UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
                        MAY 2 8 1991
MEMORANDUM

SUBJECT:  Rel

FROM:
TO:
                   f Pilot Study
          Assistant
          Office of
             inistrator
           esticides am
            n R. Clay
            sistant Adm
          Office of Sol
 bstances  (TS-788)
6
                     or
                      and Emergency Response (OS-100)

Regional Administrators
     The Office of Pesticides and Toxic Substances (OPTS) and the
Office of Solid Waste and Emergency Response (OSWER)  have
released the "Fluff Pilot Study" (attached).  The I-i_ot Study was
initiated in 1988 to determine the extent of PCB, lead and
cadmium contamination in appliance and auto shredder residue,
commonly known as "fluff."

     The results of the pilot study indicate that shredders may
generate waste contaminated at levels above regulatory concern.
However, differences in raw materials, shredding operations, and
the presence of conflicting data suggest that not all shredders
are generators of regulated waste materials.  Information
obtained to date indicates that the levels of constituents of
concern associated with the generation of fluff are not uniform.
The potential risk depends on the constituent makeup of the fluff
and the characteristics of the sites at which the fluff is
generated or disposed.

     EPA supports metal recycling activities when conducted in an
environmentally sound manner.  As such, both OPTS and OSWER have
identified a need to evaluate the current disposal requirements
as they relate to fluff.  OPTS will be publishing an Advanced
Notice of Proposed Rulemaking to solicit public comment on an
amendment to the PCB regulations, which could authorize
alternative methods of disposal for fluff.  The fluff issue may
also be discussed during RCRA reauthorization.

-------
     In the interim, because shredding operations that are well
managed and conducted in an environmentally sound manner provide
valuable environmental benefits, OPTS and OSWER are recommending
that enforcement*be focused on shredding facilities which pose
significant environmental problems.  Such operations could
•include ir.prcpcr use of fluff as fill material in wetlands or
other environmentally sensitive areas, or in residential
settings.  Other criteria to evaluate sites that generate fluff
include location (e.g., sites in 100-year floodplains, near
surface water or surface water discharge, over a drinking water
aquifer or wellhead protection areas); and operation  (e.g., the
absence of activities designed to address blowing fluff piles,
the absence of worker protection measures, the absence of run-
on/run-off controls, and leachate generation).

     This list is  intended only as a guide to indicate when
shredding operations and fluff disposal may warrant enforcement.
It is not an exhaustive list of situations where enforcement may
be appropriate.  Further guidance to assist the regions in
determining where  to focus enforcement will be developed by.the
Office of Compliance Monitoring and the Office of Solid Haste and
Emergency Response.

Attachment

-------
PROJECT SUMMARY
PCB,  Lead, and Cadmium Levels  in Shredder
Waste Materials:   A Pilot  Study
                         Authors and Research Team
Principal authors included Dan Reinftan. John Scalera, Brad Schultz. Cindy Siroup. and AY
Breen of ihe Exposure Evaluation Division, Office of Toxic Substances.   Ttiis project
summary was abstracted from a report on the Fluff Pilot Study written hy Wesiat.
  «
Field work was conducted hy Westat. Midwest Research Institute, and Battelle Columbus
Laboratories.  Chemical analyses  were performed at Midwest Research Imtiiuie.  EPA '.v
National Enforcement Investigation Center, and EPA's Environmental Systems Laboratory
1/1 Las Vegas. Nevada.

EPA's Office of Solid Waste. Characterization and Assessment Division co-sponsored ihe
ejfon  and provided deiiim. sampling, and analysis consultation for lead and cadmium
portion of the pilot study.  Alexander McBride was the principal contributor.
Summary

Prior to this pilol study, ihe United Slates
Environmental Protection Agency (EPA)
received information from stale and  local
environmental agencies which indicated the
shredding of automobiles and  other pro-
duels Tor metal recycling may  produce
waste-materials contaminated  with  poly-
chlorinated hiphcnyls (PCBs). lead, and
cadmium. The inlormulion available was
insufficient to establish  the sources and
extent of the contamination or what  regu-
latory action, if any. Mould he  approriate.
Consequently. EPA's Olfkc of Toxic Sub-
stances and Office of Solid Waste planned
aad  conducted this Mudy to  gain  more
knowledge about .«hrci(
-------
 The results of this study allow EPA to
 make  a  preliminary assessment  of
 potential PCB. lead, and cadmium
 contamination and to provide* valu-
 able information  Tor the design  of
 future studies.   It  is important  to
 acknowledge  thai this  was  a pilot
 study and the study results  may not
 necessarily be  representative  of the
 shredder recycling industry as a whole.
 Only  seven  shredder  sites  were
 included  in  the  study,  and  some
 numerical estimates are based on a
 limited  number  of samples.   For
 practical  reasons, some restrictions
 were imposed on the random selec-
 tion of sites,  although  EPA has no
 reason  to  believe  that bias  was
 introduced by the sampling plan.
INTRODUCTION

The United Stales Metal Shredding
Industry generates approximately 12
to  14 million  tons of steel scrap for
recycling each year.   About W3- of
the steel output is from the 8 to 10
million cars, trucks, and vans which
are  disposed  of every year.    The
remaining steel salvage results from
the  recycling  of  several   million
discarded household appliances and a
variety   of   other   industrial.
commercial, and household scrap.

EPA recognizes the major environ-
mental  benefits  of recycling  as a
national environmental policy  and
wrongly fosters  and   supports  all
recycling efforts which  are environ-
mentally  sound.    Metal  recycling
results  in  a  two-thirds  to  three-
quarters reduction in the volume of
space required in landfills to deposit
waste automobiles awl appliances, a
substantial  reduction   in   energy
required to  recycle metal instead of
producing it from  raw ores,  and a
reduction in air pollution associated
with metal production. The commer-
cial value of recycled metal, over S1.S
billion per year, is considerable.

In addition to recycled metal, shred-
der  operations  produce 3  million
tons of non-metallic waste material
each year. This non-metallic waste is
usually referred to as "fluff" or "auto
shredder residue* (ASR). The shred-
ding of a car. for example, produces
about 5(X) pounds of fluff on average.
Ruff is  typically composed  of a
variety   of   materials,   including
plastics, rubber, foam, fabric, wood.
insulation, glass, road  dirt, and small
metal  fragments.  Little, if any. of
this  material  is presently  recycled.
Most fluff is disposed of in municipal
landfills.

Preliminary and anecdotal  informa-
tion received by the EPA before this
study indicated thai PCBs. lead, and
cadmium  arc dispersed during  the
shredding of various scrap materials.
resulting  in  the contamination of
fluff by these substances.  Some of
the  reported  contamination  levels
exceeded  the   Federal  regulatory
levels set under the Toxic Substances
Control   Act  (TSCA)  and   the
Resource Conservation and Recovery
Act (RCRA).  Shredder wastes which
contain these contaminants in con-
centrations   exceeding  prescribed
TSCA and  RCRA regulatory levels
must, under  Federal regulations, be
managed in approved disposal sites.
This would  result  in  considerably
greater cost to the rccycler (shredder
operator).   In addition, the TSCA
landfill capacity  would  quickly be
filled  il  j   large   proportion  of
shredder  llulf  proved to  be PC&-
contaminated.    Similar  capacity
concerns exist lor  RCRA disposal'
facilities.

Since PCBs were commonly used as   '
dielectric   fluids    in   electrical
transformers and capacitors, many
scrap  metal   shredder  operators
assumed  that capacitors in moiori/cd
consumer appliances (called  'white
goods") were the primary source of
PCB contamination in shredder Huff.

These  operators, therefore, stopped
accepting appliances  for  recycling.
This decision not to accept and  pro-
cess "while  goods*  created a solid
waste disposal predicament in several
slates when refrigerators, stoves, wash-
ing machines, and  other appliances
frequently were abandoned and began
accumulating  along  streets  or  in
vacant lots.

Due to the lack of general knowledge
about shredder operations and conclu-
sive information regarding contami-
nation sources,  the  EPA's Office  of
Toxic Substances (OTS) and Office of
Solid Waste (OSW) undertook this
pilot study.   Of specific interest was
the  examination of PCB.  lead, and
cadmium levels in  shredder  output
streams;  the teachability ol these sub-
stances: and the identification of con-
tamination  sources,  if possible.  The
results of the pilot study will be used
to  evaluate the need lor additional
Agency action  and  10  design future
studies if they are required.
 OBJECTIVE

 The specific objectives ol the study
 were:

 •  To estimate ranees ol PCB. lead.
    and cadmium levels  in  Hull, the
    metallic  outputs, and  in  soil
    collected  from  where   llulf  Is
    stored by the shredder:

 •  To determine how rcadilv  PCBs.
    lead, and  cadmium will leach  itr
    dissolve oui bv percolation) Iroir
    fluff to pose a  poicnu.il  threat
    to human hcalih and the environ
    ment:

-------
 •  To examine the relationship be-
   tween shredder input  materials
   and levels or PCBs.  lead,  and
   cadmium  in   the  resulting  fluff
   output; and

 •  To develop and lest procedures
   for field  sampling, sample prep-
   aration, and   laboratory analysis
   which yield   more  precise  and
   accurate  measurement  of PCB.
   lead,   and cadmium  levels  in
   shredder output materials.
PROJECT

METHODS ANI) DESIGN

Site Selection and Description

Based  on  statistical and cost con-
siderations. EPA decided thai seven
shredder  sites  from  geographically
diverse  regions  of the  continental
United States would he  included  in
the Fluff Pilot Study.  Because of the
lime and expense required to relocate
a sampling crew in the event that a
shredder operator would not or could
not participate in the  program  (e.g.
due to breakdown), it was essential
that EPA  prearrange conveniently
located alternate shredder sites prior
to the commencement of sampling.

To implement random procedures for
site selection, to the extent possible.
and to have substitute sites readily
available. EPA began the process  of
selecting  the  seven  sites  to  be
included  in the pilot  study  by first
identifying clusters of  shredder sites
throughout the country.  Each of the
seven geographic clusters chosen for
the pilot study consisted of three  or
more sites and all sites,  within  each
cluster, were within about 100 miles
of one another.  Each cluster of sites
was located in a separate  EPA region
(there arc ten EPA regions).  From
within each geographic cluster, one
primary and two alternate sues were
randomly selected.

EPA  sent  advance  letters to  the
owner/operators of selected shredder
sites,  asking for their cooperation
with the Pilot Study, promising them
anonymity  if they participated.   In
addition, the metal  recycling trade
association, the  Institute  of Scrap
Recycling Industries (ISRI). provided
the  sampling  teams  with  letters
endorsing  the  study  and  soliciting
cooperation from its members.  (All
shredder  sites   visited were ISRI
members.)  The field sampling team
gave  these letters  to site owners/
operators at the beginning of sam-
pling visits.

Four of the seven primary sites parti-
cipated in the study, while three sites
were unable or unwilling to take part
and were replaced with alternate sites
from  the same  geographic cluster as
the primary site. While the site se-
lection process imposed some restric-
tion  on the  random  selection  of
shredder sites. EPA has no indication
that bias was introduced by this sam-
pling plan.

Shredder Operation

Shredder operations invariably have
several   important   features   in
common  (see   Figures  1  and  2).
Automobiles, appliances, and other
objects  are  fed to  a hammermill
which  shreds  them  into  Fist-sized
pieces.    Powerful   magnets   and
conveyor  belts  then  separate  the
ferrous (iron-containing) metals from
non-ferrous components. Next, using
either air cyriimc or water 'flotation* •
separation, metallic components are •
segregated  from the.  generally less
dense, (lull.  In this way. all shredder  ,
output is divided into piles of ferrous
metal, non-ferrous metal, and fluff.
EPA developed standard procedures
for collecting specified quantities of
shredder output  (fluff*  ferrous,  and
non-ferrous metal, etc.) before  the
start of sampling.

Input Kuns

One of the  objectives  of the pilot
study was to  investigate the relation-
ships between the  input  materials
being shredded  and  the  concentra-
tions of PCBs.  lead, and cadmium
measured  in  the  shredder output
streams. In order to investigate this
relationship,  shredder   operators
segregated  their input materials  into
three groups and made separate 'runs'
of the shredder, by material upe. lor
the sampling team.  Each  "run"  con-
sisted of the shredding of a predeter-
mined quantity of scrap material from
one of three categories of input.

The three categories of input material
were:

•  Automobiles, including trucks and
    vans;

•  White  Gixxls.  which  included
    refrigerators,  washing  machines.
    and other  similar appliances; and

 •  Mixed  Inputs, which  included  a
    variety of mixed  scrap materials.
    such as  those which come  Irom
    demolition sites  and  may  have
    contained parts  or   all of  some
    automobiles or white goods.

 What consumed a 'run* depended on
 the category ol input maienul  being
 processed.  For example, the shred-
 ding of two cars was delmeil as one
 "run"  for automobiles,  whereas  eicht
 home appliances equaled  .1 "run*  lor
 white  goods.   One  5-i!.illi'n bucket
 (sample) ol Hull" »;is  norm.illv col-
 lected  alter  each  "run"  usmu the
 sample collection pmiiH.nl developed
 for the studv (described N.l<«* i

-------
                                      •r •»«•»• 2
                        oer SM«cooiif6 srsTcn
                        Figure 2
   Schematic illustration of the Shredding Process
                                       •ir eytlor*
«tMl

-------
Sample Collection

Because of the heterogeneous nature
of ihc (luff ouipul stream, consider-
able effort was devoted in this study
lit the development  of methods  for
sampling fluff. Standard procedures
Tor collecting specified  quantities of
fluff  and   other  shredder  output
(ferrous or non-ferrous metal, etc)
were carefully developed prior to the
commencement of sampling.   For
fresh  duff sampling,  a   front-end
loader caught the fluff ouipul stream
as  it  tumbled off  the end of a
conveyor belt  or dropped from an air
cyclone funnel. The sampling team
spread this fluff over a tarp to create
a r deep 9* x 9' square pile. This pile
was subdivided into a nine square grid
of 3' by .V squares.   A five gallon
sample of  the fluff was collected by
compositing a roughly equal poriion
of fluff from near the center of each
of the nine squares.

F    developed other sampling proce-
dures  for piles of stored fluff,  fluff
that spilled off conveyor bells (spill-
over), and soil from the  vicinity of
fluff  piles.    Descriptions  of  all
sampling protocols are detailed in the
full report.

The following types of  samples were
collected at each shredder site:

•  Fresh fluff One sample of "fresh"
   fluff, fluff as it was produced by
 •"" the shredder, was collected from
   each run  of the shredder.   De-
   pending on what type(s) of input
   material  were processed,  there
   were generally between eight and
   twelve runs at each site.

•  Ferrous metals Two ferrous meiai
   output   samples  were  collected
   from each site, one Irom each of
   two input runs.
•  Non-ferrous  metals   Two non-
   ferrous metal samples were col-
   lected  Irom each site, one from
   each of two input runs.

•  Spillover Ruff which fell or blew
   off conveyor belts during process-
   ing  and  accumulated on  the
   ground around shredding machin-
   ery and conveyor belts was sam-
   pled.  Two samples of this spill-
   over  fluff were  collected from
   each site.

•  Stored Fluff  Five of the seven
   shredder sites visited had piles of
   stored fluff that had accumulated
   during normal operation prior to
   the arrival of ihe sampling team.
   At each of these sites, the sam-
   pling team collected four samples
   of stored fluff.

•  Soil The sampling team collected
   four soil samples from each site
   from  locations where fluff typi-
   cally accumulated. These samples
   were used to investigate the po-
   tential for migration of contami-
   nants  from fluff to soil.   These
   samples were also taken from be-
   neath stored fluff piles,  if  they
   were present.

Sample Analyses

Sample Preparation

 Before analysis,  each  five   gallon
 (.sample)  bucket of fluff was  divided
 into approximately eight to ten 450
 to  5(K)  gram  "representative  sub-
samples"   fur   chemical analysis.
These  subsamptes  were carefully
 constructed   such  that   every
 suhsamplc  contained all the basic
 physical components of  fluff (glass.
 foam, plastics,  fabrics, dirt,  etc) in
 proportions nearly identical to those
 found in the original sample.  The
 goal was  to create suhsamplcs which.
 in their  nhvsical composition,  were
  very similar to  the other suhsamplcs
  (from the same bucket) and to the
  original sample.   The  actual steps
  involved  in the  creation  of  the
  "representative  suhsamplcs*   arc
  described in the full report.

  Each 450 to SOU gram subsamplc was
  placed  into a  I-gallon  large mouth
  glass jar for storage. Depending upon
  the  quantity  of sample  material
  required  for  chemical  extraction/
  analysis, the suhsamplcs were split
  further, sieved  and/or milled.  Addi-
  tional   details   describing   sample
  preparation for each type of chemical
  extraction/analysis can be found in the
  full report.

  Development   of  PCB  (extraction
  Procedures

  Solvent Extraction

  Existing laboratory procedures for the
  preparation and  chemical analysis ol
  fluff samples for PCBs were judged to
  be deficient.  The conventional pro-
  cedure for measuring PCB concentra-
  tions prescribes that a relatively small
  quantity of material (ofien  20 grams
  or less) be subjected to solvent ex-
  traction before  instrument analysis.
  Using  such a small  quantity of  a
   heterogeneous material (such us fluff)
   has  historically  resulted  in  high
   measurement   variability  between
   subsamples from the same sample, as
   reported by many state and indepen-
   dent laboratories.  In practical  terms.
   this means that the actual estimate of
   the PCB concentration for any sample
   depends  to a great  extent on the
   specific  aliquol  of  flulf  used  for
   extraction  and  analysis.   Diilerent
   portions of fluff from  the same sam-
   ple  often  produce   \cr>   dillerent
   results which  make overall estimates
1   for each sample highly variable and
   potentially inaccurate.

-------
 Two methods were developed in  the
 pilot program to reduce this source of
 sampling error: (I) the technique for
 creating "Representative Subsamplcs"
 from the initial 5-gallon buckets, and
 (2)  the  quantity  cif  the  suhsample
 material subjected to chemical extrac-
 tion/analysis was greatly increased.

 Two  innovative   procedures   Tor
 increasing  the quantity of material
 subjected to hcxane/acctonc extraction
 were developed, tested, and compared
 for the Pilot Study. These techniques
 were (I) a tumbler (slurry) extraction
 using an agitation apparatus, and  (2)
 a large-volume Soxhlet (500 cc Soxh-
 ict) capable of extracting PCBs from
 up to 100 grams of fluff.

 A systematic comparison of measure-
 ments from matched subsamples ana-
 lyzed by the two procedures was con-
 ducted.   On the basis  of this  com-
 parbon.  the tumbler  (slurry) proce-
 dure was selected on  the basis  of its
 overall superior' }nro  he used as the
 standard extraction method  for the
 remainder of the PCB analyses.  This
 new technique allowed  extraction of
 fluff samples weighing  between 450
 to 500 grams, as opposed to 20  gram
 samples  used  in  the  conventional
 soxhlet. or the  100 gram samples used
 with the large  Soxhlet.

 Water Extraction (PCB  teachability)

 In order  to evaluate the teachability
 ofl'CBs from fluff using water as the
 solvent,  two  additional  extraction
 techniques  were  developed.   One
 technique for room temperature water
 used a slurry extraction apparatus and
 the other technique for "hot" water
extraction used a  Soxhlet extractor.
The  fluff samples  used in  the  "hot"
water extraction were milled to 9.5
mm.  A portion of the non-millable
 fraction  of the Huff was included in
each sample, the quantity added being
based upon  the  appropriate weight
fraction   of   non-millable  versus
millable portions (from the original
sample).

For the room temperature extraction
(22°C). 80 grams of Huff (particle size
£ 9.5  mm) was placed with 2 liters
of high purity water into the slurry
extraction   apparatus   (described
previously) and tumbled  for eight
continuous days. For the hot water
extraction (65°C). a similar 80 gram
sample was placed in a large Soxhlet
extractor and extracted over a period
of eight days with high purity water
at a temperature  of 65°C  Alter
eight days, the extract from each of
the two procedures was filtered  and
analyzed  for PCBs.

Chemical Analyses

All chemical analyses were based on
EPA methodology.  The  inorganic
analyses  were done using Methods
213.1 or  7131 for  Cd and 239.1 or
7421 for  Ph.

Unless   otherwise   stated.   PCB
analyses  were  performed  using  a
modified   gas   chromatography/
electron capture detector (GC/ECD)
EPA Method 8080.

The modified analytical method as
well as all other analytical methods
used in the Pilot Study can be found
in the  Appendix section of the full
report.    Summaries  are  provided
here.

•  Total PCB  Concentration   The
   total concentration of PCBs in
   each of the subsamptes analyzed
   was determined  by  extraction,
   using a hcxancfacctcmc solution..
   then analysis using a gas chroma- •
   tography/clcctron   capture
   detection   (GC/ECD)   method.
   The sample  particle si7.es  were
   not  reduced  for  any  of  the
   tumbler   (slurry)   extractions.
however  the samples  undergoing
Soxhlct extraction were milled to
< 9.5 mm.

PCB Concentrations in Individual
Ruff Components   Flulf  subsam-
ples were divided into their physi-
cal  components  (glass,  plastics.
fabrics, etc.). Each component was
individually analyxcd for total PCB
concentration  by EPA's National
Enforcement Investigation Center
(NEIC) in Denver. Colorado. The
analytical method  employed  was
EPA Method  600 The  Determi-
nation   of   Polychlorinatcd
Biphenyls in Oil. Soil  and Surface
Samples".   This   analysis  was
conducted to  determine  whether
PCB  contamination  was  more
closely  associated  with  specific
components of fluff.

PCB Leachahilitv EPA measured
the extent to  which  PCBs leach
from fluff, using water as a solvent.
to estimate how likely they arc 10
be released from shredder wastes
into the  environment. To repre-
sent a "worst case"  scenario. EPA
performed a hot water extraction
of size-reduced fluff in a Soxhlet
extractor.   Samples  were  also
subjected to a nxrni-icmpcrature
water  extraction,   to  represent
something closer to a "real world*
scenario.   These   samples were
extracted using a slurry extraction
apparatus. The extracts from both
techniques were analyzed using the
GC/ECD method.

Total Lead and Cadmium Concen-
trations  The  total concentrations
of lead and cadmium were deter-
mined by digesting the sample  in
acid and analyzing the dieesuie by
Flame Atomic Adsorption Spec-
troscopy (FLAA).  Samples  \\ith
lead or cadmium concentrations so
low that they could nm be de-
tected bv the FLAA nuMhml were

-------
    analyzed  by  Graphite  Furnace
    Atomic Adsorption Spectroscnpy
    (GFAA). The sample particle size
    used Tor the digestion and analysis
    was < 9.5 mm.

•   Lead  and Cadmium teachability
    EPA Method 1310 Extraction Pro-
    ccdurc  Toxicily  Test  (EPTOX)
    was used to measure how readily
    lead and cadmium will leach from
    fluff  to estimate the  potential
    release  of these substances from
    shredder   wastes   into   the
    environment.     The  EPTOX
    extracts  were analyzed for lead
    and cadmium concentrations using
    the FLAA und GFAA methods.
    The  sample  size  requirements
    were  that the particle  size  be
    < 9.5 mm and/or have a  surface
    area to weight ratio of > 3.1  cm
    squared per gram.

    •  PCBs in Ferrous and Non-fer-
    rous  Metal  Metal samples were
    analyzed for total PCB cor •   .ra-
    tion  by extraction  followed  by
    analysis with the  GC/ECD meth-
    od. Subsamplcs of metal samples
    were  also analyzed  for  quality
    assurance purposes and archived.

    •   Analysis of Soil  Sample  Soil
    samples  were analyzed for total
    PCB concentration. PCB composi-
    tion, and total lead and cadmium
    concentrations.   Subsamples of
    soil samples  were  analyzed in
    accordance   with   the   quality
    assurance  program  and   others
    were archived.

Quality Assurance

The Quality Assurance  Project Plan
(QAPjP) presents the features of the
quality assurance design for the pilot
study.  The QAPjP was  developed in
three phases: Phase l-Field Sampling.
Phase Il-Chcmical Analysis and Phase
Ill-Statistical Data  Processing  and
Analysis.      A   more   detailed
description  of  what  each  phase
included follows.

Field  Sampling     As  previously
described,   a   grid-type  sampling
scheme was employed to increase the
likclihixKl of obtaining representative
samples. A standard operating  pro-
cedure (SOP) was also developed for
the sampling of piles of stored fluff.
The  QAPjP  also  stated  specific
details  for  the  tracking  of  field
samples  including the  use of log
books for sampling details and diain-
of-cus tody sheets for sample tracking.
Duplicate .samples  were  taken  in
order   to   assess  field  sampling
variability.

Chemical   Analysis     Before  the
laboratory analysis of any samples.
chemical  analysis  methods  were
carefully reviewed Tor their adequacy
in meeting the project's data quality
objectives (DQOs).  The DQO for
accuracy was equal to or greater than
60% recovery  for spiked samples.
The DQO  for  precision was + or -
50',7  relative standard deviation for
replicate samples. The measurement
of  the experimental accuracy and
precision was  done through labora-
tory quality control  samples which
included method  blanks, replicate
samples. Held  duplicates, and matrix
spike samples.  All standards used for
spiking were traceable to their manu-
facturing source.  The  data quality
objectives were met for  all hut a few
samples. Spins were obtained  from
UW ol the samples, the .splits being
sent  to  an external laboratory for
analysis (F.PA  Environmental Moni-
toring  Svstems  Laboratory -  Las
Vegasj (EMSL-LV).

Statistical   Data  Processing   und
Analysis A treat deal of effort was
expended in order to assure that the
 data generated by the laboratory vtcrc
 correctly transferred to the contractor
 conducting the statistical analysis of
 the data.  The correctness of data
 values generated by the laboratory was
 cross  checked  by  the  contractor
 conducting  the  statistical  analyses
 once they were keyed into  a  matrix
 file.

 One  other  important aspect  of the
 project's  quality assurance  program
 was the use of audits. Three types of
 audits were conducted  during the
 project:   system  audits  to  assure
 standard  operating procedures were
 being followed,  performance  audits
 using performance audit samples so
 that the laboratory could demonstrate
 its ability to accurately analy/c for the
 analytc(s) of interest, and data audits
 which reviewed portions of the data
  for error.
  RESULTS

  The results of the Pilot Study provide
  a preliminary evaluation ol the waste
  characteristics of fluff and valuable
  information for the design of  future
  studies.  Caution must  be  exercised
  when generalizing from these findings.
  As  noted  earlier,  the  Pilot  Study
  results are  based on limited data and
  do not necessarily represent the metal
  shredding industry as a whole.

  Total  I'CH Concentrations in Huff.
  Metuls and Soil

  PCBs  were detected  in all -hreddcr
  output materials analv/cd  Over w;
  of PCBs arc estimated  to end up in
  the  fluff waste stream.  f'CR lonecn-
  trations  in  both  the  lerrou*  and
  nonlerrous metals  were  ver\
  (means of 0.21  ppm  tor the
1  metal  and 0.90  ppm lor  iiu  non-
  ferrous metal). The nu..m IH B u«n-
  ccntralion for all IrcNh Hull w.i- 4*

-------
ppm.  Using a bootstrap resampling
procedure,  an  approximate  95%
confidence interval for this mean was
calculated to he 22 ppm to  120 ppm.
Table  I  displays the average PCB
concentration in  parts  pur million
along with the standard deviation Tor
each category or sampled  material.
Also presented are median, minimum.
and maximum concentrations and the
number or samples  and sites upon
which these statistics are based.

Table  I shows that  fresh fluff from
mixed inputs had higher PCB concen-
trations than fresh fluff resulting from
white g(xxJs or automobiles, and this
difference is statistically  significant

PCB  concentrations in  the non-
ferrous waste metals  were roughly 50
times lower than those in Ruff. PCB
concentrations   in   ferrous  waste
streams were approximately 2UO limes
lower than those in fluff.  No samples
from  the  ferrous  and  non-ferrous
metal streams  had  PCB concentra-
tions anywhere approaching the EPA
TSCA disposal threshold of 50 ppm.
(The highest PCB concentration for
ferrous metal was 0.42 ppm and for
non-ferrous metal. 2.6 ppm.)   It Is
notable that  ferrous  metal output is
very "clean.*   In contrast to non-
ferrous metal which  cannot be sepa-
rated primarily with magnets, ferrous
metal output contains practically no
fluff. The non-ferrous metal typically
contains much higher proportions of
ndh-metallic waste (fluff), which may
explain the somewhat  higher PCB
levels in the non-ferrous output

PCB concentrations  in soils were in
the same range as   those in stored
fluff.   Some  soil samples had PCB
concentrations exceeding the 50 ppm
EPA  disposal  threshold,  but. soil
samples had PCB concentrations, on
the  average,  slightly   below   the
disposal threshold. How PCBs got
into the soil is unclear.  The PCB
concentrations  found  in  these soil
samples may reflect the migration of
PCBs  from  fluff which  regularly
accumulated nearby, or'soil* samples
may contain a substantial amount of
fluff material  which  had  become
mixed with the soil over time.  The
soil  sampling  protocol  prescribed
that if the demarkation between soil
and  fluff residues was not distinct.
"sort" samples should  contain   a
minimum of 50% soil.   From  the
accounts of sampling crew members.
the distinction between soil and fluff
sometimes was unclear. Soil samples
collected  in accordance  with  the
sampling protocol may often have
contained some portion of fluff.

PCB Composition In Huff

The fluff samples were analyzed for
the  specific concentrations of PCB
Aroclors 1242. 1254.  and  1260. to
explore the possibility  that sources of
PCI?  ,/ntaminaiion could be identi-
fied by  PCB  composition.  This
approach was  considered feasible
because different PCB Aroclors were
developed  for  specific uses.   This
analysis  for specific  PCB Aroclors
revealed that Aroclor 1242 was domi-
nant in almost every sample, making
up more than  half of the PCBs  in
each sample category.   While  this
part of the pilot study yielded some
general information on  the propor-
tions of the three most commonly
used Aroclor  mixtures,  it produced
no definite information on sources of
PCB utciaminauon.

PCR Concentrations  In Individual
Huff Components

Suhsamples  from four  fresh fluff
samples   (two   resulting  from
automobile shredding, one resulting
from the shredding of white gtxids.
and one resulting from the shredding
  of mixed inputs) were divided into the
  following components*

  •  Metals, wire and glass;

  •  Soft plastics, foams, stilt rubber.
     and vinyl;

  •  Fabrics, paper and wood:

  •  Hard materials, hard plastics, and
     hard rubber;

  •  Rnc   materials  too  small  to
     classify, dirt, and dust: and

  •  Other, not classifiable, materials.

  All  components,  except   for  the
  "Other, not classifiable" materials were
  analyzed  separately  for total  PCB
  content.  Table 2 shows the percent
  by weight, of each component in the
  four  samples,   and  the   PCB
  concentrations   found  in   each
  component

  Total  PCB concentrations  in  Iresh
  fluff samples, as  well as the rclamc
  concentrations of the  PCBs in dif-
  ferent components, varied with input
  material.  They also varied between
  the  two samples produced Irom the
  same input material (i.e. automobile).
  In one sample from automobile in-
  puts, the highest PCB  concentration
  occurred  in materials  in  the  "Soft
  plastics, foams, soft rubber and vinyl"
  category.   In  the  oihcr  (matching)
  automobile sample,  the hichcst  PCB
  concentrations were in materials  in
  the  "Fine  materials  too  small  tit
  classify, dirt, and dusi" category
  The highest  PCB concentrations  in
   fluff from white goods were also inund
   in materials in the "Fine materials too
  small  to  clussif).  din.  and   dust"
  category, while Hull Irom mixed inputs
1  showed  the  highest PC'B  tonicmra-
   tions  in the "metals, wire .uid  "l.ivs"
   category.

-------
Table  1.   Summary  of Total PCS  Concentrations  (ppn)  by Sample Type
Sample Input
Type Type
Fresh fluff Auto
Fresh .fluff White Goods
Fresh fluff Mixed input
Stored fluff
Spillover
Ferrous
Non-ferrous
Soil
Mean
32
80
180
68
28
0.2
1
44
Standard
Deviation
43
190
170
43
25
0.11
1.1
36
Median
13
21
88
52
28
0.21
0.9
32
Minimum
1.7
0.67
12
16
4
0.1
0.13
0.13
Number
of
Maximum Samples
210
760
500
ISO
65
0.42
2.6
100
28
IS
9
10
5
8
5
8
Number
of
Sites
7
5
3
5
5
6
3
4
Table 2.  Total  PCB Concentrations  (ppm) in  Five Fluff Components
                                                          Input Material
Automobile
Samole 1


Concentration
Component
Metals, tare,
and glass
Soft plastics.
foams, soft
rubber, vinyl
Fabrics, paper,
and wood
Hard materials.
hard plastics,
hard rubber
Fines too small
to classify,
dirt, dust
Other, not
analyzed
Total sample
weight (gm)
% of
Total
Sample
(by weight)
11%
17%
17%

9%

40%
6%
1090
PCB
Concen-
tration
(PP«)
13
66
37

11

43


Automobile
Sample 2
% Of PCB
Total Concen-
Sample tration
(by weight) (ppm)
2% 9.9
14% 7
28% 12

2% 24

38% 29
16%
1260
White Goods
% of
Total
Sdfflplo
(by weight)
3%
8%
9%

10%

65%
5%
859
PCS
Concen-
tration
(PP»>
0.6
35
24

5.5

£2

*
Mixed In
% of
Total
Sample
(by weight)
2%
17%
26%

5%

45%
5%
1080
DUtS


PCB
(ppm)
390
260
63

46

140



-------
                                                     10
 TCB Kxtractnbilily From FlufT

 Suhsamplcs from seven different fresh
 Duff samples found to have high total
 PCB concentrations  were extracted.
 using hoi water as solvent, to estimate
 how readily PCBs  migrate from  the
 fluff waste stream to the surrounding
 environment.  The hoi water (65°C)
 extraction  provides   a  theoretical
 'worst  case"   estimate  of  PCB
 cxtraciahility. An average of 0.0073%
 of  the  PCBs  in  the  samples was
 extracted   using   the  hot   water
 extraction described earlier.

 Using  other suhsamples  from  lite
 same seven high PCB fluff samples,
 an average of 0.0050% of the PCBs in
 the samples was extracted  using  a
 room  temperature   (22°C)  water
 extraction.  These results suggest that
 PCBs are less likely to leach (dissolve
 out hy percolation) from  fluff than
 from a wide range of soils.

Total   Lead  And  Cadmium
 Concentrations In  Huff and  Soil

Total lead concentrations   in most
 flutf samples ranged  trom   1.0UO to
 10.000 ppm.   Total cadmium con-
 centrations in most fluff samples were
 substantially lower, falling between 10
and I(X) ppm.

Table 3 presents total lead  concen-
 trations for each type of fluff and soil
sample" analy/ed.   The Table  shows
 the mean, standard  deviation, median.
 minimum,   and   maximum   lead
concentration  values,  as well as  the
 numbers' of samples and  sites  on
which the  results were  based.  The
 total lead concentration data  for fresh
 fluff from automobiles,  white  goods
and mixed inputs were combined to
 produce an average for all fresh fluff
which  was then   compared with
spillover and stored fluff. The mean
 total lead concentration in all types of
fresh fluff (combined) is 2.800 ppm.
The  approximate 95%  (bootstrap)
confidence interval for this mean is
1.800 ppm to 4.100 ppm.  Total lead
concentrations in spillover fluff are
greater than in stored fluff, which in
turn are greater than in  all types of
fresh fluff combined.  These differ-
ences  arc  statistically  significant
Lead concentrations in soil are statis-
tically significantly lower than in all
types of fluff combined.

Table 4 presents the total cadmium
concentrations in each type of fluff
and soil.  It gives the  mean, standard
deviation, median, minimum,   and
maximum lead cunccairatioa values.
as well as the numbers of samples
and sites on which the results were
based.

The  data  for fresh fluff from auto-
mobiles, white goods, and mixed in-
puts were combined to compare total
cadmium  concentrations  in  fretfs.
spillover, and stored fluff. The tnw*.i
cadmium concentration for all types
of fresh fluff combined  is 47 ppm.
The  approximate  95%  (bootstrap)
confidence interval for this mean  is
31  ppm  to 65  ppm.   Differences
between cadmium concentrations in
the different types of fluff are not
statistically   significant.     Total
cadmium  concentrations in soil arc
statistically significantly lower than in
all types of fluff combined.

IjtaA and  Cadmium teachability
From Ruff

Of  considerably  greater  interest.
environmentally, than total lead and
cadmium   concentrations  is   how
readily lead and cadmium leach from
fluff to contaminate the environment.
Lead and cadmium concentrations in
Icachate  were  measured  using the
EPTOX procedure lor samples from
all categories of fluff.
The   EPTOX  procedure  was  ihe
standard  EPA  mcituxl  for deter-
mining leachability at the time these
lead   and  cadmium  analvscs were
conducted.   In March  Two. EPA
replaced the EPTOX wiih the TCLP
(Toxicity   Characteristic   Leaching
Procedure) as the standard method for
determining   leachuhility.     EPA
comparison analyses have shown little
difference between the results of the
EPTOX and TCLP methods.

Table 5 summarizes  the results for
lead  in  Icachate from  the EPTOX
extraction.  The Table presents the
mean, standard deviation, median.
minimum and  maximum concentra-
tion of lead in the EPTOX extract for
fresh fluff from automobiles, white
goods and mixed input: as well as for
stored and spillover fluff.  Table ft also
presents the  number of samples and
sites represented in the calculation of
each statistic.

Lead concentration  values  in  the
EPTOX extract ranged  from 0.8 to
220  ppm. with  an average ol uhove 6
ppm for every type of fluff.  While the
highest EPTOX lead concentrations
were associated with stored  fluff, fresh
fluff from mixed input and *pilli«vcr
fluff, the differences between avenue
concentrations  as  presented  in the
Table are not  statistically MemUcani.
The mean EPTOX lead concentration
 for  all types of fresh fluff combined is
7.2  ppm.   The  approximate  l>5r-
 (boot-strap)  confidence intcnul lor
 this mean is 4.8 ppm to  13 ppm.

 Table 6 summarizes the results ol the
 EPTOX  cadmium extraction.   Thi-»
 Table gives  the mean,  approximate
 95*? confidence interval lor the mean.
 standard deviation, median, minimum.
 and maximum concentrations lor the
 different types ol  Hull: a*  »ell .1* the
 numbers  of   sample*,   .md   Mies
 represented by these NI.HIMH>

-------
                                                                II
Table  3.   Summary of Total Lead Concentrations  (ppm) by Sample Type
Sample
Type
Fresh fluff
Fresh fluff
Fresh fluff
Stored fluff
Spillover
soil
Input
Type
Auto
White Goods
Mixed input



Table 4. Sumnary of Total
Sample
Type
Fresh fluff
Fresh fluff
Fresh fluff
Stored fluff
Spillover
Soil
Input
Type
Auto
White Goods
Mixed input



Table 5. Summary of EPTDX
Sample
Type
Fresh fluff
Fresh, fluff
Fresh fluff
Stored fluff
Spillover
Input
Type
Auto
White Goods
Mixed input


Mean
2.700
3.100
4,600
3.900
6.100
2.200
Standard
Deviation
2,200
3.200
3,500
3.500
5.600
3.900
•
Cadmium Concentrations
Mean
47
48
46
35
32
22
Standard
Deviation
36
19
14
13
11
24
Median
2.400
1.800
3,600
2,600
4,300
1.100
(ppn) by
Median
40
47
46
35
33
18
Minimum
570
1.300
1.100
1.300
2.800
8.1
Sample Type
Minimum I
14
23
29
16
18
10
Maximum
12.000
14.000
12.000
13.000
21,000
16,000

Maximum
200
87
70
59
59
100
of
Samples
28
IS
13
20
9
16

Number
of
Samples
28
IS
12
20
9
16
Of
Sites
7
5
3
4
5
5

Number
of
Sites
7
5
3
5
5
4
Lead Concentrations (ppn) by Sample Type
Mean
6.9
6.1
23
22
IB
Standard
Deviation
5.5
5.0
24
47
12
Median
5
3 2
U
9 5
20
Minimum
.8
1.6
1
1 6
1.7
Maximum
21
14
78
220

Number
of
Samples
28
IS
12
20
9
NunbBr
of
Sites
7
5
3
5
5

-------
                                                       12
Table 6.  Summary of EFTQK Cadmium Concentrations  (ppm) by Sample Type
Sample
Type
Fresh fluff
Fresh fluff
Fresh fluff
Stored fluff
Spillover
Input
Type
AutO
White Goods
Mixed input


Mean
0.81
1.3
1
0.73
0.45
Standard
Deviation
0.67
0.77
0.27
0.41
0.26
Median
0.7
1.3
1
0.61
0.3
Minimum
0.35
0.45
0.48
0.2
0.18
Maximum
4
1.3
1.4
2
0.81
Number
of
Samples
28
15
12
20
9
Number
of
Sices
7
s
3
5
S
 The mean EPTOX cadmium concen-
 tration  Tor  all types of  fresh fluff
 combined is O.H4 ppm. The approxi-
 mate 95%  (bootstrap)   confidence
 interval for this mean is 0.53 ppm to
 1.2 ppm.
 CONCLUSIONS

 In this  pilot study.  EPA's  research
 team:

 •  Determined that  PCBs were pre-
    sent in all sampled materials ai all
    seven pilot study sites and  that
    over  98% of  the PCBs  in all
    shredder output  were associated
    with  fluff: PCB concentrations in
    fluff ranged from 0.67 to 760 ppm.

 •  Determined that  in  the  "worst
    case* scenario  of teachability, a
 ^- Hot water extraction, only .0073%
    of the PCBs present leached from
    the  sample on average.   In a
    situation more  closely resembling
    'real  world*   conditions,  room
    temperature    water   extraction
    leached .0050% of the PCBs. on
    average, from  the fluff.  In  both
    cases, the observed Icachabiliry of
    PCBs from fluff  was lower  than
    usually  found in a wide  range of
    soils;
Could  not  conclude  thai  any
particular input material was the
source of  the  PCBs.  lead  and
cadmium found in shredder out-
puts.    Cross-contamination  of
samples  within sites may  have
masked the relationship between
input material and contamination
of resulting output materials.

For   example, if   PCBs  were
released  onto shredder surfaces
during  the  shredding of PCB-
containing items, fluff produced
for some time after the initial
release may haw been contami-
nated as it came into contact with
pans of  the shredder apparatus.
although no PCBs existed in the
input  material  associated with
this fluff:

Found  that  lead  and cadmium
leachatc concentrations in fluff, as
determined  by  the  EP TOX.
ranged Irom O.X to 220 ppm and
O.JX eo 4 ppm. respectively.

Developed and tested  field sam-
pling  and  sample  preparation
procedures  to  obtain represen- •
talive samples and subsamples of
fluff,  ferrous  and  non-ferrous  ,
metals  and soil  from  shredder
sites: and
Developed and tested  laboratory
protocols to analyze very  large
fluff samples (S(X) grams) for PCB
content.   This technique reduced
the sampling variability associated
with conventional PCB extraction
and analysis of fluff, and resulted
in more reliable estimates of PCB
concentration.

Determined that '><: limned size of
the sample pr«.;.iudcs using the
analytical results from  this  Pilot
Study to characterize the shredder
industry as a whole: also identified
the need to collect  and evaluate
additional analytical data  gene-
rated by State agencies and indus-
try sources subsequent to comple-
tion of the Pilot Study.  Care will
be taken to review sampling proce-
dures and analytical methods used
in collecting data.

Identified the need to obtain a bet-
ter understanding ol the economic
viability of the shredder industry
and  to   assess  ihc  economic
impacts,  if any.  rcMiliim: irom
various   approaches  lo  residual
waste management.

-------