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
i i
uaEi
\»,/
For
U.S. Environmental Protection Agency
Region II
and
U.S. Army Corps of Engineers
Kansas City District
Book 2 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
S&)
prO&°
For
U.S. Environmental Protection Agency
Region II
and
U.S. Army Corps of Engineers
Kansas City District
Book 2 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
TABLE OF CONTENTS
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)
TAMS/TetraTech
-------
Federal
(LRC - LF)
-------
U.S. DEPARTMENT OF COMMERCE VJ'
National Oceanic and Atmospheric
Administration
National Ocean Service
Office of Ocean Resources Conservation and Assessment
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 nearshore
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
-------
NOAA comments on Hudson River Low Resolution Sediment Coring Report, July 1998
(8/28/98)
sediment PCB concentrations in the Thompson Island Pool has important implications that will
need to be addressed in both the Baseline Modeling and the Ecological Risk Assessment
An estimated 28% loss in PCB sediment inventory from the TIP occurred between 1984 and 1994
was attributed to release to the water column. A minimum loss of a 3,200 kg PCBs was estimated
for hot spots below the TID. NOAA recommends that EPA calculate whether the deposits
identified in the TIP and elsewhere in the river could have provided the amount of PCBs necessary
to maintain the apparendy steady-state concentrations in fish in the river.
Depth to wood chip or cellulose-type material was identified in the field. The authors relied upon
the association between the molar ratio of carbon to nitrogen (C/N ratio) as an indicator of wood
cellulose (Section 2.4.3). The C/N ratio ranged between 11 and 82 with a median of 40 and a
mean of 39; hence, there was a suggestion that wood chips occur throughout much of the Upper
Hudson sediments. It would have been useful if all core segments identified with wood chips or
cellulose-type material were tabulated along with the depth to wood chip or cellulose-type material,
and PCB concentrations. This would have permitted an independent assessment of the relationship
between presence and depth of wood chips and PCB maximum concentrations. The problem with
relying entirely on the C/N ratio is that only 26 samples were examined with nine samples from the
surface segment, seven from the second segment and ten from the third segment Since a total of
170 cores (371 samples) were collected and analyzed for PCBs, a much larger data set is available
that could be stratified by segment depth and sampling location.
To assess sample homogeneity, two subsamples (sample splits generated in the field) were
collected from 9 cores (23 paired samples). The average relative percent difference (RPD) for 1.
these split pairs was 36%. It should'be noted that this calculated RPD includes analytical
variability as well as sample homogeneity.
Thank you for your continual efforts in keeping NOAA apprised of the progress at this site. Please
contact me at (212) 637-3259 or Jay Field at 206-526-6404 should you have any questions or
would like further assistance.
NOAA Coastal Resource Coordinator
cc: Michael Clemetson, DESA/HWSB
Gina Ferreira, ERRD/SPB
Robert Hargrove, DEPP/SPMM
William Ports, NYSDEC
Charles Merckel, USFWS
Anne Secord, USFWS
Anton P. Giedt, NOAA
-------
State
(LRC - LS)
-------
New York State Department of Environmental Conservation
Division of Environmental Remediation
Bureau of Central Remedial Action, Room 228
50 Wolf Road, Albany, New York 12233-7010
Phone: (518) 457-1741 FAX: (518) 457-7925
Johrt P. Cahill
Commissioner
August 31, 1998
Mr. Douglas Tomchuk
United States Environmental Protection Agency
Region II
290 Broadway - 20dl Floor
New York, NY 10007-1866
Dear Mr. Tomchuk:
Re: Hudson River PCBs Reassessment RI/FS
Site No.: 5-46-031
The New York State Department of Environmental Conservation (NYSDEC) and the New York
State Department of Health (NYSDOH) have reviewed the July 1998 Hudson River PCBs Reassessment
RI/FS reports entitled "Volume 2C-A Low Resolution Sediment Coring Report Addendum to the Data
Evaluation and Interpretation Report," and "Phase 2 Human Health Risk Assessment Scope of Work."
This letter provides the State's comments on the two documents.
The Low Resolution Sediment Coring Report (LRSCR) presents four major findings. Following
are the State's general comments corresponding to each of these findings.
Finding 1
"There was little evidence found of widespread burial of PCB-contaminated sediments by clean
sediment in the Thompson Island Pool. Burial was seen at some locations, but more core sites
showed loss of PCB inventory than showed PCB gain or burial." [Page ES-3]
State Comment
The State agrees that, based on the data contained in the LRSCR, much of the PCB-contaminated
sediments in the Thompson Island Pool are not being buried with significant amounts of clean
sediment.
Finding 2
"From 1984 to 1994, there has been a net loss of approximately 40 percent of the PCB inventory
from the highly contaminated sediment in the Thompson Island Pool." [Page ES-4]
-------
State Comment
The State agrees that, based on the data contained in the LRSCR, there has been an identifiable
PCB inventory loss from the sediments of the Thompson Island Pool. However, based on the
data contained in the report, it is difficult to closely quantify the degree of sediment losses. It
may be more appropriate for the report to present a range of estimates rather than a single
number. This same concern was discussed at the Scientific and Technical Committee meeting on
August 18, 1998.
Finding 3
"From 1976-1978 to 1994, between the Thompson Island Dam and the Federal Dam at Troy,
there has been a net loss of PCB inventory in hot spot sediments sampled in the low resolution
coring program." [Page ES-4]
State Comment
The State agrees that, based on the data contained in the LRSCR, there has been an identifiable
PCB inventory ioss from the hot spots between the Thompson Island Dam and the Federal Dam
at Troy.
Finding 4
"The PCB inventory for Hot Spot 28 calculated from the low resolution coring data is
considerably greater than previous estimates. This apparent "gain" in inventory is attributed to
significant underestimates in previous studies rather than actual deposition of PCBs in Hot Spot
28." [Page ES-4]
State Comment
The State agrees with this finding based on the data contained in the LRSCR. This inaccuracy in
past data gathering efforts may also be present in the PCB inventory estimates in other areas
where the core depths were not sufficient in the past. However, NYSDEC believes the USEPA
evaluation of sediment PCB inventory gain or loss is valid, and not impacted by the earlier data
gathering efforts.
The State also has the following specific comments regarding two other fmdings of the LRSCR:
1. Page ES-5 and Section 4.1.4 second paragraph — The finding that areas within the Thompson
Island Pool (TIP), outside the known hot spot areas of the TIP, have exhibited a large net gain in
PCB inventory (up to a 100% increase) is significant because the PCBs are more readily
available to fish and other biota.
2. Section 4.4.3 The revised, sediment PCB concentration estimates for the near shore areas are
noteworthy. This portion of the river environment has not been well characterized in past
investigations, and this information will be useful to both the ecological and human health risk
assessments for the site.
The following are the State's comments , including the NYSDOH, on the Phase 2 Human
Health Risk Assessment Scope of Work:
1. The first sentence of the first full paragraph of page 9 refers to a hypothetical study population
being defined as any individual who would consume self-caught fish from the Hudson River "in
Page 2.
-------
the absence of a fishing bail." This passage should be revised for accuracy to read "...in the
absence of a fish possession ban and health advisory."
2. The number of years that a person may eat contaminated fish from the Hudson River is estimated
in Section II,2.D entitled "Risk Characterization from the Consumption of Fish." Data on how
long people live in a county along the river before moving are used to estimate the number of
years a person may eat contaminated fish. A significant number of people are likely to move
from one county along the river to another county along the river, thus increasing their length of
exposure. The number of years that a person may eat contaminated fish from the Hudson River
will be underestimated if this possibility is not considered in estimating exposure. Furthermore,
a lifetime exposure should be considered in the exposure distribution.
3. In evaluating risks, both cancer and non-cancer, the reference dose or cancer potency factor for
the Arocior (e.g. Aroclor 1016, Aroclor 1260, etc.) that is most similar to the PCB mixture in the
environmental samples should be used. This approach is more scientifically defensible than
automatically using default values as suggested in the Integrated Risk Information System
guidance.
4. Non-cancer risks are evaluated by comparing exposures to reference doses (ingestion exposure)
or reference concentrations (inhalation exposure).Since reference concentrations are not
available for the Aroclors, inhalation exposures should be evaluated using reference doses. The
risk characterization section of a risk assessment includes a discussion of the uncertainties and
limitations of the risk assessment and the uncertainties and limitations, if any, of using reference
doses instead of reference concentrations should be included in that section.
As additional information becomes available to the parties, the State would welcome the
opportunity to provide comments. The State views the completion of the LRSCR and the Risk
Assessment Scope of Work as important Hudson River Reassessment milestones, and is pleased that
USEPA is adhering to its Reassessment schedule.
cc: John Davis, NYSDOL
Robert Montione, NYSDOH
Jay Fields, NOAA
Lisa Rosman, NOAA
Anne Secord, USF&WS
William T. Ports
Remedial Section A
Bureau of Central Remedial Action
Page 3.
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
Local
(LRC - LL)
-------
LL-1
SARATOGA COUNTY
ENVIRONMENTAL MANAGEMENT COUNCIL
PETER BALET
CHAIRMAN
GEORGE HODGSON
DIRECTOR
August 28, 1998
Mr. Douglas Tomchuk, Project Manager
Hudson River PCB Reassessment
USEPA Region 2
290 Broadway, 20th Floor
New York, N.Y. 10007-1866
Dear Mr. Tomchuk:
Enclosed you will find comments from the Saratoga County Environmental Management
Council relative to Hudson River PCB Reassessment Phase 2, "Volume 2C-A Low
Resolution Sediment Coring Report" dated July 1998.
As you will note, many of the enclosed comments reinforce concerns stated by some
members of USEPA's Scientific & Technical Committee which met August 18, 1998 in
Latham, NY that this report does not adequately substantiate the "alarming" statements
recently made by EPA regarding the possible need for immediate remedial action in the
Upper Hudson River PCB hot spot areas.
The Saratoga County Environmental Management Council, consistent with our July 27,
1998 correspondence to USEPA ERRD Deputy Director McCabe, strongly recommends
that the enclosed comments, as well as ail other pertinent scientific information and public
comments relative to the Hudson River PCB Reassessment, be subject to the ERG
scientific peer review process which is currently underway. The Hudson River is too
important a resource not to conduct a comprehensive scientific peer review to provide a
sound basis for reaching the important decisions which lie ahead!
Thank you for the opportunity to comment.
Attn: LRC Comments
PB/gh
Chairman
-------
cc: Ms. Carol Browner, Administrator, USEPA
Ms. Jeanne Fox, Regional Administrator, Region 2, USEPA
Mr. Richard Caspe, Director, ERRD, Region 2, USEPA
Mr. William McCabe, Deputy Director, ERRD, Region 2, USEPA
Ms. Ann Rychlenski, Public Affairs Specialist, Region 2, USEPA
The Honorable Gerald Solomon
The Honorable Alphonse D'Amato
The Honorable Daniel Moynihan
The Honorable George Pataki
Mr. John Cahill, Commissioner, NYSDEC
Mr. Stuart Buchanan, Region 5 Director, NYSDEC
The Saratoga County Board of Supervisors
Mr. David Wickerham, Administrator, Saratoga County
Hudson River PCB Liaison Group Chairs
SCEMC members & staff
-------
COMMENTS ON VOLUME 2C-A LOW RESOLUTION SEDIMENT CORING
REPORT, JULY 1998;
Prepared for the Saratoga County Environmental Management Council
By D. D. Adams, Member-at-Large
1. P. 1-6 & P. 1-7: The 1988 NYSDEC report concluded there was no major change in
PCB distribution in the Thompson Island Pool (TIP) sediments from 1977 to 1984 u
despite the 1988 report showing a PCB inventory less than half of the 1977 estimate. The
1988 report concluded that "most of the differences between the 1977 & 1984 PCB mass
estimates were due to differences in calculation methods and assumptions". This EPA
report in later sections cites an inventory decrease of 30% from the estimated 1984
inventory when compared to the 1994 inventory estimated by EPA. EPA should either
provide its explanation as to why the 1988 NYSDEC conclusion of "no change" is invalid
or provide EPA's rationale for why the Hudson River behaved differently from 1977 to
1984 vs. 1984 to 1994, causing no change over the first period and a decrease over the
second period.
2. P. 1-7 & P. 2-1: Before any conclusions can be drawn from the low resolution coring
(LRC) data, a statistical analysis must be done which predicts the ability of the small \ 2
number of samples taken by EPA in only a few locations to adequately estimate the PCB
mass over a large area with very large spatial variations in PCB concentrations. This
analysis should include consideration of the fact that the mean values from the NYSDEC
data are based upon an order of magnitude greater number of samples than the EPA data.
Page 1-7 states the LRC program wasn't designed to duplicate the extensive spatial
coverage of the NYSDEC program. Page 2-1 dramatically quantifies this statement by 1.3
showing that EPA had only 60 core sampling sites in the TIP in the vicinity of the 1984
NYSDEC locations vs 1200 NYSDEC core sites. The meager number of EPA cores
hardly seems sufficient to quantify the PCB inventory (and therefore changes to the
inventory as put forth later in the report) in the face if existing data which show very large
spatial variation in PCB concentrations over very short distances in the TIP. EPA's data
also shows this variation as can be seen in Plate 4-23 which indicates a 100:1 difference in
concentration in nearby samples and variations of 10:1 in most, if not all, of Plates 4-21 to
4-28. Certainly, the scope of the EPA sampling program is not sufficient to justify the
"alarming" statements made by EPA to the public which accompanied the release of the
report and called for a study of immediate remedial action.
3. P. 2-3: What is the area (in square feet, acres, or square mile) that the cores below the 14
Thompson Island dam and the near shore areas represent?
-------
4, P. 2-10: What is the significance to the modeling effort of the loss of
reduction/oxidation potential data and total carbon/total nitrogen data?
5. P. 2-15: The first paragraph on this page states that cores taken for the LRC report "do
not comprise a spatial coverage sufficient to calculate PCB inventories for these areas
directly", yet later in the report these data are used to do precisely that since calculation of
a "change" in inventory implicitly carries with it the calculation of the inventory itself.
This inconsistency reinforces comment "1."
6. P. 2-16: The statement that the presence of PCB maxima in the top-most core layer
shows that PCB burial is not occurring is not justified. High resolution core profiles have
shown relatively low PCB concentrations in the first few inches of sediment with the PCB
concentration then rising rapidly to a peak before declining to a low level or zero.
Inspection of Fig. 2-7 shows that most of the PCB mass in both of these cores would
occur within the top 9 inches (about 23 cm) of the cores but yet relatively low PCB
concentrations within 2 plus inches of the sediment-water interface. Burial may or may
not be occurring (the presence of "hot spots" would seem to indicate it is) but reasoning
from the presence of a maximum PCB concentration in an upper 9" core segment cannot
be used to make such a conclusion.
7. P. 2-18: Reference in the last complete paragraph should be to Fig. 2-6 and not Fig.
2-4. Also, the units on the abscissa of the upper figure in Fig. 2-6 appear to be incorrect.
P. 2-18: An average difference of 36% between replicate samples from the same core
sample seems very large. Discussion of why this large difference is unimportant should be
provided, especially when the data are later used to estimate a change of only 30%,
8. P. 2-19: The discussion here at the end of Section 2.4.1 again raises the question of
the adequacy of the LRC data set to draw conclusions about PCB inventory change. The
difference of 4:1 in one of the outlier replicate samples and the statement that
heterogeneity in concentration as well as the ability to homogenize samples will probably
be the main source of analytical uncertainty for the PCB results reinforces the concern
about whether the number of LRC samples is sufficient to make the conclusions later
stated in this report and in the Executive Summary of the report.
9. P. 3-13, Section 3.2: If the variation of parameters was expected to occur over a
narrow range, why weren't these parameters measured on the high resolution cores?
While limited in number, these cores could have provided some data to show if the
assumption of variation over a narrow range is true.
-------
3
10. P. 4-5, top of page: Discussion in Appendix E raises serious questions as to the
ability to make a valid comparison of 1984 and 1994 data due to the differences in j j2
analytical methods. Rather than go through manipulations in Appendix E, 1994 samples
should be analyzed by 1984 methods (or as close as possible to 1984) and a correction
factor utilized based upon these results obtained for the 1984 data. If possible, the reverse
should also be done (analyze 1984 samples using 1994 methods) and the correction
factors compared.
II P. 4-7: I do not agree with the statements presented in the last paragraph. Referring
back to comment 6 , if the top 9 inches and bottom 9 inches of Core 19 (Fig. 4-2) were ^
homogenized, (Note: units on ordinate of Fig. 4-2 should be cm and not inches), the
results would be expected to be similar to Core LR-09E and not Core LR-05D. This
re-emphasizes the point made in comment 6. that the LRC core profiles do not provide a
basis forjudging burial or no burial. The homogenized profiles (using same sample layer
thickness as LRC cores) of all the high resolution cores should be calculated and plotted
and compared to the LRC cores.
12. P 4-8, First Paragraph: To conclude that 8 profiles show scour should be
substantiated by a table showing the separation distances between these 8 profiles and the j 14
1984 profiles to which they are compared. To assist in the overall review of this report, a
table should be included which shows the separation distances between each LRC core
and the 1984 core to which it is being compared. This information is important
considering that available data indicate significant PCB concentration changes occur over
distances of only a few feet.
13. P 4-10: In the discussion of Fig. 4-7, the references to the upper and lower figures of \ 15
Fig. 4-7 should be reversed. Also, are the units on the upper figure correct'7
14. P. 4-18, Last Paragraph of 4.1.3: The implication that more scour occurs in cohesive
sediments than in non-cohesive sediments seems to be at odds with what would be j ^
expected. How does EPA reconcile this other than uncertainty in the data (which might
well explain other differences, put forth by EPA in this report, between previous data and
1994 data)? If EPA believes scour is occurring, does EPA plan to revise the scour model
included in the DEIR which predicted little if any scour9
15 P 4-21: The discussion on P. 4-21 about corrections needed for the grab samples is j j.
another illustration of uncertainty in the EPA comparison of 1994 data to earlier data.
There is no way of knowing if the extrapolation to 12"is valid or not. It is suggested the
-------
4
grab samples not be included in the analysis. Also, using the concentration of the second
layer of the 1994 LRC samples to extend 9" cores to 12" is questioned. Based upon the
sharp PCB concentration gradient in the range of 9-12" seen in the high resolution cores,
this method of extending from 9" to 12" may greatly underestimate the LWA in the 1994
data.
16. P. 4-22: When choosing between two data sets for the same set of samples based on
"convenience" rather than an understanding of reason for the differences is unsound
science and casts doubt on any conclusions from the analysis. EPA should investigate the
reasons for the differences and if they are unable to justify a choice based upon this
investigation, perform the analysis using both sets of data and compare the results.
17. P. 4-22: .Any comparison of PCB inventory based upon the 1976-78 and 1994 data is
made uncertain by the lack of solid specific weight (SSW) data for the 1976-78 data. The 1.19
statement here that total PCBs and SSW show very strong correlation is in contradiction
to the statement on P. 3-18 that SSW showed a weak trend with PCB concentration.
Also the range of variation in the data in Fig. 3-15 and 4-17 is very large making
conclusions about correlation's uncertain and no 2c or 3a values are given for the SSW
to PCB values shown in Table 4-3. EPA needs to do more to show why the SSW values
used for the 1976-78 data are reasonable to use, including a level of uncertainty, to make
comparisons of PCB inventories between 1976-78 and 1994 meaningful.
18. P. 4-24: Fig. 4-18 does not substantiate the statement that the agreement between the
NYSDEC and side-scan sonar classifications of sediment is good. In the "fine sand" bin,
which is the largest and most important bin for PCB accumulation, the agreement is less
than 50% and no discussion is provided about this difference.
19. P. 4-25: References to Fig. 3-28 should be to Fig. 3-27 and to USEPA results and
not to NYSDEC. Also, comparison of Fig. 3-27 and Fig. 4-18 does not appear to back up
the statement that division between fine and coarse-grained sediment being consistent
between NYSDEC data and EPA data. Fig.3-27 shows a high percentage of fine-grained
samples for the EPA data vs. 45% for NYSDEC data.
20. P. 4-26: The paragraph at the bottom of the page highlights another uncertainty in
comparing 1994 inventory data, namely the uncertainty in the boundaries of the hot spots.
EPA needs to gather all the uncertainties in all areas of the report's analysis and provide
an overall error estimate and then evaluate whether any estimate can be made of PCB loss
or gain with any degree of confidence.
-------
5
21. P. 4-27: References should be to Figs. 4-19 and 4-20 and not to Figs. 4.2-3 and 1-23
4.2-4.
22. P. 4-33: The conclusion at the bottom of this page regarding loss of PCBs requires
that the 1994 data set accurately replicates the 1976-78 data set. For reasons set forth in
other comments in this report, conclusions such as set forth here are not justified by the
analysis presented to date as set forth in this report.
23. P. 4-34: The statement at the bottom of page 4-34 and at the top of page 4-35 that
Profile 2 of Fig. 4-25 shows burial is not occurring, completely ignores the profiles shown
by Fig. 4-24 of the high resolution cores definitely which clearly show very low PCB
concentrations in the top several inches of sediment, definitely indicating burial. The
"peak" in the upper 12 " of Profile 2 is undoubtedly due to the mixing of the very high
concentrations of PCBs buried several inches below the surface with the low
concentration sediments near the surface. The low resolution core profiles provide na
basis for making any conclusion about whether burial is or is not occurring and statements
such as "'strong PCB profile evidence for long-term sediment loss or lack of burial" should
be deleted from a revised issue of this report.
24. P. 4-35: Statements in the second and third paragraphs on this page seem to be 1.26
biased toward a conclusion that PCB loss is occurring. Why is it that "long-term storage
is clearly not assured" but long-term loss apparently is assured? Why is a PCB inventory
increase the result of bad data but all losses are absolutely true?
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
Community Interaction
Program (LRC - LC)
-------
LC-1
JOHN E. SANDERS
33 Sherman Avenue
Dobbs Ferry, NY 10522
31 August 1998
Via FAX transmission to 637-4284; original sent Priority Mail
Douglas Tomchuk
USEPA - Region 2
290 Broadway - 20th Floor
New York, NY 10007-1966
Attn: LRC Comments
Dear Doug,
I think that the main point about the low-resolution coring is that the results totally destroy the
basis for the model GE has been trying to ''sell" (i.e. "clean" sediments are covering "dirty"
sediments and thus solving the Hudson River PCB-pollution problem). Even GE admits that the
river picks up PCBs as it flows down'the Thompson Island Pool (TIP), but the key issue still j j
remains to what depth does the flowing water actively interact with the sediments and thus add to
or subtract from the sediment load of PCB's. GE may be correct that the deep-lying hot spots are
not part of the day-in-day-out game of "put and take" that the river plays with the bottom
sediments. The problem may be with all the sediments in the TIP, not merely the hot spots that
keep dominating the discussion. What a pity that my repeated suggestions about making relief
peels from the vertical faces of cores cut longitudinally (not wafered, as is needed for geochemical
analyses) have not attracted any enthusiasm. With a decent set of relief peels in hand, the details
of the laminae would show how the sediments and water are interacting. Despite the lack of
peeis. I think the LRC report provides a solid scientific basis for dispelling GE's myth about
"covering with a clean blanket." Too bad we are not loaded up with data about how deep down
into the dirty carpet the river keeps thrashing out all those PCB's that go downriver.
Respectfully submitted,
ohn E. Sanders, Ph. D
John E. Sanders, Ph. D.
Geologist
JES/s
-------
THIS PAGE LEFT BLANK INILMiuinal^i
-------
Department or Earth and Atmospheric Sciences
College of Arts and Sciences
Earth Science 35!
Albany. New York 12222
518/442-4466 or 4556
Fax: 518/442-5825 or 4468
Chair@atmos.aibany.edu
http://www.atmos.albany.edu
University at Albany
LC-2
STATE UNIVERSITY OF NEW YORK
August 29, 1998
Mr. Douglas Tomchuk
USEPA - Region 2
290 Broadway - 20th Floor
New York. NY 10007-1866
ATTN: LRC Comments
Dear Mr. Tomchuck:
In retrospect of the Science and Technical Committee (STC) meeting of August 18. I wish to
append a few comments directed at both the DEIR and LRC. For many STC members, the Hudson
River reevaiuation is not a principal activity and a thorough review of each report in the sequence
begs both a revisit to previous work as well as a consideration of current work and comments not
included in the report. For this the STC meeting is invaluable, but in my case, a further digestion is
necessary.
1) As I noted at the meetine, an evaluation of all contributors to variation is necessarv in
comparing the 1984 and 1994 sediment cores in order to test or verify any conclusions. —1
If the General Electric Co. (GE) is prepared to run the different GCMS analysis methods
for PCB on the same samples this can be accomplished, but 1 recommend that an
independent statistician be engaged to do the required analysis of variance.
2) Variance estimates for various data used in the DEIR. such as water column sediment
and PCB loadings, are needed in order to derive confidence limits for model predictions. 2.1
Since such estimates are not cited in the DEIR, they may have to be obtained from other
data sources if a direct estimation cannot be made.
3) Water column PCB and sediment concentration data used in the DEIR are regressed on
River discharge or flow for the purposes of model calibration. As I have previously
commented, this assumption calibration introduces an added variance since neither of
these quantities has a simple relationship to River flow. Further, the Hydro Qual, Inc.
study of the Thompson Island pool (TIP) (1997) has pointed out the importance of Moses
Kill and Snook Kill discharges to annual sediment in the TIP deposition; a factor not
addressed in the DEIR.
4) Water column PCB data, and interpretations thereof, used in the DEIR need a thorough
overhaul before any meaningful model calibration is possible. The point I have raised
about data feedback is particularly pertinent here for it is clear from an inspection of the
1990-1998 data furnished to the NYSDEC by the G.E. Co., that the 1994 EPA sampling
data used for the DEIR is not representation. There are several reasons for this, viz:
10.0999
-------
a) 1994 water sampling was conducted without considering the variability of water
column PCB content in space and time at the sample point. This omission is
particularly acute at the Rogers Island station where the PCB mass flux into the TIP
pool is to be determined. The problem was initially pointed out by Tofflemeier
(1984), and further evaluated by O'Brien and Gere (1993) via cross channel and
depth sampling, and dye release tests. In this regard, most of the 1990-1998 data was
obtained from composited water column samples from both channels at Rogers
Island, the bulk being taken on a regular (weekly) basis.
b) The 1994 sampling occurred during a period of documented DNAPL containing PCB 2.3
released from workings and by seepage above the TIP. These releases temporarily
reversed the previous trend of declining annual PCB loading at low flow in title TIP;
• however, this effect appears to have ended by 1996 with significantly lower PCB
contents subsequently being observed.
c) The bulk of Hudson River water column sampling has been during low to very
moderate flow conditions, and was not regularly conducted on an annual basis prior 2.4
to the 1990-1998 data. Limited observations during past high discharge events in
particular have shown that PCB contents are very variable, and that erratic transient
"pulses" of high PCB content may occur during both high and low flow. Because of
the lack of close time interval sampling during high discharge events, the PCB mass
flux of the event at Ford Edward (and event loading to the T.I. pool) cannot be
estimated. However, if the lower PCB contents observed for events in 1996-1998
(Post DNAPL; including a 50-year event, Jan. 8-12, 1998) are confirmed as the norm,
then the problem of model calibration becomes much simpler.
5) PCB mass loading of, and discharge from the TIP are estimated critical to an independent
evaluation of the 1984-1994 low-resolution core comparisons, and to a proper 2.5
interpretation of the operative mechanisms thereto. The lack of data on PCB loadings in
high discharge events presents a major difficulty because it appears no distinction can be
made between partially dechlorinated PCB released from recently deposited sediment
(i.e. <3 years old) and that presumed to be released from older "hot spot"' reservoirs.
While seasonal (water temperature viscosity) and river discharge (less dilution at low flow?)
effects in water column PCB may be inferred from the 1990-1998 data, the rate of decrease in 2.6
concentrations shown in 1996-1998 (including non-detectable levels at the TIP dam, January 6, and
27; December 29, 1997) is inconsistent with an inference that the bulk of PCB loading, 1991-1995,
at the TIP dam was by diffusional loss from hot spot sediments deposited prior to the 1976-1978
core sampling. With the apparent "shut-off of PCBs entering the River above Fort Edward, we can
now, possibly for the first time, examine conditions and processes in the TIP without this
complication.
Yew truly yours,
George W, Putman. Ph.D.
Emeritus Facultv
cc: R. Sloan, NYSDEC
W. Nicholson, STC
J. Haggard, GE
-------
LC
August 30, 1998
Douglas Tomch.uk
USEPA - Region 2
Attn: LRC Comments
290 Broadway - 20th Floor
New York, NY 10007-1866
Re: Copy of a Letter to Administrator Browner offered as Comment to Low Resolution
Coring Report
to: Carol Browner. EPA Administrator
Washington. DC
Dear Ms, Browner:
As chairman of the Agricultural Liaison Group in the Hudson River PCB
Reassessment Process. I have been active in this process for the past 8 years. After
meeting on August 11th, our Liaison Group members have agreed that I should express
our dissatisfaction with the Low Resolution Coring Report and the Community Interaction
Program process directly to you.
We must object to suggesting that the results of this report indicate a "crisis"
situation that may require accelerating the process. This is ridiculous. The data used is
already four years oid. Why is it only now becoming a crisis? The methodology of the
Core Report is quite controversial and certainly should be peer reviewed. I understand
that even the Science and Technical Committee was critical of this report. A number of
broad assumptions are made in this while completely ignoring the fact that water column
and fish studies continue to show a decrease in PCB levels in the Hudson. How can we
possibly declare the river in a "crisis" situation?
There is no emergency! Fish and water column measurements should obviously be
considered more accurate than the sediment studies in determining the immediate risks
associated with the river. Calling this an emergency should not be used as an excuse to
ignore peer review!
As one of the Community Interaction Groups participating in this process from the
start, we were told that we would be part of the process for determining the remediation
of the Hudson River. We would be made aware of new information and allowed to
comment.
We joined in this process in good faith but have become alarmed at how this
process has been handled in the last few years. We learn more about EPA activities on
this issue in the newspaper than we do at EPA sponsored meetings. It doesn't appear that
we have any influence on this process. Many of our comments go unanswered.
-------
The release of the last two reports has again been a media event. You have
publicly stated that PCBs are •likely" human carcinogens. Those of us who have followed
Jiis issue know this is not true. Yes, there is some scientific disagreement but the large
lumber of human health studies done certainly indicate that PCBs are NOT carcinogenic.
We are appalled that such statements are made by EPA officials before the peer review
process has been completed. Early in the process we were promised peer review of the
science involved in this issue. These reckless statements certainly point out the continued
leed for the review process. Conclusions drawn before peer review is completed should
aot be material for news releases.
This was supposed to be an open and public process. You even told Congress that
public input is important. Well, this is public input from a group that has stayed with this
process for eight years. We are frustrated and disgusted with secret landfill siting surveys
and news releases promoting reckless conclusions drawn from reports that no one else has
seen or reviewed.
Our active participation in this long process should warrant us a response to these
comments. EPA handling of this process seems to be much more political than scientific
and we are frankly not going to sit by quietly as if we have approved of this process!
Sincerely,
Thomas A. Borden, Chairman
Agricultural Liaison Group
2841 Valley Falls Rd.
Scahaghticoke. NY 12154
(518) 753-4341
-------
Merriiyn Pulver
Councilwoman.Town of Fort Edward
R.D.1, Box 222 ^ „
Fort Edward, N.Y. 12828 LC-4
747-4985
August 31, 1998
Mr. Douglas Tomchuk
U.S. EPA — Region 2
290 Broadway — 20th Floor
New York, NY 10007-1866
Attn: Low Resolution Coring Report Comments
Dear Mr. Tomchuk:
Below are my comments regarding the U.S. Environmental Protection Agency's (EPA) Low
Resolution Coring Report iLRCR;. dated August 1998. which was presented by TAMS. Inc. at
the July 23, 1998, meeting of the Hudson River Joint Liaison Committees. I respectfully
request that these comments be made available to any and all interested parties and made part
of the administrative record for the Hudson River Superfund site.
I. First and foremost, there is no emergency.
I must stress that the people of the Upper Hudson River in no way believe that there is an
"emergency" situation occurring in the Thompson Island Pool.
I have been instrumentally involved in this issue for more than 20 years, during which time I
have attended numerous meetings, read thousands of pages of documents and met with many
members of the concerned and interested public. Never have 1 heard mention of an "emer-
gency" situation in the river.
it's reuiiy no wonder. By every measure, the Hudson River is better than it has been in
decades. Fish levels are declining dramatically; PCB levels in water are also lower than ever
before. These improvements certainly are no secret to the public at large. Tourism dollars are
once again flooding into communities that lie on the Hudson's shores. Recreational fishermen
are traveling from far and wide to fish the Hudson's trophy fishery. New touring boat compa- •
nies are sprouting up, taking travelers on trips to view the scenic vistas. And I'm sure you are
aware that the bald eagle, once considered an endangered species, is now returning and pro-
creating on the banks of the Hudson. The bald eagle is only one example of the health and
vitality of the Hudson River's wildlife.
This does not sound like an emergency situation. Contrary to Regional Administrator Fox's
comments, it doesn't sound "startling" either.
-------
If, as EPA concluded from the LRCR, there has been a net loss of 40% of the PCB inventory
in the hot spots from 1984 to 1994, we would certainly be seeing it in the fish, the water and
the wildlife. This, as stated earlier, just is not the case. Fish and water levels are not increas-
ing; they are decreasing. And, the PCBs in fish do not look like the PCBs buried in the hot
spots; they look more like the PCBs that recently entered the Hudson from GE's Hudson Falls
plant site. Finally, to my knowledge, no mechanism has been identified that would drive this
much material from the hot spots into the water column. As such, EPA should have gone back
to the drawing board to see where they had erred, instead of needlessly scaring the public with
claims of an "emergency" situation in the Thompson Island Pool.
II. Politics is prevailing over good science.
Last year. I discovered that EPA had conducted, unbeknownst to the public, a siting study to
identify potential sites in the upper river for a PCB landfill. Need I remind the Agency of the
public uproar that occurred as a result? In response. EPA revised its schedule for the Hudson
River Reassessment to allow for greater public participation, responsiveness summaries and
peer review (see below).
The lower river environmental groups, who have been pushing for a destructive dredging and
landfilling project for years, strongly criticized the Agency for the schedule revisions. Scenic
Hudson and Sloop Clearwater met with the press and mailed hundreds of letters to lower river
elected officials in an attempt to persuade EPA to reverse its new schedule. Initially, and
thankfully, I thought their efforts had faiied. After all. who could argue with greater public
participation and independent, scientific oversight on a matter of such grave importance?
Now, however, we hear that the Agency is most likely going to proceed with a remedial action
this fall — read, dredging and landfilling. Looks like the lower river environmental groups
won after all. I can't help but wonder if this call to action is some misguided attempt by the
Agency to appear hard-hitting and aggressive on the Hudson River, in essence, to satisfy the
Agency's political agenda.
Throughout the years I've been participating in this process, I have always been reassured that
good science would prevail. Today, I have serious doubts.
III. The Low Resolution Coring Report fails to tell us anything new about the
Hudson River.
There are obvious limitations to the analysis EPA conducted in its LRCR.
A) EPA collected only 60 core samples to characterize what is happening in the Thompson
Island Pool hot spots. That's approximately one sample every three acres! In its 1984 effort,
the New York State Department of Environmental Conservation (DEC) collected more than
400 — more than six times as many. Conditions in the Thompson Island Pool vary greatly,
-------
even within the hot spots. Sixty samples is nowhere near the appropriate number of samples
for this type of evaluation.
B) In 1984, because of limitations in technology, the DEC evaluated core samples for PCBs
with three or more chlorines. Today, laboratory detection methods have improved significant-
ly, allowing analysis of all, or total, PCBs. In the report, EPA discussed the importance of
correcting its methodology to ensure that the same factor (PCBs with three or more chlorines)
was being measured in 1984 and 1994. Yet, the analysis in 1994 included all PCBs, including
those with one or two chlorines. Essentially, EPA conducted an apples-to-oranges compari-
son, where PCBs with three or more chlorines are the apples and total PCBs are the oranges.
At the recent meeting of the EPA Science and Technical Committee, Ed Garvey of TAMS,
Inc., acknowledged that a comparison of PCBs with three or more chlorines in 1984 to PCBs
with three or more chlorines in 1994 could lead to the conclusion that 73% of the hot spot
PCBs have been lost. Now this is a "startling" discovery — one that everyone knows has not
occurred. Therefore, this analysis must be flawed.
C) EPA analyzed its cores by slicing them into 9-inch slices. As a result, the Agency con-
cluded that the highest PCB levels are within the top 9 inches of sediment and implied that the
highest PCB concentrations are now available to fish and other wildlife. From my understand-
ing of the river, fish do not burrow down deep into the river sediment to eat. Only the PCBs
in the very top inches of sediment are available to the fish. Therefore, to conclude that the
highest PCB concentrations are within the top 9 inches of sediment is not sufficient. EPA
must continue its analysis to discover more specifically where, within the top 9 inches, the
highest levels of PCBs reside.
IV. The Low Resolution Coring Report must be peer reviewed prior to being used as
justification for a remedial action.
EPA w isely chose to incorporate scientific peer review into its Hudson River Reassessment.
This process will be instrumental in providing the public with an independent, technical
review of the science behind EPA's reports. I fully support this development, although I do
continue to urge the Agency to provide peer reviewers with relevant critiques, prepared by GE
and the public, of the reports under-evaluation.
That being said, it is alarming that the Agency is planning to sidestep this critical process.
You announced that EPA has begun a "'quick-term study" to evaluate if, based on the results of
the LRCR. an immediate remedial action is required in the Thompson Island Pool.
As stated earlier, the report has several deficiencies — some of which are criticai to the
Agency's mam conclusions. To depend upon this report, without first peer reviewing it, is
nonsense and does not withstand the "laugh test." EPA must continue with the process it
committed itself to — first, peer review. Then, if necessary, remedial action. Anything else is
foolhardy, rash and not scientifically sound.
-------
V. Even if dredged spoils are not dumped locally, dredging isn't the appropriate answer for
the Hudson River.
Finally, you stated that, if EPA determines chat dredging is required in the Thompson Island Pool, only
"existing permitted landfills" would be considered. You said, "We would not look to site a landfill in
the Hudson Valley. We would only go to existing facilities." 4,9
DEC estimated that the Thompson Island Pool hot spots encompass approximately 1.3 million cubic
yards of material. To the best of my knowledge, that is much more material than any currently-
licensed facility could store. Therefore, EPA must already have an idea of how much material they are
planning to recommend for removal by dredging.
This is a conclusion I cannot, and will not. stand for. I need not remind the Agency that more than 60
village, town and city local governments, as well as chambers of commerce and local farm bureaus,
unanimously approved resolutions opposing dredging and dumping of Hudson River PCBs. The
Agency's attempt to placate the public by casting doubt on the need for a local dump is, unfortunately,
transparent. The people of Fort Edward will not suddentiy support a dredging project because dredg-
ing spoils will not be dumped on oar farmland. We wiil continue to oppose such a project because it is
not the correct answer for the Hudson River.
Sincerely,
oJtSu
Merrilyn Pulver
-------
General Eitctr"*
(LRC - LCi)
-------
CHICAGO
DALLAS
LOS ANGELES
Sidley & Austin
A PAlTNtllHIP INCLUDING NOH!IIO«H CO I POl A T 10 N s
1722 Eye Street, N.W.
Washington, D.C. 20006
Telephone 202 736 8000
Facsimile 202 736 87!1
Founded 1866
LG-1
NEW YORK
LONDON
SINGAPORE
TOKYO
VttTtl'S DlftlCT NUMBII
(202)736-S27I
August 31, 1998
Douglas Tomchuk
USEPA - Region 2
290 Broadway - 20th Floor
New York, NY 10007-1866
Re: LRC Comments
Dear Mr. Tomchuk:
We submit herewith the comments of the General Electric Company ("GE") on the Low
Resolution Sediment Coring Report which the U.S. Environmental Protection Agency released
for public comment in July, 1998.
Our Executive Summary in short form and our comments at greater length focus on the
deficiencies and weaknesses of the Report, in particular the fact that the very limited amount of
data collected in 1994 is insufficient to permit any useful or reliable extrapolation to conditions
in the so-called "hot spots" or the Thompson Island Pool generally. We will not repeat in this
letter the analysis set out in the comments.
This is the appropriate place to address the gist of the comments which we understand
that EPA made at its press conference at the time the Low Resolution Coring Report was
released. GE was excluded from hearing what EPA, through the press, had to say to the public,
but we understand that at least three points were made. EPA stated it was startled by the
conclusion TAMS had reached that there had been a net loss of approximately 40% of the PCB
inventory from the "hot spots" of the Thompson Island Pool between 1984 and 1994. EPA
analogized losses of this magnitude to discovering a leaking drum. Consequently, EPA was
considering whether some form of emergency action should be taken prior to completing the
Reassessment.
The image of the leaking drum is a powerful one. It summons to mind high
concentrations originating in the so-called "hot spots" in the Thompson Island Pool, moving
downstream out of the Pool and causing substantial PCB body burdens of fish and an
-------
Sidley & Austin
Washington. D.C.
Douglas Tomchuk
August 31, 1998
Page 2
unacceptable health risk. EPA has also conveyed the idea that conditions are worsening. The net
result is the belief that the situation is far more risky and threatening than EPA has believed it to
be at any time since the Reassessment began in 1990.
This vivid tale of imminent danger on the Hudson does not withstand scrutiny.
Let's start with the 40% net loss of PCBs from the so-called "hot spots" of theThompson
Island Pool. If the comparison of the 1984 PCB inventory in the Pool to the 1994 PCB inventory
in the Pool is carried out in the manner TAMS argues is correct, this number doubles to 80%.
The Report points out that in 1984 PCBs were measured as tri-chlorinated and higher
homologues; monos and dis were not measured. In 1994 total PCBs were measured. The
appropriate comparison between 1984 and 1994 is to compare the tri- and higher PCBs for each
year. Although TAMS argues persuasively that this is the appropriate method for comparison,
no such comparison is presented in the Report. In fact, at the August 1998 meeting of the
Science and Technical Committee, Dr. Edward Garvey, who presented the findings of the
Report, responded to a question by saying that he had never performed the comparison. GE
found that statement startling. GE performed the comparison of the 1984 PCB inventory to the
1994 inventory on a tri- and higher basis using the data TAMS had selected for comparison. GE
did this despite the fact that the company contends there is insufficient reliable data to produce
useful or statistically meaningful results. When the comparison is done as TAMS advocated, the
conclusion follows that there was a net loss of 80% of the PCB inventory from the TIP between
1984 and 1994.
This result is not startling; it's amazing.
Of course, in considering whether something should be done about this situation, one
presumes this course of events continued from 1994 to 1998 (if the course of events has stopped,
the analysis of the Report would provide no basis for action today). If 80% of the inventory was
lost in ten years, what has happened after another four years? If the loss was at the rate of 8% a
year, all the PCBs would have disappeared from the so-called "hot spots." If the losses were
declining in the manner of a half-life, something like 10% of the 1984 inventory would remain
and would be lost at a rate of 1 Vi% of the 1984 inventory a year. Both of these scenarios result in
a low rate of loss at the present time. Rather than increased risks, we would see low and
continually decreasing risks in the Upper Hudson. Put another way, if the so-called "hot spot"
deposits in the TIP have been scoured out since 1984 so that only somewhere between 0 and
10% of the inventory remains, there isn't anything left in the Thompson Island Pool that is worth
chasing. Any action allegedly addressing increased risk from PCBs in the sediments of the
Thompson Island Pool cannot rest on the analysis of this Report. It would have to rest on a
repudiation of this Report.
Let's sober up and be frank. Nothing remotely like an 80% loss of the 1984 PCB
inventory in the TIP has taken place. Since 1991 GE (and before that the U.S. Geological
-------
Sidley & Austin
Washington. D.C.
Douglas Tomchuk
August 31, 1998
Page 3
Survey) has had in place a water monitoring program that measures PCBs entering and leaving
the TIP. We know that it is fully capable of detecting substantial additions of PCBs to the water
column. That's how GE found the Allen Mill releases in 1991. We also know that if there has
been a bias in the reporting of PCBs leaving the TIP, as measured at the Thompson Island Dam
from 1991 to 1997, it has been a bias resulting in an overestimate of PCB loss from the TIP. If
there had been major losses out of the Pool, we would have known it. In reality, the losses were
probably significantly lower than what EPA and GE believed them to be from 1991 to 1997.
The next obvious point is that when you go out and look for the so-called "hot spots" they
are still there. GE did a limited coring program this summer reoccupying a number of sites
where EPA had taken cores in 1994 and NYSDEC had taken cores in 1984. If there had been a
loss of 80% of the PCB inventory from 1984 to 1994 that continued to 1998, there would, of
course, be virtually no PCB inventory left. GE found the PCBs where NYSDEC found them in
1984 and EPA found them in 1994.
The real problem here is the one that GE warned EPA of in 1992 and which the Report
instinctively recognized in its examination of the so-called "hot spots" below the TIP. In a large
and highly heterogeneous mass one cannot extrapolate from a very limited number of data points
to calculate the amount of one constituent with any useful degree of reliability. Nevertheless, the
Report made such extrapolations. Above the Thompson Island Dam, the extrapolation produced
results showing substantial losses of PCBs; below the Dam, the extrapolation produced results
showing substantial gains of PCBs in some of the so-called "hot spots," particularly "hot spot"
28. The TAMS Report elected to believe the results of substantial loss and dismiss the
substantial gain in PCB inventory in "hot spot" 28. "The PCB inventory for Hot Spot 28
calculated from the low resolution coring data is considerably greater than previous estimates.
This apparent 'gain' in inventory is attributed to significant underestimates in previous studies
rather than actual deposition of PCBs in Hot Spot 28." The reality is that the results above the
Dam are just as implausible as those below the Dam and should be disregarded.
We look forward to discussing this matter with you when you have had the opportunity to
consider our comments and to soberly and prudently weigh the reliability and plausibility of this
Report against the mass of data and evidence on the fate and transport of PCBs in the Upper
Hudson collected by GE and the federal and state governments over many, many years.
We are confident you will reach the conclusion that Region 2's Director of the
Emergency and Remedial Response Division did in his cover letter to the Report:
As with the previous Phase 2 Reports, it is important to recognize that the conclusions in
this report, although significant, do not indicate whether or not remedial action is
necessary for the PCB-contaminated sediments of the upper Hudson. The numerical
analysis (computer modeling) of fate and transport of PCBs, the associated ecological and
-------
Sidle y & Austin
Washington. D.C.
Douglas Tomchuk
August 31, 1998
Page 4
human health risk assessments, and a feasibility study must be completed before any such
conclusion can be reached.
If you should feel that there is any basis, relying on this Report, to take action on the
Hudson prior to completion of the Reassessment, it is essential to conduct the peer review to
which EPA is committed before action is taken. The ultimate point of the peer review is to
assure sound science in decision making. That requires peer review before the decision is made.
We are interested in what the peer reviewers will make of the Report's advocacy of comparing
the 1984 inventory to the 1994 inventory on the basis of tri- and higher PCBs and its unexplained
omission of that comparison from the Report.
cc: Richard Caspe
William McCabe
Melvin Hauptman
Douglas Fischer, Esq.
John Cahill
Frank Bifera, Esq.
Very truly yours,
Angus Macbeth
Albert DiBernardo
Edward Garvey
Jonathan Butcher
Victor Bierman
Walter Demick
Jay Field
Anton Giedt, Esq.
AJD
-------
COMMENTS OF THE GENERAL ELECTRIC COMPANY ON
Phase 2 Report - Review Copy
Further Site Characterization and Analysis
Volume 2C Low Resolution Sediment Coring Report
Addendum to the Data Evaluation and Interpretation Report
July 1998
August 31, 1998 -
Melvin B. Schweiger
John G. Haggard
General Electric Company
Corporate Environmental Programs
1 Computer Drive South
Albany, NY 12205
518.458.6648
John P. Connolly, Ph.D., P.E.
James R. Rhea, Ph.D.
Quantitative Environmental Analysis, LLC
305 West Grand Avenue
Montvale. NJ 07645
201.930.9890
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
TABLE OF CONTENTS
Page
SECTION I. INTRODUCTION AND EXECUTIVE SUMMARY 1
A. Introduction 1
B. Executive summary 2
B.l Limitations of the data preclude the generation of useful results. 3
B.2 Implausible results are contradicted by relevant evidence. 6
SECTION II. ACCURATE ESTIMATION OF LOSS OF PCBS FROM
THE TIP SEDIMENT IS NOT POSSIBLE USING THE LOW
RESOLUTION CORE DATA 13
A. The PCB concentration measurements have substantial error. 13
B. The full vertical extent of PCBs was not always captured in the 1994 cores. 14
C. Grab samples taken in 1984 cannot be compared to core samples taken in 1994. 15
D. The same sediment was not sampled in 1984 and 1994. 15
E. Changes observed at sampled locations cannot be extrapolated to other locations. 18
SECTION III. THE METHODS EMPLOYED CONTAIN ERRORS AND
OMISSIONS THAT MASK THE IMPLAUSIBIL1TY OF TAMS'
MASS LOSS ESTIMATE 20
A. The sum of PCBs with three or more chlorines (1984) was compared to the
sum of all PCBs (1994). 20
B. The geometric mean was used to estimate average mass loss. 23
C. A bias was introduced by eliminating locations where the PCB inventory in 1984 was
less than 10 g/m:. 24
-------
SECTION IV. THE CORRECTED ESTIMATE OF MASS LOSS IS
IMPLAUSIBLE 26
A. TAMS' approach results in 80% loss of PCB>«.. 26
B. No fate and transport mechanism can account for the mass loss estimated using TAMS'
method. 28
C. There is no evidence of TAMS'mass loss in water column PCB data. 31
D. The loss rate indicated by comparing PCBj+ in 1984 and 1994 is not supported by the
change in PCBj+ mass between 1994 and 1998. 32
D.l Overview of the 1998 Sediment Coring Program 32
D.2 Temporal Changes in PCBJ+ between 1984, 1994, and 1998 33
E. If the loss rate indicated by comparing PCBJ+ is real, little of the PCBs that are
bioaccumulated remain in the Thompson Island Pool. 34
SECTION V. WIDESPREAD BURIAL OF PCB-CONTAMINATED
SEDIMENT BY CLEAN SEDIMENT DOES OCCUR IN THE
THOMPSON ISLAND POOL 36
A. Detected 7Be indicated deposition ranging from 0.2 to 3.0 cm/yr. Non-detect 7Be
indicated deposition less than 0.5 cm/yr. 36
B. Cores collected in 1998 consistently exhibit a buried peak PCB concentration with
a decline at the sediment core surface, a pattern consistent with burial. 40
C. Data analysis and modeling of the 1994 spring flood indicates widespread net
deposition in fine-grained sediments. 42
D. Sediment transport modeling indicates that the areas with non-detectable 7Be are
depositional, but the deposition rate is less than that in areas with detectable Be. 43
SECTION VI. EXISTING DATA SHOW NO EVIDENCE OF EXPOSURE OF
PREVIOUSLY BURIED PCBS VIA EROSION 44
A. PCBs increase in a nearly linear fashion as water passes through the TIP, indicating
a nearly uniform area! flux from sediments within the Pool. 44
B. The composition of the TIP load is consistent with the surface sediment PCB
composition considering equilibrium partitioning and sediment pore water
exchange processes. 46
C. The composition of PCBs in fish is consistent with exposure to relatively undechlorinated
PCBs found in surface sediments and not the dechlorinated PCBs found in buried
sediments. 47
D.
The PCB composition in the water column during erosion events is consistent with the
relatively undechlorinated PCBs found in surface sediments and not the dechlorinated
PCBs found in the buried sediments. 48
-------
E. l37Cs levels in the low resolution cores give no indication that scour sufficient to account
Tor a significant loss of PCB mass has occurred. 48
SECTION VII. ACCURATE ESTIMATION OF THE CHANGE IN PCB MASS
IN "HOT SPOTS" BELOW THE TIP IS NOT POSSIBLE
USING THE AVAILABLE DATA 50
A. None of the mean mass changes is statistically different from no change. SO
SECTION VIII. DECHLORINATION ANALYSES ARE FLAWED AND .
INSENSITIVE 52
A. Dechlorination occurs at meta and para positions only. 52
B. The mean mass loss (of PCBs on a mass basis) is less than 10% assuming Aroclor 1242
was the original mixture of PCBs. 52
C. The degree of dechlorination increased with the log of the PCB concentration. 53
SECTION IX. ESTIMATION OF THE FATE OF SEDIMENT PCBS
REQUIRES INTEGRATION OF ALL OF THE DATA AND
APPLICATION OF THE QUANTITATIVE MASS BALANCE
MODELS 54
APPENDIX A. COMMENTS OF DR. PAUL SWITZER
APPENDIX B.
DESCRIPTION OF THE GENERAL ELECTRIC CO. SEDIMENT
TRANSPORT MODEL
-------
List of Figures
la-1. PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores.
2a. Comparison of 1984 and 1994 Estimates of PCB Mass Within TAMS Low Resolution
Zones 1 Through 6.
2b. Comparison of 1984 and 1994 Estimates of PCB Mass Within TAMS Low Resolution
Zones 7 Through 12.
3. Semi Variogram of 1990/91 GE Data Collected from the H7 Site ("hot spot" 5)
Illustrating Variance of Data Versus Separation Distance.
4. Regression Analysis of 1984 and 1994 PCBj^ Mass per Unit Area Estimates for Paired
Sediment Cores Separated by 5 Feet or Less.
5. Probability Distribution of Estimated PCB Mass for Cores Collected from TIP in 1984.
6. Probability Distribution of Estimated PCB Mass for Cores Collected in 1984 from Within
1976-1978 NYSDEC Defined "Hot Spot" Areas of TIP.
7. Comparison of "Versar" Packed Column and Capillary Column PCB Quantification
Techniques a) Packed Column Total PCBs vs DB-1 Total PCBs and b) Packed Column
Total PCBs vs DB-1 Derived PCBs > Trichlorobiphenyls.
8. Frequency Distribution of APCB and log(APCB+2) Between 1984 and 1994 within
Thompson Island Pool Calculated by TAMS.
9. PCB Composition of TIP Sediment Deposition in Approximately 1968 Assessed from
Core Sections Collected in 1983 ana 1991 (Total Concentration > 500 ppm).
10. Comparison ofTIP Sediment PCB3+ Mass Loss Attributable to Fate and Transport
Mechanisms with that Estimated using TAMS' Method.
11. Calculated PCBj+ Mass Per Unit Area for Colocated 1984, 1994, and 1998 Sediment
Cores in Thompson Island Pool.
12. Changes in TIP Sediment PCB3+ Mass Estimated from Differences in Calculated MP A
for Colocated Sediment Cores Collected in 1984, 1994, and 1998.
13. Temporal Profiles of Mean (-<-/- 95% Confidence Interval) PCB3+ Mass per Unit Area for
Co-Located 1984, 1994 and 1998 Sediment Cores Within Four "Hot Spot" Areas.
14. Computed Depth of Sediment Deposition Necessary to Produce Observed Levels of Be
in the Top 1" Section of the 1994 Cores.
-------
15. Estimated Relationship Between 7Be Concentration and Depth of Sediment Deposited.
16. PCB Molar Dechlorination Product Ratio Depth Profiles for Colocated 1994 TAMS and
1998 GE Sediment Cores Collected in Thompson Island Pool.
17. Ratio of DB-1 Peaks 46:32 Depth Profiles for Colocated 1994 TAMS and 1998 GE
Sediment Cores Collected in Thompson Island Pool.
18. TIP Center Channel PCB Concentrations.
19. Comparison of PCB Peak Compositions for Calculated Diffusional Sediment Source
(1997 Summer Average) with (a) Surface (b) Deep Sediments from 1992 EPA High
Resolution Cores Collected from TIP.
20. PCB Congener Ratios Calculated for Fish and Sediment Collected from the Thompson
Island Pool.
21. Temporal Trends in DB-1 Capillary Column Peak Ratios from Three Stations on the
Upper Hudson River During the January 1998 High Flow Event.
22. Average i37Cs and PCB Profiles Calculated from Thompson Island Pool High Resolution
Sediment Cores (HR-019, 020, 023, and 026).
-------
1
2
3
4
5
6
7
8
9
10
11
List of Tables
Ratio of PCB Levels in Paired Samples from the H7 ("Hot Spot" 5) Site at Various
Distances of Separation
GE 1998 Hudson River PCB Sediment Samples Analyzed by both Capillary Column GC
and Packed Column GC Methods
Comparison of PCB Analytical and Quantitation Methods Used for 1984 and 1998
Sediment Samples
Northeast Analytical^Inc. Packed Column PCB Quantitation Scheme
Calculation of PCB Mass Change in Sampled Location Using TAMS Approach
Average PCB Load Increase Across TIP During the Summer 1997 Low-Flow Period
Water Column Derived Estimates of Annual PCB3+ Loading from TIP Sediments
Estimated Amount of Deposition Occurring During 1994 High Flow Event (4/17/94) to
Obtain Observed Detectable-7Be Values
TIP Organic Carbon Normalized Surface Sediment PCB Concentrations
Parameters used in Calculation of Surface sediment PCB Source Signature
Calculated Mean and Range (95% confidence limits) of estimated Mass Changes in "Hot
Spots"
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
COMMENTS OF THE GENERAL ELECTRIC COMPANY ON THE EPA LOW
RESOLUTION CORING REPORT
SECTION I
INTRODUCTION AND EXECUTIVE SUMMARY
A. Introduction
The General Electric Company ("GE") is pleased to submit these comments on EPA's
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" ("Report")- These comments
supplement GE's comments on EPA's 1997 "Data Evaluation and Interpretation Report"
("DEER"). The DEER focused on interpretation of high resolution sediment cores and water
column data collected by EPA's prime contractor, TAMS Consultants, Inc. ("TAMS"). The
1998 Low Resolution Sediment Coring Report provides TAMS' analysis of low resolution
sediment cores obtained in the Upper Hudson River1 in 1994 and purports to show a 40 percent
reduction in PCB mass in the "hot spots" in the Thompson Island Pool ("TIP" or the "Pool")
between 1984 and 1994. As these comments demonstrate, the data collected by TAMS and used
in the Report cannot be used to determine such a mass change. Further, the methodology and
statistics used by TAMS are fundamentally flawed. Because of the substantial uncertainty
associated with the data, which is inadequately treated in the report, no judgments about mass
changes can be drawn from the low resolution cores. TAMS very substantially overestimates the
loss of PCBs from the TEP. The available evidence, including data collected by GE in 1998,
demonstrates that burial of PCBs through sediment deposition, rather than loss through erosion,
is occurring in the Upper Hudson River.
1 GE disputes the statement on page 1-3 of the Report that the Hudson River PCBs Superfund Site "encompasses the
Hudson Rjver from Hudson Falls to the Battery in the New York Harbor." The documents in the administrative
record for the addition of the site to the CERCLA National Priorities List explicitly limit the reach of the site to the
area above the Federal Dam at Troy, and EPA's post-rulemakme comments to the contrary cannot change this fact
1
-------
B. Executive summary
The EPA low resolution coring program is part of a larger effort whose goal is to develop
an understanding of the fate of PCBs in the Upper Hudson River. The low resolution coring
program used a small subset of the relevant data to infer the fate of PCBs within selected
sediments in and downstream of the TIP. Through analysis of these limited data, TAMS drew
sweeping conclusions:
• A significant portion of the contaminated sediment in and below the TIP is not being
buried; rather the PCBs in these sediments have been lost through erosion or other
processes.
• Forty percent of the PCBs in the "highly contaminated" sediments in the TIP were lost
between 1984 and 1994.
• More than 50 percent of the PCBs in three of the major "hot spots" downstream of the
TIP were lost between the mid-1970s and 1994.
These conclusions are unsupportable. In fact, these conclusions are contradicted by the vast
amount of relevant data that were not considered, as well as by other analyses conducted by EPA
and by GE.
TAMS' conclusions are incorrect because its analysis suffers from two fundamental
defects, as well as numerous other deficiencies and errors. First and foremost, the analysis is
based on limited data that are subject to uncertainty sufficient to preclude identifying any trends
that may exist within the sediment. Second, the analysis was conducted using incompatible
measurements of PCBs; that is, PCBs with three or more chlorines (PCBw) in 1984 were
incorrectly compared with total PCBs in 1994. Using compatible measurements (i.e., PCB3. in
both years) results in an implausibly large mass loss, unsupported by the relevant evidence or by
the identification of a mechanism likely to cause such a loss.
-------
B.l Limitations of the data preclude the generation of useful results.
The central objectives of EPA's low resolution sediment coring program were to:
• Obtain new estimates of sediment PCB inventories at selected locations in the
Thompson Island Pool to compare against the existing PCB sediment database
constructed from the 1984 NYSDEC survey.
• Refine the PCB mass estimates for a limited number of historic hot spot locations
defined by the 1976-1978 NYSDEC survey in the Upper Hudson below the
Thompson Island Dam. (Report, at ES-2).
The starting point for estimating the PCB inventory in the Thompson Island Pool,
particularly the "hot spots," is the extensive sediment coring and sediment grab sampling which
NYSDEC performed in 1984. NYSDEC collected approximately 1,200 cores and grab samples
in the Pool. Despite this extensive data collection, NYSDEC and EPA made three estimates of
the PCB mass reflected in the cores and grab samples which varied by 46%, between 23.2 metric
tons ("MT") and 14.5 MT. This uncertainty in the actual size of the inventory was largely the
result of the heterogeneity of PCB deposition: samples taken in close proximity to each other
differed substantially in the concentration or mass of PCBs which they exhibited. This
introduced substantial uncertainty into the final results when these PCB numbers were
extrapolated to larger areas.
The uncertainty as to what the inventory was in 1984 meant that any comparison of a
1994 inventory estimate to the 1984 inventory estimate would inevitably have a substantial
measure of uncertainty. Given the known heterogeneity of the PCB concentrations in the Pool, a
substantially more tightly bounded 1994 inventory estimate could only have been achieved by
the collection of very substantial numbers of samples, well in excess of those taken by NYSDEC
in 1984. EPA elected not to follow such a course.
The Agency decided to follow an approach that would attempt to duplicate a limited
number of the cores and grab samples taken by NYSDEC ten years before and to extrapolate
3
-------
from those matched pairs of cores to conditions throughout the "hot spots" of the Pool. The
matched pairs comparison needed to meet five rigorous conditions in order to provide credible
and persuasive results:
1. There could only be insignificant error in the PCB concentration measurements;
2. The full vertical extent of PCBs had to be captured in both 1984 and 1994 cores;
3. One had to demonstrate that grab samples taken in 1984 could be fairly compared
to core samples taken in 1994;
4. The same sediment had to be sampled in 1984 and 1994; and
5. One had to demonstrate that changes in PCB concentrations at sampled locations
could be extrapolated to other locations.
These conditions were not, and probably could not be, met.
The measurements of PCB concentrations had substantial error. The Report shows split
sediment samples differing by up to 120% with a mean difference of 36% (Report, at 2-18).
The full vertical extent of PCBs was not captured in all of the 1994 samples. By TAMS'
calculation, 19% of the cores taken in the Pool failed to capture all of the PCBs (Report, at 2-20).
The 1984 grab samples cannot be fairly compared to the 1994 core samples. The Report
acknowledges that the "lack of good depth control adds uncertainty associated with core to grab
comparisons" (Report, at 4-7). In fact, both the 1984 and 1994 cores exhibit substantial vertical
concentration gradients which make plausible extrapolation from grab depth to core depth
impossible; consequently, fair comparison of grab samples to core samples is also impossible.
The same sediment was not sampled in 1984 and 1994. For example, the 1994 locations
were known to a precision of ±3 feet. Assuming the same precision for the 1984 locations, the
location matching has an error of ±6 feet. Analysis done by GE on data collected in 1990-91 at
"hot spot" 5 showed PCB concentration in samples taken 2 to 5 feet apart differed by more than
-------
a factor of 5. EPA's own closely grouped 1994 cores show significant variation in PCB mass
and concentrations.
Extrapolation from sampled locations to other locations cannot be shown to be plausible.
The 1994 sampling was at a density of one sample every three acres. This was ten times less
dense than the 1984 effort. This simply magnifies the substantial uncertainty already present in
the 1984 estimates.
EPA was given ample warning of the fundamental defect of collecting insufficient data to
allow for a sound estimate of the PCB inventory in the "hot spots" of the Pool. In its comments
on the plan for the Low Resolution Sediment Program in 1992, GE stated:
The fundamental problem with this approach pertains to the feasibility of defining the
PCB mass in sediments of a small area of the River ("hot spots ") based on a "small"
number of cores. The problem of estimating the mass of PCBs in any area of the River by
use of a small number of samples in the Upper Hudson River is well known (Tofflemire
and Quinn, 1979, Brown; et ai. 1988). The basic problem is that the distribution of PCB
concentrations is highly heterogeneous. This is illustrated by the data collected by GE at
the site referred to as the H-7 site, (the site of the Hudson River Research Station). These
data have already been supplied to EPA. GE extensively analyzed the H-7 site in 1990
by employing capillary column PCB analysis of samples collected on an approximately
12-foot by 12-foot sampling grid The data showed order-of-magnitude changes in PCB
concentrations from one location to the next. This indicates the need to obtain a fairly
large number of samples to properly characterize the PCB mass in any given area.2
Finally, these defects and limitations in the data are compounded by the use of a number
of dubious or erroneous statistical analyses. These are largely addressed in the body of our
comments. They are also set out in the attached report, prepared for GE by Dr. Paul Switzer of
Stanford University, on the statistical analyses employed by TAMS. Professor Switzer found
* Comments of the General Electric Company on the June 1992 Review Copy of the Phase 2 Work Plan and
Sampling Plan for the Hudson River PCB Reassessment RITS (July 24. 1992) at page 73.
5
-------
that many of conclusions in the Report are "not well supported by the reported statistics" and
"seem to be drawn from interpretations and conjectures that do not have a statistical inferential
basis." Dr. Switzer has been a professor in the Statistics Department at Stanford since 1965 and,
among numerous other forms of public and professional service, has been, since 1995, a member
of EPA's National Advisory Council on Environmental Policy and Technology.
The conclusion is inevitable that the uncertainty in the NYSDEC's 1984 estimate of the
PCB inventory in the Pool, compounded by the limited data collected by TAMS in 1994 and the
difficulty or impossibility of collecting data which would meet the rigorous conditions necessary
for plausible and persuasive conclusions, has led to results which are not supported and cannot
be relied on. The actual uncertainty which surrounds the inventory estimates is so great that they
cannot be used for meaningful estimates of PCB loss or gain in the Pool.
B.2 Implausible results are contradicted by relevant evidence.
The results of the analysis undertaken by TAMS are implausible and contradicted by a
wide array of relevant evidence.
There are three conditions which the comparison of the 1984 PCB inventory to the 1994
PCB inventory of the "hot spots" in the Thompson Island Pool should meet before analytical
information can be useful in increasing our knowledge of the fate and transport of PCBs in the
Hudson River and aiding in selecting the appropriate course of action at the end of this
Reassessment. We have already addressed the fundamental benchmark for the generation of
useful analysis: the estimates of the PCB mass in 1984 and 1994 need to be reliable so that one
can have confidence that the comparison of the two reflects actual conditions rather than
unfounded speculation. The estimates do not meet that standard.
There are two further benchmarks by which to judge the Report:
1. The information obtained from the comparison of the 1984 inventory to the 1994
inventory in the paired samples needs to be consistent with other methods of
6
-------
measurement which are of equal or greater reliability. If more reliable methods of
measurement produce materially different results, little weight can be accorded the results
of this program; and
2. If the comparison of reliable 1984 estimates of PCB mass to reliable 1994
estimates of PCB mass indicate a trend of increase or decrease, identification of the
mechanism(s) which drive the trends adds great value to the information because it
identifies the factors, if any, which should be enhanced, diminished, or changed in order
to reach an identified remedial goal.
The central and most startling conclusion of the TAMS Report is that "from 1984 to
1,4A
1994, there has been a net loss of approximately 40 percent of the PCB inventory from highly
contaminated sediments in the Thompson Island Pool" (Report, at ES-2).
The validity of this conclusion turns on a simple point. In order to have comparable,
reliable estimates of PCB mass in 1984 and 1994, one needs to take account of the fact that, in
1984, only PCBs with three or more chlorine atoms ("PCB}-") were accurately measured while,
in 1994, total PCBs could be accurately measured. Obviously, the comparison of the PCB mass
in 1994 with that in 1984 must be made using a common measurement of PCB mass. The
TAMS Report persuasively argues that the proper comparison of PCB mass is made by
comparing PCB\- in 1984 to PCBw in 1994. Appendix E to the Report is an analysis by Tetra
Tech in support of this method.
Any reader of the Report who did not approach it with a trained and highly critical eye
would believe that the Report followed what it laid out - quite accurately - to be the appropriate 1 4B
method for comparing PCBs in 1984 to PCBs in 1994. For reasons which are not explained in
the TAMS Report, this was not done. Inexplicably, PCB)- for 1984 was compared to total PCBs
for 1994. When the method recommended in the Report is in fact used, the net loss of PCB
inventory in the Pool from 1984 to 1994 is approximately 80%.
-------
A method of calculation that concludes that there was an 80% loss of PCB inventor.- in
the Pool between 1984 and 1994, approximately 10.8 MT, cannot withstand scrutiny.
First, this result must be compared to other reliable methods of calculation. The obvious
1.4C
comparison is to loss of PCBs from the Pool measured by PCBs in the water column downstream
of the Thompson Island Dam. Routine measurements of PCBs in the water column at
Schuylerville from 1984 to 1993 show a total PCB}, loss of approximately 1.0 MT out of the
Thompson Island Pool. Thus, the results of the TAMS' analysis are contradicted by a far more
extensive and reliable data set which directly measures loss of PCBs out of the Pool.
Second, the Report does not identify any mechanism that would result in an 80% or 10.8 ' ^
MT loss of PCBs out of the Pool between 1984 and 1994. GE has identified four mechanisms
which could possibly lead to mass loss out of the Pool: dechlorination, diffusion, scour, and
groundwater advection. Taking all of these mechanisms into account results in an 18% loss of
PCB inventory from the Pool from 1984 to 1994. The conservatism of this estimate is
underscored by the fact that the direct water column measurements result in an inventory loss of
approximately 7%.
Third, when the trend established by an 80% loss of the PCB inventory in the Pool 1.4E
between 1984 and 1994 is brought forward to the present, it is apparent that the trend is
contradicted by present conditions:
• GE conducted a coring program in 1998, reoccupying coring sites used by EPA in
1994. Twelve of GE's coring sites were at locations which, under the T.AMS
method of analysis, exhibited losses of approximately 80% between 1984 and
1994. Between 1994 and 1998. these locations exhibited an increase of 100% in
PCB inventory.
• If an 80% loss of PCBs took place from 1984 to 1994. there are various ways of
estimating what the loss over the last four years should have been. If the loss
were a consistent mass annually, there would be no PCBs left in the Pool. If the
8
-------
rate of loss were declining in the nature of a half life, the annual loss of PCBs
would now be at the approximate rate of V/i% of the 1984 inventory and would
have been preceded in the mid-1980s by annual losses that would have been an
order of magnitude greater. Neither of these results is discernible from the
measurements of PCBs in the Upper Hudson.
Fourth, there are two obvious corollaries to the proposition that there was an 80% loss of
PCB inventory from the Pool between 1984 and 1994: there would not be widespread deposition
of fresh sediment across the Pool; there would be evidence of widespread erosion. These are
simple points. In 1984, the PCBs in the Pool, particularly in the "hot spots." were largely buned
- substantial masses of PCBs occurred several centimeters below the surface of the core. For
80% of the PCB inventory in the Pool to reach the sediment surface and be released into the
water column, erosion by some means would be required. Conversely there should not be
widespread deposition of new sediment. These corollaries which would support and lend
credence to the conclusion of the TAMS Report are not borne out by the facts.
As to burial. GE's 1998 high resolution coring shows the highest PCB concentrations ] .5a
occurring well below the surface of the core. Be has a half-life of 53 days and is laid down
during periods of higher flow in the Upper River. In 1994, the high flow period occurred in
April, and TAMS' cores were analyzed in August. Seventy percent of the surface sediment
samples collected indicated the presence of Be and. therefore, recent deposition (Report, at 2-
21). Moreover, the passage of time between the deposition of Be and laboratory- analysis of the
cores and the depth of TAMS' surficial core segment have the effect of limiting detection of Be
to deposition levels above approximately 0.5 cm. Consequently, the failure to detect Be in the
remaining 30% of the core samples does not lead to the conclusion that recent bunal did not
occur. In short, the 1994 Be evidence viewed properly is inconsistent with the proposition that
widespread deposition of fresh sediment did not take place at the Thompson Island Pool "hot
spots."
In 1998, GE found that 11 out of the 12 cores it took, which were finely segmented in the | 53
surficiai 5 cm, showed a steep gradient from lower surficial concentrations to higher
9
-------
concentrations at depth. Additionally, many of the cores in 1998 had the highest PCB
concentrations at greater depth than the matching TAMS' core taken in 1994. Once again, this
recent empirical data supports burial, rather than erosion, as the dominant process at the "hot
spots" in the Thompson Island Pool.
Turning to erosion, GE looked at the difference in PCB congener composition between j
the more dechlorinated PCBs which entered the river in the past, approximately more than seven
years ago, and the less dechlorinated PCBs which entered the river more recently. In short,
dechlorination is a result of exposure to the biological processes active in the river; PCBs
entering the river recently exhibit materially less dechlorination than the historic PCBs. These
historic dechlorinated PCBs would have to make up a very large proportion of the PCBs in the
surficial sediments and the water column if 80% of the "hot spot" inventory in the Pool was lost
between 1984 and 1994.
Three independent tests of relevant data are inconsistent with the corollary which would
1.5D
support the TAMS report. First, TAMS took high resolution cores in the Pool in 1992; the PCBs
leaving the Pool best match the PCB composition of the surficial 2 cms of the TAMS' cores and
do not match the composition found below 8 cm. Second, NOAA sampled fish in the Pool in
1993. The composition of the PCBs in the fish indicates that the PCBs were not highly
dechlorinated and that the fish had not been exposed to the PCB congener distribution present at
lower elevations in the 1984 cores. Finally, the flood event of January 1998 was examined since
it was potentially an erosional event. Once again, the PCB composition found in the water
column at that time was consistent with the PCBs in the surficial sediments in 1991 and not with
the dechlorinated sediments which were present at lower depths. In sum, none of these
independent tests of the data yields a result which supports the central conclusion of the TAMS
report.
The result of this analysis is plain: when the 1984 PCB mass in the Pool is compared to
the 1994 PCB mass in the Pool, using TAMS' method of comparing PCB-3 to PCB~3 in each
year and using the data TAMS selected from the 1994 coring program, a mass loss of 80% is
obtained. That result is utterly implausible. The result is contradicted by direct measurement of
10
-------
PCBs in the water column. The 80% loss is not supported by the identification of any
mechanism which would lead to that result; in fact, the combination of all feasible mechanisms
of loss operating at once lead to a total loss of approximately 18% of the PCBi* estimated in the
TIP sediments compared to 80% by the method advocated in the Report. When the trend is
extended from 1994 to 1998, it is contradicted by the cores GE took this summer and by the
actual rate of PCB loss from the Pool which is taking place now. Finally, the broad proposition
of erosion in the Pool and absence of sediment deposition which would be necessary to achieve a
loss of 80% of the PCBs in the Pool are inconsistent with the empirical data.
The Report's treatment of possible loss or gain in the PCB inventory in the "hot spots"
below the Thompson Island Pool illustrates the defects which we have discussed in the context
of the Pool. The data collected and employed in the Report were extremely limited with the
result that extrapolation from the sample data to conditions throughout the "hot spots" was
highly uncertain. However, in a (lumber of instances, the extrapolation resulted in apparent
increases in the PCB inventory in the "hot spots." In particular "hot spot" 28 showed a very
substantial mass increase. TAMS found these inventory gains to be implausible, particularly that
at "hot spot" 28. The lesson to be drawn from this is not that large increases in PCB inventory in
"hot spots" is implausible, but that extraordinarily broad conclusions built on a foundation of
extremely limited data are suspect and. if not consistent with other more reliable data and
analysis, should be disregarded.
In short, the results of the Report do not pass muster under the benchmarks which test
their value for the purposes of the Reassessment. The conclusion drawn by Richard Caspe in his
cover letter to the Report largely hits the mark:
...it is important to recognize that the conclusions in this report, although significant, do
not indicate whether or not remedial action is necessary for the PCB-contaminated
sediments of the upper Hudson. The numerical analysis (computer modeling) of fate and
transport of PCBs, the associated ecological and human health risk assessments, and a
feasibility study must be completed before any such conclusion can be reached.
11
-------
We fail to understand in what regard the conclusions of the Report are significant; the rest of the
statement is not disputable.
12
-------
SECTION II
ACCURATE ESTIMATION OF LOSS OF PCBS FROM THE TIP SEDIMENT IS NOT
POSSIBLE USING THE LOW RESOLUTION CORE DATA
The approach used to estimate the loss of PCBs from the TIP sediments makes the
following assumptions:
A) The PCB concentration measurements have insignificant error;
B) The full vertical extent of PCBs was captured in the 1984 and 1994 cores;
C) Grab samples taken in 1984 can be compared to core samples taken in 1994;
D) The same sediment was sampled in 1984 and 1994; and
E) Changes observed at sampled locations can be extrapolated to other locations.
All of these assumptions are invalid and render the mass change estimates highly uncertain and
useless in estimating the fate of PCBs in the sediments of the TIP. The reasons why the
assumptions are invalid are summarized below:
A. The PCB concentration measurements have substantial error.
1.6
The report presents data indicating that PCB measurements of split sediment samples
differed by up to 120 percent and that the mean difference was 36 percent (Report, at page 2-18).
Despite these data, TAMS made no effort to account for analytical error in its statistical analyses.
Instead, its mass change calculations assume no error in the 1984 and 1994 mass estimates. In
fact, the analytical error is sufficiently large to render the reported mass changes not statistically
different from no change.
The report acknowledges that analytical error compromises the utility of the data:
Examining the differences in PCB inventory on an absolute basis was not particularly
fruitful, in part because the magnitude of the PCB change can represent both analytical
variability as well as real change. (Report, at 4-15).
13
-------
The report goes on to suggest that expressing change in relative terms diminishes the
analytical variability. This is not true. No mathematical calculation can diminish analytical
uncertainty. Therefore, dividing the absolute difference by the 1984 mass estimate does nothing
to reduce the component of the change that is due to analytical variability.
B. The full vertical extent of PCBs was not always captured in the 1994 cores. i
The vertical extent of PCBs in the 1994 sediment cores was estimated by visual
inspection of each core (Report, at 2-3) or by the depth of penetration. To determine whether
the entire PCB inventory had been captured, the bottom segment of each core was analyzed for
13 Cs. The absence of detectable 13 Cs was considered confirmation that the entire PCB
inventory had been captured, and the presence of 13'Cs was presumed to indicate that the entire
inventory had not been captured (i.e., the core was classified as 'incomplete"). Based on this
criteria, the full inventory of PCB contamination was not sampled in 19 percent of the cores
collected in the TIP and 40 percent of the cores collected below the TIP (Report, at page 2-20).
Indeed, GE collected sediment cores m 1998 from 12 of the same locations TAMS sampled in
1994 which indicate that many of the 1994 cores may have failed to capture the full PCB profile.
In 4 of the 12 cores collected in 1998, the PCB maxima were deeper than m the corresponding
TAMS' core (Figure 1 a-1). These data suggest either:
1) the TAMS coring technique failed to capture the full PCB inventory, or
2) spatial heterogeneity in sediment and PCB deposition can produce profound
changes in PCB profiles within the same region.
Because the full PCB inventory was not sampled in all cases in 1994, a bias toward
greater mass loss is introduced into the mass change calculation. The significance of this error
cannot be determined because it is not known whether the 1984 samples captured the full
. i ns>Q
14
-------
inventory.3 Nonetheless, this uncertainty introduces an unknown error into the mass change
calculations.
C. Grab samples taken in 1984 cannot be compared to core samples taken in 1994. 1.8
In the Report, fifteen of the 60 1994 sediment cores are compared to grab samples taken
in 1984. The 1984 mass at the location of a grab was calculated by assuming that the
concentrations measured in the grab could be applied to a depth of 17 or 12 inches, depending on
sediment type. The Report acknowledges the weakness of this assumption:
The lack of good depth control adds uncertainty associated with core to grab
comparisons. Nonetheless, the core-grab pairs can still provide some useful information
on the change in sediment inventory. (Report, at 4-7, Underline added for emphasis).
Despite the acknowledged uncertainty, the core-grab pairs are used along with the core-
core pairs in calculating the average percentage change in mass.
Both the 1984 and 1994 cores exhibit substantial vertical concentration gradients. These
gradients indicate that it is not appropriate to extrapolate concentrations measured in grab
samples to sediments below the depth sampled. The PCB profile at depth cannot be predicted
from the grab sample data. As such, it is inappropriate to compare a 1984 grab sample and a
1994 core sample.
D. The same sediment was not sampled in 1984 and 1994.
1.9
The report indicates that the locations of the 1994 samples are known to a precision of ±
3 feet (Report, at 2-5). The 1984 locations are likely to be known with equal or less precision.
Thus, location matching has an error of ± 6 feet or more. In addition, reoccupation of the 1984
3 The vibraconng apparatus used to collect sediment m 1984 was larger and more powerful than the apparatus used
in 1994, and likely achieved greater penetration. In fact, the penetration depths in 1984 were about 63% greater than
the penetration depths in 1994. Thus, it is likely that the 1984 data provide a better representation of the inventory
than do the 1994 data.
15
-------
locations was not perfect. Fifty percent of the 1994 locations were more than 3 feet from the
1984 location and 30 percent were more than 5 feet from the 1984 location. The combination of
the uncertainty of location and the reoccupation error results in location matching errors that
likely average about 10 feet or greater.
Given that PCB concentrations vary considerably over short distances, location error
introduces significant uncertainty and error into the calculations. This variation is exemplified
by the fact that 1984 samples located within zones delineated to minimize spatial variability
varied by a factor of two to three (Report, at 2-2). Further, regression analyses of the 1994 data
indicate that PCBs were only weakly correlated to sediment properties (Report, at Tables 3-4 to
3-8). The best correlate (percent solids) could account for only 36 percent of the variation in
PCB mass per unit area among locations.
The lack of correlation between the matched 1984 and 1994 data is illustrated by cross-
plots of these data within individual zones (Figure 2a and b). No relationships are evident. In
other words, within individual zones, the changes in mass between 1984 and 1994 are random.
Given the proximity of the samples within a zone and TAMS' efforts to define zones so that they
represent a homogeneous area (Report, at 2-2), the lack of correlation demonstrates that small-
scale spatial variability severely compromises the assumption that the same sediment was
sampled in 1984 and 1994.
The only data set available to estimate the correlation of PCB levels on the spatial scale
of the 1984-1994 location matching is the H7 site data collected in 1990 and 1991 by GE. At
this site in "hot spot" 5, core samples were collected at a spacing of several feet. These data
were used to assess the correlation between closely spaced locations. A semi-variogram was
computed relating the variance of PCB concentration as a function of distance between samples.
Correlation is indicated by a variance that is less than the variance of the full data set. The semi-
variogram showed that PCB levels were correlated only within a distance of 5 feet (Figure 3).
Beyond 5 feet, the variance was near that of the population. Pairing of data based on location
indicated that even within the distance of 5 feet, the correlation was weak (Table 1). Samples
mi
16
-------
within 2 feet of each other differed in concentration by about a factor of 2 on average. Samples
within 5 to 10 feet of each other differed in concentration by about a factor of 20 on average.
The lack of correlation at small spatial scales does not imply that correlation does not
exist at larger spatial scales. It is possible to have random variation within a contiguous area,
such as the H7 site, and still have larger scale correlation, For example, concentrations may be
generally higher in one area than another, even though within each area the individual samples
show little correlation. For this reason, correlation at larger spatial scales, such as the 100 to 200
foot scales of the 1984 sample grid, cannot be used to infer correlation at the 1 to 10 foot scale at
which the 1984 and 1994 data are matched.
Removal of samples more than 5 feet apart will eliminate comparisons for which no
correlation can be claimed. However, it must be recognized that even within a distance of 5 feet,
the variation between samples is large and imparts considerable uncertainty to any comparisons.
This uncertainty must be earned through the estimation of mass change.
When the data set is corrected so that it is limited to pairs of complete cores that are less
than 5 feet apart, 24 pairs of cores remain (Figure 4). A regression of the logarithms of PCEh.
mass has a slope of 0.26 with 95% confidence limits of -0.03 to 0.55. Neither the slope nor the
correlation coefficient are significantly different from zero [t(slope-standard error of slope 1=1.S8
< t(.05,22)=2.07; r=0,37, 0.05
-------
PCB concentrations as well as spatial variability in the physical processes controlling PCB fate
in the Hudson River. Examples of areas are the zones delineated in the Report as "hot spots."
An advantage of the areal approach is that more data can be brought to bear on the
problem. In 1984, there were 24 complete cores in the Thompson Island Pool that matched a
1994 core that was less than 5 feet away (Figure 4). In contrast, there were 298 total cores
collected in 1984, of which 210 were in "hot spots."
However, the area-based approach is also subject to severe limitations, some of which
were described above:
(1) Analytical errors may be large enough to preclude identification of real changes in
mass.
(2) The sample sizes are still small relative to the variability in mass.
(3) Within each "hot spot," the distribution of sampling locations is not spatially
random, so that the 1994 sampling does not represent "hot spot" conditions. As
described above, the 1994 cores were collected in the vicinity of 1984 cores that
exhibited higher-than-average PCB masses.
(4) Statistical problems are associated with estimating the proportional change in
PCB mass within each area.
E. Changes observed at sampled locations cannot be extrapolated to other locations.
The Report uses the average of the percent mass change estimates at 40 locations in the
TIP to infer the percentage mass change in all of the areas of "highly contaminated sediments" in
the TIP. These 40 locations are equivalent to a sampling density in the TIP "hot spots" of about
1 sample every 3 acres. Given the spatial heterogeneity discussed above, such extrapolation
cannot be made.
Spatial heterogeneity is so great that the PCB mass in TIP sediment in 1984 cannot be
determined precisely, despite a sampling density more than ten times greater than that used in
18
-------
1994. Mass estimates of 23.2 MT (Brown et a/., 1988), 19.6 MT and 14.5 MT (USEPA, 1997)
have been calculated using different methodologies. The relative percentage difference among
these values is as high as 46 percent.
In addition, the 1994 sampling program was not designed to provide an unbiased
representation of "hot spot" PCB mass. Samples were not collected randomly, but rather in
clusters, and the data used in the analysis selectively focused on areas of elevated concentration.
The locations of the low resolution cores collected in 1994:
...were selected to represent a range of sediment types and sediment PCB inventories,
with emphasis placed on areas of greatest PCB contamination. (Report, at 2-5).
The resulting data are not representative of the Thompson Island Pool as a whole. The
PCB masses in 1984 cores that were matched with 1994 cores are considerably higher than the
PCB masses in 1984 cores as a whole (Figure 5). The data are also not representative of PCB
masses within "hot spots;" the median PCB mass in matched 1984 cores within "hot spots" is 6-
fold greater than the median PCB mass in all 1984 cores within "hot spots" (Figure 6).
The Report acknowledges the insufficiency of the data for evaluation of the TIP as a
whole:
... the 76 cores analyzed in the Tl Pool ana 94 cores taken downstream ot the Tl 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. (Report,
at 2-15).
The Report then contradicts itself by concluding that the calculated average mass change
is applicable to the TIP as a whole.
19
-------
SECTION III
THE METHODS EMPLOYED CONTAIN ERRORS AND OMISSIONS THAT
MASK THE IMPLAUSIBILITY OF TAMS' MASS LOSS ESTIMATE
TAMS' methods contain the following errors and omissions:
A) The sum of PCBs with three or more chlorines (1984) was compared to the sum
of all PCBs (1994);
B) The geometric mean was used to estimate the average mass loss; and
C) A bias was introduced by eliminating locations where the PCB inventory in 1984
was less than 10 g/m2
These errors and omissions invalidate the Report's mass loss estimate of 40 percent. Correcting
the mistakes yields two alternate mass loss estimates presented below. One corrected estimate is
implausibly large (see Section IV). The second estimate is not significantly different from zero,
which demonstrates the overwhelming uncertainty of the TAMS approach.
A. The sum of PCBs with three or more chlorines (1984) was compared to the sum of
all PCBs (1994).
The 1984 sediment PCB measurements included only PCB3+. Consequently, PCB.w is
the only appropriate measure by which to compare the 1984 and 1994 PCB data.
TAMS devoted considerable effort to developing the appropriate means for comparing
the PCB analysis performed by NYSDEC in 1984 with the analysis it performed in 1994:
As part of the Phase 2 investigation, a study was made of the differences between the two
[PCB analytical] techniques [1984 NYSDEC and 1994 EPA]. This is documented in
Appendix E which describes the quantitation issues relating the 1994 Phase 2 and 1984
NYSDEC PCB data. The recommendation of this analysis was to use the 1984
20
-------
quantitation of total PCBs as representative of the sum of congeners in the trichloro
through decachloro homolog groups... (Report, at 4-5).
Indeed, Appendix E of the Report, which documents the analysis of the 1984 sediment PCB
quantitation, concludes that:
"Total PCBs" reported for the 1984 sediment data (calculated by NYSDEC as a sum of
Aroclors) provide a good representation of the sum of tri- and higher-chlorinated
congeners. They do not accurately reflect the total of all congeners.
The Report's assessment of the analytical issue of how to compare 1984 NYSDEC and
1994 Phase 2 PCB data indicates that the most appropriate way to compare the two data sets is
on a PCBj* basis. This is largely the result of the packed column techniques employed by the
NYSDEC contract laboratory, Versar. Versar analyzed sediment PCBs using a packed column
gas chromatograph ("GC") method standardized using commercial Aroclors (Brown et ai,
1988). Generally, packed columns cannot consistently resolve mono-chlonnated congeners.
Additionally, Versar only quantified and reported .Aroclors 1242, 1254, and 1260. It did not
attempt to quantify the lighter-chlorinated Aroclors. As a result, mono- and dichlorinated
congeners resulting from the reductive dechlorination of sediment PCBs were missed in the
analysis of 1984 sediments.
Tetra Tech reached its conclusion that the 1984 data represent predominantly tri- and
higher homologs by performing what it termed an "as if numerical experiment (Report, at
Appendix E). In this numerical exercise. Tetra Tech interpreted the congener data generated
from 241 high resolution sediment coring samples from 1992 as if they were analyzed according
to the procedures used by Versar. That is, congeners detected in the 1992 samples that
represented the packed column peaks used by Versar (Aroclor 1254 and 1260) and NYSDEC
(Aroclor 1242) were summed to generate a representation of the 1984 analysis. While this
provides strong evidence that the 1984 analysis scheme represented predominantly PCB3-,
potential biases not documented in the 1984 reports may not be adequately represented in this
numerical experiment.
21
-------
GE undertook a laboratory experiment to further evaluate what portion of the PCB
spectrum is represented by the 1984 NYSDEC sediment data. This study was designed to
analytically evaluate the results of the numerical experiment presented in the Report. GE's
laboratory experiment consisted of reanalyzing sample extracts generated during the Focused
Sediment Coring component of the 1998 TIP Sediment Coring Program (QEA, 1998a) using
packed column GC techniques. Twenty-one sample extracts representing a broad range of PCB
concentration and composition were used (Table 2). The extracts were analyzed by Northeast
Analytical, Inc. on a packed column GC system equipped with a 6.0-foot by 0.25-inch ID glass
column. While this set up deviated slightly from Versar's (Table 3), the resulting chromatograph
was generally consistent with Versar's (Table 4).
GE's results support Tetra Tech's conclusion that the 1984 PCB quantitation method
represents the PCBj+ in the sediment samples. The results of the reanalysis appear in Table 2
and Figure 7. The Versar packed column total PCB method grossly underestimates the total
PCB concentration of the samples (Figure 7, panel a). However, comparison of the total PCB
quantified by Versar with PCB3- generated on the DB-1 capillary column system yields a nearly
perfect linear relationship (r=0,98) with a slope of unity.
Both Tetra Tech's numerical experiment and GE's 1998 analytical experiment support
the Report's conclusion that the 1984 NYSDEC total PCB results must be compared with PCB?~
from the 1994 analysis. Comparison on any other basis is technically invalid.
Despite this conclusion. TAMS performed its analysis using total PCBs from 1994. This
analysis is therefore, flawed, because the 1994 data account for mono- and dichlonnated
biphenvls not measured in 1984. Although PCB dechlorination may have transformed a limited
number of PCB3+ to mono- and dichlorinated biphenyls between 1984 and 1994, the
preponderance of evidence presented in the technical literature indicates that PCB dechlorination
occurs rapidly (Rhee et al„ 1993; Fish and Principe, 1994) and that PCBs present within the
Hudson River sediments were highly dechlorinated by 1984 (Brown et ai. 1984; McNulty,
1997).
22
-------
Performing the analysis presented in the Report using PCBj* in 1984 and PCB]- in 1994
results in an estimated mass loss of 80% of the PCBs in Thompson Island Pool (n=37, geometric
mean of DeltaPCB+2).4 This result ts implausible, as described in Section IV, further
demonstrating the Report's flawed methodology.
B. The geometric mean was used to estimate average mass loss.
The Report estimated mass loss within the Thompson Island Pool and in selected "hot
spots" below the Pool using two different statistical methods. In the Thompson Island Pool,
TAMS used matched core pairs. TAMS computed percent loss of PCB mass in each core pair
and then averaged the percent losses in all pairs throughout the Pool. The statistic used by
TAMS was "log(APCB~2)," in which APCB is given by;
An^D \994MPA - 1984A/P.4
A/>(_ ft
1984AIP A
TAMS added a value of 2 to APCB to ensure that all values were greater than zero, to calculate
the logarithms. TAMS used the geometric mean of APCB as the average change in sediment
PCB inventory in the "highly contaminated" sediments of the TIP.
Below the Thompson Island Pool, T.AMS computed "hot spot" average mass for 1976-78
and 1994, using the minimum variance unbiased estimator of the arithmetic mean (MVUE;
Report, at 4-30). Then, TAMS computed mass change in the "hot spot" as the difference
between the 1976-78 and 1994 averages.
These two methods are inconsistent, and in fact can result in very different estimates of
mass change. The arithmetic mean of the proportional change is more appropriate than the
geometric mean. Indeed, when discussing mass change below the Pool, T.AMS states that the
4 The sample size reported here (37) differs from that of TAMS (40). because PCB data for 3 cores could not be
found in the Hudson Raver Database Version 3.5.
23
-------
arithmetic mean must be used to compute the average mass in a "hot spot" (Report, at 4-28).
Similarly, the arithmetic mean should be used to compute the total proportion of the PCB mass
that has disappeared.
In a log normally distributed group of numbers, the geometric mean is equal to the
median, that is, the value such that 50 percent of the data are higher and 50 percent are lower.
The geometric mean reduces the contribution of high values to the estimate of the mean. For
example, consider two hypothetical estimates of the ratio of 1994 mass per unit area
("MPA")/1984MPA: 0.1 and 0.9. The arithmetic average of these two numbers is 0.5; the
geometric mean is 0.3. The geometric mean is lower than the arithmetic mean, which is generally
the result in a skewed group of numbers. Both APCB and log(APCB-^2) calculated by TAMS are
skewed (Figure 8), and indeed the geometric mean of log(APCB+2) using the Report's 40 paired
values (-40%) is lower than the arithmetic mean (-30%). The arithmetic mean is the more
appropriate estimate to use.
Using the arithmetic mean of PCB3- rather than the geometric mean, results in an
estimated mass loss of 79% of the PCBs in the Thompson Island Pool (n=37, 95% confidence
limits 73% to 85%).5
Thus, even after correcting these errors, the resulting mass loss from 1984 to 1994 is
implausible, as described in Section IV.
C. A bias was introduced by eliminating locations where the PCB inventory in 1984
was less than 10 g/m2.
In the Report, the PCB mass loss in the Pool was computed using only those sample pairs
in which the 1984 core contained >10 g/m2. TAMS explained its selection of data by asserting,
without support, that:
5 The sample size reported here (37) differs from that of TAMS (40), because PCB data for 3 cores could not be
found in the Hudson River Database Version 3.5.
24
-------
...the greater-than-I0-g/'m' group corresponds to sediments typically found in hot spot
areas (Report, at 4-13).
In fact, cores with concentrations less than 10 g;m2 comprise approximately 55 percent of
all cores collected within "hot spots" in the Pool (Figure 6). Thus, by excluding cores with <10
g/m2, the Report's analysis did not represent the "hot spots" as a whole.
In addition, excluding 1984 samples with <10 g/m2 creates a bias in the analysis.
According to the Report, the analysis must include a demonstration that the 1994 PCB mass is
significantly different from the 1984 PCB mass. In other words, PCB mass did change between
1984 and 1994. TAMS removed all cores with less than 10 g/m: from the 1984 core set, but not
from the 1994 core set, implying that cores with less than 10 g'm: were not representative of "hot
spots" in 1984, but were representative of "hot spots" in 1994. There is no basis for this
assumption, and the removal of those cores caused the 1984 data to be biased high relative to the
1994 data, resulting in an inflated estimate of mass loss.
To illustrate the result of this bias. GE calculated mass loss using the arithmetic mean of
PCB3~, including matched samples regardless of concentration. As descnbed in Section II, only
complete cores separated by less than 5 feet (n=24) were included. This resulted in an estimated
PCB mass loss of 21% with 95% confidence limits ranging from a gain of 31% to a loss of 73%.
However, this value, like the 80% figure, is not an unbiased, correct estimate of the mass loss:
analytical error was not included in the analysis, the sampling locations were not random, and
the data were not truly paired.
To summarize, using the statistical methods presented in the Report, with the correct
measure of PCB mass as discussed by TAMS (PCB3» in 1994), the calculated mass change
(80%) is implausible. In addition, applying several corrections to the analysis results in a
radically different conclusion: the 1994 mass estimate cannot be differentiated from the 1984
estimate. This illustrates our primary conclusion: the data set is so limited and the variability is
so great that the 1994 low resolution core data cannot be used to estimate PCB mass loss with
any confidence.
25
-------
SECTION IV
THE CORRECTED ESTIMATE OF MASS LOSS IS IMPLAUSIBLE
Having established that PCB3* is the appropriate basis for comparing the 1984 and 1994
PCB analytical data, TAMS should have presented its PCB loss estimates on this basis. Had it
done so, it would have recognized that the loss of PCB3+ between 1984 and 1994 (80%) is not
supported by other independent estimates of mass loss, particularly water column PCB3- loading
estimates. Based on this, TAMS should have concluded that it is not possible to estimate mass
change accurately by comparing the results of sediment sampling and analysis conducted in a
spatially heterogeneous system by using different methodologies ten years apart.
A. TAMS' approach results in 80% loss of PCB3+.
TAMS' approach for estimating the change in "hot spot" inventory within the TIP was
based upon sediment core MPA calculations:
MPA = ]Tci L, pb,
i=i
where, the product of concentration (C, [M/L3]), sediment core section thickness (L, [L]), and
dry bulk density (pb,[M/LJ]) is summed for all core sections (i=l to n) over the entire core length.
To estimate MPA for the PCB3+ ("MPA3J'), C is based upon:
1984 C3- = 0.934*Aroclors, Total (Report, at Appendix E)
1994 C3+ = Total PCB - (Mono PCB + Di PCB)
TAMS calculated the change in PCB mass between paired 1984 and 1994 sediment cores using
the following equation:
26
-------
A PCB =
1994 MP A -1984 MP A
1984 MP A
TAMS incorrectly used MP A3-, for 1984 data, while the 1994 MPA was based upon total PCB
data. As outlined in Section III, the appropriate comparison of the two data sets is to calculate
APCB using MPA3+ calculations for both the 1984 and 1994 cores (i.e., A PCBi,.).
TAMS biased its data set by designating "hot spot" sediment cores as pairs in which the
1984 MPA3,. was greater than 10 g'm:. Change in "hot spot" mass was therefore estimated using
a measure of the average APCB. TAMS used the variable log (APCB-2) for combining
calculated mass changes for the set of "hot spot" core pairs. The factor of 2 was added to avoid
taking the logarithm of numbers less than or equal to 0. TAMS' mass loss estimate was
calculated by taking the arithmetic mean of (log (APCB ->-2)), and transforming it back to a
percentage:
MASS LOSS = (jo Al(' - 2)x 100%
This is equivalent to a geometric mean of (APCB-2i. The inappropriateness of using a
geometric mean to calculate mass changes is discussed in Section IV. It is used here only for
comparison with TAMS' estimate of mass change.
TAMS used this approach to estimate that PCB mass loss in the TIP from 1984 to 1994
was approximately 40%. However, if the calculation is made using the more appropriate
comparison of PCB3., for the same set of 1984 and 1994 core pairs, the PCB;. mass loss is
estimated at 80% (Table 5) - an implausible result.
Based upon knging analyses. EPA estimated the entire TIP PCB inventory to be
approximately 14.5 MT (EPA. 1997). Because this estimate was made using 1984 NYSDEC
sediment data, the PCB?- mass within the TIP can be estimated using Tetra Tech's regression
27
-------
analysis (Report, at Appendix E). Applying the factor of 0.934 results in a PCB\~ inventory of
13.5 MT within the TEP. Assuming that most of the sediment PCB inventory exists within the
"hot spots," the 1984 to 1994 PCBj* mass loss estimated using TAMS' approach would exceed
10 MT, a conclusion contradicted by all other available data. Mass loss of this magnitude cannot
be supported by known fate and transport mechanisms, or by water column PCB loading data
collected over the 1984-1994 period.
B. No fate and transport mechanism can account for the mass loss estimated using 1.15
TAMS' method.
Fate and transport mechanisms that result in a net loss of sediment PCB mass include
dechlorination, diffusion to the water column, scour, and groundwater advection.
Approximations for each of these mechanisms are developed below to estimate the magnitude of
expected PCB 3+ mass loss in TIP sediment from 1984 to 1994,
Dechlorination
Microbially mediated dechlorination of PCBs in sediments occurs when meta- and para-
substituted chlorines are substituted by hydrogen under reducing conditions. As a result of
dechlorination, higher molecular weight ("MW") PCBs are transformed to PCBs that have a
lower MW. Based upon the Report's AMW analysis (Report, at 3-9), TAMS estimated that the
average total PCB mass loss in the Upper Hudson attributable to dechlorination is 12%.
An estimate for the PCBj+ mass loss due to dechlorination was developed using data
presented in McNulty, 1997. McNulty compared high resolution sediment cores taken from the
same location within the TIP in 1983 and 1991. Core sections were first matched using ,3'Cs
dating so that sediments deposited at the same time could be compared between the two cores.
Congener PCB data were then used to evaluate composition changes over the 1983-1991 period,
PCBj~ dechlorination mass loss from the McNulty data was based upon sediments deposited in
approximately 1968, having total PCB concentration in excess of 500 ppm. The homolog
composition for this sediment layer, as measured in 1983 and 1991, is plotted in Figure 9. Also
28
-------
plotted is the change in homolog weight percent. Decreases in di-, tri-, and tetrachlorinated
biphenyls are accompanied by a large increase in monochlorinated biphenvls. Based upon this
limited data set, the PCB;- mass loss due to dechlorination between the 1980s and 1990s was
10%, which is consistent with TAMS' total PCB average of 12%. Therefore, dechlorination can
be expected to account for approximately 10% of the PCB3+ mass loss, or 1.4 MT, between 1984
and 1994.
Diffusion
Diffusion results in the transport of PCBs from the sediments due to a concentration
gradient between the surface sediment pore water and the overlying water column. Diffusion is
the primary sediment-water column exchange mechanism during low flow periods. Mass
transfer by diffusion is temperature dependent, increasing during warmer periods.
An estimate of the 1984 to 1994 PCBj- mass loss attributable to diffusion was
determined using the observed water column loading across the TIP during the summer 1997 low
flow period. Data from 1997 were considered because 1996 data collected at the TID-WEST
station were influenced by a sampling bias (QEA, 1998c). Calculation of the summer 1997 TIP
water column PCB loading (Wwc) from paired Fort Edward (Cte) and unbiased TID water column
PCB data (C,1d) was based on the following equation:
H''wc-Ole(C„a-C,J
where Qfe is Fort Edward flow (L3 T'1).
As shown in Table 6, the average summer 1997 PCB load gain across the TIP was
approximately 0.5 kg/d. The PCBj- portion of this load is approximately 44% or 0.2 kg/d.
Using this value as a conservative estimate of the daily PCB?. sediment PCB flux throughout the
29
-------
year, the total mass transfer from TIP sediments during 1984 to 1994 is 730 kg. This total
accounts for only 5% of the estimated 13.5 MT PCB 3, inventory in the TIP.
Scour
Sediment scour occurs as an event-driven process (Ager, 1981; Lick, 1992). As flows
become elevated within the river, shear stress at the sediment-water interface increases. When
the critical shear stress is reached for a given sediment deposit, the surficial layers become
resuspended into the water column and are transported downstream.
Because the largest resuspension events occur at the largest flow rates, an upper bound
estimate of sediment PCB scour within the TIP is based upon a 100-year flood event.
Predictions developed from EPA's Hudson River depth of scour model (EPA, 1996) were used to
predict the PCB mass eroded from-cohesive sediments within the TIP (i.e., "hot spots") during a
100-year flood event. EPA estimated that approximately 25 kg of PCBs would be mobilized
during such an event (EPA, 1996). If this value is used as an upper bound estimate of the yearly
TIP "hot spot" sediment PCB scour, the total PCB scour from 1984 to 1994 is approximately 250
kg. This total mass represents only 2% of the total estimated TIP inventory of 13.5 MT. This
loading is conservatively high since it is based upon total PCBs in the TIP surface sediments; the
PCB;,. component of this is approximately 20-30% less (based upon 0-2 cm data from the EPA
high resolution cores).
Groundwater Advection
The upward flow of groundwater through surface sediments and into the TIP water
column results in a net loss of sediment PCBs. As groundwater travels through PCB-
contaminated sediments, the interstitial pore water reaches an equilibrium with PCBs in the
surficial sediment. The net PCB mass transport due to groundwater advection depends on the
rate of upward groundwater flow and the surface sediment pore water PCB concentration.
30
-------
As documented in QEA, 1998c. field sampling data indicate that spring 1997
groundwater-induced PCB flux is approximately 30 g/dav. Assuming this loading represents an
average for the entire year, groundwater advection from 1984 to 1994 results in the transport of
an estimated 110 kg of PCBs from the TIP sediments to the water column. This represents less
than 1% of the initial 1984 TIP inventory estimated in the DEER. This estimate is conservatively
high due to the assumption that the spring 1997 measurements are representative of groundwater
flux for the entire year, even though they were collected during a period in which the hydraulic
gradient between surface water and groundwater was expected to be at its greatest. Furthermore,
this estimate was developed for total PCBs, and, as with the diffusion estimate, the PCB>
fraction of this mass loss would be 44% less.
Total of Mass Loss Mechanisms
When the upper bound estimates of sediment PCB mass loss for the fate and transport
mechanisms discussed above are compared with the PCB3* inventory estimated from the DEIR.
the total mass loss attributable to these mechanisms can account for only 18% of the inventory.
This value is graphically compared in Figure 10 with the value of 80%, which was estimated for
PCB3,. mass loss using TAMS' approach. In conclusion, known mechanisms responsible for
PCB3- mass transport from sediment "hot spots" cannot account for the 1984 to 1994 mass loss
estimated using TAMS' approach for paired sediment core data.
C. There is no evidence of TAMS' mass loss in water column PCB data.
TAMS' estimated 10.8 MT of PCB3- mass loss from the TIP was compared with that
developed from water column data collected between 1984 and 1993. PCB analytical data
generated by the USGS (1984 - 1990) at the Schuvlerville Station and GE (1991 - 1993) at the
Thompson Island Dam station were used to develop this estimate. The packed column method
employed by USGS represents a PCB;,. analysis (HydroQual. 1998). Therefore, the USGS data
(1984-1990) were used as reported in the EPA database to develop an estimate of the PCBi-
mass loss. The 1991 - 1993 data were corrected to PCB;,- by subtracting the mono- and
dichlorinated biphenvls from the reported totals (O'Brien & Gere. 1993).
31
-------
The water column loading from the TIP was estimated as the difference between PCB
loading measured at the Schuylerville/Thompson Island Dam station and the Roger's Island
station (Table 7). These estimates are considered upper bound estimates of the loading across
the TIP because load gains between the Thompson Island Dam and Schuylerville (QEA, 1998c)
are also included in this estimate. From 1984 to 1993, the upper bound estimate of PCBj-r mass
loss from the TIP sediments is approximately 1 MT. This represents only approximately 7% of
TAMS' estimate of total PCB3+ mass in the TIP in 1984 and is inconsistent with the 80% mass
loss calculated using TAMS' analysis of the 1994 low resolution sediment coring data.
D. The loss rate indicated by comparing PCI^ in 1984 and 1994 is not supported by
the change in PCB3+ mass between 1994 and 1998.
The PCB3+ loss rate from the TIP is not supported by changes in PCB.^ mass between
1994 and 1998. TAMS states that the correct basis for comparing the 1984 NYSDEC and 1994
TAMS data is on a PCB3* basis. This results in an estimated 80% change in PCB3+ mass, a
PCB 3+ half-life of approximately 4 years within the TIP. While there may not be an identifiable
mechanism to account for such a loss (see Section IVB), if the loss was real, the trend in PCB3+
mass loss should continue. GE conducted an analysis in 1998 to evaluate the current status and
trend in PCB3-. within the TIP. If the trend in PCB3- observed between 1984 and 1994 was
continuing, the 1998 PCB3+ concentration should be half that observed in 1994. The opposite
was, in fact,-occurring; there was mass gain or no change between 1994 and 1998.
D.l Overview of the 1998 Sediment Coring Program 1.17
The principle objective of the 1998 sediment coring program (QEA, 1998a and b) was to
evaluate the validity of postulated changes in PCB3.. between 1984 and 1994, using sampling and
analytical techniques employed by TAMS in 1994. The 1998 program included 12 coring
locations within the TIP sampled both in 1984 and 1994. Coring stations were located in the
field using Differential Global Positioning System (GPS) technology. This location system was
tested in the field and shown to be accurate to less than 1 foot. The field crew was able to
32
-------
position the coring vessel to within 2 feet of the target coordinates reported in the EPA database
for the 1994 cores. Therefore, the separation distance between the 1994 TAMS core and 1998
GE core was no greater than 3 feet.
In contrast to the 1994 cores collected by TAMS, the surface sediment portion of the
1998 cores were finely sectioned. While TAMS segmented its cores into approximately 9-inch
segments, the 1998 GE cores were segmented into 1-cm segments within the first 0-5 cm of the
sediment core. This finer segmentation near the sediment-water interface provides data
necessary to directly assess PCB burial. Below the 0-5 cm segment, the 1998 GE cores were
segmented into 23-cm sections to correspond to the 1994 9-inch segments. All core sections
were analyzed for total organic carbon, moisture content, bulk density, and capillary column
PCBs in accordance with the work plan (QEA. 1998a and b). Additionally, the 1-cm segments
within the 0-5 cm portion of the core were analyzed for 13,Cs and Be.
D.2 Temporal Changes in PCBj- between 1984, 1994, and 1998 1.17A
For the 12 coring stations sampled within the TIP in 1998, changes in PCB;,- between
1984, 1994, and 1998 were calculated using TAMS' methodology (see Section IVA). The
results are presented as average mass change estimates for the periods 1984-1994 and 1994-1998
(Figure 11). The PCB;,- mass change estimate between 1984 and 1994 for this subset of 12
locations (approx. 80%) is consistent with the entire 1994 TAMS data set. Thus, these coring
stations are representative of the entire data set used by TAMS to estimate PCB mass changes
within the TIP. In contrast to the mass change observ ed between 1984 and 1994, PCB3- mass
was observed to increase by about 90% between 1994 and 1998 (Figure 12). The inconsistency
in this trend (mass loss between 1984 and 1994 and mass gain between 1994 and 1998) is
indicative of the inherent uncertainties undermining TAMS' approach to calculating PCB3- mass
loss using this type of matched coring data.
The inconsistency and variability in PCB3- mass changes between 1984 and 1998 is
further supported by the temporal trends in MPA}- for each of four "hot spot" areas (Figure 13).
In "hot spots" 8. 9, 14. and 16. a steep decline in MPA?- appears to occur between 1984 and
33
-------
1994. However, this trend reverses between 1994 and 1998. In "hot spot" 16, the 1998 estimate
of MPA3+ appears to be consistent with TAMS' 1994 data, suggesting no apparent change in this
"hot spot" between 1994 and 1998.
The apparent change in conditions within the 12 locations from mass loss between 1984
and 1994 to mass gain or no change between 1994 and 1998 can not be supported by any known
PCB fate and transport process (see Section IVB). If these areas represent potential scour areas,
as TAMS has contended, then the scour process should have continued between 1994 and 1998.
That the data do not support a continuing scour process demonstrates the inherent uncertainties
associated with trying to quantitatively compare sediment core sampling and analysis results
from approximately the same locations over a 10-year period. Sources of errors in this analysis
are of unknown magnitude because they most likely relate to spatial heterogeneity in surface
sediment PCB concentration and unverifiable differences in sampling and analysis protocols.
Consideration of only the two data sets in which these uncertainties are minimized (1994 and
1998) suggests that there has been a mass gain in PCBs between 1994 and 1998. TAMS' report
inappropriately glosses over the uncertainties that pervade its analysis.
E. If the loss rate indicated by comparing PCB3~ is real, little of the PCBs that are l. 17B
bioaccumulated remain in the Thompson Island Pool.
The fish in the Upper Hudson River contain almost no mono- and dichiorobiphenyl, a
consequence of the relatively low bioaccumulation potential of these PCBs. Fish PCB body
burdens consist almost entirely of tri-, tetra-, penta- and hexachlorobiphenvl. Thus, changes in
the PCB levels in fish are controlled entirely by changes in their exposure to PCB3+. The core-
pairs indicate that 79% of the PCB3+ inventory disappeared from the "highly contaminated"
sediments between 1984 and 1994. Extending this rate to 1998 results in a total loss since 1984
of about 90%. If this loss is real (and we do not believe that it is), little PCB3+ would be left in
the sediment, indicating a high rate of natural recovery and little benefit to PCB levels in fish
from remediation of the sediments.
34
-------
PCB levels in TIP fish are now the lowest they have ever been (based on 1997 data), but
it is not clear that this reduction is the result of a substantial inventory reduction in the sediments.
Levels peaked in 1992 or 1993. Since that time, there has been a rapid decline in concentration.
Between 1993 and 1997, largemouth bass lipid-based PCB concentrations declined by 75 percent
from about 2,500 ppm to 600 ppm. Similar declines have occurred in pumpkinseed (80%) and
brown bullhead (75%). The rapidity of these declines is partially the consequence of a recovery
from high loadings in the early 1990s. Because current levels are about 60% of the levels
observed in the years immediately preceding the high loadings, the decline likely also reflects a
decline in surface sediment PCB levels in response to virtually eliminating residual loadings
from the Hudsons Falls area.
35
-------
SECTION V
WIDESPREAD BURIAL OF PCB-CONTAMINATED SEDIMENT BY CLEAN
SEDIMENT DOES OCCUR IN THE THOMPSON ISLAND POOL
Rigorous data and modeling analysis indicate that widespread burial is occurring within
the TIP. While there may be regions, particularly along the edges of the sediment deposits, that
may undergo occasional scour, based upon EPA's sediment transport modeling, such loss
processes are limited both in frequency and magnitude (EPA, 1996). The widespread occurrence
of burial is supported by TAMS' surface sediment 7Be data and surface sediment PCB
concentration and GE's 1998 composition profiles.
A. Detected 7Be indicated deposition ranging from 0.2 to 3.0 cnvyr. Non-detect Be
indicated deposition less than 0.5 cm/vr.
'Be is a naturally occurring isotope with a half life of 53 days. It is produced by cosmic
radiation entering the earth's atmosphere. 7Be enters the water column through atmospheric
deposition in the form of precipitation with higher concentrations generally occurring in spring
or early summer (Olsen et al„ 1986). Due to its relatively short half life, the presence of "Be in
surface sediments generally indicates that suspended matter has been recently deposited.
TAMS found that ' Be was present in 70% of the surface sediment samples (0-1"):
Of the 169 cores analyzed for Be, 119 cores [70%] indicated the presence of Be and,
therefore, recent deposition... (Report, at 2-21)
These data indicate, as TAMS stated, that these areas are depositional. However, due to the 7Be
decay rate and the time between probable Be deposition and TAMS' measurement of the
element, TAMS cannot conclude that lack of 'Be in 30% of the coring sites is evidence for a lack
of burial. Moreover, it is possible to assess the magnitude of sediment burial indicated by the
36
-------
'Be detected in each of the sediment cores and the upper bound estimate for burial in cores with
non-detect 7Be based on:
1) the laboratory's ability to quantify 'Be,
2) an estimate of 7Be concentrations in settling particles, and
3) our understanding of sediment transport processes.
Deposition Rates Indicated bv 7Be Presence
The deposition rate necessary to produce the 7Be concentrations within TAMS' 1-inch
surface sediment segment was calculated for each of the cores with detectable Be as follows:
, [ S€f ] d lee
" J'P I In.?
[ Ben] e i,:'
where:
d^ep is the depth of sediment deposition occurring during the spring high flow period,
['Be,] is the detected Be concentration reported in the EPA database.
dseg is the depth of the Low Resolution Sediment Core segment used for Be analysis
(2.54 cm),
['Beo] is the concentration of Be on particles deposited during the high flow event,
t!,2 is the half life for Be (53 days), and
t is the time elapsed between the high flow period in 1994 (April 17) and the date of Be
analysis as reported in the EPA database.
To perform this calculation, a number of assumptions were required. First, it had to be
assumed that all significant sediment deposition occurs dunng the spring high flow period, which
is consistent with GE's understanding of sediment fate and transport processes (QEA, 1998c;
Appendix B). However, limited amounts of deposition may occur during other periods of the
year. Second, the concentration of Be on particles deposited during the spring high flow penod
37
-------
had to be estimated. Water column particulate phase 7Be concentrations cited in several sources
of literature range from an average of approximately 45,000 pCi/kg in Lake Constance (Vogler
et al., 1996) to an average of approximately 5,500 pCi/kg in the Hudson-Raritan Estuary (Olsen
et al., 1986). A value of 20,000 pCi/kg was selected as a conservative estimate of the water
column particulate phase 7Be concentration (conservative in the sense that this value will
produce a lower deposition rate).
The results of this calculation are presented for each of the low resolution cores with a
detectable 7Be concentration (Table 8). The depth of deposition required to produce the 'Be
detected in the surficial 1 inch of the cores ranged over an order of magnitude, from 0.2 to 3.0
cm. These estimates were developed into averages for each of the sampled "hot spots" within the
Upper Hudson River (Figure 14). On average, this analysis indicates that the "hot spot" areas
were subject to significant deposition ranging from 0.5 to 1.5 cm/year in 1994. This estimate is
limited by the above assumptions but it indicates that in 70% of the sampled areas, significant
7Be and, therefore, sediment deposition, occurred in 1994.
Lack of Surface Sediment 'Be
The fact that 30% of the top 1-inch segments of the cores lacked detectable Be cannot be
used to indicate a lack of burial in these areas, as TAMS indicated. The data can only support
n
the conclusion that burial has not occurred in sufficient amounts to produce detectable Be
concentrations. This is due to several factors, including:
1) the 7Be decay rate,
2) the elapsed time between Be deposition and sample analysis, and
3) mixing of deposited 'Be and non- Be-containing sediments within the 2.54 cm
segment collected by TAMS.
GE calculated estimates of the minimum required depth of sediment deposition necessary
to achieve detectable levels of Be in the surficial 2.54 cm section of the low resolution sediment
cores. The bulk of sediment deposition occurs during the spring high flow period (on or around
April 17, 1994), and 'Be concentration of depositing particles was 20,000 pCi/kg. Accounting
38
-------
for radioactive decay, the concentration of 'Be in the top 1 inch of sediment was calculated and
plotted as a function of depth of sediment deposited using the following equation:
[ Be
-------
B. Cores collected in 1998 consistently exhibit a buried peak PCB concentration with a 1.18A
decline at the sediment core surface, a pattern consistent with burial.
One of the lines of evidence TAMS presents to support its conclusion that PCB deposits
are not being buried is that the PCB maxima occur within the top 9 inches of the sediment cores.
Sediment-water interactions and fish PCB exposure through the food chain occur in the top few
centimeters, so TAMS' core segmentation scheme was too coarse to support definitive
conclusions regarding the importance of burial in sequestering PCBs from the water and biota.
To remedy this situation, GE collected sediment cores in 1998 from many of the same locations
where TAMS collected cores and segmented them into 1-cm segments within the surficial 5 cm
(see Section IVD.l for further details). This allowed the assessment of the burial process on a
much finer scale than that of the 1994 TAMS' cores.
The results of the 1998 sediment coring program are presented graphically in Figures la-
11. Most of the 1998 data (11 out of 12) indicate the presence of a steep PCB concentration
gradient between the 0-5 cm segment and underlying segments. This trend indicates that the
higher PCB concentrations at depth in the cores are being buried by solids containing much
lower PCB concentrations. Moreover, in many of the 1998 cores, the PCB maxima occur well
below TAMS' maxima. This suggests that TAMS' coring program did not capture the full
extent of the PCB bearing sediments at these sites, or that spatial heterogeneity on the scale of
several feet may produce profound changes in sediment PCB profiles. In any event, the 1998
data indicate that PCBs are being actively buried within the "hot spots" of the Upper Hudson
River.
The vertical profiles of PCB composition further support the conclusion that PCBs are
being actively buried within sediment deposits of the Upper Hudson River. Vertical trends in
PCB composition indicate that surface sediments (0-5 cm) contain PCBs which more closely
resemble the Aroclor 1242 congener pattern than deeper sediments which have undergone
extensive reductive dechlorination. Figure 16 presents vertical sediment profiles of the PCB
molar dechlorination product ratio (MDPR; Report). The MDPR is defined as:
1 A 1 A C A
40
-------
MDPR = Z ''';y
I BZ,
where:
BZ refers to the numbers assigned PCB congeners by Ballschmiter and Zell (1980).
PCBs that have been completely dechlorinated and contain only congeners 1, 4, 8, 10, and 19
possess an MDPR of 1.0. This ratio has been used by EPA to determine the dechlorination status
of sediment bound PCBs (EPA, 1997, 1998).
The MDPR of surficial sediment (0-5 cm) collected in 1998 is consistently lower than
that corresponding to the 5-23 cm depth or the 0-9 inch depth of the 1994 cores. Surface
sediment PCBs general possess an MDPR of between 0.2 and 0.4. In contrast, PCBs within the
5-23 cm segments and TAMS' 1994 0-9 inch segments have an MDPR which generally falls
within the 0.6 to 0.8 range. This indicates that the PCBs at the sediment surface have not
undergone dechlorination as extensively as the deeper sediments. Therefore, PCB-containing
sediments being deposited within these regions of the river more closely resemble the Aroclor
1242 pattern associated with the discharges from the vicinity of the Hudson Falls plant site
(O'Bnen & Gere, 1994; EPA, 1997).
Similar to the MDPR, the ratio of DB-1 peaks 46 to 32 can be used as an indicator of the
dechlorination status of sediment PCBs. Peak 46 contains a tetrachlorobiphenvl that is sensitive
to dechlorination (BZ# 74) and peak 32 contains a tetrachlorobiphenvl which is relatively
resistant to dechlorination (BZ# 49). Therefore, the ratio of these peaks decreases with
dechlonnation. The profiles presented in Figure 17 generally depict a decreasing peak 46 to 32
ratio with sediment depth, indicating that the surface sediment PCBs are less dechlorinated than
deeper sediments. This observation is important because, unlike the surface sediment gradient
in MDPR, the peak 46 to 32 ratio cannot be produced by simple elution of the lighter chlorinated
biphenyls from the surface sediment. This is because both peak 46 and peak 32 are
tetrachlorinated biphenyls with similar chemical properties, subjecting them to similar physical-
41
-------
chemical fate processes within the sediments. Therefore, the changes in the peak 46 to 32 ratio
are the direct result of dechlorination of recently deposited PCBs.
1.18B
C. Data analysis and modeling of the 1994 spring flood indicates widespread net
deposition in fine-grained sediments.
Construction of a mass balance for the 1994 spring flood in the Thompson Island Pool
using data collected from March 30 through April 29 indicates that 457 MT of sediment were
exported from this reach. This type of data analysis can only be used to estimate global losses
due to net erosion from the TIP sediment bed (or global gains due to net deposition). Results
from this analysis cannot be used to infer net erosion or deposition in specific bed types, e.g.,
cohesive or non-cohesive, or areas of the TIP.
A sediment transport model, which has been extensively calibrated and validated, is an
effective diagnostic tool for quantitatively evaluating net deposition and erosion from various
areas in the TIP during the 1994 spring flood (Appendix B). The model predicts a total of 345
MT of sediment were exported from the TIP during the 30-day period considered here, which
agrees with the measured value. Closer examination of the model results, however, show that
net erosion did not occur in all areas of the TIP during this flood. The non-cohesive portion of
the TIP sediment bed, which comprises approximately 80% of the total area in this reach,
experienced a net loss of 1,244 MT. This net erosion corresponds to a decrease in the mean
elevation for the non-cohesive bed of 0.08 cm. Conversely, 899 MT of net deposition occurred
in the cohesive bed of the TIP, which is equivalent to an average increase of 0.26 cm of the
cohesive bed elevation. Prediction of net deposition in the TIP cohesive bed area during this
flood is consistent with observed depositional patterns in fine-grained areas of the Upper
Mississippi River during major flooding in 1993 (Barber and Writer, 1998).
42
-------
1.18C
D. Sediment transport modeling indicates that the areas with non-detectabie Be are
depositional, but the deposition rate is less than that in areas with detectable Be.
The TIP sediment transport model was coupled to a 7Be fate and transport model to
7 **
investigate the relationship between Be concentrations and deposition rates in the Pool. The Be
fate and transport model, which was calibrated using 7Be concentration data collected in 1994,
can predict the spatial distribution of 7Be in the TIP with reasonable accuracy. These models
were then used to examine sediment deposition and 7Be concentration distributions in the TIP for
the 6-month period prior to 7Be core sampling in July 1994.
A majority of the non-detectable Be samples in cohesive bed areas of the Pool were
collected in "hot spot" 14; the rest of the non-detectable 7Be cores were randomly distributed in
areas with detectable 7Be. The predicted mean deposition for the non-detectable Be core
locations in "hot spot" 14 was 0.16 cm (for the 6-month period from January to July 1994), with
deposition occurring at all of the core locations. The model predicted a mean deposition of 0.69
cm for the detectable 7Be samples in the TIP during this 6-month period, which is 4.3 times
higher than the deposition that occurred at the non-detectable Be cores in "hot spot" 14. Thus,
non-detectable Be concentrations in cohesive bed areas are found in areas of relatively low
deposition, at least for the time period under consideration.
43
-------
SECTION VI
EXISTING DATA SHOW NO EVIDENCE OF EXPOSURE OF
PREVIOUSLY BURIED PCBS VIA EROSION
1.19A
A. PCBs increase in a nearly linear fashion as water passes through the TIP, indicating
a nearly uniform areal flux from sediments within the Pool.
Surveys conducted within the TIP revealed that PCBs increase in a nearly linear fashion
as water passes through the TIP, indicating a nearly uniform areal flux from sediments within the
Pool. These surveys were conducted within the TIP to detect potential areas that may be
contributing a disproportionate quantity of PCBs to the water column load (O'Brien & Gere,
1998). In summary, the surveys consisted of sampling along lateral transects established every
0.25 to 0.5 miles between Roger's Island and the Thompson Island Dam, with sampling stations
at three positions across each transect: east shore, west shore, and center channel. Transects
were sampled from upstream to downstream so as to correspond with the flow of river water.
Stations along each transect were sampled simultaneously and consisted of vertically stratified
composite samples collected from three depths and were analyzed for PCBs and TSS.
The surveys exhibited similar spatial trends in total PCB concentration within the center
channel (O'Brien & Gere, 1998; Figure 18). PCB concentrations were generally at or near the
method detection limit of 11 ng/'L at the Roger's Island sampling station and increased gradually
to approximately 30 ng/1 over the first 2 miles of the TIP, to River Mile 193. Over the four-mile
section of the TIP between River Mile 193 and 189, center channel PCB concentrations
increased by approximately 40 to 60 ng'L. At average flows of approximately 4,000 cubic feet
per second (cfs) observed during the surveys, this increase represented a mass loading rate of 0.4
- 0.6 kg day*1. These mass loading rates represent sediment areal flux rates of approximately 0.3
to 0.4 mg m"2 day"1 across this region of the TIP. These data do not contain any evidence of
enhanced PCB flux within the sediment "hot spot" areas. The mass loading rate is generally
consistent with the load expected from the entire TIP sediment area, considering 1991 surface
sediment PCB concentrations.
44
-------
Spatial patterns in water column PCB concentrations depend upon spatial variations in 1.19B
sediment PCB flux. The flux of PCBs from surface sediments to the water column depends on
the organic carbon normalized PCB concentration, the sediment-water exchange coefficient, and
the PCB partition coefficient (QEA, 1998c). Regions of the river with equal surface sediment
organic carbon normalized PCB concentrations and composition contribute equally to the water
column PCB load. Data gathered by NYSDEC in 1984 indicate that mean organic carbon
normalized PCB3+ concentrations are similar inside and outside the sediment "hot spot" areas
(Table 9).6 Moreover, organic carbon normalized total PCB concentrations were similar for both
coarse-grained and fine-grained sediments collected in 1991 from the TIP (Table 9). Therefore,
coarse-grained and fine-grained sediment areas and "hot spot" and "non-hot spot" areas are
expected to have similar sediment pore water PCB concentrations and, through the process of
sediment diffusion, similar areal PCB fluxes. Such conditions would produce the pattern of
gradually increasing water column PCB concentrations observed within the center channel
during the surveys (Figure 18). The sediment data suggest that the "non-hot spot" areas
dominate water column PCB loadings because they constitute the vast majority of the river
bottom.
The surveys did not reveal any localized regions of elevated surface sediment PCBs
within the Pool that are disproportionately contributing to the water column PCB load. PCB
loadings characterized using center channel data uninfluenced by localized hydrodynamics
depict an approximately uniform increase in PCB mass loading that is consistent with surface
sediment exchange processes and the 1991 surface sediment PCB concentrations.
6In this analysis. 1984 organic carbon concentrations were estimated as 40% of the reported volatile solids
concentration.
45
-------
B. The composition of the TIP load is consistent with the surface sediment PCB i .20
composition considering equilibrium partitioning and sediment pore water
exchange processes.
The composition of the TIP load is consistent with known and understood sediment-water
exchange processes and the composition of surface (0-5 cm) sediments. The composition of the
summer low-flow (June - August 1997) average TIP load was calculated as the difference in
water column derived PCB peak loading across the TIP using unbiased data (QEA, 1998c)
collected from Fort Edward and the vicinity of the Thompson Island Dam. The source of this
loading was assessed by calculating the required composition of a surface sediment source,
assuming equilibrium partitioning between sediments and pore water and a diffusive mass
transport mechanism. Specifically, the approach included:
1) Calculation of TIP water column PCB peak (based on a DB-1 capillary column)
loadings from paired Fort Edward (Cfe) and unbiased TID water column PCB data (Ct,d)
in accordance with the following equation:
Wwc = Ok(C„!i-Cfj
where:
Qfe is Fort Edward flow (LJ T'1), and
2) Calculation of the sediment-phase PCB composition assuming the load calculated
using the above equation originates from surface sediments and is transported to the
water column via diffusional processes. This load was calculated on a DB-1 peak basis
in accordance with the following equation:
46
-------
_ Wwcfock„c
/I
kf At (I Wldoc kdncJ
The parameters used in the calculation are summarized in Table 10.
Back calculating the particulate-phase PCB concentration of surface sediments yields the
PCB DB-1 congener peak distribution in Figure 19. This PCB source best matches the surface
sediment PCB composition as represented by the 0-2 cm sections of the EPA high resolution
cores collected from the TIP in 1992 (Figure 19). In contrast, the source of the TIP load does not
appear to match the composition of PCBs found at depths greater than 8 cm (Figure 19). This
analysis indicates that the source of the TIP PCB load is surface sediments as expressed through
a diffusive flux mechanism.
C. The composition of PCBs in fish is consistent with exposure to relatively 1.21
undechlorinated PCBs found in surface sediments and not the dechlorinated PCBs
found in buried sediments.
Fish sampled in the TIP in 1993 by NOAA were analyzed for PCBs by congener. These
data were used to investigate whether the fish had been exposed significantly to dechlorinated
PCBs. The approach involved using the ratio of dechlorination sensitive and dechlorination
resistant congeners that had similar bioaccumulation potential to fingerprint the PCB. Figure 20
shows ratios for two congener pairs: BZ 56:BZ 49; and BZ 60:BZ 49. Results for BZ 74:BZ 49
were similar; data is not shown. These pairs have ratios in TIP sediments that are sensitive to the
extent of dechlorination (Figure 20). In these panels, the congener ratios in sediment samples
collected by EPA as pan of the High Resolution Core Sampling in 1992 are plotted against the
level of chlorination. The horizontal dashed lines on the panels indicate the congener ratio in
Aroclor 1242. In the least dechlorinated sediments (highest number of chlorines per biphenvl)
the ratio is about equal to that of Aroclor 1242. As the chlorination level decreases (moving to
the right on the panels), the congener ratio decreases dramatically. The congener ratios in six
species of fish are shown in the right panels of Figure 20. In all of the species, the ratios are
47
-------
similar to that seen in the undechlorinated sediment. There is no evidence that the fish were
exposed to dechlorinated sediments.
D. The PCB composition in the water column during erosion events is consistent with 1.22
the relatively undechlorinated PCBs found in surface sediments and not the
dechlorinated PCBs found in the buried sediments.
A large flood event occurred in the Upper Hudson River in January of 1998. That flood
had a peak flow of about 36,000 cfs at Fort Edward, equating to a 12- to 14-year return period.
The estimated flow in the TIP tributaries suggests that they experienced rarer flows than the
main stem, possibly equating to a 50- or 100-year return period. Samples taken at the TI Dam
and Schuylerville at the peak flow and on the downward limb of the flood were analyzed for
PCBs. The PCB composition of these samples was estimated by comparing the ratio of
dechlorination sensitive and dechlorination resistant chromatogram (DB-1 Column) peaks.
These peak ratios were compared to those found in surface sediments in 1991 (0-5 cm) and in
deeper dechlorinated sediments. For all three of the peak ratios evaluated (50/32; 43/32; 74/61),
the values were consistent with the surface sediments and inconsistent with the dechlorinated
sediments (Figure 21). Thus, these data exhibit no evidence that previously buried dechlorinated
sediments have become exposed through erosion.
E. Cs levels in the low resolution cores give no indication that scour sufficient to 1--J
account for a significant loss of PCB mass has occurred.
If a significant portion of the deeply buried, highly contaminated PCBs has been scoured
out of Thompson Island Pool "hot spots," then the !37Cs concentrations observed in the low-
resoiution cores should provide an indication that the sediments near the surface are of sufficient
age to have lost the appropriate mass of PCBs.
Figure 22 illustrates average temporal profiles of both total PCBs and 13 Cesium for the
high resolution Cores HR-019. HR-020, HR-023. and HR-026 located in the TIP. The top and
48
-------
bottom of each core section was dated by first computing a deposition rate for each core
according to the following equation:
Dcs
Deposition Rate = -—
T cof T Pk
where Dcs is the 137Cs peak depth [cm], TCOi is the year in which the sample was collected (1992).
and Tpk is the assumed year in which the l3?Cs peak occurred (1963). Using these deposition
rates, the top and bottom of each core section was dated using the following equation:
Section Date = 1992 -
Deposition Rate
where Dsect10n is the top or bottom depth of each core section [cm]. The top and bottom dates
were then averaged to determine a composite date for each section.
The data were grouped in five-vear increments and plotted at the midpoint. Everything
pnor to 1950 was plotted at 1950. The dotted horizontal lines represent the Cs levels in the
surface sediments (0-1 inch) of each low-resolution core exhibiting "evidence for sediment
scour," according to Table 4-1 of the Report.
All of the l3'Cs levels measured in the low resolution "scour" cores are consistent with
post-1975 levels measured in the high-res cores. None of the "scour" cores lie within the error
bars of the pre-1975 hi-res data. However, the bulk of PCB mass resides before 1975. Thus, the
l37Cs values provide no indication that scour down into the bulk of the PCB mass has occurred.
Such scour would be required if 40 to 80 percent of the PCB mass had been lost between 1984
and 1994.
49
-------
SECTION VII
ACCURATE ESTIMATION OF THE CHANGE IN PCB MASS IN "HOT SPOTS" i\24
BELOW THE TIP IS NOT POSSIBLE USING THE AVAILABLE DATA
In estimating the mass change in seven "hot spots" below the TIP, TAMS log-
transformed the data in order to use parametric statistics to test for significant differences
between the 1976-78 and 1994 PCB mass estimates. For those "hot spots" having statistically
significant mass changes, absolute mass changes were estimated from the differences in the
arithmetic mean estimates of the masses. In performing this later calculation, TAMS failed to
consider the uncertainty of calculated mass changes. The standard error of the calculated mass
change is the square root of the sum of the squares of the standard errors of the mean mass
estimates. The results of such a calculation are presented in Table 11. As the table shows, the
mean mass change has an uncertainty range that crosses no change for all of the "hot spots."
A) None of the mean mass changes is statistically different from no change.
The uncertainty of the mass change calculations is, in part, due to the small number of
samples. Because of the extreme spatial heterogeneity of sediment PCB levels, a large number
of samples is necessary. The 1984 study of the TIP sediments was conducted using a sampling
density of 3 to 4 samples per acre. As was noted in Section II. the estimate of PCB mass in TIP
sediments in 1984 has considerable uncertainty. The Low Resolution Coring Study involved
collecting between 4 and 14 samples in each of seven "hot spots." Based on the areas of these
"hot spots," the sampling density averaged about 0.5 samples per acre, almost an order of
magnitude less than that of 1984. This crude sampling makes estimation of mass problematic.
Further, the sampling conducted in 1976-78 was also relatively crude, so that the estimate of the
mass present at that time is highly uncertain.
In addition to the quantified uncertainty, EPA's mass estimates contain additional
uncertainty resulting from two assumptions. The impact of these assumptions on the overall
50
-------
uncertainty was not included in the comparison of 1976-78 and 1994 mass estimates. These
assumptions are as follows:
1) PCB measurements in grab samples can be extrapolated to deeper depths to
estimate inventory at the locations of the grab samples, and
2) The sediment density in 1976-78 samples can be estimated from the relationship
between PCB concentration and sediment density found in 1994 samples.
Both of these assumptions are made with no independent data from which they may be tested.
They impart unknown uncertainty to the results which undermines the significance of any
observed differences.
Perhaps the best indication of the overall uncertainty of the approach TAMS used is the
result obtained for "hot spot" 28. The implausibly large increase in mass in this "hot spot" is
dismissed. Arguments are presented that reflect the inherent uncertainty in comparing data from
dissimilar sampling programs. These arguments apply in general and accurately indicate the
futility of the approach. Further, the small sample sets contribute to difficulty in interpreting the
result in "hot spot" 28. In fact, the mean mass increase in "hot spot" 28 has uncertainty bounds
that cross no change (Table 11).
51
-------
SECTION VIII
DECHLORINATION ANALYSES ARE FLAWED AND INSENSITIVE
The report makes three major conclusions regarding PCB dechlorination in the Hudson
River:
A) Dechlorination occurs at meta and para positions only.
B) The mean mass loss [of PCBs on a mass basis] is less than 10% assuming Arolor
1242 was the original mixture of PCBs.
C) The degree of dechlorination increased with the log of the PCB concentration.
A. Dechlorination occurs at meta and para positions only.
The report does not acknowledge the dramatic effects of these processes on
physicochemical and toxicological properties of PCBs as described in Appendix D of GE's
comments to the DEIR (GE, 1997) including:
• Reduced toxicity;
• Reduced carcinogenicity;
• Reduced exposure via aerobic degradation; and
• Reduced bioconcentration.
B. The mean mass loss (of PCBs on a mass basis) is less than 10% assuming Arolor
1242 was the original mixture of PCBs.
The approaches to measuring mass loss in the Report are identical to those used in the DEIR,
thus the flaws and insensitivities are identical to those indicated in GE's comments to the DEIR.
They are briefly discussed here.
52
-------
The simplistic dechlorination indices developed in the DEIR and used in the Report
ignore known processes that can dramatically affect the PCB congener distribution. These
include partitioning of lower-chlorinated congeners to the water column, evaporation,
biodegradation and uncertainties regarding the mixture of Aroclors originally adsorbed onto the
sediment. Therefore, these approaches do not provide definitive information about PCB
dechlorination and have no scientific merit.
In addition, the MDPR is insensitive due to the selection of only "terminal dechlorination
products" to assess the extent of dechlorination. It is only capable of detecting extensive
dechlorination in sediments that carry out nearly, but not complete, removal of meta and para
chlorines. It ignores dechlorination of the most abundant congener in Aroclor 1242 (2,4'-
dichlorobiphenyl) which represents -12% on a mass basis.
C) The degree of dechlorination increased with the log of the PCB concentration. 1.27
Samples with PCB concentrations >30 ppm exhibited various levels of dechlorination,
while samples with concentrations <30 ppm were relatively unaltered.
An outcome of the flaws and insensitivies discussed above is the enormous uncertainty
associated with regressions of dechlorination with PCB concentration, leading to the erroneous
conclusion that PCBs are not predictably dechlonnated at concentrations less than 30 ppm. This
uncertainty is graphically represented in the regressions of Figures 4-22 and 3-8 of the DEIR and
the Report, respectively. The 95% confidence limits encompass 50% of the dechlorination
range. This conclusion misrepresents the data, which clearly show that the majority of Hudson
River sediments show some degree of dechlorination on careful examination. In addition, the
"removal of cross-contaminated samples" from the data set examined in the Report is not
scientifically valid and misrepresents the heterogeneity of microbially mediated processes in the
Hudson River.
53
-------
SECTION IX
ESTIMATION OF THE FATE OF SEDIMENT PCBS REQUIRES INTEGRATION
OF ALL OF THE DATA AND APPLICATION OF THE
QUANTITATIVE MASS BALANCE MODELS
To allow objective, quantitative evaluation of potential remedial measures in the Hudson
River, GE and EPA have sponsored the development of state-of-the-art PCB fate, transport, and
bioaccumulation models. It is through the application of such tools that sound remedial action
decisions regarding PCB-containing sediments can be formulated and quantitatively evaluated.
In the absence of such tools, decisionmakers are forced to rely on potentially biased conceptual
models that may be supported by a subset of the database, but can not be accurate due to their
inconsistency with other portions of the data record. For example, while comparison of the 1984
and 1994 PCBj-r data would suggest that approximately 80% of the PCB?- were lost from the
TIP between 1984 and 1994, this loss is inconsistent with PCB}* loading estimates based upon
water column monitoring over the same period (Section IV). Because models are developed and
calibrated against the entire data record, numerous data constraints are applied to minimize
inaccurate assessments of the river conditions and dynamics.
54
-------
REFERENCES
Ager, D.V. 1981. The Nature of the Stratigraphical Record, John Wiley and Sons, New York.
Ballschmiter, K. and M. Zell. 1980. Analysis of Polychlorinated Biphenyls (PCB) by Glass
Capillary Gas Chromatography. Fresenius Z. Anal. Chem. 302, 20-31.
Barber, L.B. and Writer, J.H., 1998. Impact of the 1993 Flood on the Distribution of Organic
Contaminants in Bed Sediments of the Upper Mississippi River, Environ. Sci. Technol,
32:2077-2083.
Brown Jr., J.F., R.E. Wagner, and D.L. Bedard. 1984. PCB Transformations in Upper Hudson
Sediments. Northeast Environ. Sci. 3:184-189.
Brown, M.P., M.B. Werner, C.R. Carusone and M. Klein. 1988. Distribution of PCBs in the
Thompson Island Pool of the Hudson River: Final Report of the Hudson River PCB
Reclamation Demonstration Project Sediment Survey. NYSDEC, Albany, New York.
Fish, K.M. and J.M. Principe. 1994. Biotransformations of Aroclor 1242 in Hudson River Test
Tube Microcosms. Applied and Environ. Microbiology, Dec. 1994, p. 4289-4296.
General Electric Company, 1997. Comments of General Electric Company on Phase 2 Report-
Review Copy Further Site Characterization and Analysis, Volume 2C. Data Evaluation
and Interpretation Report. Hudson River PCBs Reassessment RI/FS, February, 1997.
HydroQual, 1998. Memorandum from J. Rhea of HydroQual, Inc. to M. Schweiger and J.
Haggard of GE Corporate Environmental Programs, Albany, NY, regarding analytical
bias in the USGS water column data set, dated 29 January, 1998.
Lick, W. 1992. The Importance of Large Events, prepared for Modeling Uncertainty Workshop,
Buffalo, New York.
-------
McNulty, A. K. 1997. In-Situ Anaerobic Dechlorination of Polychlorinated Biphenyls in Hudson
River Sediments. Master of Science Thesis Submitted to Rensselaer Polytechnic Institute,
Troy, New York, August, 1997.
O'Brien & Gere Engineers, Inc. 1998. 1996-1997 Thompson Island Pool Studies. Prepared for
General Electric Company, Corporate Environmental Programs. Albany, NY. January, 1998.
O'Brien & Gere Engineers, Inc. 1994. Bakers Falls Operable Unit 3, Remedial Investigation
Report: Syracuse, N.Y. O'Brien & Gere Engineers, Inc. Prepared for the General Electric
Company Corporate Environmental Programs. Albany, N.Y. January 1994.
O'Brien & Gere Engineers, Inc. 1993. Temporal Water Column Monitoring Program Report.
Hudson River Project, 1991-1992 Sampling and Analysis Program. Prepared for General
Electric Company Corporate Environmental Programs, Albany, NY. May, 1993.
Olsen, C. R., I.L. Larsen, P.D. Lowry, and N.H. Cutshall. 1986. Geochemistry and Deposition
of 7Be in River-Estuarine and Coastal Waters. Journal of Geophysical Research. Vol. 91,
pp. 896-908
Quantitative Environmental Analysis. LLC, 1998a. Thompson Island Pool Sediment Coring
Program. Prepared for General Electric Company Corporate Environmental Programs.
Albany, New York. June, 1998.
Quantitative Environmental Analysis, LLC, 1998b. Addendum to the Thompson Island Pool
Sediment Coring Program. Prepared for General Electric Company Corporate
Environmental Programs, Albany, New York. August, 1998.
Quantitative Environmental Analysis, LLC, 1998c. Thompson Island Pool Sediment PCB Sources.
Prepared for General Electric Company Corporate Environmental Programs. Albany, New
York. March, 1998.
56
-------
Rhee, G.Y., R.C. Sokol, B. Bush, and C.M. Bethoney. 1993. Long-Term Study of the Anaerobic
Dechlorination of Aroclor 1254 with and without Biphenyl Enrichment. Environ. Sci.
Technol. 1993, 27, p. 714-719.
Tofflenine, T.J. and S.O. Quinn. 1979. PCB in the Upper Hudson River: Mapping and Sediment
Relationships. NYSDEC Technical Paper No. 56. March 1979. NYSDEC Albany, NY.
U.S. Environmental Protection Agency, 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 developed for the USEPA Region 2 by TAMS Consultants et al. July
1998.
U.S. Environmental Protection Agency, 1997. Phase 2 Report - Review Copy, Further Site
Characterization and Analysis, Volume 2C - Data Evaluation and Interpretation Report,
Hudson River PCBs Reassessment RI/FS developed for the USEPA Region 2 by TAMS
Consultants et al. February 1997.
U.S. Environmental Protection Agency, 1996. Phase 2 Report - Review Copy. Further Site
Characterization and Analysis. Preliminary Model Calibration Report. Hudson River
PCBs Reassessment RI/FS. U.S. EPA, October, 1996.
Vogler, S., M. Jung, and A. Mangini. 1996. Scavenging of 234Th and 7Be in Lake Constance.
Limnol. And Oceanogr., 41(7), 1996, pp. 1384-1393.
57
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
TABLES
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
Table 1
Ratio of PCB Levels in Paired Samples from the H7 ("Hot Spot" 5) Site at Various
Distances of Separation
Distance Between Paired Samples
i Average Ratio of the Paired Samples
(feet)
1 (Hieh/Low)
0-2
2.1
2-5
i 5.5
5 - 10
21.5
-------
Table 2
GE 1998 Hudson River PCB Sediment Samples Analyzed by both Capillary
Column C.C and Packed Column GC Methods
Versar/NYSDEC2
FS
Depth Interval
DB-I Capillary Col.
DB-1 Capillary Col.
Total Chlorines
Packed Column
Sample ID
(cm)
Total PCBs (ug/g)
PCBs'^ug/g)
per Biphenyl
Total PCBs (ug/g)
FS-08-1
46-67
1055.50
177.32
1.85
137
FS-08-2
23-46
785 24
140.24
1.87
104
I S 08-3
0-1
27.06
12.42
2 34
12.9
FS-08-3
3-4
82 61
27.28 *
2 09
29.4
FS-08-3
5-23
683 47
133.41
1.91
121
FS-08-4
0-1
27.16
14.64
2.49
15.7
FS-08-4
23-46
680.70
127.84
1,90
115
FS 09-1
3-4
4.11
1.J5
2.01
1.42
FS-09-3
1-2
8.68
.28
1.78
2,56
FS-09-3
23-46
69 61
13.94
1 92
12
FS-09-4
46-69
1 18.42
26.12
1.94
27.8
FS-09-2
0-1
2 46
1.05
2.22
1.07
FS-09-2
23-49
514.73
79.37
1.81
73.7
FS 14-1
3-4
99.92
18.73
1.83
21.1
FS-I4-1
4-5
375.49
61.84
1.84
70.3
FS-16-3
0-1
6 14
3.26
2.53
4.16
FS-16-3
4-5
13 17
6,17
2.37
8.54
FS-16-3
5-23
60 73
11.33
1.95
13.1
FS-16-1
0-1
3 13
1.70
2.55
2.3
FS-16-2
4-5
1,63
1.07
2.8)
1.33
FS-16-2
5-23
21.03
7.02
2.22
7.88
'Summation of tri- through decachlorinated bipliuiyls
"Total PCB is the sum of Aroclors 1242 + 1254 » 1260 where 1242 was calculated by NYSDEC's 1984 method
using peak H's 28,47 and 58 and Aroclors 1254 and 1260 were calculated using the chromatogram divisions
reported in Webb & McCall( 1973).
-------
Table 3
Comparison of PCB Analytical and Quantitation Methods Used for 1984 and 1998 Sediment Samples
Date
Laboratory
Column
Quantitation of Total PCBs
1984
Versar, Inc.
1.8m x 2mm or 1 mm ID glass column
Quantitation of 1254, 1260 (Webb & McCall, 1973)
Springfield, VA
3%OV-l on 100/120 Gas Chrom Q.
NYSDHC recalc. 1242 by scaling average of the
Ramp 180-210 C
weighted responses of three peaks:
5% methane in argon, 50 ml./min
T-PCB= 1242(N YS DF.C) f 12 54(Versar)+1260( Versar)
1998
Northeast Analytical, Inc.
2m x 4mm ID glass column
Followed same quantitation scheme used by NYSDEC
Schenectady, NY
1.5% SP-2250/1.95% SP-2401 on
for Aroclor 1242. Used Webb & McCall Wt% to calculate
100/120 mesh Supelcoport
1254 and 1260
Ramp 155-225 C
53 ml./min Nitrogen
30m x 0.25 mm II) capillary column
Summation of individual PCB congeners represented in 118
DB-1 (J&W bonded polydimethylsilicone)
DB-1 peaks.
Ramp 50-220 C
30 cm/sec. Helium
-------
Table 4
Northeast Analytical, Inc. Packed Colun
-------
Table 5
Calculation of PCB Mass Change in Sampled Locations Using TAMS Approach
Total MPA (g/m2)
MPA3.
(g/m2)
log (DPCB +2)
ID
1984
1994
1984
1994
TAMS
pcb3.
LR-01A
32.1
6.2
30.0
2.9
0.08
0.04
LR-01B
26.1
4.8
24.4
2.4
0.08
0.04
LR-02A
221.3
12.2
206.7
4.1
0.02
0.01
LR-02B
55.0
22.7
51.4
10.3
0.16
0.08
LR-02C
157.3
0.8
146.9
0.5
0.00
0.00
LR-03A
17.0
0.1
15.9
0.1
0.00
0.00
LR-03B
26.6
12.2
24.8
2.9
0.17
0.05
LR-03C
19.6
0.2
18.3
0.1
0.01
0.00
LR-04A
71.7
24.7
66.9
7.3
0.14
0.04
LR-04B
89.2
67.3
83.3
2M
0.26
0.09
LR-04C
222.1
88.3
207.4
23 7
0.15
0.05
LR-04D
42.2
82.2
39.4
23.6
0.49
0.20
LR-05A
26.7
36.0
24.9
10 1
0.39
0.15
LR-05B
56.5
23.2
52.8
5 2
0.16
0.04
LR-05C
28.7
39.4
26.8
9.2
0.39
0.13
LR-05D
55.2
J 19.3
51.5
32.9
0.52
0.21
LR-06A
15.8
15.7
14.8
5.6
0.31
0.14
LR-06B
40.5
96.4
37.8
30.5
0.55
0.26
LR-06C
15.9
0.6
14.9
0.3
0.02
0.01
LR-07A
44.0
5.3
41.1
2.2
0.05
0.02
LR-07B
21.3
9.3
19.9
3.5
0.17
0.07
LR-07C
50.1
29.3
46.8
11.4
0.21
0.09
LR-07D
18.6
3.5
17.4
1.8
0.08
0.04
LR-09A
11.9
14.4
11.1
5.5
0.36
0.17
LR-09B
13.8
0.1
12.9
0.1
0.00
0.00
LR-09C
25.0
24.3
23.4
6.1
0.31
0.10
LR-09D
79.7
6.3
74.4
2.0
0.04
0.01
LR-09E
45.2
14.0
42.2
4.1
0.12
0.04
LR-09F
10.8
7.9
10.1
3.5
0.25
0.13
LR-10A
50.1
42.7
46.8
8.2
0.28
0.07
LR-10B
100.7
42.6
94.0
11.6
0.16
0.05
LR-10C
83.6
37.1
78.1
8.5
0.17
0.05
SLR-10D
58.7
30.5
54.8
11.7
0.19
0.08
LR-11A
113.2
186.9
105.7
60.3
0.44
0.20
LR-11B
66.5
58.7
62.1
13.2
0.29
0.08
LR-11C
51.4
74.6
48.0
11.8
0.41
0.10
LR-18A
69.0
20.7
64.4
5.4
0.12
0.03
AVERAGE
57.6
34.1
53.8
9.8
0.20
0.08
CORRESPONDING MASS CHANGE
-40%
-80%
Notes:
Data set based upon core pairs within TIP having 1984 Tn~ MPA > I0g/m2
TAMS log (DPCB + 2) is based upon 1984 PCB J-*- and 1994 Total PCBs
Tri+ log (DPCB + 2) is based upon PCB3+ for 1984 and 1994
-------
Table 6
Average PCB Load Increase Across TIP During the Summer 1997 Low-Flow Period
Fort Edward
Flow icfsi
Total PCBs (ng/L)
TIP Load |
Date
Fort-Edward
TI Dam
Total (kg/d)
% PCB?.
PCB,. (kg/d) |
6/17/97
2800
31.7
105.3
0.5
34.1
0.2
6/30/97
2800
17.7
175.1
1.1
33.1
0.4
7/14/97
2000
13.7
91.8
0.4
52.4
0.2
7/28/97
1500
18.6
66.7
0.2
52.1
0.1
8/13/97
1600
14.6
57.4
0.2
46.1
0.1
AVERAGE
2140
19.3
99.3
0.46
43.6
0.18
Notes:
Data Source: GE Database
Flows are USGS Daily Average
Fort Edward is Rt. 197 Bridge Station
TI Dam is TIP-18C station
-------
Table 7
Water Column Derived Estimates of Annual PCB3+ Loading from TIP Sediments
Estimated Loading Past
Estimated Loading from
Thompson Isl Dam
Ft. Edward
TIP Sediments
YEAR
Metric Tons
Metric Tons
Metric Tons
1984
0.45
0.36
0.09
1985
0.19
0.15
0.05
1986
0.19
0.18
0.01
1987
0.20
0.19
0.01
1988
0.10
0.07
0.03
1989
0.15
0.10
0.05
1990
0.25
0.20
0.04
1991
0.71
0.41
0.31
1992
0.85
0.58
0.27
1993
0.43
0.25
0.18
TOTAL
3.11
2.25
0.86
Notes:
1) 1/84-4/91 based on PCB concentration and nver flow rating curves for Fort Edward and Schuylerville stations.
2) 4/91-12/93 based on interpolation of approximately weekly data collected from FE and TID stations.
-------
Table 8
Estimated Amount of Deposition Occuring During 1994 High Flow Event (4/17/94)
to Obtain Observed Detectable Be Values1,2
TAMS
Low Resolution
Core ID
"Hot Spot"
Day 7Be
Counted
7Be
Correction
(oCi/ke)
Estimated Depth of Material
Deposited fcml
?
LR-01D
16
8/23/94
1520.0000
1.0
LR-02A
16
8/29/94
2190.0000
1.6
LR-03B
t6
9/2/94
2080.0000
1.6
LR-03C
16
8/30/94
771.0000
0.6
LH-25A
25
9/15/94
1840.0000
1.7
LH-25B
25
9/17/94
617.0000
0.6
LH-25E
25
9/14/94
596.0000
0.5
LH-25G
25
9/18/94
824.0000
0.8
LH-25H
25
9/17/94
907.0000
0.8
LH-25I
25
9/18/94
583.0000
0.5
LH-25J
25
9/19/94
875.0000
0.8
LH-28B
28
9/18/94
720.0000
0.7
LH-28C
28
9/12/94
1850.0000
1.6
LH-28D
28
9/12/94
991.0000
0.9 1
LH-28F
28 •
9/10/94
1680.0000
1.4
LH-28G
28
9/17/94
472.0000
0.4
LH-28I
28
9/12/94
957.0000
0.8
LH-28J
28
9/12/94
1510.0J00
1.3
LH-28K.
28
9/9/94
589.0000
0.5
LH-28L
28
9-14/94
978.0000
0.9
LH-28M
28
9/16/94
1220.0000
1.1
LH-28N
28
9/16/94
966.0000
0.9
LH-31A
3!
9/9/94
1441.0000
1.2
LH-31B
31
9/13/94
1537.0000
1.4
LH-31E
31
9/14/94
i520.0000l
1.4
LH-31F
31
9/11/94
991.0000
0 9
LH-31G
31
9/12/94
3143.0000
2.7
LH-31 H
31
9/13/94
875.0000
0.8
LH-31 J
31
9/14/94
2517.0000
2.2
LH-34C
34
8/31/94
973.0000
0.7
I LH-34E
34
9/2/94
548.0000
0.4
LH-34G
34
9/8/94
2348.0000
1.9
LH-34I
34
9/10/94
640.0000
0.5
LH-34J
34.
9/10/94
3577.0000
3.0
| LH-34K
34
9/9/94
1116.0000
0.9
LH-34L
34
9/8/94
637.0000
0.5
LH-34M
34
9/7/94
922.0000
0.8
LH-35C
35
9/9/94
1196.0000
1.0
LH-35D
35
9/12/94
794.0000
0.7
LH-37E
37
9/20/94
503.0000
0.5
LH-37F
37
9/19/94
1352.0000
1.3
LH-37G
37
9/19/94
1957.0000
1.9
LH-37J
37
9/20/94
779.0000
0.8
LH-37JC
37
9/19/94 j 884.0000
0.8
Page 2 of 3
-------
Table 8
Estimated Amount of Deposition Occuring During 1994 High Flow Event (4/17/94)
to Obtain Observed Detectable Be Values1'2
TAMS
7Be
| Low Resolution
Day 7Be
Correction
Estimated Depth of Material
I Core ID
"Hot Spot"
Counted
foCi/ke)
Deposited fern')
LH-370
37
9/20/94
501.0000
0.5
LH-39A
39
9/16/94
859.0000
0.8
LH-39B
39
9/19/94
565.0000
0.5
LH-39C
39
9/20/94
1719.0000
1.7
LH-39D
39
9/14/94
815.0000
0.7
LH-39E
39
9/21/94
494.0000
0.5
LH-39F
39
9/17/94
1242.0000
1.2
LH-39G
39
9/16/94
1205.0000
1.1
LH-39H
39
9/22/94
679.0000
0.7
LH-39I
39 •
9/21/94
1070.0000
1.0
LH-39J
39
9/23/94
776.0000
0.8
LH-39K
39
9/24/94
1670.0000
1.7
LH-39L
39
9/22/94
610.0000
0.6
LH-39M
39
9/22/94
412.0000
0.4
LH-39N
39
9/30/94
889.0000
1.0
LH-41A
41
9/21/94
1802.0000
1.8
LH-41B
41
9/22/94
1302.0000
1.3
LH-41 C
41
9/24/94
U 00.0000
1.1
LH-42A
42
9/21/94
2657.0000
2.6
LH-43A
43
9/21/94
1346.0000
1.3
LH-43C
43
9/23/94
793.0000
0.8
LR-17A
Near Shore
8/22'94
748.0000
0.5
LR-17B
Near Shore
8/19/94
349 0000
0.2
LR-17D
Near Shore
8/20/94
1610.0000
1.0
LR-17E
Near Shore
8/19/94
589.0000
0.4
LR-16A
WC GRJFFIN ISL.
8/5/94
813.0000
0.4
LR-16B
WC GRIFFIN ISL.
8/30/94
2490.0000
1.8
LR-16C
WC GRIFFIN ISL.
8/23/94
411.0000
0.3
LR-16D
WC GRIFFIN ISL.
8/29/94
2750.0000
2.0
LR-16E
WC GRIFFIN ISL.
8/23/94
1420.0000
1.0
| LR-19A
WC GRIFFIN (SL.
9/6/94
789.0000
0.6
Notes:
' Particulate phase 7Be concentration at time of deposition assumed to be 20,OOOpCi/kg
:,Be decay coefficient is 0.013 days"' (half-life of 53 days)
Page i
-------
Table 9
TIP Organic Carbon Normalized Surface Sediment PCB Concentrations:
1) Both Inside and Outside NYSDEC "hot spots" in 1984 (0-2.5 In.), and
2) for Coarse Grained and Fine Grained Sediments Collected in 1991 (0-5 cm)
Sediment Survey
Location/
Sediment Type
n
Observations
Mean PCB
Concentration
(mg kg oc"1)
Std, Deviation
(mg kg oc"1)
1984 NYDEC
Inside
"hot spots"1
155
2045
2069
Outside
"hot spots"
117
2030
1827
1991 GE
Coarse j 16
Sediments ]
2941
1824
!
Fine
Sediments
41 j 2185
2265 :
I) These statistics excludes one sample collected in 1984 which contained 331.000 mg PCB-"kg oc ',
-------
Table 10
Parameters Used in the Calculation of Surface Sediment PCB Source Signature
Parameter
Description
Value
Units
Source
Kf
Sediment-water
Exchange Coefficient
2
cm day
GE Model
Calibration
As
Surface Sediment
Area
2xl06
m2
GE Hudson River
GIS
m
-------
Table 11
Calculated Mean and Range (95% confidence limits) of Estimated Mass Changes in
"Hot Spots"
"Hot Spot"
Mean Mass Change
Range of Mass Change (g'm") j
25
0
! +28 to -28
28
175
+347 to -3
31
-43
! -i-5 to -90
34
-10
j +4 to -24
35
2
! +16 to -12
37
1-0
CN
1
O
7
i
39
27
1 -67 to-13
Negative values indicate mass loss
-------
FIGURES
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
Figure la
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMS Core LR-IOA
1998 GE Co-Located Core FS-08-4
0 T
-i—i—i—|—i—r
I i
_L
I I ' I I L.
200 400 600
Total PCBs (mg/kg Dry)
800
10
•5 15
Ol
o
Q
20
25
30
_L
J i i i I
200 400 600
Total PCBs (mg/kg Dry)
800
-------
Figure lb
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMS Core LR-IOB
1998 GE Co-Located Core FS-08-3
200 400 600
Total PCBs (mg/kg Dry)
800
10
S 20
a.
4)
Q
30
40
X
J—i i i L
200 400 600
Total PCBs (mg/kg Dry)
800
-------
Figure lc
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMSCore LR-IOC
_! , f 1 , , , 1 , r , 1-
200 4(K) 600
Total PCBs (mg/kg Dry)
800
1998 GE Co-Located Core FS-08-2
10
t 20
U
Q
30
40 Li
r
_L
_L__J 1 I L
200 400 600
Total PCBs (mg/kg Dry)
800
-------
Figure Id
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
10
S 15
D.
O
Q
20
25
1994 TAMS Core LR-10D
0(1 1 I' I ' ' ' I ' '
0 200 400 600 800 1000 1200
Total PCBs (mg/kg Dry)
1998 OE Co-Located Core FS-08-1
10
¦S 15
D-
4>
Q
20
25
30
X
H' 1 ' I ' ' I I I I t I I I i ) i I i j r
1. I L_J L
J I I I I I I
0 200 400 600 800 1000 1200
Total PCBs (mg/kg Dry)
-------
Figure le
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMS Core LR-09A
1998 GE Co-Located Core FS-09-4
±
±
_L
40 60 80 100
Total PCBs (mg/kg Dry)
120
10
S
s.
o
Q
20
30
T I | J—|—|—|—|—]—|—|—|—I—I—p
"i—|—n—r
_L
±
_L
J i i i L
20 40 60 80 100
Total PCBs (mg/kg Dry)
120
-------
Figure If
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
J 994 TAMS Core LR-09C
1998 GE Co-Located Core FS-09-3
10
•S 20
a.
CI
a
30
40
1—i—r
—i—,—r
200 400 600
Total PC Ms (mg/kg Dry)
800
JL
I
200 400 600
Total PCBs (mg/kg Dry)
800
-------
Figu.e lg
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMSCore I.R-09K
1998 GE Co-Located Core FS-09-2
q nj11111111111111111[11111111111111 M 11 n ) 11111 ii 111II111111111
50 Ii I I I I I I I I I I I I I I I I I I I I I I I I ¦ I ¦ ' I I 11 I I I I I I I I I I I I I I I I 1 I I I I 111
0 100 200 300 400 500 600
Total PCDs (mg/kg Dry)
10
20
cx
Q
40
t IJ11111 i 1111 ¦ r 11 ii 111 j 11 ^ 11 f 1111111 f 111T11ITM 1ITTIJT
50 Ciixx 11111111111111111 111111111111111111111 n
0 100 200 300 400 500
Total PCBs (mg/kg Dry)
600
-------
Figure lh
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMS Core LR-09F
1998 GE Co-Located Core FS-09-1
¦ I 1111111111111111111111111111111
20 30 40
Total PCBs (mg/kg Dry)
50
OfT1
I I ll M I T I I I I I I I I I I I I I I I
r
111111111111111111111
10
(X
o
Q
20 -
30
lLu
lL
I' 11111111
10 20 30 40
Total PCBs (mg/kg Dry)
50
-------
Figure li
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMS Core I.R-04A
10
-5 20
Q.
30
40 Li
i I i i i I i i i [ i i i I "i i i | i ' >
0 200 400 600 800 1000 1200
Total f'CEJs (mg/kg Dry)
1998 GE Co-Located Core FS-14-1
200 400 600 800 1000 1200
Total PCBs (mg/kg Dry)
-------
Figure lj
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMS Core LR-03A
1998 GE Co-Located Core FS-16-1
[i1111ii i i [ i m i i ii i i [ i r 11 111111iii 11 t i i r |
i i i i I i iiii11 'i I i i ¦ i i i i i i Iiiiiti ii 11
0 12 3 4
Total PCBs (nig/kg Dry)
0 M 1 1 ' ' ' ¦ 1 ' 1 |' ' " ' ¦ i i i i i i i i rj-
10
¦5 15
o.
u
Q
20
25
30 I 11 11111 11111 111111111111 1111111111 1111 11
0 12 3 4
Total PCBs (mg/kg Dry)
-------
Figure lk
PCB Depth Profiles for 1994 and Col oca ted 1998 Sediment Cores
1994 TAMS Core LR-02B
1998 GE Co-Located Core FS-16-2
| ( I I [ TT i | i t I "| I i i | i i i ] I I I [
25 I ¦ ¦ > 1 i i i 1 ¦ i > 1 » i » I i ' i Lt-x,» 1 » i i I
0 20 40 60 80 100 120 140
Total PCBs (mg/kg Dry)
10
-C
D.
U
Q
15
20
11 i ' ) \ i i [ i i i ) i i i ) i i i [ ii i j i i i [
25 1 I > i I I i i I < i i I i i i I ¦ I i I i i
> I. i
0 20 40 60 80 100 120 140
Total PCBs (mg/kg Dry)
-------
Figure II
PCB Depth Profiles for 1994 and Colocated 1998 Sediment Cores
1994 TAMSCorc LR-02A
1998 GE Co-Located Core FS-16-3
qP—i—i—i—|—i—i—i—rt-i—i—i—|—i—i—r
20 40 6(
Total PCBs (nig/kg Dry)
J3
•-*
Cu
Q
20 40 6(
Total PCBs (mg/kg Dry)
-------
Figure 2a
Comparison of 1984 and 1994 estimates of PCB mass within TAMS low resolution zones 1 through 6.
Note: 1) Only samples matched by TAMS are presented
Zone LR-Q1
Total PCB*
10 15 20 25 30
1984 Mass (g/rn*)
35
Zone LR-02
Total PCBs
50 100 150 200
1984 Mass (g/m*)
250
Zone LR-03
Total PCB*
5 10 15 20 25
1984 Mass (g/m*)
Zone LR-04
Total PCB«
50 100 150 200
1984 Mass (g/m*)
250
Zone LR-05
Total PCB*
20 40 00 80 100
1984 Mass (g/m*)
120
100
ba
m
m
m
38
o>
Oi
Zone LR-06
Total PCBa
0 20 40 00 80
1984 Mass (g/m*)
100
-------
Figure 2b
Comparison of 1984 and 1994 estimates of PCB mass within TAMS low resolution zones 7 through 12.
Note: 1) Only samples matched by TAMS are presented
Zone LR-07
Total PCB.
10 20 30 40 50
1904 Mass (g/m*)
60
Zone LR-08
Total PCBa
5 10 15 20
1904 Mass (g/m*)
25
Zone LR-09
Total FCBb
m 60
S 40
20 40 60
1984 Mass (g/m*)
em
9)
(ft
fl3
CO
OS
Zone LR-10
Total PCBa
20 40 60 80 100
1984 Mass (g/m*)
120
a 200
M
m
m
<8
3
Qi
Oi
150
100
Zone LR-11
Total PCB*
50 100 150
1984 Mass (g/m")
200
Zone LR-12
Total PCBi
5 10 15 20
1984 Mass (g/m1)
25
-------
Figure j
Semi-variogram of 1990/91 GE Data Collected from the H7 site ("Hot Spot" 5) illustrating variance of
data versus separation distance. Dashed line indicates variance of full 1990/91 GE data at H7 site.
3.Ox 10
2.5x 10 -
S 4
a 2.Ox 10
(X
u1
m
u
Cu 4
1.5x10
0)
o
C
« 4
C 1.0x10
rt
>
5.0x10° -
~i—i—i—|—i—i—i—i—i—i—i—i—i—|—i—i—i—i—i—i—i—i—i—|—i—i—i—i—i—i—i—r
o -f
i T I I 1—I—I—I—L.
i i i i i i i i j i i_
I ¦ I I I I I I I I I I I 1 I I I I L I I I I ¦ I I I I I I I
0
10
20 30
Separation Distance (ft)
40
50
-------
Figure 4
Regression analysis of 1984 and 1994
PCB3+ mass per unit area estimates
for paired sediment cores separated by 5 feet or less
1000
100
r-i
<
E
fib
C/5
in
c3
2
CQ
U
cu
Os
Os
10
V ~
0.1
0.01
Slope: 0.26
95% CL: -0.03 to 0.55
0.01
0.1 1 10 100
1984 PCB Mass (g/mA2)
1000
-------
Figure 5
Probability distribution of estimated PCB mass for cores collected from TIP in 1984.
1000.OOfe 1 1 1111111 p r
bO
CO
CO
CO
s
PQ
CJ
pu,
co
CD
100.00
10.00
1.00
0.10
0.01
rn mi 11 i i i—inn111 i d
i i i nun i i i 111 m
OOOO
Legend:
Pilled circle
Open diarao
Open triang
i: all data (n = 298)
ids: TAMS-matched data (n = 42)
ea: Un-matched data (n = 256)
III II I I I I
0.1
1
10 20 50 80 90
Probability
99 99.9
-------
Figure 6
Probability distribution of estimated PCB mass for cores collected in 1984
from within 1976-1978 NYSDEC defined "hot spot" areas of TIP.
bfl
en
CO
cd
u
Oh
oo
a
100.00
10.00
1.00
0.10
0.01
0.1
i iiiiii i i i 111 in
1 1 I
1 1 1
iiiiii i i i mini i i -
o •
O
o
o
o
o ° ° °
ooOO°°
> •
*>•
>.
>
1 II Mil
0
o o o °
-
o
o
0
o
«
<
:t
—
• A
Legend:
Filled circle
Open diamo
Open triang
i: all data (n = 210)
ids: TAMS-matched data (n = 26) ~
ea: Un-matched data (n = 184)
i iiiiii i i i i 11 in
1 1 1
1 1 1
mi i i i i i iiiiii i i i
10 20 50 80 90
Probability
99 99.9
-------
Figure 7
Comparison of "Versar" Packed Column and Capillary Column
PCB Quantification Techniques a)Packed Column Total PCBs vs
DB-1 Total PCBs and b)Packed Column Total PCBs vs DB-1
derived PCBs ^ Trichlorobiphenyis
1200
1000 -
c
E
2
o
o
~o
0)
JSC
o
•SP
Ol
150
125
100
R = 0.9801
0
25
75
100
125
150
DB-1 Capillary Column PCB3+ (ug/g)
-------
Figure 8
Frequency distribution of DeltaPCB and log(DeltaPCB+2)
between 1984 and 1994 within Thompson Island Pool
calculated by TAMS
3 0.2
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
DeltaPCB
0.4
0.3
= 0.2
0.1
0.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Log(DeltaPCB+2)
Note: Only cores with mass > 10 g/m in 1984 were included.
N=40
-------
Figure 9
PCB Composition of TIP Sediment Deposited in Approximately 1968 assessed from
Core Sections Collected in 1983 and 1991 (Total Concentration > 500 ppm).
40%
35%
30%
25% -
20% -
15%
10% -
5% -
0%
~ 1983 Composition
| S 1991 Composition
Mono Di Tri Tetra Penta Hexa Hepta Octa Nona Deca
PCB Homoiogs
X
.2?
'5
*
u
w
s
01
J=
U
20%
15%
10%
5% -
0%
-5%
-10%
Tri and Higher Sum = - 10%
Mono Di Tri Tetra Penta Hexa Hepta Octa Nona Deca
PCB Homoiogs
Data Source: McNully. 1997
-------
Figure 10
Comparison of TIP Sediment PCB3+ Mass Loss Attributable to Fate and Transport Mechanisms
with that Estimated using TAMS' Method
Dechlorination Diffusion Scour GW Adveetion Total of 4 Estimated with
Mechanisms USEPA Approach
Potential Loss Mechanisms
-------
Figure 11
Calculated PCB,. Mass Per Unit Area for Colocated 1984,1994, and 1998 Sediment Cores in Thompson Island Pool
Note: Data Represent Cores with Separation Distance Less than 5 ft; Posted Mass Change is Based upon TAMS' Approach
1984 - 1994 Core Comparison
1000.0
100.0
10.0
1.0
0.1
r t i 11 mi| i i i i
-nij- i i i i i m| i f i II n ij
Average Mass Change =
Log (Delta PCB + 2)
-80% / :
r
/
•
/ i i i i 11 > J i iii
¦ •¦I i i i i 111il i i i i 11ii
0.1
1.0 10.0 100.0
1984 PCB,. MPA (g/m')
1000.0
1000.0
E
3$
<
Dm
2
CQ
u
Oh
OO
&
ON
100.0
10.0
1994 -1998 Core Comparison
rrm
Average Mass Change = +89%
Log (Delta PCB + 2)
• jl
j i i 111
0.1
1.0 10.0 100.0
1994 PCB,. MPA (g/m')
1000.0
-------
Figure 12
Changes in TIP Sediment PCB3+ Mass Estimated from Differences in Calculated MPA for Colocated
Sediment Cores Collected in 1984,1994, and 1998
Note: Only Cores with 1984-1994 Separation <= 5 feet Considered for Calculations
150%
100%
50%
¦ Delta PCB
~ Log (Delta PCB +2)
0%
-50°/t
-100%
-150%
1984-1994
1994-1998
-------
Figure 13
Temporal Profiles of Mean (+/- 95% Confidence Interval) PCBJ+ Mass per Unit Area for Co-Located
1984, 1994, and 1998 Sediment Cores Within Four "Hot Spot" Areas
Hot Spot 8
& 50
1982
IW 1986
1988
1990 1992
Year
1994
1996 1998 2000
Hot Spot 9
1982 1984 1986 1988
1990 1992
Ye»r
1994 1996 1998 2000
Ye*r
Note. Averages only include co-located cores wilhin 5 feel
-------
Figure 14
Computed Depth or Sediment Deposition Necessary to Produce Observed Levels of 7Be in the Top 1" Section of the 1994 Cores
3
E
u
V
4)
w 1
r 2
i/i
o
a
«
a
u
c£> in
— CNJ
CO
CV)
o
C\> o
o
CS1
Paved Number* KcpiCKnl Number ol Sirr.pt* I
Hot Spot
_aj o
Notes Water C olumn Particulate Chase 'Be Concentration Assumed to he 20.000 pCi/Vg
U.ne of Sediment Deposition Assumed to be April 17, 1994
-------
Figure 15
Estimated Relationship Between 7Be Concentration
and Depth of Sediment Deposited
3500
<1 3000
Relationship Between Surface Sediment (0-1") ?Be
and Depth of Sediment Deposited
8 2000
o
U
« 1500
1000
Mean +/- 95% Confidence Interval for Be Detection Limit
500
0
0 0.5 1 1.5 2 2.5 3
Depth Deposited [cm]
Notes; Assumed Date of Deposition: April 17, 1994
Assumed Initial 7Be Concentration of 20,000 pCi/kg
-------
Figure 16
PCB Molar Dechlorination Product Ratio (see text for definition) Depth Profiles for Colocated 1994 TAMS
and 1998 GE Sediment Cores Collected in Thompson Island Pool
0.0 0.2 0.4 O.e 0.6 1.0
Molar Dechlorination Product fUUo
it* vnr
40
60
ittt M
IOC
•-I
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
1
3
P.
&
a
00
00
n IH4 raffi Cm
a CI Cm
0.0 0.2 0.4 0.6 0.6 1.0
MoUr Dechlorination Product Ratio
!M4 UVVTi C*r«
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
21904 Dltfi Cm*
I Ml CI C«r»
i ....
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
or"
20
~
80 L
SIN4 ORFi Cere
IMS GC Cm
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
n lt#4 (J9tPi Cor*
tIM CI
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
IH4 U91PA cm
IIM Ct Cm
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
fl
a.
20
40
00
00
Itf4 PltPA Cm
1IW Ot Cm
or-
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
40 -
60
60
S1H4 cam Cm
It* Ct Cm
MA
ft-If-J
0.0 0.2 0.4 0.6 0.6 1.0
Molar Dechlorination Product Ratio
ioS| [jn»i Represent Aroclor 1242
-------
figure 17
Ratio of DB-1 Peaks 46:32 Depth Profiles for Colocated 1994 TAMS and 1998 GE Sediment Cores
Collected in Thompson Island Pool
Iff# M Qir*
N-M-S
0,« 0.4 »• 0,0
lUUo of D8-I P.*k* 4*32
Of 0.4 04 Ot 1.0
Hollo of DB-1 Patka 40:32
0 0
O.t 0.4 0 4 0.0
Ratio of DB-l Potk* 44:32
M
40
*0
so
T-r' JLJL '
• * r-»-r~»"Y
~ /
-
n i«*4 Car*
A liM N Cot
. 1 . . . » . . . 4
IM
IMi-l
o.e
0.1 0.4 0.0 0.0
Ratio of DB-1 Fmka 40:32
10
| 10
3
A
£ 40
¦m*
a
to
r—» *¦ »' t
•H
X IM* « Cot
f0-##—4
• • * - - *
0 0
0.2 0.4 oa o.i
Ratio of DB-1 Pt&ki 44:32
1.0
20
I
I
60
AO
~/
~ tlH UWt Car*
A IfN Ot Cot
.• . . . *
MC
n-oo-i
0.0
0.2 04 o.t o.a
Ratio of DB-1 Poftks 40:32
1.0
0.0 0.2 8.4 0.0 O.i
Ratio of DB-1 Pt«k« 40:32
0.2 0.4 0.0 0.6 1,0
Ratio of DB-i Nik« 4«:3f
q im W9TA i
A i*m ei r
0 0
0.2 0.4 0.0 O.t
Ratio of DB-1 PoaJci 40:32
20
00
oo
ItM 099A Cm«
ItM Ot Cot
eo
0.0
0.2 0.4 0.0 o.a
Ratio of OB-1 Paaka 40:32
20
40
00
0
1 "r * * » r
ir
¦
"
0K
/
t
/
A
-
« 40
/
.
\
fl
/
\
1
d~ 4 o
-
I •"
-
O IM
4 ONfl Cot
MB
Q 1M4 OfBTA COT
«u
(A ...
• OS Cot
00
I ItM M Cot
r»-u-i
0.2 0.4 0.0 0.0
Ratio of DB-1 PaaJrt 40:32
1.0
0.0 0.2 0.4 0.0 0.0
Ratio of DB-1 Poalta 40:32
Note: Verticil Lioei Represent Aroclor 1242
-------
Figure 18
TIP Center Channel PCB Concentrations
140
Flow = 4,600 cfs
September 24,1996
194
193
192
191
190
189
188
River Mile
140
120 ¦
100
Flow = 5,200 cfs
September 25,1996
® T5.
C 0)
C c
<0
£ «
o o
u U
m q.
+*
£ 2
o o
194
193
192 191
River Mile
190
189
188
-------
Figutf 9
Comparison ofPCO Peak Compositions for Calculated DifTusiooal Sediment Source (1997 Summer Average)
with (a) Surface and (b) Deep Sediments from 1992 EPA High Resolution Cores Collected from TIP
40%
35%
30%
S 25%
O
L.
%i
a.
£
Of
Z
z
20%
15%
10%
5%
0%
20
40
(a) Surface Sediments
x 1997 Calculated Summer Average
1992 Hi Res 0-2 cm
»»•»«
60
DB-1 PCB Peak
80
100
120
(b) Deep Sediments
x
& t i^i»»«t\«
20
40
60
80
k 1997 Calculated Summer Average
1992 Hi Res >8 cm
100
120
DB-1 PCB Peak
-------
Figure 20
PCB congener ratios calculated for fish and sediment
collected from the Thompson Island Pool
SEDIMENT
FISH
56:49
LMB PKSD RBRS SMB TESS YP
60:49
10 1
NUMBER CHLORINES
PER BIPHENYL
, LMB PKSD RBRS SMB TESS YP
SPECIES
-------
Figure 21
Temporal trends in DB-1 capillary column peak ratios from three stations
on the Upper Hudson River during the January 1998 high flow event.
w
CO
aj
t)
CU
\
00
M
-------
Figure 22
c
o
• in
¦*-»
CS
C
03
0)
+
Average n'Cs (filled symbols) and PCB (open symbols) Profiles Calculated from
Thompson Island Pool High Resolution Sediment Cores (HR-019,020,023, and 026)
GO
S 13
U £
Q, <0
c
u
vj »n
K On
U
O
G
<0
T3
CS
o
U
1000 ^
t -• ! ! ! 1 , |~rt 1 ii&i 1 1 1 1
1950
1960
1970 1980
Approximate Year
1990
2000
Notes:
1) Data Binned in 5 Year Intervals and Plotted at Midpoint
2) Horizontal dashed lines represent surface sediment (0-1") '"Cs concentrations detected in cores TAMS identified as indicative of scour
between 1984-1994 (03A»04A, 09D, 10B, 10D, UB, 11C, 12C)
-------
APPENDIX A.
COMMENTS OF DR. PAUL SWITZER
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
75 Pearce Mitchell Place
Stanford, CA 94305
August 24, 1998
John Haggard
General Electric Company
1 Computer Drive South
Albany, NY 12205
RE: HUDSON RIVER
Dear John,
I read the report material which you sent to me relating to Hudson River PCBs, viz..
Volume 2C-A Book 1 and Book 2 Low Resolution Sediment Coring Report by TAMS et
al., dated July, 1998, prepared for the U S E.P. A., together with portions of Volume 2C
Data Evaluation and Interpretation Report, by the same authors, dated February, 1997,
containing material related to geostatistical analyses of the 1984 Hudson River PCB data
I also have seen the two reports you sent to me on geostatistical analyses prepared by
Davis and Olea for General Electric Company, dated December 14, 1990, and February,
25, 1991 .
In my view, there are a number of important lapses both in the design of the Hudson
River PCB study and the analyses of the data that are briefly described in my report
accompanying this letter. In short, many of the stated conclusions of the July, 1998,
report by TAMS et al. are not well supported by the reported statistics. Rather, these
conclusions seem to be drawn from interpretations and conjectures that do not have a
statistical inferential basis.
Sincerely,
/
Paul Switzer
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
COMMENTS ON
1998 U.S.E.P.A
LOW RESOLUTION SEDIMENT CORING REPORT
August, 1998
prepared by
PAUL SWITZER
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
ISSUES OF DATA SELECTION
I. Since not all core locations from the earlier surveys were to be resampled in 1994, { 29
care should be taken that those selected for the 1994 survey are representative of the
zones for which PCB inventory estimates are desire^ For example, a form of random or
systematic selection or areally stratified random selection would remove the possibility of
selection bias. Instead, the sampling method used is a poorly documented form of what
statisticians call "purposive" sampling - here a misguided attempt to create zones of
- reduced sediment variability at the expense of unbiased representation of the 1984 sample
locations.
2. Seven hot spots were defined by combining certain of the original twelve dredge 1.30
locations from the 1984 survey. It is not clear whether the decision to combine dredge
locations was made before or after the 1994 data were available. If after the data were
available, then another source of selection bias is created because combining data can be
used to purposively exaggerate statistical significance. This effect can be seen, for
example, by comparing Figure 4-21 with Figure 4-22.
3. Not all of the 1994 data that were collected were used in the statistical estimates. An 1.31
important issue is the removal of nearly 40% of the data because of potential cross-
contamination from higher to lower segments of a core. The severe data rejection rule
conforms to a priori ideas of what a core sequence should look like rather than relying on
explicit evidence of cross-contamination. Because the fraction of data removed from the
analysis is large, the susceptibility to selection bias is also large. Calculations showing
that some statistics are not seriously affected by this form of data screening do not
adequately address this issue.
4. It was reported that the lowest concentration data were excluded from "many of the 1 -32
subsequent analyses" [page 2-16] so that the remaining data would look like a lognormal
distribution. The fact that the low concentration data are hard to quantify does not
change the fact that they are nevertheless informative. Trimming data to achieve a
desired distributional shape violates important statistical principles related to selection
bias.
1 3^
5. As an example of how purposive data selection can create an impression of a ' J
statistical relationship compare the weak associations in Figure 3-2 with the strong
associations indicated by the selected data subset of Figure 3-8. This example shows
clearly the perils of censoring data rather than modeling data.
-------
CALIBRATION ISSUES
1. The SSW correction [see Figure 4-17 and Table 4-3] applied to the 1984 data is itself
based on the PCB measurements. The net effect is to introduce an implicit non-linear
rescaling into the PCB data which is difficult to analyze. This dry soil weight SSW
factor should be based on some physical parameter other than PCB itself. Furthermore,
stronger evidence is needed to demonstrate that such a recalibration of the 1976-78 data
makes them comparable to the 1994 data.
2. The measurement methods for the 1976-78 survey, the 1984 survey, and the 1994
survey are different [see comments on page 4-21, for example]. However, there are no
cross-calibration data presented which analyze the same samples by both methods. By
implication, the subsequent statistical analyses assumed that relative calibration errors
were strictly zero. How could the 1994 study have been undertaken without such
calibration information7
3. The extrapolation of grab samples to make them comparable to 12-inch core segments
is not well documented in the report. The "factor' used for extrapolation appears to be
something like the mean PCB ratio between the top four inches and the top 12 inches
seen in core samples [page 4-21], Of course, such extrapolation introduces additional
error beyond the measurement error associated with core sampling but the extrapolation
error is not accounted for in the analysis.
-------
THE REGRESSION FALLACY
1.37
Surprisingly, the report does not recognize the "regression" fallacy The report makes
much of the observation that lower 1984 concentration locations tended to be followed by
later increases while higher 1984 concentrations tend to be followed by decreases This
kind of analysis is found throughout the report without any recognition of the
"regression-toward-the-mean" effect: for any collection of paired measurements which
are positively correlated, which these data emphatically are, higher initial concentrations
will tend to show a decrease on average and lower initial concentrations will tend to show
an increase on average — in the absence of any distributional changes! [See Fig 4-11 to
Fig 4-16 for example, as well as much of Section 4 and statements in the Executive
Summary] This fallacy may be understood by simply supposing that the true
concentrations were exactly the same in 1984 and 1994 and that reported concentration
differences were due entirely to analytic variability Even in this case, splitting the data
into two parts as the authors have done would demonstrate that higher 1984
concentrations tend to be followed by lower 1994 concentrations and vice-versa.
The geostatistical evidence from the 1984 survey shows that very short-scale spatial
variability is often comparable to total variability [see Figures 4-9 to 4-12 and 4-14 to 4-
16 in the 1997 report] This fact, by itself, can be used to show that even small location
errors induce a large spurious regression effect that would account for most of the
reported 1994 PCB decrease at locations with high 1984 PCB
-------
ISSUES RELATED TO STATISTICAL CALCULATIONS
1. The MVUE method for estimation of means requires an exact lognormal model. 1.38A
Otherwise the method is biased. For example, the MVUE is sensitive to procedures for
handling low concentration data. Censoring and BDL data create special problems for
the MVUE method. The straightforward sample average is less sensitive to prior
assumptions about the distribution of the data.
2. The lognormal unbiasedness correction was inconsistently applied. For example, 1.38B
compare formulas on page 4-28 and in Table 4-13. In the latter case, the supposed
MVUE ["minimum variance unbiased estimate"] for the near-shore sediments is neither
unbiased nor minimum variance, even if the data were sampled from a precisely
lognormal population distribution.
3. While data were collected on covariate parameters such as grain size, the estimates of 1.38c
PCB inventory change did not incorporate the covariate information in any statistical
model. It is not clear where or how the relationship between fine grain sediments and
elevated PCB was used to improve estimates of PCB inventory [page 3-34],
4. Simply reporting correlations between individual covariate parameters and PCB 1 •380
measures, without regard to the multivariate nature of the data, does not take advantage
of the potentially more powerful but widely used multivariate statistical methods such as
partial correlation or multiple regression analysis. See the discussion on page 3-15 as an
example. Failure to examine the multivariate statistics can also lead to mistaken
interpretations such as that found on 3-19 "The correlation of AMV and MDPR values is
largely a ramification of the correlation with total PCB" . as well as a similar confused
statement about correlations on page 3-25.
1 38
5. Hot spot PCB quantity differences between the 1976-78 and 1994 surveys are
assessed by the ratio At described on page 4-40. The explicit claim is that differences "of
more than a factor of two are considered significant and likely to be beyond the level of
uncertainty" I could find absolutely no statistical basis for this claim and no supporting
analysis1
6. Upper 95% upper confidence limits were used to compare sediments in Table 4-13. L •381
Differences among these upper limits may be due to mean differences, variability
differences, or differences in the number of samples, or some combination of these. It is
misleading to make such comparisons.
4
-------
7. The report contains no statistical analysis of variance components that could be used l. 38G
to make better judgements regarding the significance of observed differences. Some
examples of large random components of variability include the following. On Page 2-18
we read that split-pair samples differed by 36% on average. Table A-3 suggests that
field replicate-pair congener analyses differ by more than 20% about half the time.
Geostatistical studies of the 1984 survey report that "it was not unusual for samples taken
only a few meters apart horizontally to exhibit order-of-magnitude differences in PCB
concentrations" [page 4-25, 1997 report]. The reported variograms indicate that very
short-scale spatial variability in sediment PCB is often comparable to the total variability
observed between different sample locations. Without adequate accounting for random
error components, it is hard to appreciate reported differences between the earlier and
later PCB surveys.
8. Statistical analyses of the Am ratio measure of PCB molar inventory change are based
on adding the arbitrary constant 2.0 to all ratios and then taking the logarithm. It is not
reported how sensitive are the analyses to this particular fudge factor 2.0 However, it is
likely that the low end of the distribution is noticeably affected by this added factor,
thereby affecting statistics in the context of a lognormal model.
9. It appears that zonal PCB inventory estimates were obtained simply by multiplying
averages of core-based concentrations [g/m2] by estimates of zonal areas. These
estimates take no account of the spatial configuration of the cores within zones and make
no use of the geostatistical considerations. Although the 1994 data are not sufficient to
develop the appropriate geostatistics, the geostatistical models based on the more
extensive 1984 survey could have and should have been used.
10. Estimates of PCB inventory, inventory changes, and dechlorination losses [see Table
4-8 as an example] are not accompanied by statistical estimates of uncertainty. Good
statistical practice dictates the quantification of uncertainty - first as a measure of the
adequacy of the underlying information and second as a tool in rational decision making.
The uncertainty associated with inventory estimates derives from sampling variability,
measurement error, and spatial interpolation errors, all of which propagate to the final
inventory estimates.
L.38H
1.381
-------
INCONSISTENCIES
1 The Be data sometimes did, and sometimes did not show association with PCB levels l • 39a
or imputed scour of PCB, depending on how the data were divided or otherwise
manipulated. See pages 3-23 and 4-38 for example. Nevertheless, the Executive
Summary states categorically that "beryllium-7 was shown to be a statistically significant
indicator of inventory loss" [ES-3] For hot spots the report states that" [7Be] evidence
for recent deposition apparently contradicts the strong PCB profile evidence for long-
term loss sediment loss or lack of burial" [page 4-35]
2 It is reported that the molar inventory changes between the 1984 and 1994 surveys 1 • 39B
[page 4-14] indicate no statistically detectable trend However, the report also claims that
a relative measure of molar change, Am , shows statistically detectable decreases,
probably reflecting the regression fallacy described earlier In this connection, the report
makes the inaccurate statement on page 4-15 that for relative measures of change "much
of the analytic variability can be diminished in importance relative to real change".
3 The high-resolution 1984 survey did not confirm that the amount of implied l • 39C
dechlorination is proportional to total PCB or that it followed the same dechlorination
pattern as the low-resolution 1994 data [pages 3-5, 3-9, 3-35]
6
-------
CONCLUSIONS VERSUS DATA
1. Statistical analyses were not used to support frequent statements regarding lack of 1.40A
burial. Indeed, high resolution core data [Figure 4-24 is an example] indicate that the
concentration peak can rise and fall sharply within a 9" or 12" segment, representing one
or more decades of deposition. The fact that the top segment of a low-resolution core
shows higher concentration than the succeeding segment does not, by itself, imply that
the peak concentration is at the surface and that burial is absent. On page 3-10 the report
itself states that "it is likely that a low-resolution core segment would span a range of two
or more orders of magnitude.. .over a nine inch interval".
2. The simple "sign test" shows that the frequency of locations with PCB increases 1.40B
versus those with decreases is not statistically different from 50%. There are many
examples, such as the comparisons in Figures 4-4 and 4-21. The point is that the limited
size of the 1994 survey makes the results look much like the results of coin tossing. The
non-statistical rationalizations in the report, however salient, should not be confused with
statistical inference.
3. Many far-reaching statements in the report are not actually derived from a statistical I • 40c
analysis of the data, even though such non-statistical hypotheses and conjectures are
intertwined with statistical reporting. Examples of such are statements referring to scour,
resuspension or redistribution of sediments, dechlorination limits, underestimation of
prior inventory, probable losses due to incomplete cores, and the relation of Aroclors to
PCB concentrations.
7
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
APPENDIX B.
1.41
DESCRIPTION
OF THE GENERAL ELECTRIC CO.
SEDIMENT TRANSPORT MODEL
-------
THIS PAGE LEFT BLANK INTENTIONALLY
-------
SECTION 1
INTRODUCTION
Understanding the fate and transport of PCBs, both qualitatively and
quantitatively, in the Thompson Island Pool (TIP) is of critical importance when
considering various remedial action scenarios in this reach of the Upper Hudson
River. Modeling has been shown to be an effective method for quantitatively
evaluating PCB fate and transport processes in a riverine system. A
comprehensive PCB fate and transport modeling framework will contain three
sub-models that are coupled together: (1) hydrodynamic; (2) sediment transport;
and (3) PCB fate and transport.
As part of a larger effort to quantitatively evaluate PCB fate and transport
in the Upper Hudson River, a coupled hydrodynamic/sediment transport model
has been developed, calibrated, validated and applied to better understand
sediment transport processes in the TIP. The model used in this study (SEDZL)
has been successfully applied to other riverine systems, including: Lower Fox
River, Wisconsin (Gailani et al., 1991); Saginaw River, Michigan (Cardenas et al..
1995); Buffalo River, New York (Gailani et al., 1996); Pawtuxet River, Rhode
Island (Ziegler and Nisbet, 1994) and Watts Bar Reservoir, Tennessee (Ziegler
and Nisbet, 1995). The current study has produced significant improvements to
SEDZL in the areas of cohesive sediment deposition and non-cohesive
suspended load transport.
The purpose of this report is to present: (1) descriptions of the
hydrodynamic and sediment transport models; (2) calibration and validation
results; and (3) an evaluation of sediment transport processes in the TIP. The
next section provides an overview of the hydrodynamic model. Section 3
discusses the development, calibration and validation of the sediment transport
1
-------
model. The final section of the report discusses some results from the modeling
effort and the implications for PCB fate and transport in the TIP.
2
-------
SECTION 2
HYDRODYNAMIC MODEL DEVELOPMENT AND CALIBRATION
2.1 Model Structure
The Thompson Island Pool is relatively shallow (approximate mean depth
of 7 ft) and the flow is unstratified. These conditions make it possible to assume
that the water column is vertically well-mixed. Thus, the two-dimensional,
vertically-averaged equations are an accurate approximation to the general
three-dimensional equations of motion for an incompressible fluid. The
hydrodynamic equations (conservation of mass and momentum) applied to the
Thompson Island Pool are (Ziegler and Nisbet, 1994):
cr? | 5{uh) djvh) Q
c( dx dy
o{uh) o(u h) c{uvh) drj c , , cu^ d(,A du\ /0
+ —— = ~gh—~ Cfqu + —\hAH—[ + — ¦ hA„ — . (2-2)
ct ox oy cx cx\ dx j cy\ dy j
c{oh) d(uvh) c[v~h) crj c ( ru ] c '¦ cu
+ = -gh L ,qu + —j hAu — ¦ * — hAu — ¦ (2-3)
5t dx dy cy dx^ " cx j cy\ " dy)
where the total water depth is h = h0 + r; h0 = equilibrium water depth; r = water
surface displacement from that equilibrium; u and v = velocities in the x- and y-
directions, respectively; q = (u2 + v2)1/2 ; Cf = spatially variable bottom friction
factor; AH = horizontal eddy viscosity; and g = acceleration due to gravity.
Equations (2-1) to (2-3) were transformed from Cartesian coordinates to
orthogonal, curvilinear coordinates in order to more accurately resolve the
complex geometry and bathymetry of the Thompson Island Pool. The resulting
equations were then solved numerically.
3
-------
The bottom friction factor in Equations (2-2) and (2-3) is dependent on the
local water depth and effective bottom roughness (Blumberg and Mellor, 1983):
Cf = MAX
In—
V 2zo j
r
' /,min
(2-4)
where k = von Karman's constant (= 0.4); Cf,mj„ = minimum bottom friction factor;
and Zo = effective bottom roughness.
2.2 Application to the Thompson island Pool
The TIP hydrodynamic model extends from the Route 197 bridge at
Rogers Island (HRM 194.4) to Thompson Island dam (HRM 188.5). This six-mile
reach has been discretized using 68 longitudinal and 10 lateral grid elements
(Figure 2-1). The hydrodynamic effects of the jetty at the entrance to the
Champlain Canal near HRM 189 have been included in the model.
Bathymetric data collected by General Electric in 1991 were used to
specify values of h0 for model input. The 1991 bathymetric survey collected
depth soundings at about 107,000 points throughout the TIP at river flows within
a range from about 1,700 to 2,700 cfs. The average equilibrium water depth (h0)
in a particular grid element was calculated using the 1991 sounding data located
within that grid element. The resulting bathymetric distribution used for model
input is presented on Figure 2-2.
The hydrodynamic model requires the specification of two types of time-
variable boundary conditions: (1) inflows from upstream and tributary sources
and (2) stage height at Thompson Island dam. Flow rates measured by the U.S.
4
-------
Geological Survey (USGS) at the Rogers Island gauging station were used as
input at the upstream boundary of the model. The mean flow rate at this location
is approximately 5,200 cfs and the 100-year flood has been estimated to be
47,300 cfs.
A 161 mi2 drainage basin contributes tributary flow to the TIP between
Rogers Island and Thompson Island dam, primarily through Snook Kill (75 mi2)
and Moses Kill (55 mi2). The remaining 19% of the basin is considered to be
direct drainage. No permanent gauging stations are located on these streams
and limited flow data exist for either tributary. A modified drainage area proration
method has been developed to estimate tributary flows to the TIP using flow data
collected at the USGS gauging stations located on Kayaderosseras and
Glowegee Creeks. This procedure is very similar to the tributary flow estimation
method implemented by thre USEPA modeling team (Limno-Tech, 1998). The
estimated mean flow rates for Snook and Moses Kills are 105 and 77 cfs,
respectively.
Stage heights measured at the Crackers Reef gauge (gauge 118) by
Champlain Canal personnel were used to develop a relationship between flow
rate anci water surface elevation at the Thompson Island dam. This rating curve
was used to specify time-variable water surface elevation at the downstream
boundary of the model.
2.3 Model Calibration and Validation Results
The hydrodynamic model contains two adjustable parameters: horizontal
eddy viscosity (AH) and effective bottom roughness (Zo). Generally,
hydrodynamic models are run with the minimum value of Ah needed to ensure
numerical stability, which was 1 m2/s for the TIP. No adjustment of Ah was made
during model calibration or validation.
5
-------
The model was calibrated and validated using water surface elevation
data collected at locations upstream of the Thompson island dam. The effective
bottom roughness was adjusted to achieve the best agreement between
observed and predicted upstream water surface elevations during the calibration
process, which used data collected during November 1990 when flow rates
ranged between 7,000 and 8,000 cfs. Spatially variable Zo values were used,
depending on local bed type, with Zo set at 1,500 um in non-cohesive sediment
bed areas (Cf.min = 0.0035) and 75 um in cohesive sediment bed areas (Cf,min =
0.0020). The distribution of cohesive and non-cohesive bed areas in the TIP is
discussed in Section 3.
The hydrodynamic model was validated using water surface elevation
obtained during the 1983 flood. These data were collected by Champlain Canal
personnel at gauges located near Crackers Reef (gauge 118) and at the
entrance to the Champlain Canal near Rogers Island (gauge 119). The 1983
flood had the highest daily average flow rate (34,100 cfs) measured at the
Rogers Island gauge since the gauge became operational in 197.7. Results of
the validation simulation (Figure 2-3) indicate that the model does a very good
job of predicting water surface elevations throughout the TIP. Successful
calibration and validation of the model demonstrates that: (1) model geometry
and bathymetry are accurately represented and (2) bottom friction factors used in
the model are realistic and accurate.
6
-------
SECTION 3
SEDIMENT TRANSPORT MODEL DEVELOPMENT,
CALIBRATION AND VALIDATION
3.1 Model Structure
Suspended sediment particles in a river have a large range of sizes, from
less than 1 um clays to fine sands on the order of 250 um. Simulation of the
entire particle size spectrum is impractical. However, the particles may be
broadly segregated into two groups: silts and clays that may interact and form
floes and sands that are transported as discrete particles. The model uses this
approach to approximate the particle size spectrum. "Class 1" particles include
ail the cohesive particles, • i.e., clays and silts, with disaggregated particle
diameters of less than 62 um, while the "class 2" particles include coarser, non-
cohesive sediments, primarily fine sands with diameters between 62 and 250
um.
A valid assumption in the Thompson Island Pool is that the water column
is well-mixed and that suspended fine-grained sediment concentrations are
approximately uniform in the vertical direction. For these conditions, the two-
dimensional, vertically-averaged sediment transport equation for size-class k (k =
1,2) is applicable (Ziegler and Nisbet, 1994):
where Ck = concentration of suspended sediment of size-class k; KH = horizontal
eddy diffusivity; Ek = resuspension (erosion) flux of size-class k; and Dk =
deposition flux of size-class k. Results from the hydrodynamic model provide
information about the transport field in Equation (3-1), i.e., u, v and h. Similar to
(3-1)
7
-------
the hydrodynamic equations, Equation (3-1) has been transformed into an
orthogonal, curvilinear coordinate system and then solved numerically.
Cohesive sediments suspended in the water column can have a wide
range of particle sizes, from clay particles smaller than 1 pm up to -50 urn silts.
In addition, the cohesive nature of these particles cause the discrete particles to
aggregate and form floes which can vary greatly in size and effective density.
Variations in concentration and shear stress affects both floe diameter and
settling speed (Burban et at., 1990). Modeling the settling characteristics, and
associated depositional fluxes, of cohesive sediments in a natural water system
can thus be difficult.
One way to model cohesive deposition is to use multiple size classes to
simulate particle/floc heterogeneity in the water column. Difficulties with this
approach include: (1) specification of composition of sediment loading from
tributaries; (2) obtaining data for model calibration/validation; and (3)
computational constraints.
Previous modeling studies (Ziegler and Nisbet, 1994, 1995; Gailani et al.,
1996) have shown that an effective approximation is to treat suspended cohesive
sediments as a single class. This approach assumes that the settling and
depositional characteristics of cohesive sediments can be represented by
average values of a distribution of properties. Using this approximation, the
deposition flux of cohesive (class 1) sediments to the sediment bed can be
expressed as (Ziegler and Nisbet, 1994):
Dl=P] WSA C, (3-2)
where Ws,i = cohesive sediment settling speed and Pi = probability of deposition
for cohesive sediments.
8
-------
Settling speeds of cohesive floes in freshwater have been measured over
a large range of concentrations and shear stresses (Burban et al., 1990). The
Burban settling speed data for cohesive floes in freshwater have been analyzed
in an attempt to develop a formulation that can be used to approximate the
effects of flocculation on settling speed. This analysis indicates that the settling
speed is dependent on the product of the concentration (Ci) and the water
column shear stress (G) at which the floes are formed, resulting in the following
relationship:
WsA =3.3(C,G)012 (3-3)
where the units of Ws.i, C-i, and G are m/day, mg/l and dynes/cm2, respectively.
See Figure 3-1 for a comparison of Equation (3-3) with Burban et al. (1990) data.
For a depth-averaged model, as used in this study, the relevant shear stress for
use in Equation (3-3) is the bottom shear stress (ib), i.e., G = xb = Cf q2.
Modeling suspended cohesive sediments as a single class, with an
effective Ws,i given by Equation (3-3), makes it necessary to utilize a probability
of deposition (P-i) to parameterize the effects of particle/floc size heterogeneity
and near-bed turbulence on the deposition rate. The complex interactions
occurring in the vicinity of the sediment-water interface cause only a certain
fraction of the settling cohesive sediments, represented by Pi, to become
incorporated into the bed (Krone, 1962; Partheniades, 1992). An experimentally-
based formulation that realistically represents the effects of variable floe size on
probability of deposition was developed by Partheniades (1992), which can be
expressed as:
2
Pt = \- (2k )~'/2 \r_x e ~dw (3-4)
where:
9
-------
\
Y = 2.04In 0.25
h 1
(3-5)
and i b.min = bottom shear stress below which Pi = 1 (dynes/cm2), see Figure 3-
2.
Class 2 particles, i.e., non-cohesive fine sands, suspended in the water
column are assumed to have an effective settling speed (Ws,2) that corresponds
to an effective particle diameter (d2). The depositional flux for this sediment class
is then:
where P2 = probability of deposition for non-cohesive sediments. Details
concerning methods for calculating Ws,2 and P2 are presented in HydroQual
(1997b). The relationship between Ws,2 and d2, which was developed by Cheng
(1997), is shown on Figure 3-3.
Only a finite amount of material can be resuspended from a fine-grained,
cohesive sediment bed that is exposed to a constant bottom shear stress. This
phenomenon, referred to as bed armoring, has been observed and quantified in
a number of laboratory (Parchure and Mehta, 1985; Tsai and Lick, 1987; Graham
et al., 1992) and field studies (Hawley, 1991; Amos et al., 1992). The amount of
fine-grained sediment resuspended from a cohesive deposit is given by (Gailani
et al., 1991):
A = W,.2C2
(3-6)
(3-7)
10
-------
where e = net mass of resuspended sediment per unit surface area (mg/cm2); a0
= site-specific constant; Td = time after deposition in days; m and n are
dependent upon the deposition environment; and iCr = effective critical shear
stress. Note that e is referred to as the resuspension potential.
Experimental results show that the total amount of sediment is not
resuspended instantaneously but it is eroded over a time period on the order of
one hour (Tsai and Lick, 1987; Maclntyre et al., 1990). Thus, the total
resuspension rate (Elot with units of mg/cm2-s) is given by.
E" =5i5o (3"8)
where Etot is assumed to be-constant until all available sediment is eroded. Once
the amount e has been resuspended, Etot is set to zero until additional sediment
is deposited and available for resuspension or until the shear stress increases
(Gailani et al., 1991). The resuspension rate of class k (Ek) sediment from the
cohesive bed is then given by:
£»=/*£« (3-9)
where fk = fraction of class k sediment in the surficial layer of the cohesive bed.
A three-dimensional model of the cohesive sediment bed realistically
simulates the effects of bed consolidation with depth and horizontal variations in
bed composition. The layered bed model conserves mass, with mass flux
occurring only at the sediment-water interface due to deposition and
resuspension. Vertical variations of sediment bed consolidation, or equivalently
porosity, are accounted for by discretizing the bed into seven layers. The time
after deposition of the layers increases linearly from one day at the surface,
which is composed of freshly deposited sediment, to seven days old in the
11
-------
bottom layer. Once deposited sediments have reached the seven-day-old layer,
their age no longer increases; all deposited sediments with ages greater than or
equal to seven days are treated as being seven days old. Previous laboratory
studies (Tsai and Lick, 1987; Maclntyre et al., 1990) indicate that consolidation
effects on resuspension are minimal after seven days of consolidation, and are
the basis for setting the maximum age of deposited sediments at seven days.
Consolidation effects on resuspension are accounted for in Equation (3-7) by the
(Td)""1 term, which causes the resuspension potential (e) to decrease as the bed
consolidates with time. The critical shear stress, icr. is assumed to be constant
in all layers of the bed. The model properly accounts for changes in bed
composition, i.e., U and f2, due to resuspension and deposition during the course
of a simulation.
A model that predicts' suspended load transport in the non-cohesive bed
areas of the TIP has been developed and used in this study. A complete
description of that model is presented in HydroQual (1997b). One of the
important aspects of the non-cohesive suspended load model is the concept of
bed armoring, which limits erosion during a flood due to grain size heterogeneity
in the surface layer of the sediment bed. The thickness of this surface layer,
called the active layer, is of critical importance in applying this model to the TIP.
An expression for the active layer thickness has been developed by van Niekerk
et al. (1992) that is a linear function of the local bottom shear stress. A modified
form of their formulation has been used in this study (HydroQual, 1997b):
2D,
50
^b <
T =
2D.
50
B
JcSQ J
(1-5)
. *6>r c
(3-10)
50
12
-------
where Ta = active layer thickness; icso = critical shear stress for initiation of bed
load, based upon the parent bed d50; and B = adjustable constant. Note that
Equation (3-10) reduces to the original van Niekerk et al. (1992) equation when B
= 1. A potential advantage of Equation (3-10) is that varying hydrodynamic
conditions affect the active layer thickness, with Ta increasing as the current
velocity (and xb) increases, which causes the amount of sediment that is
available for resuspension to increase. The dependence of Ta on ib is not well
known and the constant (B) in Equation (3-10) was adjusted during model
calibration to account for local conditions in the TIP.
3.2 Application to the Thompson Island Pool
A side-scan sonar study was conducted by USEPA in the TIP (Flood, 1993).
Information from the side-scan sonar survey has been used to determine the
distribution of various sediment bed types, e.g., fine, coarse and rock, in that
reach of the Upper Hudson River (Figure 3-4). The TIP sediment transport
model requires as input a bed map that separates the bed into three types of
sediment: (1) cohesive; (2) non-cohesive; and (3) rock or hard bottom. The
side-scan sonar information on TIP bed types was used to generate this bed
map, which is shown on Figure 3-5.
A review of bulk bed property data for cohesive sediments in the TIP
indicated no definite spatial trends. Average values, based on available data,
were then used to specify the dry density (0.87 g/cm3) and initial composition (fi
= 0.32) in the cohesive bed areas.
A field study was conducted during November 1990 to measure in situ
resuspension potential of cohesive sediments in the TIP (HydroQual, 1995). An
analysis of that data indicated that an appropriate value for the exponent n in
Equation (3-7) is 2.94. The data also exhibited a spatial variation in the site-
13
-------
specific constant, a0, with lower a0 upstream of HRM 191 (approximately). Thus,
a0 = 0.035 for cohesive sediments upstream of HRM 190.7 and a0 = 0.107 for
cohesive sediments downstream of that location (Figure 3-6). Cohesive
sediments immediately downstream of the jetty at the entrance to the Champlain
Canal, at about HRM 189, were more easily resuspendable and a0 = 0.239 in
that small area. All of these a0 values are for e in mg/cm2. Values of icr = 1
dyne/cm2, m = 0.5 and Tdimax = 7 days were used in these calculations of a0
(HydroQual, 1995) and all model simulations.
An average dry density of 1.1 g/cm3 was applied to non-cohesive sediments
in the TIP based on bulk property data. Similar to the cohesive sediments, no
evident spatial trend in non-cohesive sediment dry density was observed in the
data.
The non-cohesive suspended load transport model requires specification of
median particle diameter (d5o) and suspendable sediment fractions (i.e., fi and f2)
in the non-cohesive bed of the TIP. The non-cohesive transport model is
sensitive to local values of dso, fi and f2. TIP data for these quantities are highly
variable. Hence, it was necessary to develop spatial distributions for dso, fi and f2
to realistically simulate non-cohesive suspended load transport in the TIP.
As would be expected, a relationship exists between median particle
diameter (dso) and fraction of suspendable sediment (fsus = fi + h) in the non-
cohesive bed. Note that fsus represents the fraction of non-cohesive bed
sediment with d < 425 um (i.e., clay, silt, fine sand and medium sand). Grain
size distribution data collected in the non-cohesive bed area of the TIP were
analyzed and the following relationship was determined (Figure 3-7):
^50 =135 fsT (3-11)
where ds0 has units of um.
14
-------
Credible spatial distributions for d50, fi and f2 could not be developed directly
from the data; the bed property data were too sparse to use various
interpolation/extrapolation methods. Spatial distributions were estimated by
hypothesizing that a relationship exists between local bottom shear stress and
dso, i.e., d50 = f(ib). The hydrodynamic model was used to predict the bottom
shear stress distribution for the non-cohesive bed at a given flow rate (30,000
cfs). The functional relationship between dso and ib was adjusted until the
predicted and measured distributions of dso agreed reasonably well (Figure 3-8).
The resulting function is given by:
where xn = normalized bottom shear stress (W-cmax); Tmax = maximum bottom
shear stress in non-cohesive bed area at given flow rate; and d5o has units of ym.
The predicted spatial distribution of dso is presented on Figure 3-9.
This process was validated as follows. Similar to Equation (3-11), a log-
linear correlation between f2 and dso was observed in the data (R2= 0.80):
where dso has units of um. Predicted d50 values, using Equation (3-12), were
used in Equation (3-13) to generate a predicted distribution of f2 in the TIP. The
resulting comparison with observed f2 values is presented on Figure 3-8. The
average predicted and measured f2 values in the TIP non-cohesive bed area
were 0.29 and 0.34, respectively. The good agreement between predicted and
measured f2 indicates that this procedure yields valid spatial distributions of dso
and f2. Non-cohesive bed data did not indicate a strong relationship between d5o
dSQ = 140e'66,: , r„ <0.45
= 9000r; °2 , t„ >0.45
(3-12)
f2 = 22 d^66
(3-13)
15
-------
and f-i. Thus, an average fi value of 0.065 was used throughout the TIP non-
cohesive bed.
The spatial distributions of d5o, U and h that were estimated using the
above procedure are initial conditions for the model. The non-cohesive sediment
bed model tracks temporal changes in fi and f2 in each grid element due to
resuspension and deposition. The data-based relationship between d50 and fsus,
Equation (3-11), was then used to dynamically adjust dso during a simulation.
Sediment loading at the upstream boundary of the model (Fort Edward)
and from the TIP tributaries was estimated using total suspended solids (TSS)
data collected at those locations. Sediment rating curves, which relate TSS to
flow rate, were developed using available data for the Hudson River at Fort
Edward, Snook Kill and M'oses Kill. To correct for bias introduced when
performing log-linear regression analyses on the data, the minimum variance
unbiased estimator (MVUE) method of Cohen et al. (1992) was used. The
resulting rating curves for Fort Edward, Snook Kill and Moses Kill are shown on
Figures 3-10, 3-11 and 3-12, respectively. Sediment loading from TIP direct
drainage was estimated using a rating curve that was the average of the rating
curves for Snook and Moses Kills.
The composition of the incoming sediment loads at the upstream
boundary and tributaries also had to be specified. The fraction of class 1
(clay/silt) and class 2 (fine sands) in the sediment loads was estimated from data.
The USGS has collected particle size distribution data at Schuylerville (32
observations), Stillwater (20 observations) and Waterford (80 observations). No
correlation between sand content and flow (or TSS) exists at these three
locations. The mean sand fractions at Schuylerville, Stillwater and Waterford
were 0.26, 0.19 and 0.16, respectively. This trend of downstream fining is
consistent with observed trends in other rivers. Based on these data, the
assumption was made that the sand content of sediment loads at Fort Edward
16
-------
and Snook Kill was 0.25 for all flow rates. The sand content of sediment loads
from Moses Kill and TIP direct drainage was assumed to be zero. Initial model
testing showed that unrealistic amounts of sediment were predicted to be
deposited at the mouth of Moses Kill whenever sand was included in the
sediment loading for that tributary. An examination of the geometry/bathymetry
of Moses Kill near its confluence with the Hudson River suggests a depositional
zone that would likely trap most suspended sands and significantly reduce the
sand load from the tributary to the river. Sediment loading from TIP direct
drainage was assumed to originate from direct runoff and very small streams.
The hydraulic characteristics of these sediment sources prevent the transport of
significant quantities of sand from the direct drainage area to the Hudson River.
3.3 Model Calibration Results
The eddy diffusivity (KH) initially was set equal to the eddy viscosity (AH)
used in the hydrodynamic model, i.e., 1 m2/s. Model validation, to be discussed
in detail in the following sub-section, using data and observations from the 1997
spring flood indicated that the model produced more realistic results when an
anisotropic Kh was used. The value of KH in the lateral (cross-channel) direction
was reduced to 0.1 m2/s to better reproduce observed patterns of suspended
sediment plumes from Snook and Moses Kills during the 1997 spring flood.
The sediment transport model was calibrated using TSS data collected
during the 1994 spring flood. The 40-day simulation period extended from March
22 to April 30, 1994. Maximum daily average flow rate at Fort Edward during this
period was 27,700 cfs (Figure 3-13). This period was unique because TSS data
were collected at Fort Edward, Snook Kill, Moses Kill and three locations in the
TIP, see HydroQual (1997a) for a detailed discussion of the data. The most
continuous set of TSS data collected at these six locations in the TIP was
17
-------
collected during the 30-day period from March 31 through April 29. Model
calibration efforts focussed on this period.
Model parameters governing cohesive sediment resuspension and
deposition were determined using TIP field data and were not adjusted during
model calibration. Settling speeds for cohesive (class 1) sediments were
calculated using Equation (3-3). The probability of deposition parameter Tb,min
used in Equations (3-4) and (3-5) was set at 0.1 dyne/cm2 and not adjusted
during calibration.
Model calibration involved determining a consistent set of values for the
following four parameters: 1) d2, the effective diameter of suspended class 2
particles; 2) B, the constant in the non-cohesive bed active layer thickness
equation; 3) d5o(x,y), the spatially variable median diameter of sediment in the
non-cohesive bed; and 4) f2(x,y), the spatially variable fraction of class 2
sediment in the non-cohesive bed.
Because suspended class 2 particles are not likely to be coarser than the
largest fine sands (i.e., diameter < 250 um), values of d2 were restricted to the
range of 62 to 250 um. Initial values for d50(x,y) and f2(x,y) were generated using
Equations (3-12) and (3-13). Determination of consistent parameter values
involved finding values of d2 and B that produced accurate erosion and
deposition fluxes for bed characteristics (d50 and f2) expected to exist at the time
of the calibration flood. The following iterative procedure was employed:
1. Use the spatial distributions of d50 and f2 that were generated by
Equations (3-12) and (3-13) as initial conditions for the non-cohesive bed
during the 1994 calibration simulation. Denote these estimated bed
property distributions as °d5o(x,y) and °f2(x,y).
2. Adjust d2 and B to achieve the best agreement between predicted and
observed sediment transport information during the 30-day period from
18
-------
March 30 through April 29 (descriptions of model-data comparisons are
provided below). Denote parameter values for the first 1994 calibration
iteration as 1d2 and 1B.
3. Run a long-term simulation (which extends from 1977 to 1998 and is
described in Section 3.4) using 1d2 and 1B. The initial bed property
distributions used in the long-term run were 0d5o(x,y) and 0f2(x,y).
4. Equation (3-11) was used to dynamically calculate dso as fsus (= f1 + f2)
changed due to resuspension and deposition during the long-term
simulation. The model was then run from 1977 to March 21, 1994 and
the predicted non-cohesive bed property distributions at the end of that
simulation, i.e., 941d5o(x,y) and 941f2(x,y), were output.
5. The model was re-calibrated using the predicted bed property
distributions ^dsofx.y) and 941f2(x,y) as initial conditions. Denote the
best-fit parameter values determined during the second 1994 calibration
iteration as 2d2 and 2B.
6. The long-term simulation (1977 to 1994) was repeated using 2d2 and 2B.
The initial bed property distributions used in the second long-term run
were 0d5o(x,y) and 0f2(x,y). Denote the bed property distributions at the
end of this simulation (March 21, 1994) as 94'2d5o(x.y) and ^ ^(x.y).
7. A validation simulation was performed by running the 1994 spring flood
using the parameter values 2d2 and 2B, with 94 2dso(x,y) and ^^(x.y) as
initial conditions. The differences in the model predictions between the
second and final iteration were small, indicating that "convergence" had
been achieved. The values of 2d2 and 2B were 90 ym and 0.02,
respectively.
The calibration process involved comparing predicted and observed TSS at
three locations in the TIP: (1) upstream of Snook Kill; (2) McDonald's dock; and
(3) Thompson Island Dam. Results of the final calibration are presented on
Figure 3-13.
19
-------
Comparisons between predicted and observed TSS are a standard method
of calibrating and validating a sediment transport model (e.g., Gailani et al., 1991;
Ziegler and Nisbet, 1994). However, this method does not necessarily ensure
that the model realistically and accurately simulates resuspension and deposition
fluxes in the TIP. The reason for this uncertainty is that external solids loadings,
from upstream and tributary sources, may dominate predicted/observed TSS in
the TIP, with deposition and resuspension causing relatively small changes in
water column sediment concentrations. Large changes in model parameters,
creating large changes in deposition and resuspension, may cause relatively
small changes in predicted TSS.
To reduce the uncertainty in model parameterization of deposition and
resuspension processes, and thus to improve the predictive capabilities of the
sediment transport model, a new calibration methodology was developed. The
authors are not aware of the proposed method being used previously to calibrate
or validate a sediment transport model. This procedure involves constructing a
sediment mass balance for the TIP using the total sediment load input from
upstream and tributary sources (Ljn) and the output sediment load at Thompson
Island dam (Lout) to calculate the net resuspension/deposition in the TIP (Lnet).
i.e., Lnet = Lm - Lout- Thus, for a given time period, net deposition occurs if Lnet > 0
and net resupension occurrs if Lnet < 0-
Sufficient data were collected during the 30-day period from March 31 to
April 29, 1994 to make credible estimates of Lnet on an hourly basis. The results
of this data-based analysis showed that net erosion occurred during this period
and 457 metric tons of sediment were transported out of the TIP. Closer
examination of data indicated that the 30-day period under consideration could
be separated into two distinct sub-periods: (1) tributary deposition and (2)
mainstem flood. From March 31 to April 10, flow rates in the Hudson River were
non-flooding (< 10,000 cfs) but high flow events occurred in the tributaries.
Snook and Mose Kills transported large quantities of sediment into the TIP during
20
-------
this sub-period and, because flow rates in the Hudson River were relatively low,
significant deposition occurred. The mass balance results indicate that 387
metric tons of sediment were deposited during the tributary deposition sub-
period. Net erosion occurred during the mainstem flood sub-period, extending
from April 11 to 29, with 844 metric tons of sediment lost from the TIP.
Note that this type of data analysis can only be used to estimate global
losses or gains due to net erosion or deposition from the TIP sediment bed.
Results from this analysis cannot be used to infer net erosion or deposition in
specific bed types, e.g., cohesive or non-cohesive, or areas of the TIP.
Comparisons between predicted and observed cumulative Let. for the final
calibration, during the entire 30-day period and the two sub-periods are
presented on Figure 3-14. ' These results show that the model can predict
temporal variations in Lnet with good accuracy. The model predicts a total of 345
metric tons of sediment were exported from the TIP during the 30-day period,
which is 25% lower than the data-based estimate. During the tributary deposition
sub-period, 523 metric tons of deposited sediment were predicted, corresponding
to a 35% over-prediction when compared to the observed value. The model is
within 3% of the data-based estimate of net erosion during the mainstem flood
sub-period.
Very good agreement between predicted and data-based estimates of
cumulative Lnet during the 1994 spring flood demonstrates that: (1) deposition
and resuspension processes have been realistically and accurately formulated in
a global sense and (2) the model is an effective diagnostic tool for quantitatively
evaluating net deposition and erosion from various areas of the TIP. While the
model and data indicate net erosion occurred during this 30-day period, closer
examination of the model results show that net erosion did not occur in all areas
of the TIP. The non-cohesive portion of the TIP, which comprises about 80% of
the total bed area in this reach, experienced a net loss of 1,244 metric tons. This
21
-------
net erosion corresponds to a decrease in the mean non-cohesive bed elevation
of 0.08 cm. Conversely, 899 metric tons of net deposition occurred in the TIP
cohesive bed, which is equivalent to an average increase in the cohesive bed
elevation of 0.26 cm. The spatial distribution of sediment deposition caused by
this flood is presented on Figure 3-15. Prediction of net deposition in the TIP
cohesive bed area during this flood is consistent with observed depositional
patterns in fine-grained areas of the Upper Mississippi River during major
flooding in 1993 (Barber and Writer, 1998).
The potential impact of the initial non-cohesive bed distribution on model
results was investigated by repeating the final calibration run with the initial bed
distributions Odso(x,y) and 0f2(x,y). The results of this simulation are presented on
Figure 3-16. Using °d50(x,y) and 0f2(x,y) primarily affected the mainstem flood
sub-period, with erosion increasing from 868 to 1,215 metric tons (40%
increase).
3.4 Model Validation Results
Three simulations were conducted to validate the sediment transport model:
(1) 1997 spring flood; (2) 1993 spring flood; and (3) 21-year (1977 to 1998)
period. No adjustments of model parameter values were made during the
validation simulations. Only model boundary conditions, e.g., flow rates and
sediment loadings, were changed to reflect the time-varying conditions during
each validation period.
The spring flood that occurred in early May 1997 had a relatively low peak
flow, with a maximum flow rate at Fort Edward of approximately 17,000 cfs.
General Electric collected TSS data at Fort Edward, Snook Kill, Moses Kill and
Thompson Island dam between May 2 and 6, which was the period when a high
flow event occurred on Snook and Moses Kills. Comparisons of predicted and
22
-------
observed TSS at Thompson Island dam during this five-day period are shown on
Figure 3-17. Model results were in excellent agreement with measured TSS on
both the western and eastern shores of the dam. Note that TSS data were
collected at 3-hour intervals on May 3 and 4 (days 29-31) and once per day on
May 2, 5 and 6. Linear interpolation was used to estimate measured TSS at the
dam for times between each data point during this period and the resulting time
series is shown on an hourly basis on Figure 3-17.
Peak flow during the 1993 spring flood was comparable to the 1994 flood,
with a maximum flow rate at Fort Edward of approximately 29,000 cfs. A limited
amount of TSS data were collected at Fort Edward and Thompson Island dam
between March 22 and May 6, 1993, which was the 45-day period simulated. No
TSS data were obtained from the tributaries, so sediment loads from Snook and
Moses Kills were estimated using the sediment rating curves discussed in
Section 3.2. Comparisons between predicted and measured TSS concentrations
at the dam are shown on Figure 3-18. The data are somewhat limited but the
model appears to perform reasonably well.
As discussed in the previous sub-section, long-term calculations were
performed in an iterative method as part of the model calibration process. A final
long-term simulation was conducted to further validate the sediment transport
model. This calculation was over 21 years long, starting on January 1, 1977 and
ending on March 24, 1998. The hydrograph at Fort Edward during this period is
presented on Figure 3-19.
Sediment loading to the TIP was determined using a combination of data
and sediment rating curves. At Fort Edward, TSS data were used on all days for
which data were available. On days that data were not collected at Fort Edward,
TSS concentration was estimated using a sediment rating curve (Figure 3-10).
Tributary sediment loads were estimated using sediment rating curves (Figures
3-11 and 3-12). The total annual sediment loads to the TIP resulting from this
23
-------
procedure for 1977 to 1997 are shown on Figure 3-20. The total sediment load
input to the TIP during the long-term simulation period was approximately
805,000 metric tons, for an average of 37,900 metric tons/year.
Predicted areas of erosion and deposition at the end of the long-term
simulation are shown on Figure 3-21. Deposition rates for this approximate 21-
year period are presented on Figure 3-22. Net erosion depths are displayed on
Figure 3-23.
The locations of three high-resolution sediment cores collected by USEPA
are also indicated on Figure 3-22 (HR cores 19, 20 and 23). Geochronologic
dating of those cores, using 137Cs, indicates average sedimentation rates of 0.9,
1.1 and 1.4 cm/yr at HR cores 19, 20 and 23, respectively. The predicted mean
sedimentation rates for the grid elements that contained HR cores 20 and 23 are
0.5 and 1.1 cm/yr, respectively. Core 23 is located at the mouth of Moses Kill
and the relative error of the predicted sedimentation rate at this location is -21%.
At core 20, the predicted rate is 55% lower than the observed rate. However, the
predicted rate for the grid element immediately upstream of this core is 0.95
cm/yr, which is only 14% lower than the core 20 rate. Net erosion was predicted
in the grid element containing core 19, but the grid element directly upstream had
a predicted rate of 1.7 cm/yr. These validation results indicate that the sediment
transport model can predict long-term sedimentation rates with reasonably good
accuracy.
Performing a sediment mass balance on the TIP, i.e., calculate Lnet for the
long-term simulation period, shows that the model predicts that about 68,000
metric tons of sediment were deposited during the period from 1977 to 1998.
This depositional mass corresponds to a long-term trapping efficiency of 8.5% for
the TIP. Most of the deposition occurs in the cohesive bed areas of the pool,
with 87% (nearly 59,000 metric tons) being deposited in those areas. This
amount of deposition translates to an average sedimentation rate of 0.8 cm/yr for
24
-------
the cohesive bed in the TIP. This sedimentation rate equates to an average
deposition of about 17 cm (6.6 inches) over the 21-year period simulated. Net
deposition also occurs for the TIP non-cohesive bed, but only at an average rate
of 0.02 cm/yr.
25
-------
SECTION 4
SEDIMENT TRANSPORT PROCESSES IN THE THOMPSON ISLAND POOL
Sediment transport processes in the TIP involve complex interactions
between external sediment loading and sediment dynamics in cohesive and non-
cohesive bed areas. Understanding these processes, both qualitatively and
quantitatively, is of critical importance when considering PCB fate and transport
in this reach. Unfortunately, the dynamics of TIP deposition and resuspension
are not clearly discernible from data analyses. A general understanding of the
processes controlling sediment transport in the TIP can only be achieved when
data are combined with a credible model, as has been done in this study.
The sediment transport model developed for the TIP is a physically
comprehensive model that simulates the resuspension, deposition and transport
of cohesive and non-cohesive sediments. Various TIP data sets have been used
to develop a model that contains only two parameters that were adjusted, within
realistic bounds, during a rigorous calibration process. Successful calibration
and validation of the model, for short-duration, high-flow events and a decadal
time-scale simulation, strongly suggests that this model realistically and
accurately simulates sediment transport in the TIP. Thus, the model can be
confidently used as a diagnostic tool to better understand TIP sediment transport
processes.
4.1 Importance of the Non-Cohesive Sediment Bed
The non-cohesive bed area is an important component of the overall
sediment transport system operating in the TIP. Suspended fine sands (class 2
sediments in the model) are deposited in the non-cohesive bed, primarily
between Rogers Island and Snook Kill, during low to moderate flows. These
26
-------
previously deposited fine sands are resuspended during high flow events and
transported downstream by river currents. The coarser suspended sediments
can be then be re-deposited in the cohesive bed areas. TIP cohesive deposits
are located in relatively low energy environments during floods and are
depositional zones for fine sands suspended in the water column. Thus, the non-
cohesive bed effectively acts as a reservoir for coarser suspended sediment;
material accumulates during low to moderate flows, it is then released back to
the water column during a flood and enhances deposition in cohesive bed areas.
4.2 Long-Term Sedimentation in Cohesive Deposits
Modeling results clearly demonstrate that the TIP is a net depositional
environment, as would be expected of a run-of-the-river reservoir, and that
widespread deposition is occurring in the cohesive ("hotspot") bed areas of the
TIP. While net erosion was predicted to have occurred in approximately 7% of
the cohesive bed area between 1977 and 1998 (with an average erosional depth
of 1 cm), significant net deposition (average rate of 0.8 cm/yr) occurred over
most of the cohesive bed (93%), These results are consistent with both the
observed decrease in surficial PCB bed concentrations since 1977 and the
occurrence of maximum PCB bed concentrations at depth, i.e., less-
contaminated sediments are entering the TIP and burying historical PCB
deposits.
4.3 Erosion During Rare Floods
The impact of a 100-year flood on the TIP was simulated using the calibrated
sediment transport model. The peak flow rate for a 100-year flood at Fort
Edward has been estimated to be 47,300 cfs, on a daily average basis. A
hydrograph for the simulated 100-year flood was developed by analyzing the
27
-------
hydrographs of nine floods in the TIP that occurred between 1977 and 1997.
The resulting flood hydrograph extended over eight days, with the peak flow
occurring on day 4 of the simulation. Sediment loadings during the flood, from
both upstream and tributary sources, were estimated using sediment rating
curves discussed in Section 3.2. Note that both erosion and deposition were
accounted for in this simulation.
The model predicted that a total of 2,166 metric tons of sediment would be
transported out of the TIP during the 100-year flood. Of this total, 952 and 1,214
metric tons of net erosion occurred in the cohesive and non-cohesive bed areas,
respectively. These masses correspond to mean erosional depths of 0.69 and
0.09 cm for the cohesive and non-cohesive bed, respectively. Maximum
erosional depths for both bed types were approximately 7.5 cm. The spatial
distribution of predicted erosional depths at the end of the 100-year flood is
presented on Figure 4-1. Note that net deposition was predicted at some
locations in the TIP.
These results indicate that an extreme flood will not scour deep enough into
the sediment bed to expose historical deposits with elevated PCB levels. The
impact of a 100-year flood on average surficial PCB concentrations in the TIP,
which are of importance when considering effects on biota, will be examined in
the near future.
4.4 Resuspension During Low Flows
The long-term simulation, from 1977 to 1998, indicates that resuspension,
from both cohesive and non-cohesive bed areas, is negligible during low flow
conditions, i.e., flow rates less than 3,000 cfs. This characteristic is caused by a
combination of two factors: (1) bottom shear stresses are less than critical
values for initiation of erosion in most of the TIP and (2) cohesive and non-
28
: 2
-------
cohesive bed armoring effects significantly limit erosion in areas where xb> xa.
Thus, increases in water column PCB loads between Fort Edward and
Thompson Island dam during low flow conditions cannot be attributed to
resuspension of contaminated sediments from the sediment bed.
29
-------
SECTION 5
REFERENCES
Amos, C.L., Grant, J., Daborn, G.R. and Black, K., 1992. Sea Carousel - A
Benthic, Annular Flume, Estuar. Coast, and Shelf Sci.. 34:557-577.
Barber, L.B. and Writer, J.H., 1998. Impact of the 1993 Flood on the Distribution
of Organic Contaminants in Bed Sediments of the Upper Mississippi River,
Environ. Sci. Tech.. 32:2077-2083.
Blumberg, A.F. and Mellor, G.L., 1983. Diagnostic and Prognostic Numerical
Circulation Studies of the South Atlantic Bight, J. Geophvs. Res.. 88(C8):4579-
4592.
Burban, P.Y., Xu, Y.J., McNeil, J. and Lick, W., 1990. Settling Speeds of Floes in
Fresh Water and Seawater, J. Geophvs. Res.. 95(C10):18,213-18,220.
Cardenas, M., Gailani, J., Ziegler, C.K. and Lick, W., 1995. Sediment Transport
in the Lower Saginaw River, Mar. Freshwater Res., 46:337-347.
Cheng, N.S., 1997. Simplified Settling Velocity Formula for Sediment Particle,
ASCE J. Hvdr. Enqr.. 123(2): 149-152.
Cohn, T.A., Caulder, D.L., Gilroy, E.J., Zynjuk, L.D. and Summers, R.M., 1992.
The Validity of a Simple Statistical Model for Estimating Fluvial Constituent
Loads: An Empirical Study Involving Nutrient Loads Entering Chesapeake Bay,
Water Resour. Res., 28(9):2353-2363.
Flood, R.D., 1993. Analysis of Side-Scan Sonar, Bathymetric, Subbottom, and
Sediment Data from the Upper Hudson River Between Bakers Fall and Lock 5,
Hudson River PCB Reassessment RI/FS, EPA Work Assignment 013-2N84.
30
-------
Gailani, J., Ziegler, C.K. and Lick, W., 1991. Transport of Suspended Solids in
the Lower Fox River, J. Great Lakes Res.. 17(4):479-494.
Gailani, J., Lick, W., Ziegler, C.K. and Endicott, D., 1996. Development and
Calibration of a Fine-Grained Sediment Transport Model for the Buffalo River, J.
Great Lakes Res.. 22:765-778.
Graham, D.I., James, P.W., Jones, T.E.R., Davies, J.M. and Delo, E.A., 1992.
Measurement and Prediction of Surface Shear Stress in Annular Flume, ASEC J.
Hvdr. Enar.. 118(9): 1270-1286.
Hawley, N., 1991. Preliminary Observations of Sediment Erosion from a Bottom
Resting Flume, J. Great Lakes Res., 17(3):361 -367.
HydroQual, 1995. The Erosion Properties of Cohesive Sediments in the Upper
Hudson River, HydroQual report, October 1995.
HydroQual, 1997a. Analysis of Sediment Loading to the Upper Hudson River
During the April 1994 High Flow Event. HydroQual report. February 1997.
HydroQual, 1997b. Modeling Suspended Load Transport of Non-Cohesive
Sediments in the Upper Hudson River, HydroQual report, June 1997.
Krone, R.B., 1962. Flume Studies of the Transport of Sediment in Estuarial
Processes, Final Report, Hydraulic Engineering Laboratory and Sanitary
Engineering Research Laboratory, Univ. of Calif., Berkeley, Calif.
Limno-Tech memorandum, 1998. Preliminary tributary flow estimates for the
Upper Hudson River between Fort Edward and Waterford for the USEPA mass
balance modeling, from Mike Erickson, to Doug Tomchuk, June 3, 1998.
31
-------
Maclntyre, S., Lick, W., and Tsai C.H., 1990. Variability of Entrapment of
Cohesive Sediments in Freshwater, Bioqeochemistrv, 9:187-209.
Parchure, T.M. and Mehta, A.J., 1985. Erosion of Soft Cohesive Sediment
Deposits, ASCE J. Hvdr. Enar.. 111 (10): 1308-1326.
Partheniades, E., 1992. Estuarine Sediment Dynamics and Shoaling Processes,
in Handbook of Coastal and Ocean Engineering. Vol. 3. J. Herbick, ed., pp.985-
1071.
Tsai, C.H. and Lick, W., 1987. Resuspension of Sediments from Long Island
Sound, Wat. Sci. Tech.. 21 (6/7): 155-184.
Van Niekerk, A., Vogel, K.R., Slingerland, R.L. and Bridge, J.S., 1992. Routing
of Heterogeneous Sediments Over Movable Bed: Model Development, ASCE J.
Hvdr. Enor.. 118(2):246-279.
Ziegler, C.K. and Nisbet, B., 1994. Fine-Grained Sediment Transport in
Pawtuxet River, Rhode Island, ASCE J. Hvdr. Engr., 120(5):561-576.
Ziegler, C.K. and Nisbet, B.S., 1995. Long-Term Simulation of Fine-Grained
Sediment Transport in Large Reservoir, ASCE J. Hvdr. Engr., 121(11):773-781.
32
-------
¦omoson isiano Cam
Figure 2-1. Numerical grid for the Thompson Island Pool.
-------
Figure 2-2. Thompson Island Pool bathymetry. Data were averaged in each grid element
for model input.
-------
130
129
128
127
Gauge 119
126
125
124
Gauge 118
'0 o-
121
120
16
14
18
20
12
22
10
24
S
8
4
2
0
Day
(April 24, 1983 - May 16, 1983)
Figure 2-3. Comparison of predicted (solid line) and observed stage heights at
gauges 118 and 119 during the 1983 flood.
-------
10000
OTAU
(mq/l ~ dynes/crrr2)
Figure 3-1. Settling speed function for cohesive (class 1) sediments (solid line)
and floe settling speed data (mean + 95% confidence interval) used to construct
function.
-------
1.00
0.90
0.80
0.70
c
o
to
o
Q_
-------
5000
1000
100
500
Particle Diameter
(microns)
Figure 3-3. Settling speed of sand particles (class 2) as a function of particle
diameter.
-------
Match Line
Match Line
V)
\\
/\/ Shoreline
/\/ Mile Markers
/\y Dams & Locks
Side Scan Sonar
¦I COARSER
™ LINEATED
1 MOUND
3 FINER
MARSH
] ROCKY
N
I
1000 0 1000 2000 Feet
Figure 3-4. Thompson Island Pool bed map based on side scan sonar information.
-------
ft
Match Line
Match Ur.e
I \
\\
/\ J Shoreline
/\/ Mile Markers
/\/ Dams and Locks
~ 1 Hard Bottom
Sediment Bed Map
Non-Cohesive
Cohesive
N
t
1000 0 1000 2000 Feet
Figure 3-5. Thompson Island Pool bed map used for model input.
-------
a o
« \ 0 20
River Mile
188
Figure 3-6. Spatial distribution of resuspension potential parameter (a0) used as
model input. TIP shaker data (mean ± 95% confidence interval) used to
determine model input are also shown.
-------
gj w
as
U
0.01
o
o
' ' I r 1 i I i I I ' i i i i i I I ! I 1 1 I 1 ' 1
10
100
1000
10000
D50
(um)
Figure 3-7. Relationship between fraction of suspendable sediment (d2 < 425
um) and median particle diameter (d50) for the non-cohesive sediment bed in the
Thompson Island Pool.
-------
D50
fTTJiitn —i i iiiiiii—
-i—i—r
—i—r- i-
—rimii i >- pirn 11 -z
.
jtfooooo I
1000
1 Kill
A
:
-
"
100
:
z
-
N=102 "
10
1 t IIHH ' ' ' """
i t i
1 1 L_
tft 11 i l i r ¦inti l i
0.1 1 10 20 50 80 90 99 99.9
Fraction Sand
. i i lint* -7 17 iTinr
1 1 1
1 '1 -r - i-
—nrnrri l eiinit t _
-
>
3
-
-
-------
Malcn une
Matcfi Lire
/\J Shoreline
/\J Mile Markers
/\/ Dams & Locks
Non Cohesive Bed D50
Fine Sands
' Medium Sands
J|| Coarse Sand
3M Very Ccarse Sancl
¦¦ Gravel
N
I
1000
1000 2000 Feet
Figure 3-9. Estimated spatial distribution of median particle diameter (d50) used as
model initial conditions for the non-cohesive sediment bed in the Thompson Island Pool.
-------
in ~
CO O)
t— E
1000,
100
10
o.i
100
• «DO « O
-1 I I ' ' 1 '
_l 1 ' . ¦ ' . ' I
1000
10000
100000
FLOW RATE
(cfs)
Figure 3-10. Sediment rating curve (solid line) for upstream boundary at Fort
Edward.
-------
1000.
100.
C/)
-------
1000
100.
CO
CO
»-
Ci
E
10
' ' t I I I I 1 ¦
1 ¦ I 1 ¦
10
100
Flow Rate
1000 2000
(cfs)
Figure 3-12. Sediment rating curve for Moses Kill.
-------
3.5x10
3.0x10*
Flow at Fort Edward
z
u 2.5x10*
=r
-=
(Ibr 2.0x10*
=-
J ^ _
J^- 1.5x10*
=-
/ -=
It. A
1.0x10
i-
»_ J* S
-=
5.0X103
Zxj—
0
E
80
TIP above Snook Kill
60
(/>Ot
¦~E
40
20
80
TIP at McDonalds Dock
60
40
80
Thompson Islond Dom
60
20
25
30
20
29
5
10
15
1
March 30 - April 29, 1994
Figure 3-13. Comparison of predicted (solid line) and observed suspended
sediment concentrations at three locations in the Thompson Island Pool during
the 1994 spring flood.
-------
3.5x10 F
3.0x10* =-
Flow ot Fort Edward
2.5xio
2.0x10
1.5xi0
1.0x10
5.0x10"* i
c
o
€>0*
>i.C
•«3LdO
ctv
m
oSe
*1
Q
1500
1000
500
0
-500
-1000
-1500
— Deposition
i_ Erosion
c
o
UO»T
•^io
tfvT*
Do-
V
a
1500
1000
500
oSE "500
-1000
-1500
— Deposition
Tributary Deposition Period
—4
4
i_ Erosion
'^aJO
its
OoE
at—'
v
a
1500
1000
500
0
-500
-1000
-1500
Deposition
Mainstem Fiooa Perioa
1_ Erosion
30 1
10
15
20
25
J
|
-4
29
March 30 - April 29, 1994
Figure 3-14. Comparison of predicted (solid line) and data-based (dashed line)
changes in TIP sediment bed mass during the 1994 spring flood. Initial
conditions for non-cohesive bed properties calculated during long term
simulation.
-------
Match Line
M
/\/ ShorsiirM
/\J Mis Markers
/ y Dams & Locks
Rocky Areas
8«d Elevation Chang® (mm)
Erosion
1000 2000 Feet
1000
Figure 3-15. Predicted deposition (in mm) that occurred during the 1994 spring flood.
-------
c
o
'rrr>
?2c
aujo
3 = o
r Oc
oSE
O—'
*)
o
1500
1000
500
0
-500
-1000
-1500
=_ Deposition
z_ Erosion
c
o
ur~-
22c
• -i ¦' o
CTsT
Egfc
u8e
a—
v
o
1500
1000
500
-500
-1000
-1500
Deposition
Tributary Deposition Period
Erosion
1500
1000
500
0
oSE "500
a -1000
C
o
*»C
22c
CTsT'
IS*
I_ Deposition
-1500 E
Mainstem Flood Period
r_ Erosion
30
j
j
10
15
20
25
29
March 30 - April 29, 1994
Figure 3-16. Comparison of predicted (solid line) and data-based (dashed line)
changes in TIP sediment bed mass during the 1994 spring flood. Initial
conditions for non-cohesive bed properties estimated as shown on Figures 3-8
and 3-9.
-------
20000
15000
£ 10000
6000
FT. EDWARD
°r ¦ ¦ ¦ i . i i i i . i i i i i i . i ¦ ¦ i ¦ i i ¦ . ¦ ¦ i ¦ i ¦ i i . i i ¦ i i
28.0 28.5 29.0 29.5 30.0 30.E 31.0 31.5 32.0 32.5 33.0
60
40
30
20
10
i i i i i i i i i i i i i i i i i i i i i i i
i i i i i i i i i i i i i i i i i i i i i i i i i i i i M i i i i i i i i i
FT. EDWARD
¦¦¦¦¦I'1 1
28.0 28.5 29.0 29.5 30.0 30.5 31.0 31.5 32.0 32.5 33.0
50
THOMPSON ISLAND DAM-EAST
40
30
2B.0
28.5
29.0
29.5
30.0
30.5
32.5
33.0
32.0
31.0
31.5
60
THOMPSON ISLAND DAM-WEST
40
30
20
29.5
30.0
33.0
28.0
28.5
29.0
30.6
31.6
31.0
32.0
32.5
DAY
Figure 3-17. Comparison of predicted (solid line) and measured suspended
sediment concentrations at the Thompson Island dam during the 1997 spring
flood (May 2 to 6, 1997).
-------
3.5*104p"
•3.0xio4|-
u 2.5x10
2.0x10*p-
I.Sxio4
u. 4
i.oxio E-
5.0x1 0^ =-
Flow ot Fort Edward
80 r~
60 -
OT\
t/)a> 40
""E
20
0
TIP obove Snook Kill
TIP ot McDonalds Dock
i/l cn 40
I/T\
i/)o> 40
Thompson Island Dam
March 22 - May 6, 1993
Figure 3-18. Comparison of predicted (solid line) and measured suspended
sediment concentrations at the Thompson Island dam during the 1993 spring
flood.
-------
40000
30000
£ 20000
10000
1979
1978
1977
1980
40000
30000
a: v>
~ 20000
10000
1983
1984
1982
1981
30000
~ 20000
10000
1987
1988
1986
1985
Year
Figure 3-19a. Daily average flow rate at Fort Edward from 1977 to 1988.
-------
40000
30000
«>
* ~ 20000
o ' '
u.
10000
1990
1991
1989
1992
40000
30000
u
J z 20000
o ' '
10000
1995
1996
1993
40000
p-—. . . — . • ¦ -• ¦ T"
30000
i-
-E
o ^
E
OC m
f 20000
o
f
—~
u_
10000
bJA
u .
0
1997 1998
Yeor
Figure 3-19b. Daily average flow rate at Fort Edward from 1989 to 1998.
-------
4.0* 10
3 0x10'
o
* {• 2 OxIO4
o w
1 0x10
Max Ft. Edward Flow Rate
1980
1985
1990
1995
8x10
¦a ia
S c
9 °
TIP Sediment Load
6x10
o 4x10
2X10
1980
1985
1990
1995
200
TIP Sediment Load
T3
150
« K 100
1990
1985
1995
1980
Year
Figure 3-20. Maximum annual flow rate at Fort Edward, total TIP annual
sediment load and relative annual sediment load for 1977 to 1997.
-------
Match Line
Match Line
/\/ Shoreline
/\/ Mile Markers
/\/ Dams & Locks
¦fc Rocky Areas
Bed Elevation Change
Erosion
Deposition
N
I
1000 0 1000 2000 Feet
Figure 3-21. Predicted deposition and erosion areas at end of long term simulation
(1977 to 1998).
-------
Match Line
HR Core 2C: 1.1 cm/yr
Match Line
A
HR Core 23: 1.4 cm/yr
^\HR Core 19: 0.9 cm/yr
/\/ Shorwne
/\/ Mil* Martcars
/S/Oams & Locks
# High Resolution Coras
Rockv Areas
Bed EJevaeon Change (cnvyr)
Erosion
0.0 - 0.5
"Z 0.5- VO
M 1 s - 2.0
¦I > 2.0
N
I
1000 C 1000 2000 Feet
Figure 3-22. Predicted average deposition rates (cm/yr) for 21-year period
(January 1977 to March 1998). Measured deposition rates at three high resolution
''ore locations are also shown.
-------
Match une
Match Une
/\/Shof»»n«
/\/ Mil* Manors
/V 0«'"* * Locks
¦¦ Booty Arau
Bad B«v»nor Cnangt (cm)
5 -2--1
-1 -0
Oapomon
N
I
1000 0 1000 2000 Feet
Figure 3-23, Predicted erosion depths (cm) for 21-year period (January 1977 to March 1998).
-------
Match Line
Match Una
. / Shotwn*
A./ MM Manor*
/ '/ Dam* !l Locks
Mated Una
Rocky Araaa
Sad Bawaoon Cftanga (cm)
¦ < -6
TZ3
-1 -o
~~ DapoaMon
1000
N
I
1000 2000 Feet
"igure 4-1. Predicted erosion depths (cm) for 100 year flood.
------- |