Samples
Fine sand
Cohesive
342
Fine sand and wood chips
Cohesive
95
Clay
Cohesive
20
Muck
Cohesive
19
Fine sand and gravel
Cohesive
10
Silt
Cohesive
6
Clay and gravel
Cohesive
4
Gravel and clay
Cohesive
3
Fine sand and clav
Cohesive
2
Gravel and muck
Cohesive
1
Silt and wood chips
Cohesive
1
503
Gravel
Noncohesive
461
Coarse sand
Noncohesive
65
Coarse sand and wood chips
Noncohesive
29
Gravel and wood chips
Noncohesive
29
Coarse sand and gravel
Noncohesive
3
FC and wood chips'
Noncohesive
1
Sand
Noncohesive
1
Sand
Noncohesive
1
Sand and wood chips
Noncohesive
1
591
Note:
1. NYSDEC's sediment texture classification is FC for this sample, but the
definition of FC is unknown.
-------
Table B-2
Previous and Revised Thompson Island Pool Sediment Total PCB Inventory Estimates
Sediment
Type
Previous Total
PCB Mass Estimate
(metric tons)1
Revised Total
PCB Mass
Estimate
(metric tons)
Tri and Higher
PCB Mass
Estimate (metric
tons)
Cohesive
8.7
8.2
Noncohesive
6.2
5.9
Total
14.5
14.9
14.1
From USEPA, 1997 - Based on the kriging analysis of the Thompson Island Pool.
Based on correction factor developed in Appendix E of the LRC (USEPA, 1998).
These values are believed to represent the most accurate inventory of the Thompson Island Pool.
This estimate represents the sum of trichloro to decachloro homologues only.
Notes:
1.
2.
-------
APPENDIX B
FIGURES
-------
Figure B-l
Cohesive Sediment Mass per Unit Area
Uroadwav
1 !>ttriseeiuwt! BbJHMt
Legend:
Blivck House Roiiu
4 tuck House Rqm\
Grin in
400 0 400 S00 1200 1600 Feet
NYSDEC 1984 Fine-G rained Samples
> Core
Grab
j ] Thiessen Polygon
Total PCB Mass/Area in Fine-Grained Area (g/sq. m)
(from Side Scan Sonar Sediment Classification)
0-3
3 - 7.5
lill 7.5 -10
ffj 10-30
» 30-100
100- 123 8
Shoreline
192 River Mile
Nmcs:
1 Shcrdinc tpJHrMltHEtf awl was pflncjpulty derived fry JAMS
Ihjm the I iinlsort fclm Survey, Wfa.lQ?? fit NVSDEtl
iMimnundcau A«iei«rt^ Ilie «heom lt>e Hiidsmi Rivti Survey^ J97t»-1977. UtiXYSDEC
fNormiiDclcau Awxjiutoi^hic 1977V. Portvon&of lite Sllurcliric
around RM 1 '/4 tuid HM < 97 were obhunod (timi N YSDEC baSSl
nir su^itnem Yitfqya juxftm&eU w >986* 1993 (imgufrllelftajjl
2. .Sediment 'jlttshilicirtton1. bvtwdan Flood I'393 Analysis
• M iIk* Svd^Sctui Somu- IUiUrMiictfics Smblwsdoin, uiwl Seaimcnt
I Jtuti mini Iht ijppff J Judwiti River Sdwti»*ii tlokcis l otfv and
I wtik 5. Kei-v^d' Fftu,miwrTon of Sunt Uiuventiy of f^ksw York
:;7^V\ / Black House Road
Black House Road
A
------- OCR error (D:\Scanspot\Jobroot\ps1\MJA0005C\tiff\P100X8GI.tif): Unspecified error
-------
Appendix C
-------
APPENDIX C
Revised Estimates of PCB and Suspended Solids Loads
in the Upper Hudson River
Estimation of 1993 Upper Hudson PCB and Suspended Solids Loads During
the Transect and Flow-Averaged Sampling Fvents
TAMS.TciraTech
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
APPENDIX C
Revised Estimates Of PCB And Suspended Solids Loads
In The Upper Hudson River
Estimation Of 1993 Upper Hudson PCB and Suspended Solids loads During
The Transect and Flow-averaged Sampling Events
TABLE OF CONTENTS
Page
Discussion C-l
List of Tables
C-l Correction to Original Transect PCB Load Calculations
C-2 Correction to Original Flow-averaged Event PCB Load Calculations
C-3 Correction Factors for the TI Dam PCB Loads
List of Figures
C-l Conceptual Model of PCB Loads Near the TI Dam
C-2 Upper River Water-Column Instantaneous PCB Loading for Transect 2 Low-Flow
Conditions
C-3 Upper River Water-Column Instantaneous PCB Loading for Transect 5 Low-Flow
Conditions
C-4 Upper River Water-Column Instantaneous PCB Loading for Transect 8 High-Flow
Conditions
C-5 Upper River Water-Column PCB Loading for Flow-Averaged 4 Low-Flow Conditions
C-6 W'ater-Column Instantaneous Total PCB Loads for Transect 1
C-7 Water-Column Instantaneous PCB Homologue Loads for Transect 1
C-8 Water-Column Instantaneous Total PCB Loads for Transect 2
C-9 Water-Column Instantaneous PCB Homologue Loads for Transect 2
C-10 Water-Column Instantaneous Total PCB Loads for Transect 3
C-l 1A Water-Column Instantaneous PCB Homologue Loads for Transect 3
C-l IB Water-Column Instantaneous PCB Homologue Loads for Transect 3 Excluding
Waterford
C-l2 Water-Column Instantaneous Total PCB Loads for Transect 4
C-l3 Water-Column Instantaneous PCB Homologue Loads for Transect 4
C-14 Water-Column Instantaneous Total PCB Loads for Transect 5
C-l 5 Water-Column Instantaneous PCB Homologue Loads for Transect 5
C-l6 Water-Column Instantaneous Total PCB Loads for Transect 6
C-l7 Water-Column Instantaneous PCB Homologue Loads for Transect 6
C-18 Water-Column Total PCB Loads for Flow-Averaged Event 1
C-l9 Water-Column PCB Homologue Loads for Flow-Averaged Event 1
TAMS'TenaTech
-------
APPENDIX C
Revised Estimates Of PCB And Suspended Solids Loads
In The Upper Hudson River
Estimation Of 1993 Upper Hudson PCB And Suspended Solids loads During
The Transect And Flow-averaged Sampling Events
TABLE OF CONTENTS
C-20 Water Column Total PCB Loads for Flow-Averaged Event 2
C-21 Water-Column PCB Homologue Loads for Flow-Averaged Event 2
C-22 Water-Column Total PCB Loads for Flow-Averaged Event 3
C-23 Water-Column PCB Homologue Loads for Flow-Averaged Event 3
C-24 Water-Column Total PCB Loads for Flow-Averaged Event 4
C-25 Water-Column PCB Homologue Loads for Flow-Averaged Event 4
C-26 Water-Column Total PCB Loads for Flow-Averaged Event 5
C-27 Water-Column PCB Homologue Loads for Flow-Averaged Event 5
C-28 Water-Column Total PCB Loads for Flow-Averaged Event 6
C-29 Water-Column PCB Homologue Loads for Flow-Averaged Event 6
C-30 Water-Column Instantaneous Total PCB Loads for Transect 8
C-31 Water-Column Instantaneous PCB Homologue Loads for Transect 8
List of Corrected Figures for the DEIR
3-32 Suspended-Matter Loading in the Upper Hudson River - Transect 1 Low-Flow
Conditions (corrected)
3-33 Suspended-Matter Loading in the Upper Hudson River Transect 3 - Transition between
Low-Flow and High-Flow Conditions (corrected)
3-34 Suspended-Matter Loading in the Upper Hudson River - Transect 4 High-Flow
Conditions (corrected)
3-35 Suspended-Matter loading in the Upper Hudson River - Transect 6 Low-Flow Conditions
(corrected)
3-38 Upper River Water-Column Instantaneous PCB Loading for Transect 1 Low-Flow
Conditions (corrected)
3-40 Upper River Water-Column Instantaneous PCB Loading for Transect 3 Transition from
Low-Flow to High-Flow Conditions (corrected)
3-43 Upper River Water-Column Instantaneous PCB Loading for Transect 4 High-Flow
Conditions (corrected)
3-44 Upper River Water-Column PCB Loading for Flow-Averaged Event 1 High-Flow
Conditions (corrected)
3-45 Upper River Water-Column PCB Loading for Flow-Averaged Event 2 Low-Flow
Conditions (corrected)
3-46 Upper River Water-Column PCB Loading for Flow-Averaged Event 3 Low-Flow
Conditions (corrected)
ii TAMS/TeiraTech
-------
APPENDIX C
Revised Estimates Of PCB And Suspended Solids Loads
In The Upper Hudson River
Estimation Of 1993 Upper Hi dso.n PCB and Suspended Solids loads During
The Transect and Flow-averaged Sampling Events
TABLE OF CONTENTS
3-47 Upper River Water-Column Instantaneous PCB Loading for Transect 6 Low-Flow
Conditions (corrected)
3-48 Upper River Water-Column PCB Loading for Flow-Averaged 5 Low-Flow Conditions
(corrected)
3-49 Upper River Water-Column PCB Loading for Flow-Averaged 6 Low-Flow Conditions
(corrected)
ill
TAMS TetraTech
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
APPENDIX C
Revised Estimates of PCB and Suspended Solids Loads
in the Upper Hudson River
Estimation of 1993 Upper Hudson PCB and Suspended Solids Loads During
the Transect and Flow-Averaged Sampling Events
DISCUSSION
In this Appendix, corrections factors are discussed and applied to the Phase 2 transect and
flow-averaged events to account for changes in the understanding of Upper Hudson River conditions
which have come to light since the release of the Data Evaluation and Interpretation Report (DEIR).
As discussed in the corrections to Section 3.2 of the DEIR (see the Responsiveness Summary for
Volumes 2A, 2B and 2C), the transect and flow-averaged event calculations required revision due
to new information pertaining to flow and loads in the Upper Hudson. As a result, two sets of
correction factors were developed for the load estimates. The development of these factors is
described below
Flow Corrections
The first corrections stemmed from a comparison of the USEPA, USGS and precipitation data
as discussed in the correction to Section 3.2.2 of the DEIR (see the Responsiveness Summary' for
Volumes 2A, 2B and 2C). A short review of the flow data issue for 1993 is presented here as a
service to the reader.
Because of dam construction activities which occurred in 1993, the regularly recorded USGS
staff gauges at Stillwater and Waterford were effectively destroyed and the long-term USGS flow
measurements at these stations were stopped. The loss of the two flow monitoring gauges occurred
just prior to the inception of the USEPA water column measurement program. Notably, the staff
gauge measurements at Ft. Edward were not affected by the construction activities.
To remedy the lack of direct flow measurements, both the USGS and the USEPA attempted
to estimate river flow based on other available data. The USEPA used NYS barge canal staff gauges
located throughout the Upper Hudson between Ft. Edward and Waterford in a correlation analysis
to develop a river flow/barge canal staff gauge relationship which could be used to discern flow at
various points in the Upper Hudson. This analysis is described in Section 3.2.2 of the DEIR. The
USGS used the limited number of tributary staff gauges in the Upper Hudson valley to estimate the
net yield of the watershed below Ft. Edward. This information was also translated into flow
estimates. Both models utilized the Ft. Edward staff gauge measurements to represent total flow to
that point. These efforts resulted in two partially independent flow estimates.
c-i
TAMS'TetraTech
-------
To discern the better estimate, a comparison was made of summer time precipitation with the
average incremental increase in flow between Ft. Edward and Stillwater as calculated by the USEPA
and the USGS relationships. This analysis showed that the USEPA estimates overestimated the
average incremental flow between Ft. Edward and Stillwater relative to the historical USGS
measurements. That is, the USEPA data indicated an relatively high runoff yield per unit of
precipitation. Conversely, the USGS estimate fell within the range of prior historical measurements
of flow and precipitation. On this basis, the USGS estimates were ultimately selected over those
calculated by the USEPA.
This comparison was not available at the time of the preparation of the DEIR and the USEPA
results were originally selected for the calculations presented in the DEIR. Since the comparison
suggests that the USGS estimates are more in line with prior measurements, the original transect and
flow-averaged load calculations were revised to reflect these flows. The USEPA database issued in
July, 1998 (release 4.1) contains the USGS flow data reflected in the revisions presented here.
In general, the USGS and USEPA flow estimates agreed to within about 10 percent at higher
flow conditions but the USGS flow estimates were 35 to 50 percent lower when total river flow was
less than approximately 5,000 cfs at Ft. Edward. When these lower flows are applied to the USEPA
PCB and suspended solids measurements, proportionately lower loads are estimated for Stillwater
and Waterford. These results will be discussed later in this Appendix. Tables C-l and C-2 represent
the correction factors for the flow estimate revision. Note that in each instance the flow estimate
correction factor (CF) is defined as follows:
CF - Flow usgs
FlOW USEpA
It should be noted as well that the USEPA flows for the Schuylerville station are also affected
by changes in flow data for Stillwater and Waterford. This is because the flow at this station was
obtained by proportioning the flow increase between Ft. Edward and Stillwater on the basis of
drainage basin area. Thus, the changes in Stillwater flow are partially reflected in the flow at
Schuylerville. Typically, the flow correction at Schuylerville resulted in a decreased flow estimate
of 25 percent or less relative to the original USEPA estimate {i.e., the revised flow at Schuylerville
was 75 percent or more of the original flow estimate).
In addition to the modifications to the mainstem station flow estimates, the flow estimates for
the major tributaries between Ft. Edward and Waterford also required modification. Specifically,
flow estimates for the Batten Kill and Hoosic River were revised in proportion to their drainage
basin contributions and to the changes in flow at Stillwater and Waterford. Flow correction factors
were always less than unity for the Batten Kill, reflecting the similar direction of change at
Schuylerville. The corrections for the Hoosic River were both greater and less than unity,
corresponding to the direction of change for the Waterford flow estimates. Note that although the
C-2
TAMS/TetraTech
-------
flow corrections for these tributaries can be fairly large (as much as a 78 percent decrease), the
correction has little impact on the river's PCB load calculation since the tributaries contribute so
little PCB mass.
Correction for Potential Bias in the TI Dam Monitoring Station
As noted in the corrections to section 3.2 of the DEIR (see the Responsiveness Summary for
Volumes 2A, 2B and 2C), recent sampling collected by GE in the vicinity of the TI Dam indicates
a consistent difference between the water column PCB concentration at the TI Dam monitoring
station and that at a center channel monitoring station nearby. (Note that GE's TI Dam monitoring
point is on the west wing wall of the TI Dam while the USEPA's monitoring station is located in
about 3 to 4 feet of water about one-quarter mile upstream of the Dam. At present, interpretations
of the GE and USEPA data suggest both locations capture the western-most portion of the flow at
the Dam.) An analysis was completed for the available data pairs covering 1997 and 1998. The
results of the analysis are presented in Section 1 of the USEPA review contained in Book 3 of the
Responsiveness Summary for Volumes 2A, 2B and 2C.
The analysis yields the correction factors shown in Table C-3 (reproduced from Table 1-3 of
Volume 3 of the Responsiveness Summary for Volumes 2A. 2B and 2C). These factors are
dependent on both river flow and the PCB concentration at Rogers Island, at the upstream end of the
TI Pool. The description of the model used to develop the correction factors is described in Section
1.1 of Volume 3 of the Responsiveness Summary for Volumes 2A, 2B and 2C. A portion of that
text is reproduced here as an aid to the reader:
To understand the relationship [between the TI Dam monitoring station and the actual
load crossing the Dam], consider the extremely simplified conceptual mode shown in
Figure [C-l], in which downstream flow through the TIP is indicated by arrows.
Discrepancy between shore concentrations (C,) and mixed concentrations at the dam
(C:) presumably arises because there is an additional load in the near-shore area (L).
which is not immediately mixed laterally. Consider a case in which transport is
laterally mixed at some point (say, the end of Griffin Island). At this point, there is a
flow of magnitude Q0 with a concentration of Cc. Downstream (i.e., in the areas of the
TID-West sampling station) full lateral mixing does not occur, and an additional load,
L, is introduced. For simplicity, assume that the flow is split into two portions, with
a flow of Q, going through the near-shore portion, and a flow of Q0-Qi going through
the main channel. These flows then mix and recombine at the dam. It is important to
realize that the concentration in the near-shore area is determined by both the upstream
concentration and the local loading, L. Under these conditions, the concentration in
the near-shore area (TID-West) would be given by
C, = C0 + L/Q,
TAMSTetraTech
-------
while the mixed center channel concentration at the dam would be given by
C: = C0 + L/'Q0
The ratio would then be
C: Co - L/Qo
C, C0 + L/Q,
This ratio depends on the relative magnitude of Q, to Q0, indicating that lateral mixing
intensity presumably increases with the magnitude of Q0. As Q increases toward Q0
(implying instant lateral mixing of L), the ratio should approach 1. The ratio also
depends on the relative magnitude of C0 versus L. As the upstream concentration
increases, the ratio should again increase toward 1 because the contributions from the
near-shore area are swamped by upstream loads.
Thus, the high bias seen in initial GE sample comparisons is a joint result of low flows
and low upstream concentrations. The bias results from incomplete lateral mixing of
what is likely (to a first approximation) a fixed local load. If this load is small relative
to the upstream load, or if mixing is high, the bias is reduced. Thus, it is entirely
inappropriate to apply the apparent bias correction observed in 1996-1997 to the entire
observed time series at TID-West. In particular, a much smaller bias correction should
apply during conditions prior to 1995 in which much higher upstream loads were
observed.
In the model described above, the GE main channel monitoring station would be represented
by the concentration C0, since this location would not "see" the additional loading introduced in the
near-shore environment. Ultimately, the actual load crossing the TI Dam (C;) lies between the near-
shore value C, and the main channel value CG. See the discussion in Section 1 of the USEPA
commentary concerning this issue.
Essentially, the data show the greatest correction at low flow conditions (less than 4000 cfs)
and when the upstream load at Rogers Island is at levels less than 17 ng/L. Higher flows and higher
Rogers Island concentrations diminish the need for a correction.
Assuming that the center channel value is closer to the '"correct" value for determining the
total load at the Dam, these correction factors were applied to the 1993 TI Dam samples as shown
in Tables C-l and C-2. Corrections for the 1993 data set for the TI Dam station were only required
in about two-thirds of the sampling events. Specifically, transects 1. 2, 4 and 8 and flow-averaged
event 1 did not require corrections due to the combination of high flow and high concentrations at
C-4
T.A.MS TerraTetb
-------
the Rogers Island station. The correction factor for the remaining transects and flow-averaged events
was 0.8 (i.e., a 20 percent decrease in concentration and load) since all events had Rogers Island
concentrations well above the threshold of 17 ng/L.
Application of the Corrections
The flow and concentration corrections described above were appropriately applied to the
various transects and flow-averaged events. Note that the flow corrections affect both PCB and
suspended solids loads. Using the revised flows, the suspended solids loads for transects 1.3.4 and
6 were revised and replotted. Figures 3-32 to 3-35, representing these sampling events in the DEIR.
have been corrected and are included in this Appendix.
In a similar fashion, the figures in the DEIR representing the PCB loads for the transect and
flow-averaged events were updated to reflect the revised flows as well as the TI Dam bias
corrections described above. Corrected versions of Figures 3-38, 3-40, 3-43, and 3-44 to 3-49 are
included in this Appendix. These figures represent transects 1,3,4, and 6 as well as flow-averaged
events 1, 2, 3, 5, and 6. New plots, representing transects 2, 5 and 8 and flow averaged event 4. are
presented here as well in Figures C-2 to C-5, respectively. These plots were developed using the
revised flows and the TI Dam correction as appropriate. Lastly, a set of diagrams describing the
Mohawk River's dissolved, suspended and total PCB loads has been added to all transect plots when
available (specifically Figures 3-38, 3-40, 3-43, 3-47, C-2 and C-3), to permit the direct comparison
of the Mohawk River loads with those of the Upper Hudson at Waterford. An additional revision to
these figures is the reporting of the total PCB load at each station in both mg/s and kg/day as an aid
to the reader.
Interpretation of the Revised Estimates in the Upper Hudson
As discussed in the USEPA corrections to Section 3.2 of the DEIR (found in the
Responsiveness Summary for Volumes 2A, 2B and 2C), the revised flow estimates change the low-
flow conditions far more extensively than the high flow conditions due to the similarity of the
USEPA and USGS high flow estimates and the larger disagreement between the low flow estimates.
This is clearly in evidence in the correction factors shown in Tables C-l and C-2.
For suspended solids, the flow revisions yield proportionately lower loads at Stillwater and
Waterford for transects 1 and 6 while transects 3 and 4 have slightly higher loads at these stations.
The revised estimates do not change the initial interpretations given in the DEIR with regard to
suspended solids loads. The suspended solids loads are relatively constant throughout the Upper
Hudson within any individual sampling event during the period of study with the exception of the
resuspension event seen in transect 3. attributed to the spring flood event on the Hoosic River. (As
noted in the Responsiveness Summary for Volumes 2A, 2B and 2C, the Hoosic River flood event
represents a l-in-3 year event and not a 1-in-100 year event as stated in the DEIR.)
C-5 TAMS'TctraTech
-------
The interpretation for the PCB loads of the Upper Hudson was more affected by the flow
revisions. An additional set of figures was developed for this Appendix to aid in the examination of
the data. Figures C-6 to C-17 represent the transect PCB loads plotted as a function of river mile.
Each transect is represented in two plots, the first showing total PCB load as a function of river mile,
the second showing homologue load as a function of river mile. Note that only monochloro- to
tetrachlorohomologues are represented on the second plot since these homologues represent the
majority of PCB mass in the water column. A similar set of plots was developed for the flow-
averaged events (see Figures C-18 to C-29), exhibiting total PCB load and homologue loads for each
sampling event. Lastly, two figures presenting transect 8 results are included as Figures C-30 and
C-31. Transect 8 was a unique sampling event and is described later in the text.
In subsequent discussions, the transects and flow-averaged events are organized based on
season and the notable features of the sampling event. In general, this organization follows the
discussions presented in Section 3.2.6 of the DEIR and the reader is referred to this section for more
detailed discussion. In the discussions to follow, differences between the original interpretation and
that based on the revised results are noted. Based on the revised results, the transect and flow-
averaged events were separated into 3 groups, specifically winter-early spring (low flow-cold water),
spring flood (high flow) and late spring-summer (low flow-warm water). PCB transport during these
three periods show different characteristics. Note that this is two less groups than discussed in the
DEIR. This does not supersede the groups presented in the DEIR but is done to simplify the
discussion of the impact of the revised flows. Specifically, transect 3 is now examined under both
winter and spring flood conditions rather than by itself as a transitional event. Flow-averaged events
2 and 3 are examined in the context of the summer sampling events.
Winter - Earlv Spring
The first period, represented by transects 1, 2 and the upstream portion of transect 3 above
Stillwater, is characterized by the typical TI Dam load consisting of a monochloro- and
dichlorohomologue-dominated mixture. However, both transects 1 and 2 have some sampling or
analytical problems associated with them. The issue with transect 1 affects only the Rogers Island
sample and prevents the calculation of a net TI Pool contribution. As discussed in the DEIR, the
Rogers Island sample in transect 1 is unlike any other sample collected during the Phase 2
investigation (See Figure 3-38). As such its concentration and congener pattern is suspect. Results
for transect 1 are plotted in Figures 3-38, C-6 and C-7.
In transect 2, analytical problems relating to blank laboratory contamination affected many
congeners, in particular, BZ#1 and #4. The measurement of BZ#4 was compromised for the TI Dam,
Schuylerville and Stillwater samples. Since BZ#4 comprises the vast majority of the
dichlorohomologue mass, the dichlorohomologue loads for these stations are suspect as well. The
homologue plots for Schuylerville and Stillwater clearly show the impact of the BZ#4 issue, since
there is essentially no dichlorohomologue mass without the congener (see Figure C-2). The
quantitation of BZ#1 was also an issue in several of the transect 2 samples. BZ#1 constitutes the
C-6
TAM S/T e traTec h
-------
vast majority of the monochlorohomologue mass, so the proportion of total mass represented by this
homologue is also suspect. As a result of the high frequency of blank contamination issues, little can
be inferred from this transect. Nevertheless, the results for this transect are presented in Figures C-2.
C-8 and C-9.
Transect 3 had no important analytical problems and the portion of the transect upstream of
Stillwater presents conditions similar to transect 1 (see Figures 3-40, C-10 and C-11). When these
two transects are examined together, several basic observations can be made. Specifically, the
homologue pattern of the TI Dam load is quite distinct from the Rogers Island load seen in transect
3 and later transects. Transect 3 yields a large gain in water-column load across the TI Pool as seen
in later transects as well. Transect 2 also suggests such a gain across the Pool although its results are
much more uncertain as described above. Downstream of the TI Dam under these conditions, the
homologue pattern is well preserved (see Figures 3-38 and 3-40). In transect 1, all homologue loads
appear to be translated relatively conservatively (to within 25 percent) all the way to Waterford (see
Figure C-7). In transect 3, there appears to be some gain in load to Stillwater but note that the
homologue patterns are largely unchanged (compare Figures 3-40 and C-10). Figures C-l 1A and C-
11B show the similarities among the homologue loads vs. river mile. This suggests the load gain
may be due to uncertainties in the flow estimates resulting from the flow transition which was
occurring during transect 3 rather than a true addition to the water column inventory. Thus, clearly
in transect 1 and most likely in transect 3, the water column load originating above the TI Dam is
transported in a near-conservative manner, for all homologucs. As will be shown, this was not the
case in summer.
Spring Flood
The high flow events were largely unaffected by the flow and TI Dam revisions. As a result
the conclusions drawn for these events remain the same. Transect 4, transect 8, flow-averaged event
1 and a portion of transect 3 all characterize this period in the river. During the earliest spring
sampling event (i.e., transect 3 between Stillwater and Waterford), the spring flood on the Hoosic
River delivers a large suspended matter load (see Figure 3-33) but resuspension from the Hudson
River bottom also adds significantly (note the difference between the Hoosic River suspended solids
load and the suspended solids load at Waterford). The additional suspended solids load is attributed
to scour of the Hudson River bottom which also serves to contribute a very large PCB load (ca. 19
kg/day). The net result of this addition is clearly expressed in the distinct change in the homologue
pattern of the water column load (see Figure 3-40). This event clearly documents the occurrence of
river bottom scour with accompanying PCB transport.
The major spring flood sampling event, transect 4, shows PCB loads in the Upper Hudson to
be transported conservatively to Waterford. This is evident in all three figures for this transect
(Figures 3-43. C-l2 and C-l3). Total PCBs as well as the individual homologues are transported to
Waterford in an apparently conservative manner. Note that the TI Dam station is not represented on
these plots due to the influence of the Moses Kill on this sampling station during this particular
TAMSTctraTcch
-------
event. See Section 3.2.6 of the DEIR text for further discussion of this issue. Evident in this event
is the increase in the monochlorohomologue load as a result of passage through the TI Pool, despite
the large overall loading from upstream. This homologue is then transported along with the others
to Waterford. Additionally, as noted in the DEIR text, there is no evidence of a scour event in the
Hudson River below Stillwater during this event despite the fact that the river flows are higher than
those noted in transect 3. Based on this observation, it appears that the scour event in transect 3 is
related to the way in which the high flow of the Hoosic River enters the Hudson, perhaps serv ing
to scour river sediments in the vicinity of the Hoosic River confluence.
Flow-averaged event 1 is essentially unchanged as a result of the revisions and describes a
condition similar to that seen in transect 4. Specifically, the principal load is derived above Rogers
Island with an additional monochlorohomologue load obtained in the TI Pool (see Figures 3-44. C-
18 and C-19). In this event, some tetrachlorohomologue load is lost and some dichlorohomologue
load is gained across the TI Pool, making this portion of the transect similar to that seen in transect
1. These two events (flow-averaged event 1 and transect 1) both suggest a partial removal of the
Rogers Island load and replacement with TI Pool-derived PCBs as a result of passage through the
Pool. This is based on the extensive change in homologue pattern which occurs during these events
coupled with the relatively minor change in total loading. Other Phase 2 sampling events generally
have minor Rogers Island loads so that evidence for this hypothesis is less clear.
The last station in flow-averaged event 1 had some issues regarding the accuracy of the
samples collected as measures of the true loading condition between the TI Dam and Waterford.
Specifically, local dam construction was causing some obvious resuspension during the first week
of sample collection. For this reason, the individual samples were composited into 2 one-week
composites instead of a single two-week composite and analyzed. Both values are represented in
Figures C-18 and C-19. The line connecting Waterford with TI Dam is based on the average of the
pair of composites at Waterford. In this figure the one-week composites are multiplied by the
average water flow for the corresponding week rather than the average flow for the two weeks. The
resulting loads for the two composites are quite different (see Figure C-18). reflecting the impact of
the dam construction on the first sampling week and yielding a substantially higher PCB load
relative to the second week. The suspended solids results also demonstrate the impact of the dam
construction, with a mean suspended solids concentration of 46 mg/L during the first week and 8.4
mg'L during the second week. The higher suspended solids concentration corresponds to the higher
PCB load shown in Figure C-18. As a result of the dam construction it is unclear what the true PCB
load at Waterford would have been during this period. Nonetheless, the homologue data can provide
some insights here.
Specifically, as shown in Figure C-19, the average trichloro- and tetrachlorohomologue loads
clearly increase downstream of the TI Dam as expected due to the resuspension of sediment caused
by the dam construction. However, the monochloro- and dichlorohomologue loads do not increase
relative to the TI Dam loads. This would be expected if the TI Dam loads were translated
downstream in a near-conservative fashion with subsequent addition of a large quantity of low-level
C-8
TAMS TetraTech
-------
PCB contaminated sediment. These sediments would have little monochloro- and dichloro-
homologue content since the concentration would be too low to support extensive dechlorination.
In fact the PCB concentration on the suspended matter during the first week of this sampling event
contained only 2 mg/kg of Total PCBs. consistent with this scenario. (An extensive discussion of
the relationship between PCB sediment concentration and the extent of dechlorination can be found
in Chapter 4 of the DEIR.) Thus despite the impact of the dam construction, the underlying
homologue distribution appears consistent with the results of transect 4 at least for the less-
chlorinated homologues, i.e., near-conservative transport of PCBs from TI Dam to Waterford during
high flow conditions. Notably, it also clear that the dam construction had a major impact on PCB
loads at Waterford, generating loads very comparable to the spring runoff event captured by transect
4.
One last sampling event, transect 8 is presented here which was not presented in the original
Report. This sampling event is a unique event in that it is neither a timed transect nor a flow-
averaged event. Instead, the samples were collected in a single day without regard to timing. This
collection effort represented a simple sampling opportunity, since flow-averaged event 1 was
commencing and the river flow was rapidly rising. The results of this transect are presented in
Figures C-4. C-30 and C-31. Because the sampling event was neither sequenced nor averaged over
time, the samples are not directly related to each other, unlike the other sampling events. For this
reason, the changes in load among the stations, particularly between TI Dam and Waterford, may
not reflect the true load changes. Nonetheless, the similarity of the homologue pattern between TI
Dam and Waterford supports the condition seen in the other spring high flow events, that is, near-
conservative transport through the Upper Hudson. Independent of the conservative transport issue,
these samples do yield the instantaneous loads at the time of sampling. The most useful information
to be drawn from this event is the individual scale of the loads, which are substantially lower than
transect 4 (50 percent or more, see Figure C-12) despite the fact that this event represents a rising
and higher water flow relative to transect 4. In fact, these loads arc very similar to the average load
for the subsequent two week period captured in flow-averaged event 1. As discussed in response
DG-1.15B to the DEIR, these results indicate that transect 4 captured the major PCB transport event
of the year since the sampling event represented the flow peak conditions of the first major flow
event of the year.
Overall, these events describe in detail the PCB loads associated with the spring floods. The
revisions of flow do not affect these events particularly and there is no TI Dam correction required.
In transects 4, 8 and flow-averaged event 1, a large load is generated upstream of the TI Pool, in the
range of 8 to 18 kg/day. Despite this large load, evidence of the TI Pool input can be seen in the
addition of monochloro- and dichlorohomologue loads across the Pool. Below this point the
homologue pattern is preserved to Waterford. In the detailed examination of transect 4 (and to some
degree, flow-averaged event 1), total loads as well as the homologue pattern are preserved to
Waterford, suggesting conservative transport during high flow conditions. Transect 3 between
Stillwater and Waterford documents a significant local load produced by scour of the river bottom.
This event documents the instability of sediment deposits below TI Dam. as suggested by the results
TAMS TetraTcch
-------
of the Low Resolution Sediment Coring Report, which documents large PCB losses from several
previously-defined NYSDEC hot spot areas. Finally, dam construction in the vicinity of Waterford
served to create a very significant PCB load, suggesting that such activities may need further control
to prevent large PCB release events.
Late Spring - Summer
The period of May to September was characterized by two transects and five flow-averaged
events. During this period, efforts by GE served to greatly diminish the scale of the loads released
upstream of Rogers Island, beginning in June. Sampling events collected by both USEPA and GE
prior to June, 1993 frequently show large loads entering the TI Pool at Rogers Island while sampling
events collected after June consistently show a greatly diminished load at this station. This period
(May to September) is also characterized by warmer water column temperatures and lower flows
relative to the earlier conditions. Thus these sampling events are the most affected by both the flow
revisions and the TI Dam bias correction.
For the purposes of examining the effects of the corrections, these sampling events can be
combined since the impacts are similar. When these events were first examined, total PCB loads
delivered to Waterford appeared very similar in magnitude to those present at the TI Dam. Notably,
the homologue patterns changed as the river moved downstream from TI Dam to Waterford despite
the consistency of the magnitude of the total load. The change in pattern became more and more
pronounced from spring to summer. The revisions did not affect these patterns since they were only
applied to flow or the total PCB concentration. The difference in the TI Dam correction factor for
Total PCBs vs. the Tri+ sum was not used here since the difference in the factors was not found to
be statistically significant. This finding may change as more data are obtained since such a difference
might be expected due to the nature of the TI Pool source (i.e., predominantly lighter congeners).
Flow-averaged events 2 and 3 represent conditions in May and June, respectively. As a result
of the revisions, the total PCB load estimates at the TI Dam have decreased by 20 percent. For flow-
averaged event 2, this still represents a large load gain across the TI Pool (see Figure C-20). For
flow-averaged event 3 (see Figure C-22), this correction results in a minor load decline across the
TI Pool, since there was a large Rogers Island load associated with this event. Nonetheless the load
crossing the TI Dam during flow-averaged event 3 is still quite different from that at Rogers Island,
again suggesting substantial replacement or modification of the upstream load during its passage
through the Pool. This is consistent with the results seen in transect 1 and flow-averaged event 1.
The absolute loading level at Waterford for flow-averaged event 2 is essentially unchanged
but due to the modification of the TI Dam load estimate, it appears that a small additional load (less
than 20 percent) occurs between TI Dam and Waterford (see Figure C-20). A similar scale loss is
apparent in the revised flow-averaged event 3 plot (see Figure C-22). However, both of these events
show a marked decline in the fraction of monochloro-homologue between the two stations,
representing about a 50 percent loss (see Figures C-21 and C-23). This change is beyond the
C-10
TAMS/TetraTech
-------
analytical uncertainty and suggests some other process may be involved beyond simple translation
of the TI Dam load. Flow-averaged event 3 also shows a decline in the dichloro-homologue load,
at roughly 15 percent. Again, the change in the proportion of the dichlorohomologue mass relative
to the trichloro- and tetrachlorohomologues is more substantial and suggests an additional process
affecting the PCB load.
Transect 5, flow-averaged event 4, flow-averaged event 5, transect 6 and flow-averaged event
6 sequentially span the entire summer of 1993. In each of these events, the load at Waterford is
consistently lower than that at the TI Dam (see Figures C-14, C-24, C-26, C-16 and C-28,
respectively). For transect 5 and flow-averaged event 4 which represent late June and early July, the
load decrease is about 40 percent. Subsequent sampling events exhibit smaller total PCB losses, in
the range of 12 to 16 percent. Notably the sampling event prior to transect 5 (flow-averaged event
3) exhibited a 20 percent loss. As seen in late spring sampling events, these losses do not occur
consistently across all homologues. Specifically, these losses are almost exclusively related to
parallel losses of the monochloro and dichlorohomologue loads (see Figures C-15. C-25. C-27, C-17
and C-29, respectively). These changes are quite substantial in terms of the proportion of these
homologues at Waterford relative to the TI Dam and are well beyond any analytical uncertainty.
Absolute declines in load are typically 30 to 50 percent for the dichlorohomologues and 80 to 100
percent for monochlorohomologues.
The trichloro- and tetrachlorohomologue groups typically show much smaller absolute
changes in load, both positive and negative and on the scale of 20 percent. However, these groups
clearly increase their importance relative to the total PCB load. In most instances as well, the
proportion of tetrachlorohomologue increases relative to the trichlorohomologue. This suggests the
addition of PCBs, perhaps from the sediments, with a higher fraction of tetrachlorohomologue
relative to the water column load. Alternatively, this may represent a minor loss of the lighter
trichlorohomologues during transit to Waterford. However, this loss would be far smaller than that
seen for the monochloro- and dichlorohomologues.
A second observation concerning the loss of the lighter homologues can be made concerning
the scale of the total transport. Over the period May to September, both the highest total loads at the
TI Dam and the greatest mass loss of monochloro- and dichlorohomologues are associated with the
sampling events occurring at the end of June and early July. This period is quite close in time to the
mid to late June peak in PCB transport seen in the subsequent monitoring conducted by GE over the
period 1994 to 1998 and suggests that this phenomenon was occurring in 1993 but was partially
obscured by the release events occurring upstream.
Lastly, but perhaps most importantly, the results showr that the trichloro- and
tetrachlorohomologues are largely transported from the TI Dam to Waterford regardless of the time
of year or rate of transport. These results suggest that transport of these homologues is largely
conservative, since the loading rate set at the TI Dam is very close to that at Waterford, regardless
of the absolute rate of loading. The relationship among the major homologue groups further supports
C-ll
TAMS TctraTech
-------
this with the ratio of trichloro- to tetrachlorohomologue mass staying fairly constant while moving
downstream while monochloro- and dichlorohomologues tend to track each other, with large
apparent mass losses occurring during warmer weather.
All of the discussions above are predicated on the validity of the TI Dam station as an accurate
measure of the total PCB load as well as the homologue pattern of this load. The fact that there is
apparent translation of the TI Dam loads during winter and spring conditions lends some credence
to this assumption. However, the conclusions concerning the low flow conditions and the loss of the
less chlorinated homologues may appear more tenuous since the interpretation of the data is less
straightforward. In this instance the measurements at Schuylerville for transects 5 and 6 provide
additional information to further support and confirm the interpretations given above. Specifically,
transects 5 and 6 can both be examined in terms of the load changes between TI Dam, Schuylerville
and Waterford. In both transects, the monochloro- and dichlorohomologue loads both peak at the TI
Dam and then decline at a similar rate per river mile from the TI Dam to Schuylerville and then from
Schuylerville to Waterford. This suggests that the loss process begins soon after the TI Dam under
warm conditions but more importantly, that clear evidence for loss of these homologues exists
independent of the TI Dam station. The flow-averaged events do not have data for Schuylerville but
are clearly consistent with this loss phenomenon, confirming that the conditions seen in these
transects apply throughout the summer. The internal consistency among all these sampling events
also serves to support the TI Dam station as a useful measure of the Upper Hudson load.
A second observation can be made from the late spring and summer sampling events
regarding the trichloro- and tetrachlorohomlogue loads. Evidence for a total PCB load gain
downstream of the TI Dam station is only present in transect 6, based on the revised results.
Specifically, this transect shows a small gain in total PCB between TI Dam and Schuylerville. All
late spring-summer events show a net mass loss to Waterford. However, beginning with transect 5
in late June, these events show a consistent but relatively small increase in the tetrachloro- to
trichlorohomolgue ratio (see Figures C-15, C-25, C-27, C-17 and C-29). In the two detailed sampling
events, transects 5 and 6, this increase in the tetrachloro- to trichlorohomologue ratio occurs between
TI Dam and Schuylerville, accompanied by net gains in their total loads. Below Schuylerville, these
loads appear to be translated in a near-conservative fashion. Evidence for the increases in these loads
below TI Dam is also apparent in the three summer flow-averaged events. Taken together, these
events suggest a small additional PCB load is generated by the sediments between TI Dam and
Schuylerville. Based on transect 6, the data suggest a slower rate of load production per mile across
this river section relative to the TI Pool. (Note the change in slope in the load plots for the two
homologues in transect 6 as seen in Figure C-17. Transect 5 could not be used in this comparison
since no data are available for Rogers Island) Nonetheless, these results suggest this region to be a
net source of the tetrachloro- and trichlorohomologues to the water column, with the region
downstream simply transporting these homologues to the Lower Hudson.
The homologue signal indicated by the water column load gains from TI Dam to Schuylerville
during summer conditions suggest a less dechlorinated source than that of the TI Pool. The results
C-12
T AMS/TetraT ech
-------
also suggest a lower flux rate per mile, suggesting a lower PCB concentration or inventory available
to drive the additional load. These observations are consistent with the historical measurements of
PCBs which show the highest concentrations and inventories in the TI Pool and lower concentrations
and inventories downstream. The higher concentrations of the Tl Pool would tend to be more
extensively dechlorinated, thus yielding a more dechlorinated PCB load.
Lastly, the late spring-summer events also demonstrate the importance of the TI Pool load to
the entire suite of homologues and not just the less chlorinated ones. This is evident in all of the
summer events in the majority of the monochloro-. dichloro- and trichlorohomologue loads for the
entire Upper Hudson are clearly produced within the TI Pool. The tetrachlorohomologue load
appears is principally generated from the Upper Hudson sediments, but the reach from TI Dam to
Schuylerville yields a nontrivial portion of this load but still less than the TI Pool. With the
anticipated, continued control of the PCB releases from the GE Hudson Falls facility to levels similar
to those seen in 1997 and 1998. the Pool is expected to continue to represent the major source of all
PCBs to the water column of the Upper Hudson.
Summary
The revisions of the conditions at the TI Dam, Stillwater and Waterford changed the
relationships among the PCB loads observed at these stations to a limited degree. However, the most
important conclusion regarding PCB loads remains intact and solidly-based. Specifically, the
sediments of the TI Pool are the major source of PCBs to the water column during low flow
conditions. Based on the level of source control at the GE Hudson Falls facility demonstrated in the
GE/QEA Modeling Report (GE/QEA. March 1998) the sediments of the TI Pool sediments have
clearly become the year-round dominant PCB source. Evidence for a sediment-based PCB source
between TI Dam and Schuylerville is suggested by the internal consistency of the late spring-
summer sampling events which is brought out more clearly by the revisions due the reduction in the
Tl Dam load estimates.
The near-conservative behavior of the total PCB load from TI Dam to Waterford discussed
in the DEIR is apparently only characteristic of winter and spring conditions and does not apply to
Total PCBs during late spring and summer. Low flow/low temperature or high flow conditions yield
near conservative transport. During late spring and summer conditions, the total PCB load is not
conservative and declines downstream of the TI Dam. However, the decline is largely confined to
the less-chlorinated homologues, suggesting the occurrence of another process which selectively
affects these homologues. Gas exchange or aerobic degradation are likely candidates for this loss.
Sediment exchange is not a viable basis for removal of these homologues due to their low partition
coefficients which largely prevent their preferential absorption relative to the higher chlorinated
homologues.
C-13
TAMS TctraTech
-------
Other Observations
It should be noted that transect 5 had an apparent sampling issue with the Rogers Island
station. Specifically, the congener and homologue pattern associated with this particular sampling
event was quite different from any other sample collected at this station. The pattern, as shown in
Figure C-3, was a monochloro-, dichlorohomologue-dominated mixture, quite similar to that of the
sediments. This sample also exhibited notably higher suspended solids concentrations relative to the
upstream and downstream stations. Lastly, the sample seemed quite high in concentration given the
prior remedial work completed earlier in the month by GE. On this basis it was concluded that the
Rogers Island sample for this transect had incorporated a portion of local sediment during the
sampling process, presumably due to a disturbance of the river bottom while the sample collector
waded into the river to fill the sample bottles. Thus this station could not be used to estimate the load
across the TI Pool. As a surrogate, the load at the remnant deposits station was substituted in the
preparation of Figures C-14 and C-15, but the true load gain across the Pool can not be obtained for
this transect. Notably, this transect was handled differently than transect 1 which had a similar
sampling issue at Rogers Island. However, in the case of transect 5, the source of the problem with
the Rogers Island samples was fairly well defined whereas in transect 1, the source of the Rogers
Island sampling issue is unknown.
Although stated in the DEIR, the issue of the Mohawk River contribution is worth reviewing
here. The presentation of the entire set of Mohawk loads in the revised and new plots shows that this
source area, i.e., the PCB load produced by the entire Mohawk water shed, is dwarfed by Upper
Hudson load. The load from the Mohawk is typically less than 5 percent of the Upper Hudson load
at Waterford under low flow conditions and less than 20 percent under high flow conditions. These
results clearly show Upper Hudson as the source of PCBs to the Lower Hudson.
A last observation concerns the Troy sampling location, near the Green Island Bridge.
Specifically, the load estimated at this location is inconsistent as an estimate of the load to the Lower
Hudson relative to the sum of the Waterford and Mohawk loads. Most likely, the location, a
shoreline sampling point, is too close to the confluence of the two tributaries and thus the samples
obtained from this point do not always represent a well mixed sample. This is most evident in
transect 4 (Figure 3-43) for PCBs and Figures 3-34 and 3-35 for suspended solids. As a result, loads
calculated for this location should not be used as they are too unreliable.
Conclusions
From the discussions above the following conclusions can be made:
• Transport of trichloro- and tetrachlorohomologues appears to be nearly conservative
throughout the year from the TI Dam to Waterford.
C-14
T AM S/TetraT ec h
-------
• Conservative transport of monochloro-, dichloro-. trichloro- and tetra-chlorohomologues
downstream of TI Dam to Waterford occurs during high flow spring conditions based on total
load and homologue pattern of downstream stations.
• Late spring-summer conditions suggest additional sediment-derived loading of a relatively
small amount of trichloro- and tetrachlorohomologues (less than 20 percent of the TI Dam
load) from the region between TI Dam and Schuylerville. No evidence exists for additional
net loads downstream of Schuylerville.
• A late-spring-1993. Tl-Pool-load maximum, similar to that seen by GE in later years, is
suggested by the Phase 2 data as well.
The anticipated decline in Waterford and Stillwater loads was partially offset by the revisions
resulting from TI Dam bias correction such that the relationship among station load estimates
did not change as much as expected. The net result of the revisions is lower overall loads
(approximately 20% lower) in the Upper Hudson under low flow conditions, with high flow-
conditions largely unmodified from those estimated in the DEIR. The revisions did yield a
more distinct and consistent decline in less chlorinated homologues at low flow t warm water
conditions, suggesting loss of these homologues via a process such as gas exchange or aerobic
degradation.
• The total proportion of the TI Pool contribution to the 1993 annual PCB budget for the Upper
Hudson declines as a result of these revisions. However, its importance over the post-June
1993 does not decline appreciably, since its load appears undiminished over time while the
Hudson Falls source has been substantially reduced.
• The TI Pool remains the major source of trichloro- and tetrachlorohomologue mass to water
column during low flow in 1993 and year-round post-June 1993.
Overall, the corrections do not require a major revision to the main conclusions of the DEIR,
with the exception of the concept of year-round conservative PCB transport. The TI Pool remains
the dominant source under low flow conditions although there is evidence to suggest some release
from sediments between TI Dam and Schuylerville. Year-round conservative transport is limited to
the higher chlorinated homologues while the less chlorinated homologues are subject to substantial
mass loss while enroute from the TI Dam to Waterford. Nonetheless, the Upper Hudson PCB loads
remain the dominant source of PCB contamination to the Lower Hudson, with post-June 1993
contamination arising principally from the sediments of the TI Pool.
C-15
TA.MS TetraTech
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
APPENDIX C
TABLES
-------
Fable CM
Correction to Original Transect I'CIt Load Calculations 1
T ransect
No.
Meat) Flow at
Fort
Edward (cfs)
Flow Adjustment2
TI Dam Load
(concentration)
Adjustment''
Batten Kill
Schuylerville
Stillwater
Hoosic River
Waterford
1
4924
0.67
0.93
NA"
0.31
0.67
None
2
4545
0.63
0.89
0.9
0.22
0.79
None
3
5103
0.29
0.79
0.75
5.924, 0.98s
1.05
0.8
4
17300
0.36
0.91
1
1.17
1.04
None
5
24(H)
0.44
0.85
NA"
0.23
0.6
0.8
6
2461
0.25
0.75
NAh
0.45
0.67
0.8
Notes:
1. Correction represents the ratio of the new llow or concentration over the original How or concentration as reported in the DHIR.
For example:
Transect fi at Water'lord
Original How = S,l(X)cls
Revised llow = 3.4(H) cts
Correction Factor (CF) = 3,400 / 5, KM) = 0 67
2. Flow ad|usltncnls are based on a comparison of the IJSGS Hows ami those developed in the DFIR See the coirections to Section 3.2
id the DFIR in the Rcponstvene.ss .Summary lor the Phase 2 Reports: Volumes 2A, 2B and 2C.
V T! Dam correction factor derivation is described in the IJSHPA's discussion in Hook 3 of the Responsiveness Summary lor the Phase
2 Reports: Volumes 2A, 2B anil 2C
4 This factor applies to the low llow condition sampled on 3/26/93.
5. This factor applies to the high llow condition sampled on 3/30/93 and note that no I'C'B sample was obtained on this date.
6 Not applicable since no sample was obtained at this station lor this transect
Hudson River Database Release 4 1
TAMS/Telra Tech
-------
Table C-2
Correction to Original Flow-Averaged Event PCB Load Calculations1,4
Mean Flow at
Flow
TI Dam Load
Flow-Averaged
Fort Edward
Correction at
(concentration)
Event No.
(cfs)
Waterford'
Adjustment''
1
18852
NA5
None
2
3385
1.05
0.8
3
2988
0.66
0.8
4
2484
0.56
0.8
5
2513
0.58
0.8
6
2515
0.5
0.8
Notes:
1. Correction represents the ratio of the new (low or concentration over the original
tlow or concentration as reported in the DEIR.
For example:
Flow-Averaged Event 6
Original Waterford flow = 7.080 cfs
Revised Waterford tlow = 3.540 cfs
Correction Factor (CF) = 3.540 / 7.080 = 0.5
2. Flow adjustments are based on a comparision of the USGS flows and those
developed in the DEIR. See the corrections to Section 3.2 of the DEIR in the
Reponsiveness Summary for the Phase 2 Repons:Volumes 2A, 2B and 2C.
3. TI Dam correction factor derivation is described in the USEPA's discussion in
Book 3 of the Responsiveness Summary for the Phase 2 Reports: Volumes 2A. 2B
and 2C.
4. Flow corrections are only presented for Waterford since this is the only
station downstream of TI Dam in the flow-averaged events.
5 Samples collected at Waterford were not applicable in this event due to local
canal construction which is believed to have influenced the samples
Hudson River Database Release 4 1
TAMS/Tetra Tech
-------
Tabic C-3
Correction Factors for the TI Dam PCB Loads
Empirical Bias Correction Factors
Total PCBs
ETri +
Low Flow,
Low Upstream Concentration
Fort Kdward Flow < 4000 cfs
Fort Fdward Concentration < 17 ng/1 total PCBs or < 15 ng/1
ETri +
0.64
0.69
Low Flow,
High Upstream Concentration
Fort Fdward Flow < 4000 cfs
Fort Fdward Concentration > 17 ng/1 total PCBs
or ¦ 15 ng/1 STri +
0.80
0.88
High Mow,
Low Upstream Concentration
Fort Fdward Flow 4000 cfs
Fort Fdward Concentration < 17 ng/1 total PCBs or • 15 ng/1
UTri +
0.78
1.0
High I'low,
High Upstream Concentration
Fort Fdward Flow 4000 cfs
Fort Fdward Concentration 17 ng/1 total PCBs
or • 15 ng/1 HTri +
1.0
1.0
TAMS/TetraTcch
-------
APPENDIX C
FIGURES
-------
dp..
Q»
^
C,=C0 + L/Q, C0
I '—^
C2 = C0 + IVQ0
~^r
o-Q,
C2, Q„
Hudson River Database Release 4.1
Figure C-l
Conceptual Model Of I'CB Loads Near the TI Dam
TAMS/TelraTech
-------
Whole-Water PCBs
Dissolved-Phase PCBs
Suspended-Phase PCBs
Fenimore Bridge
River Mile = 197.6
Total Load = 0.04 mg/s
(0.003 kg/d)
Flow = 130 mVs
Rogers Island
River Mile = 194.6
Total Load = 2.0 mg/s
(0.17 kg/d)
Flow = 130 mVs
Thompson Island Dam
River Mile = 188.5
Total Load = 7.0 mg/s
(0.61 kg/d)
Flow = 130 mVs
Batten Kill
River Mile = 182.1b
Total Load = 0.05 mg/s
(0.004 kg/d)
Flow = 25 ntVs
Schuylerville
River Mile = 181.3
Total Load = 4.0 mg/s
(0.34 kg/d)
Flow = 160 rnVs
Stillwater
River Mile = 168.3
Total Load = 3.9 mg/s
(0.33 kg/d)
Flow = 180 mVs
Hoosic River
River Mile = 167.5b
Total Load = 0.03 mg/s
(0.002 kg/d)
Flow = 10 mVs
Waterford
River Mile = 156.5
Total Load = 3.4 mg/s
(0.30 kg/d)
Flow =190 nvVs
Mohawk River
River Mile = 156.2C
Total Load = 0.08 mg/s
(0.01 kg/d)
Flow = 70 rnVs
W)
S,
73
n
o
J
E,
T3
M
o
J
OX)
B
X)
«
©
J
s,
•u
n
©
J
C = 0.29 ng/L
Total ™
~53d
B
•a
at
©
hJ
C =16 ng/L
Total "
C =52 ng/L
Total "
C = 1.9 ne/L
Total "
25 ng/I
C =22 ng/L
Total
4
3 -
2
1 -
0
4
3 -
C = 2.6 ng/L
Total
V--9—9 "9 ?
- rt n a a
Q. o g ^
£ ° z Q
C =18 ng/L
Total
r\
-f=
o
c
o
S
T I
—r • >-
n rz
8 1
C = 1.2 ng/L
Total
"9—9-
£ u -r
(— a. I
o Z Q
PCB Homologue
C =0.11 ng/L
Dissolved
-v ¦?
Z Q
C = 3.7 ng/L
Dissolved
C =44 ng/I
Dissolved v
C = 0.56 ng/L
Dissolved
C = 20 ng/I
Dissolved ™
= 18 ng/L
-v-
c
s
H cl,
=Y—9- 9—9—9—
rt rt cj rt
u S. o g
= £ ° z a
C = 2.0 ng/L
Dissolved
9
g c
O
-9—9 ¦ 9 9
r" « « «
h g c g
r° t) 4;
Ho.1
jV—?.
O £ o
C = 4.5 ng/L
Dissolved
C = 0.6 ng/L
Dissolved
-9—9
g 5
o
S
i!
5 0
C = 0.17 ng/L
Sispended
C =12 ng/L
Suspended
C = 8.4 ng/L
Suspended
. 9--V. "°^-
9- 9 -9—<9
O
2
Q
H g c £
r° «T«
H a.
3 2
a. u
£ 0
0 £
z a
C =1.4 ng/L
Suspended
9 9 -9—9 -
£ ° Z O
C = 4.4 ng/L
Suspended
¦9—9—V—9—
3 S 2 g
C = 3.8 ng/L
Suspended
C = 0.65 ng/L
Suspended
C =14 ng/L
Suspended
C = 0.61 ng/L
Suspended
PCB Homologue
I h i 1 ^ o z a
PCB Homologue
Hudson River Database Release 4.1 TAMS/Tetra Tech
Notes:
a. Suspended-phase PCB concentration in ng/L calculated as function of dry weight concentration (ug/kg) and total suspended solids concentration (mg/L).
b. Tributary river mile designations correspond to point of confluence with the Hudson River.
c. Transect 2 samples were collected during the period of February 19 to February 23, 1993.
Figure C-2
Upper River Water-Column Instantaneous PCB Loading for Transect 2 Low-Flow Conditions
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.04 mg/s
(0.003 kg/d)
Flow = 80 m3/s
Rogers Island
River Mile = 194.6
Total Load = 25.6 mg/s
(2.2 kg/d)
Flow = 80 m3/s
Thompson Island Dam
River Mile = 188.5
Total Load = 14.6 mg/s
(1.3 kg/d)
Flow = 76 m'Vs
Batten Kill
River Mile = 182.lc
Total Load = 0.02 mg/s
(0.002 kg/d)
Flow = 15 m3/s
Schuylerville
River Mile = 181.3
Total Load = 13.7mg/s
(1.2 kg/d)
Flow = 85 m3/s
Hoosic River
River Mile = 167.5°
Total Load = 0.03 mg/s
(0.002 kg/d)
Flow = 10 m3/s
Waterford
River Mile = 156.5
Total Load = 8.2 mg/s
(0.7 kg/d)
Flow = 102 m3/s
Mohawk River
River Mile = 156.2C
Total Load = 0.13 mg/s
(0.01 kg/d)
Flow = 41.1 m3/s
lb
B
-o
M
O
J
1i>
B
•v
a
o
J
Whole-Water PCBs Dissolved-Phase PCBs Suspended-Phase PCBs'
B
o
J
Is*
B
-o
o
J
C = 0.55 ng/L
Total
t/>
B
T3
«
O
J
Ik
S
¦o
o
J
ia>
B
"O
«
o
J
"Bb
B
-o
a
©
J
C = 320 ng/L
Total "
C = 192 ng/I
Total
C = 1.3 ng/L
Total
C = 160 ng/L
Total
C = 3.0 ng/L
Total
~9—9—9—9—9—9——¦
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 2.1 mg/s
(0.18 kg/d)
Flow = 790 m3/s
Rogers Island
River Mile = 194.6
Total Load =110 mg/s
(9.5 kg/d)
Flow = 790 mYs
Thompson Island Dam
River Mile = 188.5
Total Load = 88 mg/s
(7.6 kg/d)
Flow = 828 m3/s
Waterford
River Mile = 156.5
Total Load = 62 mg/s
(5.3 kg/d)
Flow = 1189 m3/s
C/>
"Bb
£
"O
©
J
Whole-Water PCBs Dissolved-Phase PCBs Suspended-Phase PCBs'
"Sl>
B
'o
©
J
60
50
40
30
20
10
0
C = 2.6 ng/L
Total
C = 0.44 ng/L
Dissolved
C = 2.2 ng/L
Suspended
60
50
40
30-
20
10
0
C = 140 ng/L
Total "
C =99 ng/L
Dissolved
/ \
O
s
"'"'I
15
£
T3
«
O
-J
60 -i
....
50-
40-
30-
1
20
I
10
0
¦. T""""
§ 5
, - - . • V V !
cq d
£ & & £ o £ Q
C =110 ng/L
Total w
"T r~¦ T" i - ?—? V "V"
O ' T"
Q .!> u Hi
C =40 ng/L
Suspended
/\
\
o
s
;r -rr ed cd cd
be*
« S 4i
H Q.
cx o
i °
?--o-
ri cd
c o
o
£ Q
H £ i £ o g Q
C =85 ng/L
Dissolved v
r
•r
Z Q
C =22 ng/L
Suspended
? ?"
o
o
2
r
Q
<=0
I fll U< if D
" i> ^5
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.04 mg/s
(0.003 kg/d)
Flow = 70 m'/.s
Rogers Island
River Mile = 194.6
Total Load = 2.6 mg/s
(0.22 kg/d)
Flow = 70 m'/s
Thompson Island Dam
River Mile = 188.5
Total Load = 12.8 mg/s
0.1 kg/d)
Flow = 74 m3/s
Waterford
River Mile = 156.5
Total Load = 7.6 me/s
(0.66 kg/d)
Flow = 96 m 7s
V.
"Sc
#¦»
c
-o
u
E
¦o
a
e
W!
"3d
E
-a
a
5£
e
¦e
3
6
5
4
3
2
1
0
6
5
4
3
2
1
6
5
4
3
2
6
5
4
3
2
1
0
Whole-Water PCBs
€ = 0.56 ng/L
Total **
H
X
%>
u
ZJ
C = 37 ng/L
Total **
s ><
H 2. —
; - z -
C = 176 ng/L
Total &
5 .3*
C = 80 ng/L
Total K
— x
r u
PCB Homoloeue
n
Hudson River Database Release 4.1
Note: Flow-Averaged 4 samples were collected during the period of July 6 to July 20. 1993.
TAMS/Tetra Tech
Figure €-5
Upper River Water-Column PCB Loading for
Flow-Averaged 4 Low-Flow Conditions
-------
Rogers Island
¦3
O.J
0.6
0.4
0.2
River Station
Tributarv
Batten Kill Hoosic River Mohawk River
200 190
180 170
River Mile
160 150
Hudson River Database Release J i
Figure C-6
Water-Column Instantaneous Total PCB Loads tor Transect 1
TAMS'TtftraTech
-------
0.8
0.7
0.6
0.5
0.4
0.3
Mono
--
— Di
- • ~
• - Tn
v-
- Tetra
•
Tributary. Mono
Tributary. Di
•
Tributary. Tri
<7
Tributary. Tetra
VVaterford
0.2
0.1
: !¦'
200
190
180 170
River Mile
160
150
TI Dam
Rogers Island Batten Kill Hoosic River Mohawk River
Hudson River Dalabase Release 4 1
Figure C-7
Water-Column Instantaneous PCB Homologue Loads for Transect 1
TAMS.TclrjTe.h
-------
o.7;
TI Dam
0.6
0.5
River Station
Tributary
0.4
0.3
0.2
Rogers Island
Batten Kill
Hoosic River Mohawk River
200 190 180 170 160 150
River Mile
Hudson River Database Release 4 1
Figure C-8
Water-Column Instantaneous Total PCB Loads for Transect 2
TAMSiTtftraTVch
-------
0.3
0.25
o.:
0.15
0.05
— Mono
-C
— Di
- - «
• - Tn
v-
--- Tetra
•
Tributary,Mono
Tributary. Di
~
Tributary. Tn
7
Tributary.Tetra
Waterford
200
190
180 170
River Mile
160
150
TI Dam
Rogers Island Batten Kill Hoosic River Mohawk River
Hudson River Daraha.se KeJea.se 4 I
Figure C-9
Water-Column Instantaneous PCB Homologue Loads for Transect 2
TAMSTeiraTe^h
-------
20
River Station
Tributary
15
10
5
Mohawk River
Rogers Island TI Dam
0
Hoosic River
Batten Kill
200 190 180 170 160 150
River Mile
Notes:
Two sets of data were collected in Hoosic River
Hudson River Database Relea.se 4 1
Figure C-10
Water-Column Instantaneous Total PCB Loads For Transect 3
TAMS/TetraTech
-------
Waterford
6 .
4 .
— Mono
-C
— Di
- - •
• -Tri
- Tetra
•
Tributary. Mono
Tributary. Di
~
Tributary. Tn
7
Tributary. Tetra
a.
~
*> ~ i * *^ *
200
190
180 170
River Mile
160
150
TI Dam
Rogers Island Bauen Kill Hoosic River Mohawk Ri\er
Note:
Two sets of data were collected in Hoosic River
Hudson Riser Database Release 4 1 TAMS-TeiraTech
Figure C-l 1A
Water-Column Instantaneous PCB Homoloizue Loads for Transect 3
-------
Mono
-=
— Di
• - ~
- -Tn
g.
*•- Tetra
•
Tributary. Mono
~
Tributary, Di
•
Tributary, Tri
Tributary. Tetra
i /
04: K /x
' N /
: ' ^
; i
0,3 1 /
; /
I
I
8
200 190 180 170 160 150
River Mile
TI Dam
Rogers Island Bauen Kill Hoostc River Mohawk River
Note:
Two sets of data were collected in Hoosic River
Hudson Rner Djubase Release 4 1
TAMS-TciraTech
Figure C-l IB
Water-Column Instantaneous PCB Homologue Loads for Transect 3 Excluding Waterford
-------
20
¦j Rogers Island
Schuylerville
15
River Station
Tributary
10
5
Batten Kill
Hoosic River Mohawk River
0
200 190 180 170 160 150
River Mile
Hudson River Database Release 4 I
Figure C-12
Water-Column Instantaneous Total PCB Loads for Transect 4
TAMS/Teira T ech
-------
W'aterford
10 ~
8 t- . .
— Mono
--
— Di
- - •
• - Tr;
— Tetra
•
Tributary.. Mono
-
Tributary. Di
~
Tributary. Tri
7
Tributary. Tetra
200 190 180 170 160 150
River Mile
TI Dam
Rogers Island Batten Kill Hoosic River Mohawk River
Hudson River Database Release 4 1
Figure C-13
Water-Column Instantaneous PCB Homologue Loads for Transect 4
TAMS/TeiraTuvh
-------
TI Dam
0 6
River Station
Tributary
0.4
Remnant Deposits
0.2
Batten Kill
Hoosic River Mohawk River
200 190 180 170 160 150
River Mile
Hudson Rner Database Release 4 I
Figure C-14
Water-Column Instantaneous Total PCB Loads for Transect 5
TAMS.TetraTeth
-------
5
"Sb
0 5
— Mono
-
- Di
* ~ •
- Tn
---.v..
-- Tetra
•
Tributary. Mono
Tributary. Di
•
Tributary, Tri
Tributary. Telia
Waterford
200
190
180 170
River Mile
TI Dam
Remnant Deposits
160
150
Batten Kil
Hoosic River Mohawk River
Hudson River Database Release 4 1
Figure C-15
Water-Column Instantaneous PCB Homologue Loads for Transect 5
TAMS.-TeiraTtvh
-------
0.6
TI Dam
River Station
0.4
Tributary
0.3
0.2
Rogers Island
Batten Kill
Hoosic River Mohawk River
200 190 180 170 160 150
River Mile
Hudson River Database Release 4 1
Figure C-16
Water-Column Instantaneous Total PCB Loads for Transect 6
T AMS-Teirj Toch
-------
Waterford
73
0.15 !
0.1
0.05
I
r'
— Mono
-C
— Di
- - »
• - Tri
f.
Tetra
•
Tributary. Mono
-
Tributary. Di
«
Tributary. Tri
V
Tributary. Tetra
200
190
180 170
River Mile
160
150
TI Dam
Rogers Island Batten Kill
Hoosic River Mohawk River
Hudson Riser Database Release 4 1
Figure C-17
Water-Column Instantaneous PCB Homologue Loads for Transect 6
TAMS/TtftraTech
-------
20 :
Waterford
* j —•— River Station
Rogers Island
0 "
TI Dam
5
0
200 190 180 170 160 150
River Mile
Note:
Value represnting Waterford is the average of 2 one-week composites.
Hudson River Database Release 4 t
Figure C-18
Water-Column Total PCB Loads for Flow-Averaged Event 1
TAMS.'TetraTech
-------
Mono
Waterford
_ _ - - - ""*?
200 190
TI Dam
Rogers Island
180 170
River Mile
160
150
Note:
Values representing Waterford are the averages of 2 one-week composites.
Hudson Ri\er Database Release 4 ]
Figure C-19
Water-Column PCB Homologue Loads for Flow-Averaged Event 1
TA.MSTciraTi.vh
-------
River Station
0.6
Rogers Island
0.4
1
200 190 180 170 160 150
River Mile
Hudson River Database Release 4 1
Figure C-20
Water-Column Total PCB Loads for Flow-Averaged Event 2
TAMS/TtftraTcch
-------
0.6
0.5
Watcrford
•Ul
0.4
0.3
Mono
— 9 Tetra
200 190
TI Dam
Rogers Island
180 170
River Mile
160
150
Hudson River Database Release 4 1
Figure C-21
Water-Column PCB Homologue Loads for Flow-Averaged Event 2
TAMS'TetrjTech
-------
Rogers Island
River Station
Hudson Riser Database Release 4 1 TAMS/TetraTejh
Figure C-22
Water-Column Total PCB Loads for Flow-Averaged Event 3
-------
0.5
0.4
0.3
0.2
0.1
Mono
Waterford
200 190
TI Dam
Rogers Island
180 170
River Mile
160
150
Hudson Ri\er Database Release 4 1
Figure C-23
Water-Column PCB Homologue Loads for Flow-Averaged Event 3
TAMS.TetraTech
-------
1.2 T
River Station
TI Dam
8
0.6
0.4
oners Island
->
0
200 190 180 170 160 150
River Mile
Hudson River Datable Release -i I
Figure C-24
Water-Column Total PCB Loads for Flow-Averaged Event 4
TAMS.TetraTt.vh
-------
0,4
0.35
0.3
0.25 .
0.2 .
0.15
0.1
0.05 -
I X
Waterford
Mono
- Telra
200 190
TI Dam
Rogers Island
180 170
River Mile
160
150
Hudson River Database Release J 1
Figure C-25
Water-Column PCB Homoloaue Loads for Flow-Averaged Event 4
TAMS/TeiraTcch
-------
0.7
TI Dam
0.6
0.5
River Station
0.2
0
200 190 180 170 160 150
River Mile
Hudson River Darubjsc: Release -i
Figure C-26
Water-Column Total PCB Loads for Flow-Averaged Event 5
TAMSTeiralYvh
-------
0.25
0.2 :
Waterford
Ml
0.15
0.1
0.05
Mono
200 190
TI Dam
Rogers Island
180 170
River Mile
160
150
Hudson River Daiabase Release 4 1
Figure C-27
Water-Column PCB Homologue Loads for Flow-Averaged Event 5
TAMS/TctnTeeh
-------
River Station
Rogers Island
Hudson River Database Release 4 1 TAMS.TelraTc^h
Figure C-28
Water-Column Total PCB Loads for Flow-Averaged Event 6
-------
0.2
Waterford
0.15
0.1
0.05 i
Mono
' Telra.
200 190
TI Dam
Rogers Island
180 170
River Mile
160
150
Hudson River Daiaba.se Release 4 1
Figure C-29
Water-Column PCB Homologue Loads for Flow-Averaged Event 6
TAMS.TetraToch
-------
River Station
River Mile
Hud-on Rsver Oatahaw Release 4 I T AM ¦vTettaTcv h
Figure C-30
Water-Column Instantaneous Total PCB Loads for Transect 8
-------
I I
TI Dam
Waterford
Mono
— Di
• • - Tn
.» Tetra
2(X)
190
180 170
River Mile
160
150
Rogers Island
Hudson River Database Release 4 1
Figure C-31
Water-Column Instantaneous PCB Homologue Loads for Transect 8
TAMS/TetraTech
-------
\
I
in
m
w
M
e
93
©
J
3
©
c/j
TJ
»/~)'/">
i
T"1 I
rl O
>n >n
River Mile
Hudson River Database Release 4 I TAMS/Tetra 'lech
Notes:
a) Tributary river mile designations correspond to point of confluence with the Hudson River
b) Fish Creek suspended matter load is estimated using the suspended solids value lor the Batten Kill and a How estimate based on drainage basin area.
c) Sample is believed to over-represent upstream main Stem Hudson River loading due to incomplete mixing ol the Mohawk River.
Figure 3-32 (corrected)
Suspended-Matter Loading in the Upper Hudson River - Transect 1 Low-Flow Conditions
-------
Ik
OJj
c
'•B
et
0
-
1/1
2
"o
C/)
T3
0>
*0
c
01
a
<«
3
C/5
6(M)
500
400
300
200
100
Legend;
0
o
Li
A
Total Measured Main-Stem Hudson River
Suspended Mailer Load
Batten Kill Contribution4
t-'ish ('reek Contribution*1''
I ioosie River Contribution'
Hoosic River Load
Measured on 3/30/93
ri O oo so
O O O ON
ri ri - -
rj-
Ov
i "T "T"T*1
ri O oo
O O oo
Tt^"P"rT""l''! T I I
Hoosic River Load
, * Measured on 3/27/93
• ' i iA' ¦ r i i i i rrrrn tti i
sO
OO
oo
(N
00
o
00
00
r-
so
r-
rt
r^
h-
O
r-
oo
so
SO
sO
sO
fN
sO
O
sO
00
l/~i
so
ID
CN
O
River Mile
Hudson River Database Release 4,1 TAMS/Tetra Tech
Notes:
a) Tributary river mile designations correspond to point of confluence with the Hudson River
b) Fish Creek suspended mailer load is estimated using the suspended solids value for the Batten Kill and a flow estimate based on drainage basin area,
c) Scour event due lo onset ol spring Hood event in lower part of the Upper River,
Figure 3-33 (corrected)
Suspended-Matter Loading in the Upper Hudson River
Transect 3 - Transition between how-Flow and High-Flow C onditions
-------
120
Legend
1 *
Total Measured Main-Stem Hudson River |
Suspended Mallei load
o
i
KM)
o
Batten Kill Contribution*
I'isli Creek Contribution'1 h
*b3d
-X
A
Hoosic River Contribution"'
Otl
B
80
,
1. )
Mohawk River Contribution''
• d
•5
C5
O
-J
1/!
s
60
"o
cn
"O
0)
¦o
c
point of confluence with the Hudson River
b( I• is 11 Creek suspended mailer load is estimated using the suspended solids value lot the Batten Kill and a How estimate based on drainage basin area,
el Sample is believed to over-represent dilution by the Moses Kill due to proximity ol sampling location to Mo.se.s Kill confluence,
d) Sample is believed to over represent upstream Main-Stem Hudson River loading due to incomplete mixing of the Mohawk River.
Figure* 3-34 (corrected)
Suspended-Matter I.(lading in the Upper Hudson River - Transect 4 High-Flow Conditions
-------
~Sk
DC
e
T3
m
©
J
i/i
2
"o
C/3
T)
4>
-o
c
a>
a.
t/3
3
Cfi
0.6
0.5
0.4
0.3
0,
0.1
0
Legend:
o
A
r l
O
r l
O
O
fN
Total Measured Main Stem Hudson River
Suspcmtcd Mailer Load
Batten Kill Contribution-1
Fish Creek Contribution"1'1'
Hixisic River Contribution"1
Mohawk River Contribution"1
O
A
oo
Ch
sO
o
OS
I
r I
OS
o
Os
OO
oo
v 1.1
i i i r i i ' i i
so <3-
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.47 mg/s
(0.04 kg/d)
Flow - 170 niVs
Rogers Island b
River Mile = 194.6
Total Load = 12 mg/s
( 1.04 kg/d)
Flow = 170m Vs
Thompson Island Dam
River Mile = 188.5
Total Load = 9.3 mg/s
(0.8 kg/d)
Flow = 170 mVs
Batten Kill
River Mile = 182.1c
Total Load = 0.06 mg/s
(0.005 kg/d)
Flow = 20 m/s
Schuylerville
River Mile = 181.3
Total Load = 8.7 mg/s
(0.75 kg/d)
Flow = 195 mVs
Hoosic River
River Mile = 167.5°
Total Load = 0.03 mg/s
(0.003 kg/d)
Flow = 19 m /s
Waterford
River Mile = 156.5
Total Load = 7.6 mg/s
(0.65 kg/d)
Flow = 220 mVs
Mohawk River
River Mile = 156.2°
Total Load = 0.09 mg/s
(0.01 kg/d)
Flow = 80 mVs
Green Island Bridge
River Mile = 151.7
Total Load = 6.6 mg/s
(0.57 kg/d)
Flow -- 285 mVs
DC
Whole-Water PCBs
C = 2.8 ng/L
Tola)
C =71 ng/L
Tola) "
"r*
(— a. -1-
C =54 ng/L
Total
10-r
4-
C =2.9 ng/L
Total "
H a.
C =45 ng/L
Total
10
iz: v "r* v
t- a. x x
Total
C =34 ng/L
Total "
2-
H a. ac
C = 1.1 ng/L
Total "
2-
O z Q
C = 23 ng/L
Total
H £
PCB Homologue
Dissolved-Phase PCBs Suspended-Phase PCBf
C = 0.25 ng/L i
Dissolved "
C = 7.0 ng/L
Dtaaolved "
C =46 ng/L
Dissolved
¦9—9"- 9 9"
o
z
C = 0.32 ng/L
Dissolved
O
s
-
H
o «
Z Q
C =31 ng/L
Dissolved "
C = 0.35 ng/L
Dissolved
= 18 ng/L
H I £ I | o | |
PCB Homologue
C = 2.5 ng/L
Suspended
C =64 ng/L
Suspended
C = 7.4 ng/L
Suspended
C = 2.6 ng/L
Suspended
C = 8.4 ng/L
Suspended
C = 0.53 ng/L
Suspended
—9—9 —
C = 3.0 ng/L
SispewJed
C =0.79 ng/L
Suspended
•r ps cq « «
£ £ £ £ ° z Q
= 5.2 ng/L
PCB Homologue
Hudson River Database Release 4.1
Notes:
a. Suspended-phase PCB concentration in ng/L calculated as function of dry weight concentration (ug/kg) and total suspended solids concentration (mg/L).
b. The homologue pattern measured for this station was unlike any seen in other Phase 2 samples and is considered suspect.
c. Tributary river mile designations correspond to point of confluence with the Hudson River.
d. Transect 1 samples were collected during the period of January 29 to February 8, 1993.
Figure 3-38 (corrected)
Upper River Water-Column Instantaneous PCB Loading for Transect 1 Low-Flow Conditions
TAMS/Tetra Tech
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.16 mg/s
(0.01 kg/d)
Flow = 94 rnVs
Rogers Island
River Mile = 194.6
Total Load = 2.4 mg/s
(0.20 kg/d)
Flow = 94 m'Vs
Thompson Island Dani
River Mile = 188.5
Total Load = 9.9 mg/s
(0.86 kg/d)
Flow = 98 mVs
Batten Kill
River Mile = 182. lb
Total Load = 0.02 mg/s
(0.002 kg/d)
Flow = 11 m'Vs
Schuylerville
River Mile = 181.3
Total Load = 8.4 mg/s
(0.72 kg/d)
Flow =110 m3/s
Stillwater
River Mile = 168.3
Total Load = 15 mg/s
(1.27 kg/d)
Flow = 300 m3/s
Hoosic River
River Mile = 167.5b
Total Load = 0.9mg/s
(0.08 kg/d)
Flow = 71 m3/s
Waterfordc,d
River Mile = 156.5
Total Load = 220 mg/s
(19.1 kg/d)
Flow = 1360 m3/s
Mohawk River
River Mile = 156.2b
Total Load = 38.4 mg/s
(3.3 kg/d)
Flow = 1830 m3/s
Whole-Water PCBs Dissolved-Phase PCBs Suspended-Phase PCBs
(/)
"53a
E,
-o
CS
O
_1
lb
B
~o
a
o
J
C = 1.8 ng/L
Total v
Dfo
b
•a
M
o
J
(/)
~St>
E,
is
et
o
J
1/5
~5b
¦v
o
-J
u>
B
T3
«
o
J
"Si
B
T3
CS
O
J
6
4
2
0
10-
8
6-
4-
2-
0
C = 104 ng/L
Total **
~~r~
Q
-9—9- 9—9—
¦a o
z a
C = 2.0 ng/L
Total
C = 76 ng/L
Tolal "
C =13 ng/L
Total
C = 160 ng/I
Total "
C =21 ng/I
Total "
M 60
C = 0.42 ng/L
Dissolved
C =16 ng/L
Dissolved
g 5
o
S
¦9"--9"
= 78.4 ng/L
A
"C «
H ^
-9—9—9-
a a n
rt c o
A O £
° Z O
= 0.51 ng/I.
A **
C =63 ng/I.
Dissolved
C =33 ng/L
Dissolved "
9~
"9—9"
I- a, 3C
C = 4.8 ng/L
Dissolved
O
s
"9—"9—9—9—9~
a £ 3 1 S
P* t, C 4)
U n
H Q. £
a ca n
M o 8
O Z Q
15 ng/I.
C =1.5 ng/I.
Dissolved "
O
5
PCB Homologue
h £ a £ O £ Q
PCB Homologue
C = 1.3 ng/L
Suspended
-9—9—9-
5
£ 0 Z Q
C = 9.7 ng/I.
Suspended
" T : ?—9-9-99-9
0 ' r- rt w rt W cQ
cuj£i37S*o7lccj
1 £ £ a £ o £ £
C =22.4 ng/L
Suspended
C =1.5 ng/L
Suspended
C =13 ng/I.
Suspended
C =16 ng/L
Suspended
C = 7.8 ng/L
Suspended
-V—9—9—9—¦?—'9—?-
Q c 2 2 s a 2
-9—<;
= 150 ng/L
o
S
~l 1 1 1 T"
*C
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.68 mg/s
(0.06 kg/d)
Flow = 580 m3/s
Rogers Island
River Mile = 194.6
Total Load = 210 mg/s
(17.9 kg/d)
Flow = 580 mVs
Thompson Island Dam*
River Mile = 188.5
Total Load = 120 mg/s
(10 kg/d)
Flow = 600 m3/s
Batten Kill
River Mile = 182.1c
Total Load = 0.28 mg/s
(0.02 kg/d)
Flow = 27 m3/s
Schuylerville
River Mile = 181.3
Total Load = 193 mg/s
(16.7 kg/d)
Flow = 630 rrvVs
Stillwater
River Mile = 168.3
Total Load = 183 mg/s
(15.8 kg/d)
Flow = 690 m3/s
Hoosic River
River Mile = 167.5L
Total Load = 0.96 mg/s
(0.08 kg/d)
Flow = 210 m"Vs
Waterford
River Mile = 156.5
Total Load = 220 mg/s
(19 kg/d)
Flow = 940 m"Vs
Mohawk River
River Mile = 156.2°
Total Load = 41 mg/s
(3.54 kg/d)
Flow = 1000 mVs
Green Island Bridge
River Mile = 151.7
Total Load = 390 mg/s
(33.7 kg/d)
Flow = 1970 mVs
200-r
IS, l60;
S 120
80-
T3
03
O
40-
0-
Whole-Water PCBs Dissolved-Phase PCBs Suspended-Phase PCBsa
C = 1.2 ng/L
Total
¦?"?¦
Q
u v ~ u o •; a
H CL 3- X z Q
200
C = 360 ng/L
T^-l "
= 190 ng/L
Total
200 -
C =10 ng/I
Tola! v
C = 300 ng/L
Tolal **
C = 270 ng/I
Tolal **
a>'60:
S 120 -
80-
73
<3
O
i-J 40
C = 4.5 ng/L
Tolal
C =230 ng/I
Total **
C =40 ng/L
Tolal "
a to o
C = 200 ng/L
Total
C = 0.16 ng/L
Dissolved
C =48 ng/L
Dissolved **
C =67 ng/L
Dissolved
C = 7.2 ng/L
Dissolved
C =1.0 ng/L
Suspended
C - 310 ng/L
Siapendcd
130 ng/L
Suspended
H 5 S | 8- o I ^
H a, J- rr. _ Z Q
C =3.1 ng/I.
Suspended
-¦?
o
V__V ....0 ..9 9
Q
? I £ ^ z a
C =94 ng/L:
Dissolved " j
— i 'T'
Z Q
C =210 ng/L
Suspended
O
C rs rt
H ~ C
,_.S
-?¦—-¦
1 o £ S
o
5
r—i—i ; Y—?—V—V—9~
c® CJ <0 09 <9 <0
a E 2 5. o g a
£ £ X £ O g q
C =1.4 ng/L
Dissolved "
C =38 ng/L
Suspended
j
J
-J
_i
C = 140 ng/L
Suspended v
C =53 ng/L •
Dissolved "
PCB Homologue
PCB Homologue
PCB Homologue
Hudson River Database Release 4.1 TAMS/Tetra Tech
Notes:
a. Suspended-phase PCB concentration in ng/L calculated as function of dry weight concentration (ug/kg) and total suspended solids concentration (mg/L).
b. Sample is believed to over-represent dilution by Moses Kill due to proximity of sampling location to Moses Kill confluence.
c. Tributary river mile designations correspond to point of confluence with the Hudson River.
d. Sample is believed to over-represent upstream load contribution due to incomplete mixing of the Mohawk River.
e. Transect 4 samples were collected during the period of April 12 to April 14, 1993.
Figure 3-43 (corrected)
Upper River Water-Column Instantaneous PCB Loading for Transect 4 High-Flow Conditions
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 1.4 mg/s
(0.12 kg/d)
Flow = 530 m3/s
Rogers Island
River Mile = 194.6
Total Load = 83 mg/s
(7.2 kg/d)
Flow = 530 m3/s
Thompson Island Dam
River Mile = 188.5
Total Load = 68 mg/s
(5.9 kg/d)
Flow = 560 mVs
!/l
s
TJ
S3
C/3
"©3d
§
¦o
¦y.
"si
T3
C
40
30
20-
10-
0-
40
30
20
10
0
40
30
20
10
0
Whole-Water PCBs
C = 2.6 ng/L
Total *
Cw ^ w
h - x
£ 53 £
H a. 3.
£ 3
cL u
X c
C U
c 1)
Z £
C = 160 ng/L
Total v
c Q
c
u 0>
p™ a.
ra
x
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.03 mg/s
(0.003 kg/d)
Flow = 96 m3/s
Rogers Island
River Mile = 194.6
Total Load = 4.7 mg/s
(0.41 kg/d)
Flow = 96 m3/s
Thompson Island Dam
River Mile = 188.5
Total Load = 14 mg/s
(1.2 kg/d)
Flow = 100 m'/s
Waterford
River Mile = 156.5
Total Load = 16 mg/s
(1.4 kg/d)
Flow = 168 m3/s
V)
r"
-a
es
o
-J
OJj
m
s
¦a
«
c
W3
"afc
c
'V.
"ex
i"
s
¦o
X
c
Whole-Water PCBs
6
4
J
6 -
4 -
T —
6 -
4 -
i -
8
6
4
C = 0.29 ng/L
Total
-¦>-
c 5
0
1
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.05 mg/s
(0.004 kg/d)
Flow = 85 m3/s
Rogers Island
River Mile = 194.6
Total Load = 15 mg/s
(1.3 kg/d)
Flow = 85 m3/s
Thompson Island Dam
River Mile = 188.5
Total Load = 14 mg/s
(1.2 kg/d)
Flow = 89 m3/s
Waterford
River Mile = 156.5
Total Load = 11.3 mg/s
(0.98 kg/d)
Flow = 125 m3/s
B
¦a
03
o
-J
"Si
b
•a
«
o
-J
"Id
B
¦o
CI
o
J
"Si
B
-o
a
o
6
4
2
0
8
6
4
2
0
Whole-Water PCBs
C = 0.61 ng/L
Total w
—io—
—c—
—9—
—o-
C3
C3
cs
c3
<3
u.
E~
C
U
O-
X
1)
a:
Cu
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.01 mg/s
(0.001 kg/d)
Flow = 65 mVs
Rogers Island
River Mile = 194.6
Total Load = 2.0 mg/s
(0.17 kg/d)
Flow = 65 m"Vs
Thompson Island Dam
River Mile = 188.5
Total Load = 6.2 mg/s
(0.54 kg/d)
Flow = 69 mVs
Batten Kill
River Mile - 182. Ib
Total Load = 0.0 mg/s
(0.0 kg/d)
Flow = 7.6 mVs
Schuylerville
River Mile = 181.3
Total Load = 6.5 mg/s
Whole-Water PCBs
Dissolved-Phase PCBs Suspended-Phase PCBs
a
(0.56 kg/d)
Flow = 74 m'/s
Hoosic River
River Mile = 167.5b
Total Load = 0.02 mg/s
(0.002 kg/d)
Flow = 9.5 mVs
Waterford
River Mile = 156.5
Total Load = 5.5 mg/s
(0.48 kg/d)
Flow = 100 mVs
Mohawk River
River Mile = 156.2b
Total Load = 0.22 mg/s
(0.02kg/d)
Flow = 63 m'Vs
Green Island Bridge
River Mile = 151.7
Total Load = 5.6 mg/s
(0.48 kg/d)
Flow = 170 m/s
WD
B
-o
o
J
lb
E
•v
«
o
J
E,
¦o
«
O
-J
"Si
B
«
o
J
cm
B
-o
es
O
-J
B
~a
a
o
J
"5b
E
T3
«
O
-J
la
J,
T3
es
O
J
oJd
B
•a
es
o
J
C = 0.11 ng/L
Dissolved
C = 0.12ng/I
Total v
c JS O i, 5
C =31 ng/L
Total "
C = 28 ng/L
Dissolved
C =92 ng/L
Tola! ^
C =88 ng/L
Dissolved
C = 0.02 ng/L
Total ^
C = 0.0 ng/L
Dissolved
C =89 ng/L
Total ^
C =84 ng/L
Dissolved
C = 2.0 ng/L
Total ^
C =1.4 ng/L
Dissolved ^
C =57 ng/L
Total v
C =49 ng/L
Dissolved
C = 3.5 ng/L
Total v
C =0.41 ng/L
Dissolved v
C =29 ng/L
Dissolved
PCB Homologue
2 H. o § jj
£ £ = £ O | Q
PCB Homologue
C = 0.0 ng/L
Suspended
o r cs ^
c a £ b c ji
s H Q. X
V—¦ —- 9 -9 -9~
o
2
L- C 7= X
i C dl
w I) .2:
f- Cl. x
C = 4.8 ng/L
Suspended
=9=
o
2
« r« ea rt *0
2 s. 3 § a
x £ 0 S a
C = 0.51 ng/L
Suspended
C = 8.6 ng/L
Suspended v
C = 3.1 ng/L
Suspended
ra rj r3
H ^ c o
H Q- 3C
° §. Q
C = 3.7 ng/L
Suspended
.9- s>-
£ a £ 0 z Q
PCB Homologue
Hudson River Database Release 4.1 TAMS/Tetra Tech
Notes:
a. Suspended-phase PCB concentration in ng/L calculated as function of dry weight concentration (ug/kg) and total suspended solids concentration (mg/L).
b. Tributary river mile designations correspond to point of confluence with the Hudson River.
c. Transect 6 samples were collected during the period of August 19 to September 1, 1993.
Figure 3-47 (corrected)
Upper River Water-Column Instantaneous PCB Loading for Transect 6 Low-Flow Conditions
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.27 mg/s
(0.02 kg/d)
Flow - 71 m'Vs
Rogers Island
River Mile = 194.6
Total Load = 2.5 mg/s
(0.22 kg/d)
Flow = 71 m'Vs
Thompson Island Dam
River Mile = 188.5
Total Load = 8 mg/s
(0.69 kg/d)
Flow = 75 mVs
Waterford
River Mile = 156.5
Total Load = 6.7 mg/s
(0.58 kg/d)
Flow = 100 m'Vs
'Si
"ok
E
•o
©
-J
!A.
"sic
s
¦o
CS
o
J
v.
"Sfc
E
•a
o
¦y.
"ait
E
¦o
c
5
4
3
2
1
0
5
4
3
2
1
0
5
4
3
2
1
0
5
4
3
2
1
0
Whole-Water PCBs
c
Total
= 3.8 ng/L
§ 5
O
H 5 5
H Q-
r3
X
OJ
O.
ra
C
C
Z
73
CJ
U
Q
C =35 ng/L
Total *
O
c
o
S S 1
E-
c
a.
r3
x
X
a.
X
C = 104 ng/L
Total ^
c 5
£ g
H
c.
w
I
C
Z —
C =69 ng/L
Total w
c —
o
f- 5
c:
CL
X
X
o
Z
Q
PCB Homologue
Hudson River Database Release 4 1
Note: Flow-Averaged 5 samples were collected during the period of August 2 to August 17. 1993.
TAMS/Tetra Tech
Figure 3-48 (corrected)
Upper River Water-Column PCB Loading for
Flow-Averaged 5 Low-Flow Conditions
-------
Fenimore Bridge
River Mile = 197.6
Total Load = 0.02 mg/s
(0.002 kg/d)
Flow = 71 m3/s
Rogers Island
River Mile = 194.6
Total Load = 2.0 mg/s
(0.17 kg/d)
Flow = 71 m3/s
Thompson Island Dam
River Mile = 188.5
Total Load = 5.8 mg/s
(0.5 kg/d)
Flow = 75 mVs
Waterford
River Mile = 156.5
Total Load = 5.1 mg/s
(0.44 kg/d)
Flow = 100 m'/s
E
¦o
CJ
o
-J
E
¦o
o
"Si
E
•o
a
o
J
"Si
E
¦o
o
4
3 -
2 -
1 -
0 <
Whole-Water PCBs
C = 0.23 ng/L
Total **
—C-
4
3
1
0
4
3
2
1
0
4
3
2
1
0
o
c
o
£
= 22
H a.
r3
x
X
a.
<
o
Q.
U
r3 rs
§ 3
z Q
C = 78.4 ng/L
Total w
\
O
c
o
s
£ 1
f-
5 x
cu -*•
f3
C-
o
r3 rs
C
z 2
C =51 ng/L
Total
O
c
o
h*
is f3
u
a- —
ra
Q.
o
X
CJ
c
S3 r3
C
o £
Z a
PCB Homologue
Hudson River Database Release 4 1 TAMS/Tetra Tech
Note: Flow-Averaged 6 water column samples were collected during the period of September 9 to September 23. 1993.
Figure 3-49 (corrected)
Upper River Water-Column PCB Loading for
Flow-Averaged 6 Low-Flow Conditions
------- |