United States
Environmental Protection
Agency •	
Enforcement And
Compliance Assurance
(2224A)
EPA 530-R-94-019
October 1994
Waste Analysis Guidance for
Facilities That  Burn
Hazardous Wastes
Draft
                  Notice
   This document is a preliminary draft. It has not been formally
  released by EPA and should not at this stage be construed to
  represent Agency policy. It is being circulated for comment on its
  technical accuracy and policy implications.

-------
     WASTE ANALYSIS GUIDANCE FOE
FACILITIES THAT BURN HAZARDOUS WASTES

-------
                            DISCLAIMER

The policies set out in this document are not final Agency action,
but are intended solely as  guidance.   They are not intended, nor
can they be relied upon, to create any rights enforceable by any
party in  litigation with the  United  States.    EPA  officials may
decide to follow the guidance provided in this manual,  or to act at
variance with the guidance, based on an analysis of specific site
circumstances.  The Agency also reserves the right to .change this
guidance at any time without public notice.

-------
                             CONTENTS

Section          -        -           ..         '  :          '   Page
1.0   INTRODUCTION  	 .................   1

     1.1  BACKGROUND	   1
     1,2  PURPOSE	 ........   1
     1.3  REGULATORY REQUIREMENTS ....... ;	   2
                                   /':       '     .;.-'•       .
2.0  WASTE ANALYSIS PLANS . . . . . . ... . . ..: . . . . ....   4

    1 2.1  SAMPLING METHODS FOR EACH FEED STREAM	.5
     2.2  METHODS OF PREPARATION AND ANALYSIS OF SAMPLES   .  .   5
     2.3  SAMPLING AND ANALYSIS STRATEGY  	   6
     2.4  SAMPLING LOCATION . . 	i. ......   6
     2.5  QUALITY ASSURANCE AND QUALITY CONTROL . . . . .  .  .7
                                                I       '     ,
3.0   SAMPLING AND ANALYSIS STRATEGIES  . . . ..;._.. . ....   7

     3.1  SAMPLING AND ANALYSIS BY BATCH  ...........   8
     3.2  QUALIFICATION OF A FEED STREAM  ...%......  10
     3.3  STATISTICAL ANALYSIS  ....... .i	  .  13

          3.3.1     Upper Tolerance Limits  ...•...'	16
          3.3.2     Statistical    Approach:   Compliance
                    Issues  ..................  18

4.0  INCOMPLETE DATA, OUTLIERS, AND DETECTION LIMITS  . .  .  .'  21

     4.1  OUTLYING DATA POINTS  . . . . . . . . i. . . ,. .  .  .21
     4.2  INCOMPLETE ANALYTICAL DATA	;	22
     4.3  DETECTION LIMITS  ..........:.......  22

5.0  MANAGEMENT OF RESIDUES	:.	23

     5.1  'DETERMINATION OF THE BEVILL EXEMPTION	  24
     5.2  FREQUENCY OF SAMPLING AND ANALYSIS  . ;.	25
     5.3  CONSIDERATIONS  . . 	 ...........  25

6.0   DOCUMENTATION TO DEMONSTRATE COMPLIANCE ........  27

Appendix            •                    "        '  •

A    Table of K Factors for Calculation of Tolerance Limits

-------
                       1.0    INTRODUCTION

1.1       BACKGROUND

On January  23,  1981,  the United  States  Environmental Protection
Agency (EPA),  pursuant to requirements of  the Resource Conservation
and Recovery Act of 1976  (RCRA), promulgated regulations governing
the combustion of hazardous waste (HW) in incinerators (HWI).  On
February  21,   1991,   EPA  promulgated  regulations governing  the
burning of  hazardous wastes  in  boilers  and  industrial  furnaces
(BIF).  The regulations are  intended to  protect human health and
the environment from  exposure to emissions from the combustion of
hazardous  wastes.     Regulations  governing  such activities  are
codified at 40 Code of Federal Regulations  (CFR)  Section (§)  264
Subpart O and,§265  Subpart O for permitted and interim status HWIs,
respectively, and at  40 CFR §266 Subpart H for BIFs.
                                                !
1.2       PURPOSE

The,purpose of this document is to provide guidance to facilities
and  EPA  Regional  and  state  personnel  regarding  .appropriate
approaches  to  sampling  and  analyzing  feed  streams to  ensure
compliance with EPA requirements for waste analysis for hazardous
waste  combustion   devices.     This   document  describes  three
alternatives  for  demonstrating   compliance.    The  alternatives
provide a uniform approach to documenting  compliance with limits on
constituent feed rates established during compliance testing or a
trial  burn.     The  three  alternatives   are  batch  analysis,
qualification of a feed stream, and statistical analysis.

The concepts presented in the following sections meet the intent of
the regulations and can be implemented in a way that is consistent
with daily operations of the facility.  Guidance also is provided
in  this  document  for analysis  of  residues  generated from  the
combustion  of hazardous wastes.   This  document does not  cover
specific methods of sampling and analysis  for units regulated under
RCRA and  does _not preclude  EPA  or  state personnel  from taking
enforcement actions related to waste analysis.
                 - DRAFTS Do Not Cite Or Quote -

-------
1.3    .    REGULATORY REQUIREMENTS1

.To ensure proper combustion of hazardous wastes,  the HWI and  BIF
regulations and individual operating permits  establish limits on
operating parameters for combustion units.  The limits ensure that
the maximum levels  of emissions of the constituents of concern from
the combustion units are low enough to ensure acceptable levels of
constituents in ambient air, as specified in EPA regulations or  the
facility's  permit   (levels protective  of  human  health  and  the
environment).  To demonstrate this, air dispersion modeling and/or
emissions testing  is  used  to  establish operating limits for  the
facility that ensure that a facility's emissions do not exceed  the
regulatory levels.   Emissions tests are conducted  during the trial
burn  for HWIs and  BIFs attempting to obtain  operating permits or
during the compliance test for  BIFs operating under interim status.

Some   examples  of   operating  parameters  contained  in   the  BIF
regulation and  some HWI  permits that  generally  are established
through  air dispersion modeling and/or  emissions testing  are:

      «    Continuous monitoring and recording of the flow rates  and
           composition  of  hazardous waste,  other  fuels,  and feed
           stocks for industrial furnaces to yield the feed rates of
           10 metals (mercury,  lead, cadmium, chromium,  barium,
           beryllium,  arsenic,   thallium,  silver,   and  antimony),
           chlorine and chlorides,  and ash

      »    Maximum and  minimum  temperature  limits  for the burning
           zone

      •    Maximum production rate (for  example,  steam)

      •    Continuous  emissions  monitoring  for  oxygen,   carbon
           monoxide, and hydrocarbons

      •    Appropriate  operating  parameters  for  air  pollution
           control equipment

Operating  permits   and   regulations   require   that  combustion
facilities maintain,  monitor,  and record established operating
parameters while burning HW to  document compliance (for example, 40
CFR  §266.102(e)(10),  §266,103(k), §264/265.73).    Of the various
operating parameters,  feed rate  limits for metals,  chlorine  and
chlorides, and  ash are  key elements  for  which  facilities must
maintain records to demonstrate compliance.  For example, the  BIF
      This guidance document discusses requirements that arc generally included in tbe regulations and/or individual permits for combustion facilities.
Because tbese requirements may vary by type of status of * combustion unit, it is necessary to consult tbe regulations and/or permit for requirements specific
to a particular uak,
                  - DRSJTs Do Not Cite or Quote -

-------
• .;   x  .        '        '     .--•.--. 3'          •  .  ;t, _'-...;

regulations , specify that the  feed  rate limits for metals,  total
chlorine and chloride, and asti are to be "established and monitored
by knowing the concentration of the substance in  each feed  stream
and the  flow rate of  the  feed stream"  (for example,  4.0  CFR
§266.103(c)(4)(iv)(D)).  The flow rate must be monitored under the
continuous monitoring requirements specified in  a permit  or the
regulations  (for example,  40 CFR §266.103(c)(4)(iv)(D)  of the BIF
regulations or 40 CFR §264.347 of the incinerator regulations),  In
other  words,  the feed rate for each metal,  chlorine  and chloride,
and ash in the total feed streams must be established and monitored
continuously   (for example,  40 CFR §266.103(c.) (3)) .   The  term
"total feed streams" includes anything that is fed to the unit (for
example,  liquid  and solid  hazardous wastes, raw materials,  fuels,
nonhazarvdous wastes, and off-gas streams from production processes,
see 56FR7176).                 ,

In addition,  compliance with all the  other limits  on  operating
parameters (for  example, operating limits on air pollution control
devices   and   temperature)   may  not   be   adequate  to  establish
compliance with  emission limits, if the feed streams fed into the
combustion unit  are not  characterized  and  monitored  properly.  It
is easy to conclude, then,  that analysis of constituents of concern
in the  feed  streams  is   the  starting  point  in  demonstrating
compliance with many requirements governing  combustion of hazardous
waste.   However,  the regulations  do not  require that  specific
methods  be used  in sampling and  monitoring the concentrations of
each substance.   Various interpretations therefore have  been put
forth  of what constitutes compliance with  requirements  for waste
analysis at such facilities.                             «

Since  waste analysis is the basis for knowing the concentrations of
constituents   and  demonstrating  compliance  with   requirements
governing feed rates, a waste analysis plan describing the specific
procedures that will be followed to obtain accurate, representative
results  is necessary to support the  analysis*  EPA's waste analysis
regulations at 40 CFR §264.13,  265.13 state that before a Waste is
treated,  stored  or disposed, the facility  must obtain  a  detailed
analysis of the  waste, which,  at a  minimum, "must contain all the
information which must be known  to treat store or dispose  of the
•waste" in compliance  with  relevant standards.  In addition,  the
regulations  governing permitted  incinerators and BIFs,  set forth
under  40 CFR  §264.341 and  §266.102, require detailed analysis for
concentrations of  constituents as  necessary  to. ensure that the
waste  feed is "within  the physical and chemical composition  limits
specified" in the permit.    Because of the  uncertainty associated
with most production processes, the Agency has found that process
knowledge alone does  not  generally  give  the type of precise
information  necessary to establish and monitor feed  rate limits.
Therefore, any facility choosing to  rely on process knowledge alone
will be  intensely scrutinized and runs the risk that the Agency's

                  - DRAFTS  Do Hot Cite  Or Quote -

-------
own sampling will  demonstrate that the facility's waste analysis
method  did not produce  the information required  to demonstrate
knowledge  of the constituent  concentrations in the feed streams.
Further,   the   requirements   for   waste  analysis  at   40  CFR
§264.13(a) (3) and (b)(4)  and at §265.13(a)(3) and (b)(4) state that
analysis must be repeated at a frequency  sufficient to ensure that
it is  accurate and up-to-date.   The following  sections  of this
document provide facilities guidance on demonstrating compliance
with the  waste analysis  requirements for  monitoring  feed rates
through:                       ..."

     «    Development of a waste analysis plan
               \                               • .                  *
     •    Selection of the appropriate frequency for sampling and
          analysis

     •    Quality  assurance  and  quality  control  of  data  from
          analysis

     •    Documentation that demonstrates compliance

Included  is a  discussion  of  the  requirements  for analysis  of
residues generated from  combustion of hazardous wastes.   Because
some of those residues have the potential to be hazardous to human
health  and  the environment, proper sampling and analysis is also
necessary to make this determination.


                   2.0   WASTE ANALYSIS  PLUMS

Waste  analysis is the  backbone  of the  RCRA  program  and  the
hazardous waste combustion regulations discussed above.  Therefore,
every facility that treats, stores, or disposes of hazardous wastes
is required to develop a waste analysis plan (WAP)  (40 CFR §264.13
and §265.13).  Elements of the WAP that are particularly applicable
to combustion facilities are discussed here.   General contents and
development  of a  WAP  will  not  be  covered;  for such  general
information  the  reader  should refer to  the EPA  guidance Waste
Analysis  At Facilities  That  Generate,  Treat,  and  Dispose  of
Hazardous  Waste,   (OSW1R [Office  of Solid  Waste and  Emergency
Response] # 9938.4-03,  April 1994).  The document is available from
the National Technical Information Services (NTIS), publication #
PB94-963-603.

In this document,  the term waste analysis plan refers to a written
document,  prepared by each regulated facility,  that  defines the
sampling and analysis protocols and  frequency  through  which the
facility determines the concentration of  regulated  constituents in
each feed stream at all times.  The WAP is not  limited to hazardous
waste  feed  streams,   but  includes  all  feed  streams,  such  as

                 - DRAFT: Do Not Cite Or Quote -

-------
nonhazardous wastes,  fossil fuels, and raw materials when they (the
nonhazardous feed streams) are cofired with hazardous waste.  The
waste  analysis  plan  should  be amended .with  the  appropriate
information  when  new units  are  added,  processes  change,  new
regulations are  promulgated,  or permit  modifications  are issued
that affect analysis of feed  streams before treatment, storage, or
disposal of those feed streams.
                                        /       .    ,       . t
Some items that may be contained in the WAP are:  a description of
treatment activities conducted at the.facility; identification and
classification of HW generated, treated or managed at the facility
and of their quantities; and descriptions of HW units and operating
procedures  (for example,  use  of  safety  equipment);   and  other
pertinent  information.    Some  specific  items  that  treatment
facilities such as BIFs and  HWIs must  include in the WAP, per 40
CFR §264.13, §265.13, and §268.7 are discussed below:

2.1       SAMPLING METHODS FOR EACH FEED STREAM

Sampling methods may be included in the WAP eithesr by reference to
sampling methods described in 40 CFR §261 Appendix I (for example,
specific methods set  forth in EPA publication SW-846 or specific
American Society for  Testing  and Materials (ASTM)  methods)  or by
specifying  an  equivalent standard  sampling procedure  for  the
selected analytical method. The WAP must  describe measures used to
ensure  that the analytical  sample(s)  is representative of  the
entire feed stream  (40  CFR §264.13  and  §265.13).  Representative
samples may  be grab  samples  or composite samples.   In general,
compositing of samples should be used only to account for spatial
variations within a single sample lot (for example,  a rail  car load
of coal or'a truck load of limestone).   Compositing should not be
used to reflect the  concentrations of constituents in a group of
waste containers that originate at any one of  several sources.  If
a facility's  regulatory limits on  feed  rates are  specified on a
time-average basis (that  is,  hourly rolling average), compositing
also may be used to account for temporal variations in the sample
lot.  In such cases, the  compositing period should not exceed the
regulatory  averaging  period  for  that  sample  lot   (that  is,
compositing of several sample lots being  burned at different times
is  not  appropriate).  -  If  the   facility   is  subject  to  an
instantaneous  constituent feed rate limit,  temporal compositing
should not be used.  Test methods in SW-846 provide'more detailed
information on sampling methods.

2.2       METHODS OF PREPARATION AND ANALYSIS OF SAMPLES

Methods of preparation and analysis of samples must be specified in
the WAP for each regulated constituent in each feed stream (40 CFR
§264.13  and §265.13).    This  requirement can  be met  either  by
reference  to  standard  methods  of  preparation  and analysis  of

                 - DRAFTS Do Not Cite Or Quote -'

-------
samples  (specific  methods in EPA publication  SW-846  or specific
ASTM  methods)   or  by  specifying  a 'step-by-step procedure  for
preparation and analysis of samples.  An SW-846 method must be used
when required by regulation.  If an SW-846 method is not specified
in the regulations, it is recommended that SW-846 methods be used
whenever the methods  are both available  and  appropriate for the
sample matrix; however, other equivalent methods generally may be
used.  In addition, any laboratory that  is to conduct the analysis
and meet requirements  for quality assurance  and quality control
(QA/QC)  procedures (testing methods,  laboratory  procedures  for
handling of the sample, and others) should be specified  in the WAP.

2.3       SAMPLING AND ANALYSI13 STRATEGY

An acceptable  strategy  is one  that,  in  combination with the data
from continuous monitoring of  the feed  rate,  provides reasonable
assurance that all constituent feeds are within allowable limits
before they are fed and that the limits on feed rates will not be
exceeded at any time while waste is being burned.  After-the-fact
knowledge of feed rates  of constituents is not  an acceptable way to
determine compliance with the regulations.   The strategy should
outline the frequency at which the feed streams will be  sampled and
analyzed.   Supporting  documentation should be  kept on record to
justify the selection of the frequency of sampling and analysis,

2.4       SAMPLING LOCATION

The location from which  a  sample is to be collected is important in
determining  the appropriate  sampling method  and assessing  the
ability to obtain a representative sample.  The location also may
influence the results of the analysis, thereby affecting the feed
rates,  as  well  as the choice  of  an  appropriate frequency  of
sampling and analysis.   Examples of appropriate  sampling locations
include:

     •    For  an on-site, continuous process  that generates  one
          waste stream,  a sample may be obtained from the pipeline
          that  feeds  hazardous  waste  to  the  combustion  unit.
          However,  such sampling should  be implemented  only  at
          facilities at which  it  is  known, through a statistical
          profile,  that none  of the concentrations of constituents
          in  the  feed  streams  is  above  the  maximum allowable
          limits.                           .    ,

     •    For a batch process, such as a  tank filled with hazardous
          wastes, a representative sample of  the entire batch in
          the  tank should be  obtained  and analyzed  before  the
          contents  of   the   tank   are   fed.     Potential  for
          stratification of the wastes should be considered during
          the   sampling  procedure.      Continuous    mixing   or

                 - DRAFT: Do Not cite Or Quote -

-------
          reeirculating  of  the  contents  of  a  tank  reduces  the
          significance of the degree of heterogeneity of the waste.

     •    For a  lot  of containerized wastes  from  the same waste
          stream,  a representative  sample  may be  obtained  by
          compositing  samples  from the containers.   ASTM Method
          D140-70 may be used to estimate the  number of containers
          within the  lot to be  sampled.   Each  sample  should be
          considered acceptable only if the particular, waste sample
          closely  resembles  all other  samples   (for  example,  in
          color).  The  composite or representative sample should be
          analyzed before the  wastes  are  fed to  the combustion
          unit.                                 ;   ;

Whatever  sampling  location is  selected,  the location  should be
identified clearly and its selection justified in the WAP.

2.5       QUALITY ASSURANCE AND QUALITY CONTROL

The facility's QA/QC procedures for sampling and analysis should be
stated in the WAP.   Sources of information on developing a QA/QC
procedure include:  1)  Chapter One of SW-846, "Quality  Control"; 2)
Guidance  on  Setting Permit Conditions and Reporting Trial  Burn
Results (EPA/625/6-89/019); and 3)  Handbook -  QA/QC Procedures for
Hazardous Waste Incineration (EPA/625/6-89/023).

All the factors discussed above can  influence the  quality of the
analytical results. Therefore, they should be addressed in a site-
specific WAP as part of a facility's demonstration that the waste
streams will be sampled and analyzed in a manner that complies with
requirements  for monitoring of  feed  rates of constituents.   In
their  WAPs,   facilities   also  should  set  forth  procedures  for
evaluating analytical data with respect to outliers, completeness,
and detection limits,  as discussed  in greater  detail in Section IV
of this document.                                ;
             3.0    SAMPLING AND ANALYSIS STRATEGIES

This section presents options for sampling and analysis programs to
ensure  compliance  with  either  the permit or  the  regulatory
requirements discussed above.

The BIF  rule and  some  permits for  hazardous waste incinerators
require  combustion facilities  to continuously monitor  the feed
rates of selected metals, chlorine and chlorides,, ash, and wastes
(40 CFR  §266.103  (c) (iv) (D). and §264 and 265.347).   As discussed
above, to  satisfy this requirement,  the feed rate  of  each feed
stream must  be monitored continuously and  the facility operator
must "know" the concentration of each regulated constituent in each

      ,           - DRAFT: Do Not Cite Or Quote -

-------
                         -       8

feed  stream.   The  "requirements  for  a  continuous monitor  are
provided in 40 CFR 266.103(c)(4)(iv)(B)(i).  A logical and coherent
sampling  and  analysis  program for regulated  constituents-,  and
continuous  monitoring  of   feed   rates   of  feed  streams,  are
fundamental  aspects  of a  compliance strategy  that ensures that
limits on maximum feed rates  of regulated  constituents  are not
exceeded.     Knowledge  of   the   concentrations   of   regulated
constituents ' in  each feed stream  should be based  on -an ongoing
sampling  and analysis program.  Fundamentally,  the "knowing", of
concentrations of regulated constituents  allows  the  calculation of
feed rates for those constituents  for  any point in time at which
hazardous waste  is being burned.   That  calculation then  can be
compared with regulatory limits.

When  a sampling  and  analysis program  is  established,  several
factors should be considered, including:  variability of the feed
stream, sampling  location,  and proximity  of levels of regulated
constituents  to   established  limits.   The  following  discussion
describes three  generally  acceptable approaches  to sampling and
analysis.   Other strategies may also  be acceptable and  will be
considered on a case^-by-case basis.
                                                     '
3.1       SAMPLING AND ANALYSIS BY BATCH

Sampling and analysis by batch, is a strategy most appropriate for
facilities that have multiple feed  streams in which  concentrations
of regulated constituents vary  greatly  and for  facilities that
receive wastes from  off-site.   Multiple storage  and  feed tank
systems  may  be  necessary  to  properly  execute  this  compliance
strategy.     The   batch   methodology   reguires   that  •once  a
representative sample has been taken from a tank and analyzed, no
other material can toe added to the  tank.   Results of  laboratory
analysis  must be known  before  wastes  are burned;  therefore,
laboratory turnaround time may  be, a consideration.   "The measured
concentrations of the regulated constituents establish a maximum
feed rate that is  at  or below the regulatory limit.   For tanks that
do not have  agitation systems,  stratification of the contents is
possible.   Therefore,  care should be  taken  to  ensure that  a
representative sample is  obtained.   The  objective  of a  batch
strategy is to enable a facility to calculate a maximum feed stream
rate based on measured  concentrations  of regulated constituents.
The facility also can calculate the actual feed rate of regulated
constituents  at  actual  feed rates  of  feed streams and  actual
concentrations of constituents  in  any   given  instance.    Batch
sampling and  analysis is a relatively simple and straightforward
methodology  for  ensuring compliance.   Examples  that  illustrate
generally acceptable and unacceptable ways of complying with this
strategy are given below:
                 - DRAFT: Do Not cite Or Quote -

-------
»    A simple example of a case in which sampling and analysis by
     batch is generally  appropriate is a  facility  that receives
     hazardous waste  from many off-site  sources and  blends  the
     wastes on site.   Such a facility may conduct some preliminary
     analysis on  each waste  stream  before it  is  accepted  and
     discharged to the storage tank system.  Wastes are accumulated
     in one  of three continuously agitated" mixing and  storage
     tanks.  When a tank  is full,  a representative sample of the
     waste  in the  tank  is  obtained  and  analyzed.   After  the
     facility receives the results  of  the  analysis,  the waste is
     fed as a batch to the combust ion, unit.  .Once characterized, no
     other material (for  example,  hazardous wastes,  used oil, or
     fuels) is added to the tank being fed,  and incoming wastes are
     accumulated in the remaining  two tanks.  Calculations of feed
     rates are based on the results of analysis of that batch.

•    A facility generates several waste streams  from  relatively
     consistent production.processes.  One or all  of the streams
     may be  piped to  a  storage  tank  at  any given time  and in
     quantities determined by production.   A sample to determine
     the concentrations of metals,  ash, and chlorine and chlorides
     is taken once from  the storage tank  for preparation of the
     certification of compliance,  and again six months  later in
     preparation for emissions testing  to revise the certification
     of  compliance.     The.  two   samples show   variations  in
     concentrations   of   constituents?    as    shown    however,
     concentrations in both samples are below limits on feed rates.

In  the  second  example,  the   facility   performed  analytical
determinations but did not consider how the results would be used
to document compliance.   For example^  to calculate a feed rate at
a given point in time, which of the  two results (if either) should
be used to determine compliance?  How can the facility prove that
the  two samples  include  the  variations   in  concentrations  of
constituents, considering that the various process streams exist in
different ratios in the burn tank  at any given time?  The facility
should consider several options that are more reliable compliance
protocols.   One option may be a batch feed  operation, in which the
three streams would  be collected  in  the tank and  an  analytical
determination made  after  preparation of  the  batch and  before
feeding.   Under this option, a  given set of analytical results
correlates directly with the period when a particular batch is fed.
Drawbacks are associated with the approach:  frequent analysis is
necessary (every tank) and installation of several new tanks may be
appropriate.                     .                       ,
                 - DRAFT: Do Not Cite Or Quote -.-

-------
                                10             .

3.2   ,    QUALIFICATION OF A FEED STREAM

As' long as a facility can ensure that the feed rates for regulated
constituents are at or below regulatory limits,  it is not necessary
to  know  the  exact  or  actual  constituent   feed  rate.    This
alternative  is a  variation  of  the  batch sampling  and analysis
strategy and may be appropriate for  facilities that have complex
feed management systems  and those that have a continuous demand for
steam  or production  rate requirements  (this does  not  imply  a
constant feed rate of hazardous waste,  since such wastes often are
cofired with other  fuels).  The  qualification strategy is similar
to  a batch  strategy in  that  all feed  streams are  sampled and
analyzed  for  all  constituents  identified  in  the  permit  or
regulations  at some point in  the feed stream management system
before they  are fed into the  combustion  device.  After-the-fact
knowledge  of   constituent   concentrations  or   feed  rates  of
constituents is not acceptable.   The qualification strategy also
can  be  an  appropriate  approach  for  facilities that  generate
multiple waste streams in various quantities and  at various times.
This  strategy  can  be implemented in  two ways.   There  are two
variables  for  the  calculation of constituent  feed rates:   The
concentration of the regulated constituent in the feed stream and
the feed rate of the stream.   If one variable is fixed, the other
variable can be adjusted to ensure that the regulatory limit is not
exceeded.  This approach is illustrated below as:

     FR =  (C)  (Q)

     where:        '              .

          FR   =    The   regulatory    fee,d  rate   limit  of .  a
                    constituent   for  the  feed  stream   (unit
                    weight/time)

           C   =    The concentration  of the  constituent in the
                    feed stream

           Q   =    The feed rate of the feed stream

The  feed rate  limit, FR, has  a  maximum, value that  cannot  be
exceeded.  (It should be noted  that the sum of all Qs must be at or
below  the  allowable hazardous  waste feed rate  for the  Qs  that
represent hazardous waste feed  streams '(40  CFR §266.103(c)(l)(i)).
It follows that both C,  the concentration of the constituent, and
Q, the feed  rate of the feed stream,  can  vary,  as long as their
product does not exceed FR.   Two options using this approach are:
                 - DRAFT: Do Not cite Or Quote -

-------
     •    Qualification based  on  predetermined feed rate  of the
          total feed system

          This option can  be  used by a facility  that  requires a
          relatively constant  feed rate,  Q,  to  meet  production
          needs or  demand  for steam.   The fixed feed  rate also
          fixes the maximum acceptable concentration, C, that can
          be present without exceeding the maximum  regulatory limit
          FR.  In other words, the concentration, C, can vary below
          its maximum  limit.   Thus,  when using  this  compliance
          strategy, a facility would analyze each batch of waste or
          feed  material   for   regulated  constituents   before
          acceptance of the waste into the feed management systems.
          If each regulated constituent is determined to be at, or
          below its maximum allowable concentration (determined by
          the  fixed feed  rate  Q so  that  thei  product   of  the
          concentration,   C,   and  Q  does  not  exceed  FR,  the
          regulatory   limit) ,   that  batch   is;  qualified  for
          combustion.  For determination of compliance, the  "known"
          concentration of a regulated constituent is the maximum
          concentration at which the material meiy be ''qualified."
          The  qualified  material  then  may be   blended  without
          restriction with other qualified feed  streams  without
          further analysis, since FR  is  at  or below limits  for the
          fixed Q.                  , .   . '                    .

     *    Qualification based on predetermined concentrations

          This approach sets a maximum limit on C (that is, fixes
          this  variable),  the   concentration  of   a   regulated
          constituent,  and allows Q to vary below a maximum value
          determined in such a way that the product of C and Q does
          not  exceed  FR,   To  use .this  compliance strategy,  a
          facility performs the required waste analysis on incoming
          batches of waste before the batches  are mixed.   Rather
          than doing another analysis of,the  blended wastes, the
          waste stream is considered to have the. concentration of
        *  regulated constituents  found in the batch having the
          highest concentrations.  The facility then calculates a
          maximum feed rate,  Q.   Compliance with the  regulatory
          limits on feed rates will be ensured as long as Q remains
          at or below its maximum value.

Implementation  of   the  qualification  strategy can  vary  widely
depending on the complexity of a' facility's feed; stream and waste
management system (for  example, presence of an interconnection and
isolation system for storage  tanks) .   It  is alisd possible for a
facility to reestablish a predetermined maximum concentration, C,
that could be  lower than a previously established level and thus
allow an increased feed rate, Q.  In any event, as is true of the

                 - DRAFTS Do Not Cite Or Quote -

-------
                                12

original limits,  a  change in limits on  concentrations should be
well documented,  justified, and  specifically associated  with a
particular time period during which the waste is burned.

Application  • of   the  compliance   strategy  based  on   use  of
predetermined  concentrations " is   illustrated by  the  following
example:

     •     A  facility  that receives liquid  hazardous  wastes from
          off-site has a series of  storage,  blend, and burn tanks,
          as shown below:
                      Combustion
                        Unit(s)
         Level 3
         Level 2
         Level 1
                      Burn tanks
                     Blend tanks
              Storage tanks, containers
              and transport vehicles
     All incoming shipments are received, sampled, and analyzed at Level
     1 and determined to be below the maximum concentration limits, C,
     which are based on a fixed feed stream feed rate limit, Q.  Thus,
     constituents past level  1  in  the system will have concentrations
     less than the limits, because the operator does not allow transfers
     into the  system of waste having concentrations above the maximum
     concentration limits.   This sampling and analysis strategy also can
     be applied effectively at Level 2.  When monitoring concentrations
     at  Level  2,  for  example,  it  is not  necessary that  the  actual
     concentrations of the individual loads delivered to  the Level 1
     tanks be below the maximum limits, C,  since all concentrations that
     pass through levels 2 and 3 to be fed to the combustion unit must
     be at or below maximum concentrations C.  However, the facility may
     not be able to apply this strategy at Level  3  since after-the-fact
     knowledge   of   constituent .  concentrations   or  feed   rates  of
     constituents is not acceptable.   Weighted averages  should not be
     used to determine levels  of concentrations.  Compositing of samples
     from  different  levels  and   tanks   is  not  acceptable.    If  a
     concentration, C, anywhere  in the system  before the  level being
     monitored exceeds a  predetermined maximum concentration limit, the
                 - DRAFT: Do Not Cite Or Quote -

-------
               •'   .           .   .   ' .13  '•      .     ','.".-'..-•'....

          contents of the tank can be reblended with, other fuels until the
          concentrations  are lowered  (and resampled  and  analyzed before
          transfer to the next level) , or  a new maximum C can be established
          and applied in future calculations  of feed rates.  Under certain
          circumstances, blending of wastes (for the batch or qualification
          strategies) may require a permit.
3.3       STATISTICM,

A statistical  approach  can characterize concentrations, of constituents in
fossil fuels, raw materials,  or wastes generated on-site.   It is appropriate
for "consistent" feed streams  (for  example,  hazardous waste generated by a
specific on-site production process,  coal  produced from a specific mine or
seam, or limestone ore produced from a specific quarry), for which there is
reasonable expectation  that  each constituent will  be normally distributed
about a mean.  It should not be used at facilities that receive wastes from
off  site.    The  approach demands that the  operator  of the  facility have
sufficient knowledge of the source  of the  feed  material to be aware of any
change that is likely to affect the  sample  distribution.  When such a change
is known to  have occurred, the  facility  operator should  not  rely on. this
approach until a statistical  profile of  the  "new" , feed stream  has been
developed.    Through  statistical  analysis,  the owner or operator ultimately
will develop a program that  specifies a  frequency at which  sampling and
analysis are to  be conducted  to  ensure,  with  an .appropriate  degree of
confidence, that feed rates are not exceeded.  It 'also' should be understood
that, with the use of a statistical approach, there is a finite probability
that a : facility can  be  found to be  out of  compliance based on sampling and
analysis.  If such a circumstance occurs,  use of a statistical sampling and
analysis strategy is not a shield against enforcement action and the adequacy
of the analysis may be considered in penalty calculations.

Because this approach should be used to characterize waste streams only as-
generated.    It should not be used after the waste has been blended with any
other  waste,  fuel,   or raw  material.    It  is   therefore generally  not
appropriate  in any  case  in  which  the hazardous  waste  is generated  at a
facility that  is  not under the same  ownership  and control as the facility
that burns  the waste. (This approach may be appropriate however,  in cases in
which raw materials and  fossil  fuels are produced  by entities other than the
facility that  burns the  waste,  provided  that  there  is a  contractual
requirement  that  the  burner   be  notified  of  changes   that  could  have
significant  effect on concentrations of constituents in those feeds.)

When using any statistical  approach, facilities  should  be guided  by the
following principles:

     •    The  statistical  analysis  should be  based,. on 'actual  analytical
          results.   As  discussed  above,  Process  knowledge  alone  is not
          generally sufficient.
                      - DRAFT: Do Not Cite Or Quote -

-------
                                     14

     •    The  operator of the  facility should  demonstrate at least  a 95
          percent probability and confidence that the maximum concentration
          of any sample will not exceed an allowable limit.

     •    A continuing sampling and analysis program should be established to
          demonstrate that the statistical distribution does not change over
          time.

Of the several approaches to conducting a statistical analysis,  the use of an
upper tolerance limit is discussed here.  This is the same method described
in 40 CFR §266 Appendix IX Section 7.2 of the BIF  regulations for Statistical
Methodology  for Bevill  Residue Determinations.   There  also  is  a useful
discussion of  the application  and calculation of upper tolerance limits in
Statistical Intervals:   A Guide for Practitioners ,  by Gerald J.  Hahn and
William Q. Meeker  (ISBN  0-471-88769-2) .   For reasons discussed below,  this
approach  is   a  recommended  approach  to  waste  analysis  at  combustion
facilities.

A general overview of statistical analysis is an appropriate starting point
in understanding the approaches to be discussed.  The underlying concept of
statistical  analysis  is  the  development of  a  mathematical model  for the
expectation (or prediction) of a random variable within a given population.
Such a model,  commonly known as a probability distribution model, gives the
probability that a random variable,  x,  lies between two values.  Development
of the model is simple and is illustrated below:
          •
For a  sample  of random  values for  a given data  set,  one  could  find the
average value  for the sample,' which  is called the arithmetic mean, X.  This
is expressed as:
                                       n
where:

     X    =    Numerical value of sample point n

     n    =    Number of samples
     X    =    Mean of X
                      - DRAFT: Do Not cite Or Quote -

-------
                •• •  '           -      15              ."          . ..  .

 Next,  the distribution of the sample values.about the mean is desired.   The
 most common distribution  is  called a "normal" or  "Gaussian" distribution.
 Graphically,  this distribution  is represented by  a bell-shaped  curve,  as
 shown  below:                              .            •
                    o
                      -4   -3   -2  - -I
                                    Value  of  X
      Solid curve:
      Dotted curve:
Normal  distribution,  with  mean  of
deviation of 1.                     .

Normal  distribution,  with  mean  -of
deviation of 1.5.            "   . !
0  and  standard
0  and  standard
As this curve shows, lower and upper values of .the data set can be calculated
with known probabilities.  The shape of the curve depends upon the scatter or
dispersion  of the values about the mean and is often referred to as a "two-
sided" distribution.  Evaluating the dispersion or scatter about the mean can
be done  by calculating the standard deviation.  The standard deviation,  s,
can be calculated as  follows:
                       s =
                                    (x-x)

                           Un-x)2
 The objective is to describe  a population represented by  the  samples,  for
 which any given sample can be found to be between  a  set  of upper and lower
"limits.   From the samples,  a confidence interval for the unknown population
 mean can be constructed.  This interval consists of  two values, the upper and
.lower limit.  Given 'certain assumptions about the population, the chance that
 these values  straddle  the unknown mean is a certain percentage.
                       - DRAFT:  Do Not Cite Or Quote -

-------
                                     16

3.3.1          Upper Tolerance Limits

One approach that  satisfies  the  criteria set forth above is based on upper
tolerance limits.  This approach is outlined in the paragraphs  below.  For a
more detailed description, see Meeker  and Hahn  (1991).

If  a  "variable  is normally distributed  and  the  sample mean,  standard
deviation, and number  of  samples are known,  it is possible to estimate the
probability that a fixed percentage of the sample population will not exceed
a certain value.  That  value  is known as an upper tolerance limit (UTL) .  Fox-
purposes of this guidance,  the minimum UTIj that  should be used in lieu of
continuous analysis of waste is  the  value of the  one-sided  upper 95 percent
tolerance bound  that exceeds at  least 95 percent of. the sample population.
In other words, we can say with 95 percent .confidence that  95 percent of all
individual, samples will  not exceed the UTL.    Therefore, if  a  facility
generates a good initial database to establish the UTL  for  the constituents
of  concern,   and  subsequent,  sampling  and  analysis   shows  that  the
concentrations are below  the UTL, the waste can  be considered the type of
waste  for which the  UTL was  calculated.    The  UTL  values  then may be
considered  the  "known"  concentration for  each  constituent  in  that feed
stream.                            .

Although concentrations of constituents in any single sample are likely to be
well  below the  UTL,  feed  rates always  should  be  calculated as  if each
constituent were present  at its UTL.   The UTL  is adjusted continually to
reflect new  information from  analysis.   The  UTL for  each constituent is
calculated as follows:

Step 1:   Using all valid analyses of the subject feed stream,  calculate the
          mean concentration (X)  and the standard deviation  (s)  for the
          samples.-

Step 2:   Using the equation below, calculate the upper  tolerance limit, UTL
           (o.95;o.9s)  •  so that there is  at least 95  percent confidence that at
          least  95 percent of all samples will not exceed the UTL.  Values
          for K are obtained  from a table for calculating one-sided tolerance
          bounds for a normal distribution (see Appendix A) .
            = X + (K (lKrip)) (s)

where:

     1-a =     The desired level of confidence that at least 100 (p) percent
               of the individual samples will be below the UTL.

     p    =    The decimal fraction of samples that will be predicted to fall
               below the UTL.

     n    =    The number of samples.


                      - DRAFTS Do Not  Cite Or Quote -

-------
Table  1  in  Appendix A  lists  values  of K  for  l-a=0.95,  with p=0.95.
Statistical  references may be consulted for other values  of i-a.  Linear
interpolation may be  applied to  obtain values of n that are not  tabulated.

This  guidance recommends  that,  if a , UTL  is  to be used to  demonstrate
compliance, i-a must be >0.95 and p must be >0.95.  A  more  conservative  (that
is, higher) UTL may be used to decrease  the necessary frequency  of sampling
and analysis, as described in the following step.

Step 3.   Determine  the  appropriate frequency  of  sampling  and  analysis
        ,  according to the following equation.

     number of samples   =  (a^ic)  (₯)
            year

•where:       ''     _                 •     "             | "-./•'

     acalc       =    One minus the level  of confidence used  to calculate  the
                    UTL; at a 95 percent confidence  level,  a^,. = (1-0.95) =
                    0.05

     Y         =    days per year on which waste is generated
                                               •                 N
For  facilities  that  meet  the  minimum requirements of  this  methodology
(estimating  concentrations based on "«„,,,.  — 0,95),  the feed  stream  should be
sampled  and analyzed on  at least  5 percent  of the days  on which  it  is
generated.  If the facility chooses to use a more conservative  UTL, where a^
>0.95, the burden of  sampling and analysis will be reduced.

In qualitative terms, as  the  statistical confidence  that an allowable feed
limit  on, constituents will not  be exceeded increases,  the frequency with
which sampling and analysis are necessary decreases.   However, at a minimum,
each feed stream should be analyzed at least once per year.  Also,  sampling
dates should be spaced evenly throughout the year.

Most statistical tests assume that the data come from a normal distribution.
The normal distribution is the assumed underlying model for such procedures
as  calculation of tolerance  intervals.    if the data  are  not  distributed
normally,  false  conclusions can result if  the  data  follow a more skewed
distribution  like lognormal.  Therefore, checking the data  for normality is
an  important  step  in statistical calculations.   EPA  has  available a useful
discussion of evaluating data for normality in  a  document  titled  Statistical
Analysis  of Ground-water Monitoring Data at RCRA Facilities, Addendum  To
Interim Final.Guidance - Draft  (EPA/500/R-93/003, July 1992).  The document
is available for sale through the RCRA docket at (202) 260-9327.  Copies cost
$0.15 per page.                                       r
                       - DRAFTS Do Not cite Or Quote  -

-------
                                     18

3.3.2          Statistical Approach:  Compliance Issues

No statistical approach can guarantee true continuous compliance with short-
term constituent feed rate limits.  There is always a finite possibility that
concentrations of constituents in any given sample will exceed the UTL.  This
fact  is  accepted  in statistical  characterization.    If the  sampling  and
analysis indicates  that the UTL has been exceeded, then the following is
recommended:

   O continue to calculate constituent feed rates using the UTL

   o immediately following receipt of an  analysis that exceeds the UTL for
     any constituent, the facility should begin daily sampling and analysis
     of that feed stream.  Daily analyses should continue until all regulated
     constituents are below their UTL for three consecutive days.
                                                          i
   O if the feed stream exceeds  the  UTL  for the same  constituent  2 or more
     times  while  conducting  the  daily  sampling,  the  facility  should
     immediately cease  using the statistical .approach  for that feed stream
     until a new  feed profile is developed  (using  data obtained  after the
     initial UTL exceedance).

It should be noted that, at facilities that  have more than one waste stream/
the maximum concentration  of different regulated  constituents can occur in
different waste streams; thus, UTLs  should  be calculated for the different
waste streams; and the UTLs are then composited for all waste streams.

The  following  example  illustrates the  calculations  for the UTL  and  for
determining sampling frequency.

     •    A facility  generates one waste stream on site from a relatively
          constant production process.. The  stream has been analyzed several
          times for metals, ash, and chlorine and chlorides.   The analyses
          revealed some .variations in concentrations  of constituents.   The
          level of chromium (Cr)  is near the allowable feed rate limit,  but
          the levels  of all other constituents are well below the limits™
          The  facility  would like   to  use  the  results  in  its  WAP  to
          demonstrate  that the  values  are  below the. concentrations  of
          constituents  used in  calculating feed  rate  calculations.   The
          facility also would like  to  specify more frequent analysis  for
          chromium than for the other constituents.

          Because  not all the  constituents are  well  below  the  allowable
          limits,  it is  appropriate in  a  case  such  as  this to  specify
          different frequencies of  analysis  for different constituents.  This
          UTL of the tolerance interval can be compared with the  feed rate
          limit for each constituent. The WAP  should specify  that the upper
          limit of the tolerance  interval  is s to . be  used  in determining
          compliance, and,  with future analyses at some reduced frequency, to
          verify that concentrations remain below the UTL.

                      - DRAFT: Do Not Cite Or Quote -

-------
                  . •                -19               ,.    •      .  .     ..

     Data from analysis for  chromium used for statistical calculations and
     calculations of feed rate are:                                      ,
                                                     !
        •  - '   Ten samples were analyzed; n — 10.

          -    The mean of the data  (average) is calculated; x = 2.39 mg/L.

          -    The standard deviation is calculated; s = 1.53. .  .

          —    The facility established a maximum acceptable concentration of
               Cr of 8.47 mg/L for calculations of feed rate.
                                          *   _  \       •
Using the procedures described earlier, UTL is calculated as follows:
                                         \             •,           .

          UTL  =    X + Ks .
                                                     :   •  - \

     where:   .     '        '•,.'.         .      '       -        '

            X  =    Mean of the samples (Cr concentration - mg/L)

            s  =    Standard deviation of samples

            k  =    2.911 f or n =10  (sample size)
                                                   \       "     ;-    "
          UTL  =    2.39 + (2.911) (1.53)

          UTL  —    6.84 mg/L               -                   .

The  UTL  of  6,84 mg/L  then  is  compared with  the  maximum  acceptable
concentration of Cr of 8.47  mg/L.   For  a  fixed feed stream feed rate at or
near its maximum, use of the UTL provides a safety margin that ensures that
the regulatory limit for Cr is not exceeded,.

The facility also should determine the' frequency of sampling and analysis for
Cr as part of its sampling  and analysis  program.   That program should be
described in detail in the waste analysis plan.  Assuming that the  facility
generates waste for 365 days  per year, the frequency  should be determined as
follows:

     Number of samples/year = [ (a^) ] (₯)

where:      . .         •          '          -  •"         • '.  •   y ,          "  •

     [(
-------
                                     20

Thus, the facility should  sample -the feed stream a minimum of 19 times per
year  (rounding up provides  an extra degree of  certainty).    The minimum
frequency of sampling should be once per year.

Site-specific factors can influence the choice of a statistical approach to
compliance and its associated  sampling and analysis strategies.  Even if a
facility follows the procedures outlined in its WAP, problems related to the
results  of  analysis  may  arise.    The issues  that arise  most  often are
incomplete data,  outliers,  and detection limits.   These  issues are discussed
further in a later section of this document.                    ,

Stiaunaarv*                '                                            ' .

Below are  listed factors  to  be considered  when selecting a  sampling and
analysis strategy  for each methodology:   sampling and  analysis  by batch,
qualification of the feed stream,  or the statistical approach.   Such factors
include, but, are not limited to:

     •    Sampling and Analysis by batch

          -    Appropriate for feed streams generated both on and off-site
          -    Appropriate for multiple waste streams produced  from on-site
               processes   -         -            '-         - •
          -    Simplicity    •
          -    Ease of documentation of compliance
          -    Waste management system  (burn tanks, blend tanks, and sample
               location)
          -    Economic factors related to laboratory analysis

     •    Qualification of the Feed Stream

          -    Appropriate for multiple feed streams generated  on site
          -    Flexibility  with  regard   to  feed  rates  or  constituent
               concentrations
          -    Possible complexity of documentation of compliance
          -    More complex methodology to establish and execute than a batch
               system                               ,                      -
          -    More appropriate for situations in which a constant production
               rate to generate steam is necessary

     •    Statistical Approach

          -    Appropriate for "as-generated" waste streams originating oh-
               site                               ..        :         '.
          -    Fossil fuels                              .      .   t   •''.•'
          -    Raw materials                        '           .
          -    Requirements for maintenance of the database
          -    Possible  requirements  for  periodic   reestablishment  of
               statistical baselines for feed streams


                               Do Not Cite or Quote -

-------
   .-,-"'     •         -"   •    '    21       •  .   .   : _  •       '-   .--..••

          -    Measurable and finite probability that a facility might be out
               of compliance
          —    Minimum frequency of sampling of once per year


           4.0   INCOMPLETE DATA, OUTLIERS, AND DETECTION LIMITS

Because it is important to have complete and accurate data on waste analysis,
it is appropriate to  discuss the issue in  further  detail.   The facility's
QA/QC procedures should be set forth in  the WAP,  The QA/QC procedure should
outline a protocol for dealing with incomplete data, outlying data points,
and detection limits.  Such information may be requested during an inspection
and will  play an, important  role  in determining compliance.   Therefore,  a
facility should maintain it.

4.1       OUTLYING DATA POINTS

In waste analysis data, an outlying data point is one that does not appear to
be within a reasonable or expected numerical range.   Such an assumption most
likely will be based on historical data  with which a comparison can be made.
When  it is  suspected that  an  outlier has  occurred,  the  facility  should
determine why it has  occurred.   The quality assurance procedures submitted
for the analytical test of the sample should include detailed and objective
rejection  criteria  for  all  outlying data points.    Those  criteria could
include procedures  for documenting  outliers and determining  why outliers
occur and what corrective action should  be taken to  prevent such events from
occurring in  the future.   Several references  are available for evaluating
outliers.   For  example,  a facility  may evaluate the  validity  of its data
using ASTM  Method  E  178-80,  "Standard  Practice for Dealing With Outlying
Observations."    in  applying  that  and  other  methods,   the  underlying
assumptions  of  the  methodology should  be kept in  mind  (for example,  ASTM
Method  E  178-80 states  that "the  criteria for outliers  are based  on an
assumed underlying normal (Gaussian) population or distribution").  Data that
are .suspected of being outliers,  but in  fact are the result of the character
of the feed stream, or data that cannot be explained  otherwise as an error in
sampling  or  analysis,  are  not outliers and  should  bet  included in the data
set.  Outliers  caused by an error in sampling should  be  corrected through
immediate resampling and reanalysis of the feed stream.  Outliers caused by
errors in analysis often are corrected through  reanalysis of  the sample..  If
the holding  time of a sample has expired, the  facility should resample and
perform the  analysis  again.   However,  it  is recommended that the facility
take  two  or  three  samples  at  the  same  time; if  one sample  exhibits  ah
outlier,  the  remaining samples  can be analyzed.  If the  facility is using
statistical .analysis and  has  an outlying data point above the  calculated UTL,
it  is suggested .the facility  use  the  value  of  that data point  in  the
calculations  of feed rate until  resampling or reanalysis  shows different
results.  All procedures for identifying and discounting outliers should be
documented in advance in the WAP.                            .     r
                      ,- DRAFT! Do Not cite br Quote -

-------
                                     22

4.2       INCOMPLETE ANALYTICAL DATA

Data from waste analysis  are  considered incomplete when results for one or
more  regulated  constituents  are  missing  from  the  analytical  report.
Incomplete  analytical  data  is  generally unacceptable  in  demonstrating
compliance.  A facility using sampling and analysis by batch or qualification
of the feed stream should  reanalyze.  However,  analyses conducted pursuant to
a  WAP  that  specifies different  frequencies  of  analysis  for  different
constituents  would  not  be considered  incomplete.    Otherwise,   for  the
statistical program, the facility should use the reported data and reanalyze
to get results for  all  constituents.    As discussed  earlier,  all quality
assurance  should  be  conducted,   documentation gathered,  and corrective
measures taken to prevent recurrence of this problem in the future.

4.3       DETECTION LIMITS

The BIF regulations and some HWI permits specify the use of testing methods
set forth in  SW-846  for some  constituents (for example, 40 CFR §266.106(a)
for  metals).   Limitations  associated  with  these  methods,  such as  the
detection  limits,  can present  problems in the effort to  use statistical
analyses to determine sampling frequencies for some facilities that generate
wastes  on-site.   When using  these methods, such  facilities may  find that
results  of  analysis  are  at  or  near  the  detection limit(s)  for  the
constituent(s).  Consequently, it may be difficult to develop a statistical
distribution  for the constituent  because most of  the distribution is below
the detection limit.   Therefore,  the  facility may not be  able to  use  a
statistical approach to determining an  appropriate  frequency of  sampling arid
analysis.   In such situations, it may be appropriate  to  specify that the
facility sample  and analyze  more  frequently  (for example, every batch).
Possible solutions to this problem  include:

     •    The facility can use/when appropriate, the SW-846 Method 6020 (see
         •58FR46052) for analysis.  (This method was promulgated in January
          13,  1995 in the second update of the third edition, second update
          of  SW-846.)   This  inductively  coupled  plasma-mass  spectrometry
          (ICP-MS)  method is a multielement,  simultaneous method  for the
          analysis of  inorganic analytes.   It is capable of  testing for
          metals at much lower levels than other SW-846 test methods [parts-
          per-billion  (ppb)  instead of  parts-per-million (ppm)].  Use of the
          method where allowed  by  the  regulations  or permit conditions may
          provide better analytical data upon  which to base development of a
          statistical distribution, because the results of analysis would be
          less likely to fall below the lower detection limits.

     •    The facility can,  when appropriate, develop a mathematical model to
          estimate the statistical distribution of .constituents  below the
          detection limit.  For  example, EPA has recommended,  in  the guidance
          document on  groundwater  monitoring  referred to above, the simple
          substitution method, under which nondetected results are replaced
          by one-half the detection limit.  A mean and variance then can be

                      - DRAFT: Do Not cite Or Quote -

-------
             :       ..           '      23     .    •  '    ;          ,

          calculated by assuming all measurements  were observable with the
          same precision*   Another  model that may be used, when appropriate,
          is the  maximum  likelihood estimator  (MLE).   Cohen  (1959,  1961)
          developed the  MLE  for  calculating the  mean and variance  of  a
         , distribution based on the mean and variance of the  detected values,
          the difference  between  the mean  of  the detected values  and the
          censoring point, and a factor that depends on the proportion of the
          data,that are nondetected results.  This approach has been found to
          work best for  small, normally distributed results.  A discussion
          this  Cohen's method can  be   found  in  Statistical  Methods  for
          Environmental Pollution  Monitoring,  J?.O, Gilbert, Van Nostrand,
          1987.  Also, detailed discussions of other approaches to handling
          nondetects  can  be found in  Statistical Analysis Of Groundwater
          Monitoring  Data At  RCRA Facilities,   Addendum  To Interim  Final
       1   Guidance (OSWER # EPA/500/R-93/003,  July 1992).

     •    Facilities  that are having  difficulty  implementing  one of the
          approaches described above should establish a frequency of sampling
          and analysis for all constituents of at least once per year.s,


                       5.0   MANAGEMENT OF RESIDUES

Management of residues generated during combustion is an important element in
the operations of facilities that burn nonexempt hazardous wastes.  Because
such.residues may be considered hazardous,  the relevant concepts are similar
to those  discussed in the previous  sections of this guidance.   For this
document, the term "residue" includes bottom ash  generated in the combustion
unit and/or fly ash that is collected in an air pollution control device.  As
is the case  in the generation of any solid waste,  a determination must be
made whether the residues are hazardous wastes.   There are three regulatory
requirements governing residues generated  from  the combustion of hazardous
wastes:        .                   '        .          ;-                 ^

     •    Listed Hazardous Wastes:, All  residues  derived from the combustion
          of  a  listed  hazardous  wastes   remain  listed  wastes  (40  CFR
          §261.3 (c) (2)) [until delisted]  and are subject to the land disposal
          restrictions  (LDR)  requirements codified  in  40  CFR  §268  in
          disposing of such residues.

     •    Characteristic  Wastes:   Residues derived from  the  combustion of
          characteristic  wastes  remain hazardous  unless  they no  longer
          exhibit  any characteristic of a  hazardous waste.   The facility
          should  sample  and  analyze  the residue  to  determine  whether it
          exhibits any of  the characteristics.  Further, if the waste was one
          classified in EPA waste code D002 or D012 through D043 at the point
       '   of  generation,   the residue  must  be  analyzed  for  "underlying
          hazardous constituents,11 as defined at 40 CFR §268.2, that can be
          reasonably expected to have been  present in the  waste at the point
          of generation  (40 CFR  §268.7(a)).  When meinaging such residues,

                      - DRAFT: Do Not Cite or Quote •--

-------
                                     24        '

          facilities also, must comply with LDR requirements set forth at 40
          CFR  §268.9,  "Special  Rules  Regarding  Wastes  That  Exhibit  a
          Characteristic."          ,                     .

     •    Bevill Exemption:   Section 3001(b)(3)(A)  of RCRA exempts certain
          types of  residual  materials (generally "high-volume, low-hazard"
          materials) from regulation under RCRA Subtitle-C; this is commonly
          referred  to  as  the Bevill exemption.   Examples  of Bevill exempt
          material  include,residues generated primarily from the combustion
          of coal  or  other  fossil fuels  and cement  kiln  dust.    The BIF
          regulations define which BIF residues are subject to this exclusion
          from the definition of hazardous waste  (40 CFR §266.112).

5.1       DETERMINATION OF THE BEVILL EXEMPTION

The promulgation of the BIF regulations specifically addressed the issue of
continued applicability of the Bevill exemption when devices burn hazardous
wastes  (40  CFR  §266.112).    This regulation  specifically  states  that "a
residue derived from the burning or processing of hazardous waste in a boiler
or industrial  furnace is not excluded from the  definition  of a  hazardous
waste under  §261.4(b)(4),(7), or  (8)  unless the device -and  the  owner or
operator meet  the requirement  (described below)."  The first requirement
states that the device must be a boiler that burns at  least 50 percent coal
or an  ore or  mineral  furnace or cement kiln that processes at  least 50
percent by weight  normal,  nonhazardous materials.   The second requirement
mandates  testing to  determine  whether  the  residues  have been  affected
significantly by the hazardous waste, thus causing them to no  longer be the
"high-volume, low-hazard" material that the Bevill exemption was intended to
cover.  That  determination  is achieved through, either of two tests that show;

     »    Test  One:   The  waste-derived  residue does  not contain  toxic
          constituents  at  concentrations  significantly  higher  than  is
          exhibited by the residue generated  when hazardous wastes are not
          burned.

     •    Test Two:  The  concentration of toxic constituents does not exceed
          health-based limits identified in the regulation.
        i                              •
The regulation at 40 CFR  §266.112  requires that the waste-derived residue be
sampled and analyzed "as  often as  necessary to  determine whether the residue
generated during each 24-hour period" meets requirements to qualify under the
Bevill exclusion.  However, no specified frequency is  identified for making
such a  determination.   Therefore,  the discussions that follow will focus
primarily on the issue of  frequency of sampling  and analysis as it affects
facilities that attempt to claim the Bevill exemption.
                      - DRAPTs Do Not Cite Or Quote -

-------
                  ,.        .           25               :    - . "            '

5.2       FREQUENCY OF SAMPLING AND ANALYSIS

The first step a facility may take in determining the frequency of sampling
and analysis is to develop a WAP that addresses the frequency of sampling and
analysis  of the  residues and  management  practices for  disposal  of  the
residues or to  include those subjects in  its overall WAP for the facility.
Two options a  facility  may consider when selecting  a frequency,  which are
similar to the options described earlier,  are sampling and analysis by batch
and  sampling  at  a  reduced  frequency,  with statistical  analysis.    Both
approaches may be appropriate for residues generated from the combustion of
characteristic wastes.

     *    Sampling and analysis  by batch:  Since the regulation governing the
          Bevill  exemption   requires  that  sampling  and   analysis   be
          representative of residues generated during a 24-hour period, daily
          sampling and analysis is  acceptable for a  batch frequency.   The
          results of  such analysis  then should be compared with the limits
          established through either of the two tests described earlier.

     «    Statistical Analysis;   If  a facility chooses* to sample and analyze
          less  frequently than  daily, the  facility  should be  prepared to
          provide a technical  justification of the  appropriateness  of the
          lesser frequency and an explanation of how  the  results of analysis
          represent the  24-hour  periods during  which  residues were  not
          sampled to determine eligibility for the Bevill exemption.   Using
          the.methods of  statistical analysis described earlier, a facility
          might  be  able  to establish  that  a less  frequent  sampling  is
          adequate.    The  facility should  consider that,  when sampling at a
          reduced frequency  based on statistical analysis, there is  some
          chance that the facility will be out of compliance.

S.S       CONSIDERATIONS

When  determining the  frequency of  sampling  and analysis, several site-
specific factors should be considered.  Some of those factors are discussed
below.

     •    Sampling:   To obtain a representative sample for  a 24-hour period,
          40 CFR §266.ll2(bj (1) (ii) and  (2) (iii) state that one or  more
          samples may be  taken,  provided  that the sampling does not exceed
       :   the 24-hour period. If more than  one sample is taken, the samples
          may be composited or analyzed separately.  The regulations do not
          specify the number of  samples that may be taken within the 24-hour
          period.   However,  the  facility  should  specify  in  the  WAP  a
          frequency that will account for  any spatial or  temporal variations
          in the residues.   The location  from which  the sample is taken is
          another  factor that  should  be   considered  in   obtaining  a
          representative  sample.  According to requirements set forth at 40
          CFR  §266.112,  the  residues must  riot conta in  toxic  compounds at
          levels  above  the  limits established  in  either of the  tests

                      - DRAFT: Do Not Cite Or Quote -

-------
                           26

discussed  earlier that  could  reasonably be  attributed to  the
hazardous  waste.   Therefore,  samples should  be collected in  a
manner that minimizes any environmental contamination that is not
attributable  to the  burning, or  processing  of  hazardous  waste.
Samples should  be taken  from a location  as  close  to the residue
outlet(s) as practicable; that location should be identified in the
WAP.   Any sampling'  conducted at a  location  other than  that
specified in the WAP may not be considered valid.

Management of Residues:  Because of the potentially large volumes
of residues generated in  any 24-hour period,  it is possible that a
facility may  have disposed of the residue after a  sample  of  the
residue had been taken but  before  the results of analysis had been
received.  The problem arises when results show that the residue is
a hazardous waste and  the residue is disposed  of in a non-hazardous
disposal  area or  unit.    A similar  problem  becomes especially
significant   for   facilities  conducting  sampling   at  • reduced
frequencies.  As the preamble to the BIF regulations set forth in
the August 27, 1991 BIF Federal Register states that "if the waste-
derived residue is sampled  and analyzed less  often than on a daily
basis, and subsequent analysis  determines that the residue fails
the  test  and is  fully  regulated hazardous  waste,  the  Agency
considers  all residue generated   since  the previous successful
analysis to be fully regulated hazardous waste absent documentation
otherwise."  In addition, residue generated after the failed test
may also  be  considered a hazardous waste  until the  next passing
test. Therefore,  in all of these,scenarios, the facility risks not
only the residue becoming subject to the RCRA  regulations, but also
the disposal  area or unit becoming subject to RCRA Subtitle  C
requirements. Resampling of the residue in the disposal area would
not generally be acceptable, because the area  would not normally be
the appropriate sampling location and the residue found there may
not be  representative of >  the residue  generated over a 24-hour
period.   To minimize the extent to  which disposal areas are subject
to  RCRA  regulations,  a  facility  may  want to  implement certain,
disposal management practices.  For example,  management practices
controlling disposal of residues into a quarry on-site  may include:

     -    Transfer by a dedicated truck for disposal
     -    Disposal in specific segregated locations in the quarry
     -    Documentation of disposal practices and locations   '  ,

«    Other Factors:   Other factors to consider when  selecting  a
     frequency of sampling and analysis include:

     -    Feed rate of wastes
     -    Levels of volatility of metals in the waste   .     '
     -    Physical form  of the waste  (for example,  solid  rather
          than liquid)
     -    Waste feed system

            - DRAFT: Do Not Cite Or Quote -

-------
                                     27       •'..'...'"

               -    Levels and types of organic constituents  in  the waste
                    (for example, difficulty of destruction or formation of
                    by-products)
               -    Levels and types of metals  regulated under RCRA, other
                    than those regulated by the BIF regulations  (for example,
                    selenium)
               -    Changes in feed streams          j    '
               -    Changes in operating conditions or equipment
               -    Operating conditions when sampling compared with those
                    when not sampling
               -    Trends in partitioning of metals in fly as compared with
                    bottom ash
                  5               .       '-"          *

               6.0    DOCUMENTATION TO DEMONSTRATE  COMPLIANCE

Documentation of  compliance  consists of detailed and complete records of a
facility's activities that  are  regulated by  either permit  conditions or
regulatory requirements (e.g., 40 CFR 266.103(j)).   Some conditions governed
by permits can be measured and recorded directly,  while others may require
indirect measurement or calculation.  For example,.real time and continuous
monitoring systems for such stack gases as CO,  O2»  and SO2 are in widespread
use.  However,  no monitors are available for effectives real-time measurement
of metals or total chlorine in the stack gases.  Therefore, compliance with
those limits on emission rates are  demonstrated  indirectly by calculation of
the feed rates of the constituents of interest.

Essentially, compliance with a permit or the regulations is demonstrated by
showing  that  a  facility  burned  wastes  only  under  certain  specified
conditions.   Those conditions  usually are  stated as  maximum conditions,
minimum conditions, or conditions over a. specified  range.  Documentation and
recording .of  those  conditions  provide the  basis  for determinations of
compliance, not only  by personnel  of EPA  or  state agencies but also by the
facility  itself.    The  operators  of  a  facility   should  have a  detailed
understanding of permit or conditions governed by permit or regulation that
determine compliance or noncompliance.  Since no two facilities are identical
in permitted or regulatory operating conditions, no all-encompassing check
list is feasible.   However, the items listed below form a basis  for a logical
and coherent  approach that will enable  a facility  to document combustion
conditions and other parameters required by permit or regulation.  The list
is not all inclusive, and each facility should tailor it to that facility's
own needs.

Facility operators should consider the following:       .              ,

     • .   Specific responsibility for compliance should be assigned.  While
          owners  and  operators  ultimately are  responsible  for compliance,
          they  may not  be   involved  in this  issue  daily. -.  Therefore,  a
          designated individual,  with the necessary number of backups, should


                      - DRAFT: Do Not Cite or Quote -

-------
                           28

be  responsible   for   ensuring  that  activities  necessary  for
compliance are carried out and documented.  (This step is the basis
of all subsequent actions).

A  written  compliance methodology  is  recommended.   It  should
identify the responsibilities  of individuals associated with the
combustion of hazardous waste.   The  document  should.be tailored to
the specifics facility, with emphasis on regular use by operators.
In addition,  it should be, available  to personnel  of EPA  or the
appropriate state agency.  Decision trees and actions to be taken
in the event of failures or other situations  should  be a principal
focus of the document.  Conditions to be satisfied before hazardous
wastes are burned should be set forth in the document.

A WAP that contains the items required by permit or regulation is
vital and mandatory (40 CFR §264/265.13).  The plan, at a minimum,
should  cpver  sampling  locations  and  frequency  of  sampling,
statistical methodology  (if  applicable),  procedures for handling
outliers and  nondetects,  methods of preparation and analysis"of
samples, and QA/QC procedures and should identify the  laboratories
to be used.  The foregoing list is not all-inclusive,  since it may
be necessary to meet other requirements. The  result  of application
of procedures  in  a properly  constructed WAP  is that .the facility
will know what it .burned and when.  As an example, for any day on
which waste is. burned,  corresponding data on waste analysis should
be available that can be linked directly with that day's activity.
The data should be available from whatever method of sampling and
analysis is used.  A facility's lack of this information can lead
to determinations of noricomplianceo

Although not specifically required,  computer hardware  and software
systems can be useful in providing a record of  the most crucial
data.   The systems  can be  tailored to  meet  almost  any  permit
condition or requirement.  They enable the inspection  of operating
records with  relative ease.   Care should be  taken  to  select a
system  that meets  the needs  of the  facility,  can  accommodate
upgrades of software, and facilitates compliance rather than merely
reporting on compliance status.

Recordkeeping  that.documents compliance is  recommended.   Records
should  include,  but  not be limited  to:    waste analysis  data;
records of continuous monitoring of  feed stream rates; statistical
data; data  on  permit or operating  limits, such as temperatures;
parameters for air pollution control devices; and any significant
operating requirements  or constraints.   It  is  important to note
that recordkeeping to  support  the sampling  and analysis strategy
used will provide data on operating conditions and limits that form
the  basis  for  enforcement actions  --for  example,  feed  stream
qualification values or UTLs.  Finally,  if the facility chooses to
change from one compliance strategy to another (for example, from

            - DRAFT: Do Not cite Or Quote -

-------
     '    •       ••;''••           29               '                     :

          sampling and analysis by batch to statistical analysis), complete
          and supportable records should be kept to document the change.  It
          also is  prudent to keep records of compliance in a single location,
          with a remote backup.  This precaution applies not only to computer
          data storage but also to paper records.

Again, these  items  should be merely a starting point  in  the documentation
effort and should be adjusted or expanded to meet the specific needs of the
facility.

The objective of maintaining these items is that the  facility will be able to
demonstrate compliance to regulatory officials  and the public.  For example,-
if  an  inspector has  a-, difficult time  determining that  a  facility  is in
compliance  with  feed rates  for  various  constituents,  the  inspector's
underlying assumption may be that the facility is having the same difficulty.
Such an assumption could  lead inspectors to believe that the facility is out
of  compliance.   Well-maintained  documentation, can   prevent  potentially
unnecessary actions.
                      - DRAFT: Do Hot Cite Or Quote -

-------
                      Appendix A
Table of K Factors foe Calculation of Tolerance Limits

-------
   TABLE 1.  K factors for Calculation of Tolerance Limits for
         95 Percent Confidence and 95 Percent Proportion
5«:*S*}SS£»aiBp.i'B«il3 iJBJB «:₯ %:W^J::-i.5;^g::g:::;::::^^^yv^:: •
• 2 . ' :' '
• 3 '
' - 4 -'
. • • . 5 ' • . •'
6 • .
7
8
.' ' 9
10
11
12
13
14
15
1.6
17
; is
19
20
21
22
23
24
25

•26
7
•5
4
3
3
3
3
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2

.260-
.656
. 144
,203
.708
.398
.187
.031
.911
.815
.736
.670
.614
.566
.524
.486
.458
.423
.396
.371
.349
.328
.303
,.292
From:     Hahn, Gerald  S.  and William Q. Meeker.
          Intervals:  A Guide for Practitioners.,
          (ISBN 0-471-88769-2)'.
 1991.   Statistical
Wiley Interscience.
                 - DRAFT! Do NotlCite Or

-------