EPA/540/2-90/005b

                                  OERR Directive 9285.5-02-2
                                       February 1990
Environmental Asbestos Assessment Manual

 Superfund Method for the Determination of
          Asbestos in Ambient Air
  Part 2: Technical Background Document
               INTERIM VERSION

-------
^efldk '7%e mention &f trad& or commerceprocttiQted$8$:

-------
                                         Contents
Figures	    v
Tables  	,	'.'.'..'.'.'.'.'.'.'.'.'.'.'.'.'.I'.',   vi
Acknowledgements	••...'.	-.......'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.   vii

1. Introduction	      1
2. Overview		       2
3. Background	    4
    3.1. Biological Activity		 .       4
    3.2. Risk Factors	    5
    3.3. Morphology of Asbestos Dusts	    7
    3.4. Published Size Distributions	    8
    3.5. Representative Asbestos Concentrations  	'.'.'.'.'.'.'.'  30
    3.6. Representative Total Dust Concentrations	'.'.'.'.'.'.   39
    3.7. Analytical Techniques Used to Monitor Asbestos	   40
    3.8. Existing TEM Analytical Methods	   42

4.  Risk Assessment Objectives and Method Requirements   	   49
    4.1. Consideration of Biological Activity	 .   49
    4.2. Precision Requirements	   50
    4.3. Sensitivity Requirements   	   52
    4.4. Reporting Requirements	.'.'.'.'.'.'.'.'.'.'.'.'.'.'.   53
    4.5. Method Specifications  	,	'.'.'.'.'.'.'.   54

5.  Overview of Method	   55
    5.1. Sample Collection	   55
    5.2. Sample Preparation  	'.'.'.'.   55
       5.2.1. Indirect TEM Specimen Preparation  	'..'.'.'.'.'.   55
       5.2.2. Direct TEM Specimen Preparation	'.',   55
    5.3. Analysis	   56

6.  Method Components  	   58
    6.1. Factors Affecting Characterization of Asbestos Structures   	   58
       6.1.1. Analytical Technique	   53
       6.1.2. Magnification/Resolution	 ',  '   61
       6.1.3. Rules for Counting, Characterization and Recording   	   62
    6.2. Factors Affecting Sensitivity 	   62
       6.2.1. Filter Loading   	'.'.'.'.'.'.'.'.'.   64
       6.2.2. Sampled Air Volumes  	[   65
       6.2.3. Area to be  Scanned for Analysis	'.'.'.'.'.'.'.'.'.'.'.'.'.   66
       6.2.4. Filter Preparation  	   67
       6.2.5. Existing Analytical Methods	   73
       6.2.6. Analytical Technique	   74
       6.2.7. Magnification/Resolution   	   74
    6.3. Factors  Affecting the Limit of Detection  	'.'.'.'.'.'.'.'.'.'.'.'.   76
       6.3.1. Filter Blank Contamination	   76
           6.3.1.1 Contamination on MCE Filters	'.  .   77
           6.3.1.2. Asbestos Contamination on Polycarbonate Filters   	   78
           6.3.1.3. Estimating Filter Blank Contamination Levels  	   79
       6.3.2. Conclusions	   81
    6.4 Factors Affecting Precision	   81
       6.4.1. Analytical Technique  	'.'.'.'.'.'.'.'.'.'.   81
       6.4.2. Magnification/Resolution   	   82
       6.4.3. Uniformity of Filter Deposit	   83
       6.4.4. Rules for Counting, Characterization, and Recording	  .   83
                                             in

-------
7, Recommended Procedures for Manipulating Analytical Data	   85
    7.1. Computing Average Concentrations from Multiple Samples  .	   86
    7.2. Computing Confidence Limits for Averages of Homogeneous Samples 	   87
    7.3. ComputingConfidence Limits for Averages of Non-Homogeneous Samples  ...   87
    7.4. Testing for Differences Between Concentrations Derived from
           Different Sets of Samples	• -	   88
    7.5. Adjusting Measured Concentrations for Contributions from Analytical
           Background Contamination  .	•  •   89

References   	•	   90
                                              IV

-------
                                              Figures
3.1:

6.1:


6.2:
Components of Respirable Particles

The Relationship Between Analytical Sensitivity, Sampled Air Volume, and the Number of
Grid Openings Counted

The Relative Analytical Sensitivities of Published TEM Methods

-------
                                               Tables
3.1



3.2



3.3



3.4



3.5



3.6
The Fraction of Asbestos Fibers in Total Fibers and Total Dust



Fiber Size Distributions Measured in Various Environments For Several Asbestos Minerals



Fiber Size Distributions Measured in Various Environments For Chrysotile



Published Asbestos Concentrations Representing Background



Published Asbestos Concentrations Representative of Environmental Concentrations



Comparison of Available Asbestos Analytical Techniques
                                                   VI

-------
                                   ACKNOWLEDGEMENTS

       We would like to acknowledge the timely and critical support for this effort provided by Jean
Chesson, Chesson Consulting, Washington D. C. and Kenny Crump, ICF Clement, Ruston, Louisiana.
Kenny Crump also assisted directly with the preparation of the Sections of this document addressing
data manipulation.  Comments and opinions were also solicited from numerous individuals active in the
field of asbestos investigation and analysis.  We would like to thank Dan Baxter, Particle Diagnostics,
Inc., San Diego; Tony Kolk, EMS Laboratories, Inc., Pasadena; and Graham Gibbs , Occupational Health
Services, Government of Alberta, Alberta, Canada for their stimulating discussions. Support for data
presentation was provided by D'Arcy Richardson and Tim Swillinger of ICF Technology.
                                              VII

-------

-------
1.
INTRODUCTION
      A sampling and analysis method for the determination of asbestos in air
is presented in Part 1 of this report, under separate cover.  This method is
designed specifically to provide results suitable for supporting risk
assessments at Superfund sites, although it is applicable to a wide range of
ambient air situations.  Considerations addressed during the development of
the method are presented in this companion technical background document.
Also, in the interest of facilitating the use and interpretation of analytical
results derived from the method presented in Part 1, recommended procedures
for manipulating such data as part of a site evaluation are provided in
Section 7 of this document.

      -Asbestos presents a complex challenge to investigators evaluating risks
at Superfund sites.  Unlike the majority of other chemicals frequently
monitored at hazardous wastes sites, asbestos exposures can not be adequately
characterized by a single concentration parameter.  This is because the
different size ranges of airborne asbestos particles, even when they are of
the same mineral variety, exhibit different dose/response relationships.  Thus
a more accurate characterization of asbestos exposure is that arising from a
family of substances•independently contributing to toxicity rather than that
of exposure to a single chemical.  Therefore, proper characterization of
asbestos exposure requires that the relative contributions from each of the
many components of exposure be defined.    .  .-.

      Existing equipment and methods used to measure asbestos are limited in
their ability to fully characterize asbestos exposures.  In addition,  the
toxicity of asbestos is currently a subject of .scientific debate.
Consequently, monitoring asbestos in a manner that satisfies the needs of a
risk assessment requires innovations that tax the limits of available
technology.  Several variations were considered during development and the
method presented in Part 1 of this report represents a workable compromise
among several technical constraints.

      The purpose for documenting the data and assumptions used to develop the
method proposed in Part 1 is to facilitate critical evaluation while
highlighting the needs for additional research and for better documentation of
existing analytical results.  Considerations addressed in this report that
have been documented in the literature are cited accordingly.   Considerations
that remain largely a subject of conjecture are_also noted.   Due to the
current level of interest and activity provoked by asbestos,  further
improvements in asbestos sampling,  analysis,  and evaluation are anticipated.

-------
2.
OVERVIEW
      Development of the method presented in Part 1 of this report began with
consideration of the kinds of data required to perform a risk assessment.
Because the primary objective is to provide analysis results that reflect
potential health risks, factors contributing to the biological activity of
asbestos were considered to identify specific asbestos exposure parameters
that relate to risk assessment objectives.

      Published risk factors are expressed in terms of airborne concentrations
as determined by phase contrast microscope (PCM) counts.  However, a range of
dimensional parameters have been shown to relate to biological activity in
addition to PCM counts.  Therefore, to address persisting controversies in the
interpretation of asbestos biological activity, several parameters were
selected for characterization by the method presented in this report.  To
maximize flexibility, sampling and analysis results will be recorded in this
method so as to allow for re-interpretation of results without the need to re-
evaluate the original sample specimens.

      The characteristics of the kinds of environmental samples likely to be
collected and analyzed by the proposed method were considered in order to
identify the sampling and analysis criteria, which are required to satisfy the
objectives of a risk assessment.  The morphologies of asbestos structures and
total dusts typically found in these samples were considered in order to
determine requirements for distinguishing and characterizing components that
potentially relate to biological activity.

      To determine the required level of analytical sensitivity for the
method, concentrations typical of environmental samples were estimated from
the limited published data.  A range of typical concentrations was defined
from measured background and concentrations expected to be observed in the
immediate vicinity of asbestos sources.   Precision criteria defined for the
method were developed by considering the requirements for delineating spatial
and temporal trends in environmental asbestos concentrations as they apply to
risk assessment.

      After the dimensional parameters to be characterized had been
established and the criteria for the sensitivity and precision of the method
had been defined, the available sampling and analysis technologies were
evaluated to determine what combinations would be capable of providing the
required information.  Because published risk factors are expressed in terms
of PCM counts, phase contrast microscopy was considered for use as the
analysis technique but this was rejected for a variety of severe, inherent
limitations.

      Despite the relationship between PCM measurements and existing risk
factors, measurement of the PCM equivalent fraction of asbestos in
environmental samples has been shown to be less  important for assessing risks

-------
than characterizing fractions of asbestos that are more directly associated
with biological activity (Chesson et al 1989a).   Therefore, transmission
electron microscopy (TEM) was selected as the appropriate tool for analysis of
environmental samples.

      Because sensitivity and precision are functions both of sample
characteristics and of rules for analysis, alternate combinations of sampling
procedures, sample preparation techniques, and analysis methods were evaluated
to identify the most cost-effective combination capable of providing the kinds
of analytical results necessary to support a risk assessment.  Published TEM
analytical methods were evaluated for their applicability to this problem.
Combined with several specific modifications,  many of the features of the
published analytical methods were incorporated into the method developed in
this report.  For sampling, alternate types of collection filters and the
degree of filter loading were addressed.  For TEM specimen preparation, both a
direct and an indirect preparation technique were considered.  For analysis,,
the area of the filter to be scanned was defined as a function of sample
loading and the required analytical sensitivity.  In addition, counting rules
and recording rules were modified to assure that data would be preserved in a
manner allowing detailed re-interpretation after analysis was completed.
Finally, a two-phased, approach for sample collection and analysis was adopted
for cost-effectiveness.

-------
3.
BACKGROUND
      Available data from published studies of asbestos exposure and
biological activity were reviewed to define method requirements as they relate
to risk assessment objectives.  Existing sampling and analysis technologies
were considered to identify an approach that best satisfies the method
requirements.

      3.1.  BIOLOGICAL ACTIVITY

      It is generally recognized that health risks posed by exposure to
asbestos dusts depend predominantly on the concentrations and physical
dimensions of the individual fibrous structures in the dust.  Fibrous
structures may include fibers, bundles, clusters..or matrices because airborne
asbestos is often found as a mix of such complex structures in addition to
single fibers.  Asbestos structures may also be found aggregated with equant
(non-asbestos) particles.  Somewhat arbitrarily, fibrous structures have been
defined as those exhibiting aspect ratios greater than 3:1 to distinguish them
from isometric particles (Walton, 1982).  Isometric particles have not been
shown to exhibit the same types of biological activity as fibrous structures
(Elmes, 1982).  However, a cutoff in aspect ratio below which biological
activity can be considered insignificant has not been formally established.

      Although it has been shown qualitatively that a relationship exists
between the physical dimensions of fibrous structures and biological activity,
the form of the relationship appears to be complex (see for example: USEPA
1986a).  The relationship between certain types of structures and biological
activity has been investigated in a series of animal studies.  Based on these
studies, asbestos toxicity varies directly with the length.and inversely with
the width  (thickness) of asbestos fibers.  Thus, the longest and thinnest
fibers tend to be the most potent.  However, there has not been general
agreement  on a minimum length below which the biological activity of asbestos
can be considered insignificant (USEPA 1986a).  Therefore, the exact shape of
the relationship between length, diameter, and potency is still a subject of
controversy.

      Unlike fibers, the biological activities of bundles, clusters, and
matrices have not been investigated directly.  Some investigators (see for
example: Nicholson, 1988) believe that biological activity correlates best
with the mass of the structure.  However, other investigators do not share
this view  (see for example: Bertrund and Pezerat, 1980).  Although, various
individuals have proposed theories concerning the relationship between
aggregates (bundles, clusters, and matrices) and biological activity, due to
the lack of data, such theories currently remain in the realm of conjecture.

      Several studies  (see for example: Bonneau et al, 1986) indicate that the
mineralogy of a structure also plays a role in determining the biological
activity of the structure.  The majority of studies relating biological
activity to physical dimensions indicate that structures bearing appropriate
dimensions all tend to exhibit similar biological activity provided that the

-------
structures are durable in vivo (See for example: USEPA 1986a).  In contrast,
the Bonneau paper is one of a small number of papers that reports that
structures with identical dimensions that are composed of different minerals
exhibit different levels of biological activity.

      Based on a review of the published studies, it is apparent that the
characteristics of an asbestos dust that best relate to biological activity
are still a subject of scientific debate.  The principle outstanding issues
are:

      (a)   whether the total structure count adequately tracks the biological
            activity of asbestos or whether a specified, minimum length can be
            defined- below which contributions to biological activity may be
            considered unimportant;

      (b)   whether asbestos aggregates (bundles, clusters, and matrices)
            should be counted as single entities or weighted in proportion to
            the number of individual fibers present in the aggregate to
            properly represent their contributions to risks;

      (c)   whether biological activity correlates with the overall mass of an
            asbestos structure;

      (d)   whether durable, non-asbestos fibers with similar morphologies to
            asbestos are biologically active;

      (e)   whether to employ different criteria for different asbestos
           -mineral types or for different exposure .circumstances (eg.
            exposure to mine tailings verses exposure .to textile wastes,
            etc.).

Ideally, the" results of an analytical method should be sufficiently flexible
to allow for interpretation based on any of the prevailing theories concerning
biological activity so that such results will retain validity even as the
understanding of asbestos dose/response relationships matures.

      3.2.  RISK FACTORS

      Although animal studies provide an indication of the qualitative
relationship between physical characteristics and potency,  they are not useful
for quantifying dose/response relationships for humans.  .Such risk factors
have been derived for asbestos from existing epidemiological studies in which
the exposures of the cohorts evaluated are based on a combination of
analytical methods including,  primarily,  midget impinger and phase contrast
microscopy (USEPA, 1986a).   However,  phase contrast microscopy (PCM) and
midget impinger measurements provide  only crude indices  of exposure and do not
necessarily track characteristics representing the biological activity of
asbestos.

-------
      The traditional methods of characterizing exposure used in existing
epidemiology studies have not proven adequate to develop a dose/response
relationship that uniformly applies under all exposure circumstances.
Published risk factors vary by a factor of 50 depending on the specific
exposure setting studied (USEPA, 1986a).  This may be due to one or a
combination of several factors that include errors and uncertainty in the
quantification of mortality, errors and uncertainty in the quantification of
exposure, differences in the size distributions of structures, and differences
in the way that the various analytical techniques respond to dusts containing
varied structure size distributions.  Due to these limitations, such risk
factors should be viewed as order-of-magnitude estimates at best.  Despite
such limitations, however, published risk factors (properly modified by
considering other factors that have been shown to define biological activity)
represent the only reasonable tools currently available for estimating,risks
from measured asbestos exposures.

      Published risk factors are generally expressed in terms of PCM counts.
When data from different types of analytical techniques were combined in the
existing epidemiology studies, results were generally normalized and converted
to PCM equivalent counts before completing the evaluation.  However, despite
the fact that published risk factors are generally expressed in terms of PCM
equivalent counts, quantification of the PCM visible fraction of asbestos in
current measurements does not provide the best index for comparing current
asbestos measurements to existing risk factors.                          ...;:,.

      It has been shown that the use of identical methods for measuring
asbestos exposure in two different settings (such as two different factories)
is not sufficient to assure the direct comparability of the two measurements,
at least in terms of comparing risks (Chesson et al 1989a).  Measurements from
two different exposure settings, where the characteristics of asbestos dusts
may vary, only indicate relative risk when the characteristics of asbestos
that determine biological activity are measured directly.  Other measures of
asbestos exposure that do not relate directly to biological activity may not
remain proportional to the characteristics that determine biological activity
in different exposure settings.  Therefore, PCM counts and other measures of
exposure that do not relate directly to biological activity do not provide
results from different exposure settings that are proportional to risk.

      The best approach for monitoring current asbestos exposures to estimate
risks is to measure the characteristics that relate directly to biological
activity and to compare the results to existing risk factors using adjustment
factors that relate PCM counts from the historical studies to proper exposure
characteristics representative of the historical exposures.  Such adjustment
factors may be estimated from published asbestos characterizations determined
using transmission electron microscopy  (TEM) by pairing characterizations from
specific exposure settings with existing risk factors derived from similar
exposure settings.  Alternately, exposure settings (selected for their
similarity to settings originally studied to derive the existing risk  factors)
could be re-characterized using the method presented in this report, to provide
improved adjustment factors.

-------
      The approach recommended in the last paragraph for estimating risks
presupposes a concise definition of asbestos characteristics that relate to
biological activity.  Because the definition of such characteristics remains a
subject of controversy, the method presented in this report is designed to
retain sufficient information from each analysis to span the range of
characteristics currently considered likely to impact biological activity.
Thus, results may be easily re-evaluated in the future.

      3.3.  MORPHOLOGY OF ASBESTOS DUSTS

      The size distributions of asbestos structures are not the only
characteristics of asbestos exposure that are potentially critical to
biological activity.  This is because asbestos structures will always be
accompanied by other materials as components of dust, even in the workplace.

      The major components of dust in all environments (occupational and
environmental) are nonfibrous, isometric particles.  Fibrous structures
consistently represent a minor fraction of total dust.  In addition, even
among fibrous structures, asbestos represents a variable fraction of the total
present.  The relationship between these fractions is depicted in Figure 3.1.

      Figure 3,1 is a Venn diagram depicting the universe of particles present
in dust.  Fibrous structures represent a minor fraction of total dust.  Long
fibrous structures represent a subset of total fibrous structures.  Optically
visible fibrous structures, those that are visible using a phase contrast
microscope (optical microscope),  represent a subset of long fibrous
structures.  The importance of the optical fraction of asbestos dusts is
addressed in Section 3.2.  Although not depicted, each of the subsets of
fibrous structures could potentially be further subdivided into individual
fibers, bundles, clusters, and/or matrices.

      Particles composed of asbestiform minerals represent an independent
subset of total particles that is distinct from fibrous materials.  The
overlap between asbestiform minerals and fibrous structures represents the
fraction of fibrous structures that are asbestos.  Correspondingly, subsets of
long, fibrous structures and optically visible, fibrous structures (the PCM
equivalent fraction) are also composed of asbestos.

      The large square in Figure 3.1 represents the fraction of respirable
particles.  Subsets of asbestiform particles,  and the three fractions of
fibrous structures are also respirable.   The large circle at the bottom of the
diagram represents the fraction of particles traditionally counted by midget
impinger.  Note the limited overlap with fibrous structures.   The shaded area
of Figure 3.1 represents the fraction of respirable structures that are
fibrous.  The cross-hatched area represents the fraction of asbestos dusts
(the long structures) currently believed to be the most biologically active.

      Figure 3.1 is a qualitative representation.  The relative size and
orientation of the various fractions depicted will change as a function of the
occupational or environmental setting considered.  For example,  the overlap

-------
between asbestiform structures and fibrous structures is expected to be much
greater in asbestos occupational settings than in environmental settings.  In
this manner, Figure 3.1 is probably more representative of an environmental
setting.

      Only a very limited number of available studies address the issue of the
fraction of dust particles composed of asbestos.   This is because in general,
either the analytical technique used in a study has been incapable of
distinguishing asbestos from non-asbestos or, if the analytical technique was '
capable of distinguishing asbestos from non-asbestos, only asbestos particles
are traditionally characterized.  Table 3.1 presents data from the few studies
where the fraction of asbestos particles in dust was considered.

      Unfortunately, much of the data in Table 3.1 provide an indication of
the fraction of total particles composed of asbestos, while it is the fraction
of fibers composed of asbestos that is of principle interest.  In some
exposure settings, a measurement of the fraction of asbestos in total
particles may simply reflect the fraction of total fibers in total particles
with no manner of distinguishing asbestos from non=asbestos fibers.

      The data presented in the table from Cherrie et al (1987) and from
Altree-Williams and Preston (1985) indicate that, while the vast majority of
fibers  in an asbestos (textile) factory may be composed of asbestos (which may
be representative of occupational settings), the fraction of fibers composed
of asbestos in other exposure settings (including environmental settings) may
vary over a wide range.

      3.4.  PUBLISHED SIZE DISTRIBUTIONS

      In any exposure setting, airborne asbestos exists as a series of simple
fibers  and complex structures of varying length, width, and breadth.
Historically, available analytical techniques were capable of detecting  only a
fraction of the population of asbestos structures existing in an exposure
setting.  Thus, only an'index of exposure could be measured.

      In the last 15 or 20 years, measurements with  the superior resolving
power and magnification of the transmission electron microscope  (TEM) have
allowed complete characterization of the distribution of sizes and
morphologies of asbestos structures that exist within an exposure population.
Comprehensive characterization of asbestos  exposures by scanning electron
microscope  (SEM) have also been reported, although the visibility of SEM
 (resulting  from the combined constraints on resolution and contrast) is  not
generally sufficient to detect the smallest and  finest -asbestos  structures
 (Walton 1982).

      Table 3.2 presents published asbestos size distributions  that were
characterized by electron microscopy.  The  first column of the  table lists the
exposure setting in which the distribution  was measured.  The second column
provides the type of asbestos monitored.  The next several columns provide

-------
                   FIGURE 3.1:  COMPONENTS OF RESPIRABLE PARTICLES
Particles Composed of
 Asbestiform Minerals
Particles Counted by
 Midget Impinger  .
 Total Particles
                                                           Fibrous Structures (aspect ratio > 3:1)
                                                                Long Fibrous Structures
                                                                 (length > 5 microns)
                                                                  PCM Visible Fibrous Structures
                                                                     (diameter > 0.25 microns)
                                                                        Respirable Particles
                                                                             Respirable Fibers
                                                                              Biologically-Active
                                                                              Fraction of Asbestos
                                                                                 D. Wayne Berman
                                                                                 ICF Technology
                                                                                 November 3,1988

-------
TABLE 3.1: PUBLISHED MEASUREMENTS OF THE FRACTION OF ASBESTOS IN DUST

Exposure Setting


Cape/Mine
Cape/Mill
Iran/Mine

Tran/Miil

Mine
Mill
Mine
Mill
Txt
Txt/Out
Clearance
I
Bldg w/Insl

Urban
Bldg

Fiber Type


Crc
Crc
Crc

Crc

Ams
Ams
Chr
Chr
Chr&Crc
PChr
PAms

PAms

No re
Chr

Analytical Method


PCM
&
K
&

TP
X Jt
1
1
1
1
1
1
TEM
1
1
1
1
1
1
1


Preparation Technique












D/M
1
1
1
1
1

1


Fraction of Asbestos in Shape Categories
(% by count)
of total particles of total fibers of optical fibers
41
51
28

59

33
55
45
52
•
•














82
21
48

15
1

	
o











94
41
59

36

0
20

Reference


ฉ









ฉ








-------
                                                      Table 3.1 (continued)
                                                               Key
Exposure Settings:
    AC      Dust from asbestos/cement sanding
    AP      Asbestos production plant'
    ATB    Dust from asbestos tape and board
    EC      Dust from a brake and clutch plant
    Bag     Dust from bagging
    Cape    Capetown, South Africa
    Cstr     Dust from construction
    CP      Asbestos/cement pipe manufacture
    Card    Dust from carding
    Crsh     Dust from crushing
    Cut      Dust from cutting
    Dump   Dust from raw material dumping
    Dry      Dust from drying
    Fin      Dust from fiber finishing
    FP      Friction products plant
    Fprp     Dust from fiber preparation
    Form    Dust from forming
    Grnd    Dust from grinding
    Insl      Site where insulation is applied
    Mill     Dust from asbestos milling
    Mine      Dust from asbestos mining
    Mix       Dust from fiber mixing
    Orest      Dust from ore storage
    Out       Outside Plant
    Pirtsl      Pipe insulation manufacturers
    Que       Quebec, Canada
    RR        Railroad factory
    Spec       Nonstandard clearance, etc.
    Spin       Dust from fiber spinning
    Tran       Transvaal, South Africa
    Twst      Dust from fiber twisting
    Txt        Dust from textile plant
    UICC      Sample of UICC standards
    Weav      Dust from weaving
Fiber Types:
    Actn
    Amph
    Ams
    Chr
    Crc
    Pr
Actrnolite
Non-amosite amphibole
Amosite
Chrysotile
Crocidolite
Predominantly    .  ..
                              Preparation Techniques:
                                   D      Direct
                                    I      Indirect
                                   M      Mixed cellulose ester filter
                                   N      Polycarbonate filter
                                   Spec    Nonstandard preparation
                               Analytical Methods:
                                   PCM    Phase contrast microscopy
                                   K       Konimeter
                                   MI     Midget impinger
                                   SEM    Scanning electron microscopy
                                   TEM    Transmission electron microscopy
                                   TP     Thermal precipitator
                               Miscellaneous:
                                  ()  Extrapolated from data
                                  (T)   Total fibers

-------
I
                                                                 TABIE 3.1   fc
                      REFERENCES
                      1.   Gibbs and du Toit,  1979
                      2.   derrie et al.,  1987
                      3.   Altree-Williams  and Preston, 1985

-------
 information on  the  techniques used  in the study  to  generate  the  size  data.
 These columns list,  respectively, the method used to prepare the samples,  the
 type of microscope  used  to  determine the size distribution,  and  the
 magnification employed.  The next two columns present  reported median lengths
 and median diameters of  the fibers  in the distribution characterized.

      Size distributions are presented  in Table  3.2 as the number percent  of
 total asbestos  structures for structures in each of a  series of  size  ranges
 that are listed in  the headings of  the  next several columns.  The column
 headed "percent with lengths less than  5 pm" represents  the  fraction  of short
 structures detected within  the particular distribution listed.   The next two
 columns list structures with lengths greater than 5 urn.  The second of these
 columns presents the fraction of structures with diameters (widths) greater
 than 0.25 /*m, which represent the fraction of PCM equivalent structures
 detected.  Since 0.25 fj.m is the limit of resolution generally quoted  for phase
 contrast microscopy (PCM),  this column  represents the  fraction of structures
 that would be detected by optical microscopy1.   The first of the  two columns
 representing structures longer than 5 /zm lists structures that are too thin to
 be resolved by  optical microscopy:  the  long, thin structures.  The sum of  PCM
 equivalent structures and long, thin structures  represent the fraction of
 structures defined  as long  structures.

      The four  columns to the right of  long, thick  structures on Table 3.2,
 provide number  percent fractions for two other ranges  of length  within the
 component represented by long structures.  These are shown because it  is
 possible that the most biologically active asbestos structures may be  more
 restricted than indicated in Section 3.1 and 3.2 and Figure  3.1.  Each length
 range is also subdivided into those resolvable and  those unresolvable  by
 optical microscopy  (PCM).

      Because chrysotile dusts have been characterized most  frequently among
 the published fiber size distributions,  the data for chrysotile  (grouped by
 exposure setting) are presented in  Table 3.3.   Table 3.3 includes size
 distributions for dusts characterized both as "chrysotile" and as
 "predominantly  chrysotile", that includes approximately half of  all of the
 distributions presented in  Table 3.2.  The format of Table 3.3 is identical to
 that used in Table  3.2.

      Although  data are compressed  and extrapolated in Table 3.3.(and  Table
 3.2) so that results from each study could be presented in a common format,
 the gross features of chrysotile size distributions can be summarized based on
 the detailed results of the studies listed.   Where  such studies provide
 sufficient detail,  similar  features are found for all  chrysotile  dusts
       Note that the limit of resolution of PCM depends on a variety factors
including the type and the condition of the instrument.  The typical ranges of
resolution quoted for PCM is between 0.2 and 0.4 pm.  The impact of such
variation on the interpretation of results is considered in later sections.
                                      13

-------
TABLE 3.2: STRUCTURE SIZE DISTRIBUTIONS MEASURED IN VARIOUS ENVIRONMENTS FOR SEVERAL ASBESTOS MINERALS

;
Cape/Mine
Cape/Bag
Tran/Mine
Iran/Bag
Que/Mine
Que/Bag
AP/Dump
Crc
Crc
Ams
Ams
Chr
Chr
Ams
j
Ins/App JAms
AP/Dump
Mill/Dry
Mill/Bag
Txt/Card
Cape/Mine
Crc
Chr
Chr
Chr
Crc
I \
Cape/Orest i I
:: •
i 1
Mill/Crsh | 1

Mill/Bag

AP/Dump
1
1

1
: .
AP/Mix ! 1
PCM
&
TEM
1
1
1
1
1
SEM
1
1
1
1
|
1
1
1
TEM
&
<5FM

&

PCM
1
I
1
i i
i i • 1
AP/Cut l!) |
D/N
1




D











JN






D/M,N
1
1
1
i
1
1
1

1

1
_





2500

2500
6000
6000
6000
6000
400
to
15,000
l
I
1
1
1

1

1
0.95
1.16
1.83
2.53
0.34
0.55
3.9

2.5
2.5
1.25
1.35
1.0
0.95
1.07

1.00

1.16

0.91

1.25

0.08
0.08
0.26
0.26
0.06
0.06
0.41

0.37
0.25
0.17
0.16
0.15
0.07
0.10

0.08

0.09

0.08

0.10

0.50 1 0.04
/Fraction of Asbestos Fibers in Size Categories (% by Count)
/ JD < 0.25pm
96 | 3.72
93 1 5.88
5
88 ! 3.85
77 \ 5.56
99 | 0.40
96 ! 2.06
54 | (3)
1
69 j (2)
79
88.9
95.2
97.7
95.9
93.6

96.4

92.9

98.1

93.9

.99
3
1.5
•i
0.40
3.33
—

—

5.89

1.41
L >5pm
D> 0.25pm
0.7
1.2
8.33
17.8
0.2
1.72
(43)

(29)
(18)
10
(4)
(2.1)
0.71
-

-

1.21

0.44

1

-
L5-10rtn
D< 0.25pm
3.02
5.1
3.42
4.42
0.36
1.56
(3)

(1.5)
(3)
(1.5)
-
(0.4)
2.57
—

_

5.11

1.27

-

-
L5-10pm
D> 0.25pm
0.57
1.09
7.03
12.6
0.7
1.31
(27)

(23)
(14)
(7)
(3)
(D
056
—

_

1.1

0.38

-

-
L>10pm
0.7
0.78
0.43
1.14
0.04
0.5
• -

-
_
-
.(D
-
0.76
L>10ttn
D>0.25/im
0.13
0.11
1.30
5.2
0.13
0.41
(16)

(6)
(4)
(3)
(D
(D
0.15
~ i i . ~

— • —

0.78 0.11

0.14 0.06

— —

-
Reference
ฉ





(2)






ฉ












-------
TABLE 3.2 (CONTINUED)
/
If
Special
Special
Special

UICC
Cape/Mine

Cape/Mill

Iran/Mine


Iran/Mill


Mine

Mill

Mine
Mill
Txt/Fprp

Txt/Spin


FP/Mix

FP/Grad
CP/Mix
CP/Fin

// 11*111

I Ams( PCM
j &
' SEM
1 '
1
1
Crc
PCM
•
Crc
&
•
Crc KON


Crc


Ams

Ams

Chr
Chr
Chr
l
i

i
1
l
1
1
1
1

i

1

l
1
1


i
1
PCM

&


TEM
I
1
1
1

D/Spec
l
1
1


—


























400
to
10,000


~_















































1.4

1.0


0.9

0.8
0.9
0.7

































V
-ฐ /

21-40
57
50

77
88

84

82



Fraction of Asbestos Fibers in Size Categories (% by Count)
L >5pm
D< 0.25pm
L >5pm lL5-10ttn
D>0.25/im |D<0.25A>m
52>81
i
sb
\
i
l






73 |


91
'
65

93
87
96

98


98

98
98
99













.








L5-10pm
ID > 0.25 Mm
26*37
I6

\
2J1
|





•:

J
>




























j






'


\
















:
L>10/ซm iL>10i'm
D< 0.25pm !D>0.25/im
26J44
J7
24






























^




























Reference
0




ฉ
^*~-l^














ฉ











-------
TABLE 3.2 (CONTINUED)
/I/

Txt/Fprep

Txt/Twst

Txt/Weav

FP/Mix

FP/Form

FP/Fin

CP/Mix

CP/Form
CP/Fin
Pinsl/Mix
Pinsl/Form

Pinsl/Fin

Que/
Out/Mine



BC

>RR/ATB „
RR/AC

! I
Chr PCM D/M
1 o
1
TEM
1 1

1
1
1

1
i



1
i i
i
i
Ams
Ams

Ams

Chr

Chr


Chr
I
I























TEM

TEM


PCM

, &
TEM
1





































I

D


D/M
1





900

to
17,000






















500

to
10,000






























































/ Fraction of Asbestos Fibers in Size Categories (% by Count)
*/
/ L<5/ซm L >5pm
/ D< 0.25m

79 7.9

83.4 8.0

81.2 9.0

90.4 [ 6.3
I
90.7
5.9
;
83.8

63

87.6 6.0

90.7 6.4
95.0 3.0
59.2 j 8.9
63.2 ] 9.4
;
65.1

92.86

84.54


44

75
78
6.6

7.14

3.54


(<6.1)

(<2.5)
L >5wn
D>0.25/im

12.8

8.6

9.7

3.4

3.4

9.9

6.4

2.9
2.1
32.0
27.4

28.4
L5-10ซn
D< 0.25 Mm

4.5

3.6

4.9

2.9

3.2

4.2

3.3

3.8
2.0
5.9
6.1

6.1
i
0.00

11.92


(49.9)
D> 0.25pm !D<0.25/ปm JD>0,25Mm
•
5.7 } 3.4 7.1

2.9

3.1

1.2

1.5

3.4

3.1

1.4
0.8
.15.0
11.9

14.4







•
(22.5)
r







(<2.4) (19.6)

4.3 5.7

4.2 6.6

3.4 2.2

2.8

2.0

2.0 6.5
1
2.7 3.3


2.5 1.6
i
0.9
2.9
3.2

0.4










1.2
16.9
15.5

14.0










Reference






















ฉ
—


^^^
GD




-------
TABLE 3.2 (CONTINUED)
JtUJw/lm
Cstr

Bag
Cut

Chr
l
1
1
1
Cut | j
Mill/Crsh \ \
Mill/Crsh
Mill/Crsh
Crsh
•
CP/Mix
Clearance

Clearance
1
1
I
1

1
PCM

&
STEM
l


:
i



PAms;
i
PAmd
Insl/Rmvl jPChr
Insl/Rmvl
:

PAms:
Mine JAmph
Mine JAmpl:
Crsh Actn
Crsh j Actn

1 .
\
I























D/M




















I




























































































•/
I/
7 I L<5/ซm
86.2

94.3
53.4

87.5
91.2
97.9
88.9
83.1

97.8
68.9

65.2
64.2
66.1
96.9
85.3
87.4
87.6



j
Fraction of Asbestos Fibers in Size Categories (% by Count)
;L >5pm >L >5^m
sD<0.25wn lD>0.25Mm
8.7
j
1.9
13.3
i
10.2
2.0
0.2
4.6
4.8

0.2
11.7

11.6
10.2
12.7
0
0
0
0


5.1

3.8
33.3

2.4
6.8
1.9
6.5
12.1

2.0
19.5
L5-10m
iD<0.25/im














23.2
25.6
21.2
3.1
14.7
12.6
12.4







|

L5-10/ปm iL>10/ซm
D>0.25/im JD<0.25jfm 	
L>10Aซm
D>0.25um







.













f
i
ฃ {
















j















•' } '•
Reference


























-------
   TABLE 3.2 (CONTINUED)
///

Laboratory jChrB! SEM
I i
1
|
1
i
1
i
:



Txt
Txt
Txt/Out
Txt/Out
Clearance
Clearance

Bldg w/Insl
t
ChrBJ TEM
Amsl SEM
Amsj TEM
i
Crc | SEM
Crc j TEM
Trmj SEM

Trm 1 TEM
;
Chr ! SEM
& ! ^,
prซ J TEM
v-iL' J
PChr| SEM
PChr! TEM
PAms] SEM
PAmsj TEM
i
PAms; SEM
1
Bldgw/Insl|PAms| TEM
Urban
Urban



None] SEM
None! TEM
i :
:- :
1/M
I
1
1
1
1
1
1
i
1
1

D/M
1





























































































/Fraction of Asbestos Fibers in Size Categories (% by Count)
/L<5pm
69

89
74
84
88
93
50

59

78 (T)
60 (T)
81 (T)
—
66 (T)
82 (T)

63 (T)
(91)(T)
69 (T)



, >5pm
D< 0.25pm





















L>5pm fL5-10pm
D> 0.25pm JD < 0.25pm


1


























L5-10pm
D> 0.25pm














:
:












.
j
i i
i 1

L>10pm
D< 0.25pm
























5 > 0.25pm
























: : i 1: ^
Reference
ฉ
























oo

-------
TABLE 3.2 (CONTINUED)
/ // 11*111

Mine Amsi PCM
Crsh

Bag
Assay
Mine
BC
Mill
Mine

Mill
l

1
1
Chr
l
1
1
Crc

Crc
Mill Chr
l
Mill i
Mill |
Mill
BC/Card
&

1
1
1
1
1
1.
1

PCM
&
TEM
1 I
1 i
PCnr i
i '
BC | 1
CP 1 |
CP
! i
Txt I |
Txt 1 '
Txt 1 !






















10,000
1
l
1
I
1
1
1
1
1
I
1
























































/ Fraction of Asbestos Fibers in Size Categories (% by Count)
ฃ 1
SI
7 / L<5pm
35.5
A ฃ A
46.4
55.8
66.1
56.6
45.5
61.4
77.1

63.7
93.7
96.1
98.5
98.6
98.7
97.1
98.6
97.5
87.5
94.4
93.7
L >5^m !L>5ซn 5L5-10rtn
D<0.25/im lD>0.25/Jm JD<0.25/im

















S

(6.2)
(3.9)










\
(0.11)
(0.02)
(1.5) 1 (0.01)
\
(1.4)
(1.3)
(2.9)
(1.4)
(2.5)
(12.4)
(0.06)
(0.04)
(0.05)
(0.03)
(0.06)
(0.1)
'







LS-10/im
D> 0.25 fim












L>10/ปm |;L>10/'m
D<0.25/im D>0.25^m
|
,

\





\



(5.5) (0.08)
(6.2) (0.1) 1










\

Reference
ฉ










ฉ











-------
                                                      Table 3.2 (continued)
                                                              Key
Exposure Settings:
    AC      Dust from asbestos/cement sanding
    AP      Asbestos production plant
    ATB    Dust from asbestos tape and board
    BC      Dust from a brake and clutch plant
    Bag     Dust from bagging
    Cape    Capetown, South Africa
    Csu-     Dust from construction
    CP      Asbestos/cement pipe manufacture
    Card    Dustfrom carding
    Crsh    Dust from crushing
    Cut     Dust from cutting
    Dump   Dust from raw material dumping
    Dry     Dust from drying
    Fin      Dust from fiber finishing
    FP      Friction products plant
    Fprp    Dust from fiber preparation
    Form    Dust from forming
    Grnd"  Dust from grinding
    Insl     Site where insulation is applied
    Mill    Dust from asbestos milling
   Mine       Dust from asbestos mining
   Mix        Dust from fiber mixing
   Orest       Dust from ore storage
   Out        Outside Plant
   Pinsl       Pipe insulation manufacturers
   Que        Quebec, Canada
   RR        Railroad factory
   Spec        Nonstandard clearance, etc.
   Spin        Dust from fiber spinning
   Tran       Transvaal, South Africa
   Twst       Dust from fiber twisting
   Txt        Dust from textile plant
   UICC      Sample of UICC standards
   Weav      Dust from weaving
Fiber Types:
    Actn
    Amph
    Ams
    Chr
    Crc
    Pr
Actinolite
Non-amosite amphibole
Amosite
Chrysotile
Crocidolite
Predominantly
                             Preparation Techniques:
                                  D     Direct
                                   I     Indirect
                                  M     Mixed cellulose ester filter
                                  N     Polycarbonate filter
                                  Spec   Nonstandard preparation
                              Analytical Methods:
                                  PCM    Phase contrast microscopy
                                  K      Konimeter
                                  MI     Midget impinger
                                  SEM    Scanning electron microscopy
                                  TEM    Transmission electron microscopy
                                  TP     Thermal precipitator
                              Miscellaneous:
                                  ()   Extrapolated from data
                                  (T)   Total fibers

-------
REFERENCES -
1.  Gilฑ>s and Hwang,  1980.
2.  Gibbs and Hwang,  1975.
3.  Hwang and Gibbs,  1981.
4.  Beckett and Jarvis, 1979.
5.  Gibbs and du Toit, 1979.
6.  Lynch et al., 1970.
7..  Dement and Harris, 1979.
8.  Sebastien et al., 1984.
9.  Marconi et al., 1984.
10. Snyder et al., 1987.
11. Cherrie et al., 1987.
12. Rendall and Skikne, 1980.
13. Winer and Cossette, 1979.

-------
analyzed.  Briefly, the distribution of structure lengths is unimodal and
skewed with the mode occurring between 0.8 and 1.2 /^m.   The tail of the
distribution extends out so that longer structures are present,  but at
decreasing frequencies.  Structures longer than 5 //m constitute no more than
25% of the total and frequently constitute less than 5%.  Structures longer
than 10 nm constitute no more than half of the concentration of structures
longer than 5 /im and frequently represent less than 2% of the total structures
counted.  Thus, representative counts of longer structures can only be
guaranteed if counting procedures direct that long structures be counted
independently from short structures and that 5 to 10 times as much, area is
scanned for the count of long structures.  Such a procedure is termed
statistically-balanced counting (Sebastien et al, 1982).

      The reported magnitude of the fraction of short structures needs to be
addressed with caution.  Because the distribution of structures peaks in the
vicinity of 1 pm and on the order of one half of the short structures are
shorter than 1 /^m, the relative number of short structures in a total
distribution is extremely sensitive to the minimum size of the structures
characterized.  For example, counting all structures longer than 0.2 /MI in
length is likely to yield a structure size distribution where structures
longer than 5 urn constitute less than 5% of the total while the same
distribution truncated at 1 pm minimum length will yield contributions from
long structures of 20%.  The lower length limit counted is not reported in the
majority of these studies. However, those where the limit is known all show
short structure contributions greater than 90% (with corresponding
contributions by long structures of less than 10%).

      Diameters for chrysotile structures also vary.  The median diameter of a
typical structure less than 5 ^m in length lies between 0.02 and 0.03 /xm and
virtually all of the structures less than 5 urn long have diameters less than
0.05 /zm.  Mean diameters increase with increasing length, but the increase is
not proportional, so that the aspect ratios of long chrysotile structures are
much larger than the aspect .ratios of short structures.  The thinnest
chrysotile structures exhibit diameters  on the order of 0.02 pm.

      The data in Tables 3.2 and 3.3 must be interpreted carefully. " Although
the majority of the listed studies employed TEM to derive structure size
distributions, lack of standardization of sample preparation techniques, of
rules for counting and characterizing structures, and of the criteria used for
establishing the mineralogy of counted structures potentially contribute to
analytical variation between the studies  (see Sections  6.1 and 6.4).  For this
reason,  comparisons of distributions within a study are more reliable.than .
comparisons between studies.  In fact, the data in Table 3.2 clearly  indicate
that at  least  2 studies are quantitatively different than the other studies.
For the  same fiber type in similar exposure settings, Rendall and Skikne
(1980)  and Marconi et  al  (1984) consistently report distributions containing a
greater  fraction of long structures  than the other studies.  Unfortunately,
none of the studies contain sufficient documentation of the methods used to
derive  size distributions to determine the cause  of the  apparent   :
discrepancies.                                                    -

                                   :   22 .       •        •  '•••'...

-------
TABLE 3.3: STRUCTURE SIZE DISTRIBUTIONS MEASURED IN VARIOUS ENVIRONMENTS FOR CHRYSOTILE ASBESTOS
/ o / / Sx l-ง *>l 1 I Fraction of Asbestos Fibers in Size Categories (% by Count)
lit MII
Que/Mine
Mine
Que
Out/Mine
Mine
Mill/Dry
Mill/Crsh
Mill/Crsh
Mill/Crsh

Crsh
Mill
Mill
Mill
Mill
Mill
Mill
Lab
Lab

Chr
i














1
ChrB
ChrB

PCM/TEM
PCM/K, TP
TEM
TEM
PCM/EM
SEM
PCM/STEM
PCM/STEM
PCM/STEM

PCM/STEM
PCM/K.TP
PCM/TEM
1
1
1
PCM & EM
TEM
SEM

D/
D/
I
D

D/N
D/M
D/M
D/M

D/M
D/





1/M.
I/M

0.34




1.25














0.06




0.17


















10,000
6000
10-70,000
I








10,000




L<5m
99
93
92.86
84.54
56.6
88.9
91.2
97.9
88.9

83.1
87
93.7
96.1
98.5
98.6
61.4
89 (T)
69 (T)

D< 0.25 Mm
0.40

7.14
3.54

1.5
2.0
0.2
4-6

4.8

(6.2)
(3.9)
(1.5)
(1.4)




D> 0.25pm
0.2

0.00
11.92

10
6.8
1.9
6.5

12.1

(0.11)
(0.02)
(0.01)
(0.06)




L5-10m
D< 0.25pm
0.36




(1.5)














LS-lOpm
D> 0.25pm
0.7




(7)














L>10pm
D< 0.25pm
0.04



















D> 0.25pm
0.13




(3)















Reference
1
5
8
8
12
2
10
10
10

10
5
13
13
13
13
12
11

11

-------
TABLE 3.3 (CONTINUED)

FP/Mix
FP/Form
FP/Fin
(FP)/BC

FP/Mix
FP/Grnd
BC/Card
BC
BC
RR/ATB
RR/AC
CP/Mix
CP/Form
CP/Fin
CP/Mix
CP/Fin
CP/Mix
CP
CP
Chr
i

















I
PCM/TEM
\



\
\
\
\
\
PCM & EM
PCM/TTEM
1

|
1
1
1
" i
PCM/STEM
PCM/TEM
PCM/TEM
D/M
D/M
D/M
D/M






D/M
D/M
D/M
D/M
D/M


D/M






On
.y
0.8








0.9
0.7























900
-to
17,000
10,000





10,000
10,000
10,000
[900
- to
17,000


i


/ Fraction of Asbestos Fibers in Size Categories (% by Count)
L<5/*m
90.4
90.7
83.8
44
98
J \j
98
98.7
97.1
45.5
75
78
87.6
90.7
95.0
98
99
97.8
98.6
97.5
L >5/ซra
D<0.25/ปm
6.3
5.9
6.3
ซ6.1)



(1.3)
(2.9)

(<2.5)
(<2.4)
6.0
6.4
3.0


0.2
(1.4)
(2.5)
D>0525%n
3.4
3.4
9.9
(49.9)



(0.04)
(0.05)

(22.5)
(19.6)
6.4
2.9
2.1


2.0
(0.03)
(0.06)
LS-Wtm
D < 0.25pm
2.9
3.2
4.2









3.3
3.8
2.0





LS-lOwn
D> 0.25 Aim
1.2
1.5
3.4









3.1
1.4
0.8





D<0.25jim
3.4
2.8
2.0









2.7
2.5
0.9





LMOwn
D>0.25/im
2.2
2.0
6.5









3.3
1.6
1.2





Reference
7
7
7
9
6

6
13
13
12
9
9
7
7
7
6
6
10
13
13

-------
     TABLE 3.3 (Continued)
A- / A.
/ &lฐ /•&<>/ q^-*?
/ 4r /w //
Bag
Que/Bag
Mill/Bag
Cut
Cut
Cstr
InsI/Rmvl
Txt/Card
Txt/Fprp
Txt/Twst
Txt/Weav
Txt/Fprp
Txt/Spin
Txt/Out
Txt/Out
Txt
Txt
Txt
Chr
\
\
\
\
I
\
PChr
Chr
\
\
\
\
I
\
\
PChr
PChr
Chr
\
\
\
PCM/STEM
TEM/PCM
SEM
STEM/PCM
STEM/PCM
STEM/PCM
STEM/PCM
SEM
TEM/PCM
\
\

\
I
\
SEM
TEM
TEM/PCM
I
\
t

D/M
D/
D/N
D/M
D/M
D/M
D/M
D/N
D/M
1
I
1
1
1
1
l
1
1
1
1
1

0.55
1.35




1.0



1.4
1.0






0.06
0.16




0.15










10-70,00(

6000
10-70,000
10-70,000
10-70,000

6000
[900
- to
17,000







/ Fraction of Asbestos Fibers in Size Categories (% by Count)
L<5Mm
94.3
96
95.2
53.4
87.5
86.2
64.2
97.7
79
83.4
81.2
96
98
81(T)
—
87.5
94.4
93.7
L >5/ซm
D< 0.25/* m
1.9
2.06
1
13.3
10.2
8.7
10.2
0.4
7.9
8.0
9.0




(12.4)
(5.5)
(6.2)
D> 0.25/1 m
3.8
1.72
(4)
33.3
2.4
5.1
25.6
(2.1)
12.8
8.6
9.7




(0.1)
(0.08)
(0.1)
LS-Wum
D< 0.25/J m

1.36





(0.4)
4.5
3.6
4.9







L5-10/im
D> 0.25/1 m

1.31
(3)




(D
5.7
2.9
3.1







L>10/ซm
D < 0.25/t m

0.50
(D





3.4
4.3
4.2







L>Wfm
D > 0.250 m

0.41
(1)




(1)
7.1
5.7
6.6







Reference
10
1
2
10
10
10
10
2
7
7
7
6
6
11
11
13
13
13
to
Ul

-------
                                                             Table 3.3 (continued)
                                                                     Key
to
Exposure Settings:
    AC      Dust from asbestos/cement sanding
    AP      Asbestos production plant
    ATB    Dust from asbestos tape and board
    BC      Dust from a brake and clutch plant
    Bag     Dust from bagging
    Cape    Capetown, South Africa
    Cstr     Dust from construction
    CP      Asbestos/cement pipe manufacture
    Card    Dust from carding
    Crsh    Dust from crushing
    Cut     Dust from cutting
    Dump   Dust from raw material dumping
    Dry     Dust from drying
    Fin     Dust from fiber finishing
    FP      Friction products plant
    Fprp    Dust from fiber preparation
    Form   Dust from forming
     Grnd   Dust from grinding
     Insl      Site where insulation is applied
     Mill     Dust from asbestos milling
Mine       Dust from asbestos mining
Mix       Dust from fiber mixing
Orest      Dust from ore storage
Out       Outside Plant
Pinsl       Pipe insulation manufacturers
Que       Quebec, Canada
RR        Railroad factory
Spec       Nonstandard clearance, etc.
Spin       Dust from fiber spinning
Tran       Transvaal, South Africa
Twst       Dust from fiber twisting
Txt       Dust from textile plant
UICC     Sample of UICC standards
Weav     Dust from weaving
                                                         Fiber Types:
                                                             Actn
                                                             Amph
                                                             Ams
                                                             Chr
                                                             Crc
                                                             Pr
            Actinolite
            Non-amosite amphibole
            Amosite
            Chrysotile
            Crocidolite
            Predominantly
                                                                                                     Preparation Techniques:
                                                                                                          D     Direct
                                                                                                          I     Indirect
                                                                                                          M     Mixed cellulose ester filter
                                                                                                          N     Polycarbonate filter
                                                                                                          Spec   Nonstandard preparation
Analytical Methods:
    PCM    Phase contrast microscopy
    K      Konimeter
    MI     Midget impinger
    SEM    Scanning electron microscopy
    TEM    Transmission electron microscopy
    TP     Thermal precipitator
Miscellaneous:
    ()   Extrapolated from data
    (T)   Total fibers

-------
REFERENCES
                                            TABLE 3.3  (continued^
1.  Gibbs and Hwang, 1980.
2.  Gibbs and Hwang, 1975.
3.  Hwang and Gibbs,  1981.
4.  Beckett and Jarvis, 1979.
5.  Gibbs and du Tbit, 1979.
6.  Lynch et al., 1970.
7.  Dement and Harris, 1979.
8.  Sebastien et al., 1984.
9.  Marconi et al., 1984.
10. Snyder et al., 1987.
11. Cherrie et al., 1987.
12. Rendall and Skikne, 1980.
13. Winer and Cossette, 1979.

-------
      The extent that the published size distributions presented in Tables 3.2
and 3.3 are representative of the exposure settings examined also needs to be
considered when comparing results between studies.  In most cases, each of the
distributions presented is derived from a single grab sample.  Since it is not
known to what extent the size characteristics of an asbestos dust in a
particular exposure setting vary over time, it is unclear how to relate single
grab samples to the general features of an asbestos dust in a particular
exposure setting.  Nevertheless, this is currently the only data available for
providing an indication of the distribution of particle size in (environmental
and occupational) asbestos dusts.

      Despite the above caveats, the data in Table 3.2 clearly indicate that
different asbestos mineral types tend to produce dusts with distinctive size
characteristics.  Amosite dusts appear to contain the greatest fraction of
long and thick structures. Averaged over the 26 distributions for amosite and
predominantly amosite dusts, 34% of total amosite structures are longer than
5 pm.  This compares with 10% of the total structures in the 53 chrysotile and
predominantly chrysotile dusts and 12% of the 18 crocidolite dusts.  The
difference between amosite and the other two mineral types is further
exaggerated if Marconi et al (1984) and Kendall and Skikne (1980) are removed
from the data set.

      NOTE

      The absolute magnitudes of the fraction of long structures  in a
      distribution all represent upper limits and may be revised  downward
      depending on the minimum lengths counted in each of the studies.
      However, the relative values  (between mineral type) are less likely to
      be affected because the observed differences are also  apparent within
      individual studies, including the Kendall and Skikne study.

      Although chrysotile and crocidolite dusts appear to contain comparable
length distributions, the data in Table 3.2 indicate that chrysotile dusts
tend to contain thicker structures  than crocidolite.  For example, the
fraction of dusts visible by optical microscopy  (those thicker than 0.25 ^m)
represents an average of 9% of chrysotile dusts with a range of 1 to 50%.
Only an average 4% of crocidolite dusts are visible by optical microscopy
(range 1 to 18%).   The range of diameters across chrysotile dusts may
actually be narrower, if inter-study variation is considered.  Removing the
Marconi et al  (1984) data and one apparent outlier from  Snyder et al  (1987),
the range of the fraction of structures visible by optical microscopy  in
chrysotile dusts is reduced to between 1  and 18% with an average  of 8%.  Note
that the Winer and Cossette  (1979)  data were not  included in the  above
averages because thick structures were counted by optical microscopy  in this
study and such results cannot be compared directly with  the  TEM data presented
for counts of  total long structures.

      As indicated previously,  amosite dusts appear  to contain the greatest
fraction of thick structures.  Between 8  and 43%  of  total amosite structures
are  thick enough to be visible by optical microscopy with an average  25%  among

                                      28

-------
amosite dusts.  Interestingly, the median diameters reported for dusts do not
appear to track the fractions of dusts represented by specific ranges of
thicknesses.  This is not surprising and simply indicates that size ranges are
heavily skewed so that the mean and the median of the distributions are widely
separated.

      Overall, the fraction of long structures in crysotile and crocidolite
typically represent less than 25% of total structures.  Approximately 20 to
70% of long chrysotile structures or 1 to 18% of total chrysotile structures
are visible by optical microscopy.  Although the data is sparse for
crocidolite, approximately 10 to 50% of long crocidolite structures are
visible by optical microscopy.  In contrast, the majority of long amosite
structures are visible optically and long structures typically represent up to
50% of the total number of amosite structures present.  In general, therefore,
the fraction of amosite dust visible by optical techniques is up to 5 times
higher than for chrysotile or crocidolite, although all three show a wide
range.

      As indicated earlier, the magnitude of the range and average fractions
of long structures and structures visible by optical microscopy reported for
the various size distribution may be revised downward subject to the minimum
size counted in the distribution.  For example, assuming published size
distributions could be adjusted so that structures longer than 0.2 pm were
included uniformly, long structures likely represent less than 10% of total
structures in chrysotile size distributions and 4% may represent a better
average fraction of structures in chrysotile that are visible by optical
microscopy.  This is in contrast to the numerical averages for these size
fractions (25% and 8%, respectively), which were calculated directly from the
data and presented above.  Similarly, only 2% of crocidolite structures may
fall within the range visible by optical microscopy.

      It must be emphasized that the published size distributions presented in
Tables 3.2 and 3.3 do not generally indicate whether aggregates are included
or excluded from the distribution presented.  However, aggregates likely
represent major components of all asbestos dusts found in the environment (see
for example: Chatfield, 1985b).   In other examples, data generated at the
Atlas and Coalinga Superfund sites in California Indicate that 50% of the
structures collected in the vicinity of the Atlas and Coalinga mines are
matrices or other aggregates.   Data from the South Bay Superfund site in
California indicate that aggregates represent between 10 and 30% of the
structures found in fibrous dust at this site.   In two studies,  Sebastien et
al (1984 and 1986) report that aggregates constitute 40% of the fibrous
structures in asbestos dusts found in the vicinity of the asbestos mines in
Quebec.

      One observation concerning size distributions that relates to biological
activity address the spread in apparent risk factors derived from different
exposure settings (see Section 3.2).   Variations in fiber size distributions
in dusts from different exposure settings have been identified as one of
several possible causes of the observed spread in reported risk factors from

                                      29

-------
published epidemiology studies (see for example: USEPA 1986a).   However, such
variations are not readily apparent in Table 3.3.  Overall, a slight trend
toward increasing length for structures in dusts from mining and milling
(except for bagging), through bagging, to textiles appears when distributions
for these exposure settings are averaged.  Structures in dusts from friction
products and asbestos/cement pipe both appear shorter than textiles, although
they cannot be distinguished from either mining, milling, or bagging.

      Within individual studies, trends relating asbestos structure dimensions
to exposure setting should be easier to distinguish since inter-study effects
do not interfere.  The clearest trend appears in the data reported by Dement
and Harris (1979) with structures in dusts increasing in length from
asbestos/cement pipe, through friction products, to textiles.  Such trends are
not apparent in the data of Winer and Cossette  (1979) or Snyder (1987).
However, the representativeness of the samples  in terms of the types of dusts
generated in the exposure settings examined has not been addressed in any of
the studies except for Dement and Harris.   None of the other studies listed
in Tables 3.2 and 3.3 contain sufficient data for a single mineral type over a
broad enough range of exposure settings to address this question.  To resolve
issues concerning the quantification of the risk associated with asbestos
exposure, additional structure size characterizations may need to be developed
using a standardized analytical method.

      3.5.  REPRESENTATIVE ASBESTOS CONCENTRATIONS

      A limited number of published studies contain measurements of ambient
concentrations of asbestos.  Unfortunately, the utility of the data from these
studies is restricted by the use of different analytical techniques and
methods of reporting so that the results from different studies generally are
not directly comparable (Chatfield, 1983b).  Another problem with the
available database is that several of the  studies report results in mass
equivalent concentrations and it has been  shown that mass concentrations do
not generally correlate with structure counts at ambient concentrations
(Chatfield, 1985c).  The high mass measurements  that sporadically occur within
any set of ambient samples are frequently  due to the presence of a single,
large aggregate on the sample filter, while the  overall structure count
remains low.

      Available ambient asbestos measurements are presented  in Tables 3.4 and
3.5.  Table 3.4 presents representative background concentrations in various
environmental settings and Table 3.5 presents several representative
concentrations near  sources of contamination.   To distinguish ranges of
concentrations that  likely represent natural background from anthropogenic
contamination, both  the range of concentrations  and  the median concentrations
reported  in each study are presented  in  the tables.  The median concentrations
probably  are better  to base judgments upon since they tend to smooth the
effects of outliers  that potentially  skew  the limits of the  reported ranges.

      To  the extent  that the data  are available in each study, concentrations
are expressed in the tables in three  formats: total  asbestos structures per

                                      30

-------
 unit air volume,  asbestos  structures  visible  by optical  microscopy (PCM
 equivalent)  per unit air volume,  and  asbestos mass  per unit air volume.   As
 addressed in later sections  of this report, each of these  formats  has
 particular advantages and  disadvantages.   As  indicated in  the  first paragraph
 of this  section,  however,  the concentration formats are  virtually  independent
 variables;  correlations  between these formats that  have  been reported  in the
 literature have been weak  at best.

       Comparisons between  results presented in each of these tables should be
 made only with extreme care.   Although the data presented  is entirely  from
 studies  employing transmission electron microscopy,  comparisons between
 studies  must address the type of preparation  (direct or  indirect)  employed and
 the specific size fraction of asbestos counted.   Unfortunately,  documentation
 for several  of the studies is insufficient to determine  precisely  what size
 fraction was counted and whether aggregates are included or excluded from the
 count.

       A  few  of the studies referenced in Tables 3.4 and  3.5 address the
 question of  aggregates and provide a  qualitative indication of size
 distributions.   Sebastien  et al (1986)  indicate that in  and around the mines
 of Quebec,  14% of structures  longer than 5 pm are aggregates.   This
 corresponds  closely to the fraction of the long structures  that represents the
 PCM equivalent fraction.   Several authors  report that asbestos  structures
 encountered  as background  at remote locations  are all short (less  than 5  /zm)
 chrysotile fibers (see,  for  example,  Chatfield,  1985c).  It is  unclear,
 however,  whether  the detection of only short  structures  is  solely  due  to
 statistical  limitations  in the number of structures  encountered (longer
 structures are one to two  orders  of magnitude  less  abundant than short
 structures in most chrysotile distributions)  or  whether  such observations
 represent a  valid phenomenon indicating that  the detection  of  contamination
 representative of a anthropogenic source can be  limited  to  searching for  long
 structures.

       Due to  the  severe  limitations of  this database, it is  difficult  to
 identify  reasonable  target concentrations  that  represent the boundary  between
 natural background and anthropogenic  contamination.   However, certain
 generalizations are  apparent.   For example, the  data  in Tables  3.4  and 3.5
 suggest that  analysis  of samples prepared by an  indirect-transfer  technique
 generally yield results  that  are higher than analyses of samples prepared by a
 direct-transfer technique.   However,  because the  concentration of  indirectly
 prepared  samples  can be optimized to  facilitate  analysis, it is  the limiting
 concentrations observed on directly prepared samples  that is of principle
 concern for development of the method.

      Of  the  results  identified as derived from directly prepared samples in
 Table 3.4, the range of median concentrations  reported for  total structures
 spans from less than 0.4 s/L  to 6 s/L.  Note that this range does not  include
 the median of  less than 0.01  reported in one study (Tuckfield et al 1988),
which appears  to  lie outside  the range of the  majority of other studies.
Results reported  in studies where the preparation technique was not specified

                                      31

-------
   TABLE 3.4: PUBLISHED BACKGROUND CONCENTRATIONS OF ASBESTOS
Environmental
Setting

Outside Schools
Outside Schools
Outside Schools

Outside Schools
Outside Public
Buildings
Urban
Urban Toronto
Urban Paris
Urban
Urban
New York
US Cities
US Cities
Canadian Cities
Canadian Cities
English Cities
German Cities

Urban Switzerland
a
Median of Reported Concentrations
s/1 PCME/1 ng/m3
1 d
25 | (0.05)
10
4

—
<0.01
2
6
—
—
—
— •
—
—
—
—
—
3g

03d
	

	
	
<2
<2h
—
— •
—
—
—
—
0.7 h>J
0.1 h'j

—
h,i
— 0.4 "
!
Upwind of
Asbestos Plants
0.2 —

?
b
Range of Reported Concentrations
s/1 PCME/1 ng/m3
''
0.12 1 0-2000 —
'h;
0.08 j 0-100 0-8 d
0.02 j 0-10
I
0.5 j -
~~
0.03
0.07
0.4
1
—
10
1
3
—
lj
—
—
	

—
i
1
0-8
045
	
—
o-io1

—
—
—
—
—
0-10
h i ^
0.75 j —

0.03

0-11
:
1
04
0-4 h
—
—
—
—
—
—
0.6-0.9h'k
0-3 h
—
—

—

	



0-8
0-0.9
0-0.07

0-100
—
0-20
0-0.3
0.1-9
1-10
—
0-100
0-50
0-15

0-6
0.1-1
—

—

0-170


Conversion
(PCME/1 to ng/rr?)
Analytical
Sensitivity c
Preparation
Technique
i -
I
e f
0.1-100, (4)
—

—
—
0.4 g
—
—
—
—
—
—
—
—
0.1
—
—



2
2
—
—
—
—
—
	
—
—
1 -
—

0.5
:

	
;
— i 0.3


1
indirect
indirect
indirect

—
direct
direct
direct
indirect
—
—
—
• —
—
indirect
—
—
—

—




Reference

0
ฉ
0

0
ฉ
ฉ
ฉ
ฉ
ฎ
ฉ
ฉ
ฉ
ฉ
ฉ
ฉ
ฉ
ฉ

ฉ

CD


NJ

-------
   TABLE 3.4 (CONTINUED)
Environmental
Setting
Suburban
Remote
Remote
Remote (FRG)
Rural Ontario
Rural
Rural Austria
a
Median of Reported Concentrations
s/1 PCME/1 ng/m3
-
0.6
<0.4
—
2
—
••
d
<0.6
<0.4
—
<2h
—
-h,j
'
0.003
—
0.3
0.002
—
—
b
Range of Reported Concentrations
s/1 PCME/1 ng/m3
0-6
0-0.4
—
0.03-0.9
0-30
0-0.3 l
—
0-2
<0.4
—
<2h
—
—
0-9
0-0.008
0.1-2
0-0.2

—
Conversion
(PCME/1 to ng/m3)
0.2 g
—
—
—
—
—
Analytical
Sensitivity c
0.6
0.4
—
2
—
—
Preparation
Technique
direct
direct
—
direct
direct
—
—
Reference
ฉ
ฉ
ฉ
ฉ
ฎ
0
OJ

-------
                                     TABIE 3.4  (CONTINUED)
TOOTNOTES:

a.
      Values in these columns represent estimated median values for the range of
      concentrations reported in the study.  In some cases, due to the form of data
      presentation, values presented in this table represent the median of a range of
      averages.  In other cases, which are marked accordingly, only mean values could
      be derived from the study.

b.    The lowest and highest values reported in each study are presented here as the
      limits of the range of reported concentrations.  In some cases, due to the form
      of data presentation, the values presented in this table represent a range of
      average values from multiple locations.

c.    Values in this column represent an estimated average of the analytical
      sensitivities reported for each measurement in the study.

d.    These values are based on PCM analyses rather than TEM analyses.

e.    This is a range derived from the manipulation of paired PCM, TEM analyses.

f .    This value is simply the quotient of the median values for the parameters
      indicated.

g.    These values are derived from a single measurement.

h.    These values represent total structures longer than 5 urn rather than PCM
      equivalent structures.

i.    These values are estimated values.

j .    These values are the mean of a range of concentrations rather than the median.

k.    This is a range of means of multiple samples from several locations.

-------
                                                            (continued)
       REFERENCES

-------
   TABLE 3.5: PUBLISHED ASBESTOS CONCENTRATIONS PROXIMAL TO ASBESTOS SOURCES
Environmental
Setting
In Buildings:
-with asbestos

-with damaged
asbestos
-with asbestos

-with no asbestos
-with asbestos

Toronto subways

Toronto subways
Near mines

Near waste piles

Paraoccupational

a
Median of Reported Concentrations
s/1 PCME/1 ng/m3
i
<5 | <1
1
0.6 —

0.4

—
s
0.1
—

	
	

25 • { —

200


— 1 40e'h
^ h
60 j 7

' —

' "
~~
:

<0.1

—

—

—
12

0.2

15
—

	 ;
b
Range of Reported Concentrations
s/1 PCME/1 ng/m3

0-40



—

. — .
—

15-150

40-3000
—
j
13-340
!


•" — '


0-2





—
— '
e
0-15
e
0-55
15-150e>i
i
0.5-65

0-100 e


0-40 d



—

—
0-750

0.015-50

0.2-200
—

—

—

Conversion
(PCME/1 to ng/m3)

.







e,f
0.3
e,g
0.9




,

Analytical
Sensitivity c














'',.




Preparation
Technique

direct



—

—
—

direct

indirect
indirect

direct

—

Reference

0

(5)

ฉ

ฉ
0

ฉ
^^
(i)
ฉ
^^
ฉ

ฉ

(jO

-------
                                      TABLE 3.5 (CONTINUED)
 FOOTNOTES:
 a.
b.
c.


d.



e.


f.

g.

h.

i.
Values  in these columns represent estimated median values for the range of
concentrations  reported in the study.  In some cases, due to the form of data
presentation, values presented in this table represent the median of a range of
averages.  In other cases, which are marked accordingly, only mean values could
be derived from the study.

One lowest and  highest values reported in each study are presented here as the
limits of the range of reported concentrations.  In some cases, due to the form
of data presentation, the values presented in this table represent a range of
average values  from multiple locations.

Values in this  column represent an estimated average of the analytical
sensitivities reported for each measurement in the study.

The highest mass concentration in this range is reported for a sample composed
largely of amphibole.  The highest mass concentration reported for a chrysotile
sample in this data set is 24 ng/m.

These values represent total structures longer than 5 urn rather than PCM
equivalent structures.

These values are derived from a single measurement.

This value is the mean of three measurements.

These values are the mean of a range of concentrations rather than the median.

This is a range of means of multiple samples from several locations.

-------
I
                                                              TABLE 3.5   (continued)
                   1.  Burdett and Jaffrey, 1986.
                   2.  Hatfield, etal., 1988.
                   3.  Sebastien, et al., 1979.
                   4.  Chatfield, 1983b.
                   5.  Sebastien, et al., 1986.
                   6.  Ruch and Serper, 1977.
                   7.  Steen, et al., 1983.
             w
             03

-------
appear to be in general agreement with this range,  as opposed to the range of
medians reported for samples prepared by an indirect technique: 4 to 25 s/L.
The range of background concentrations in urban environments appear higher
than exhibited at rural locations.  As indicated by the data presented in
Table 3.5, the lower range of concentrations reported in the vicinity of
asbestos sources appears similar to the range reported in Table 3.4 for
background.

      Deriving a range of background concentrations for PCM equivalent
structures is more problematic.  For directly prepared samples from urban
settings, the reported medians range between 0.1 and less than 2 s/L.
However, for rural locations, PCM equivalent structures were largely
undetected on directly prepared samples.  Interestingly, median concentrations
for PCM equivalent structures reported for indirectly prepared samples appear
to span a lower range than for directly prepared samples: 0.05 to less than
2 s/L.  Thus, the true range of background concentrations for PCM equivalent
structures may not have been identified in this dataset.

      Rounded to the nearest half order of magnitude, a reasonable range for
background asbestos concentrations appears to span 0.5 to 5 s/L for total
asbestos structures.  The -lower end of the range appears to be closer to the
median background for rural locations.   Although a representative range for
background concentrations of PCM equivalent structures (or long structures) is
difficult to estimate from the available literature, such values may be
estimated from the range reported for total structures given the data
presented in Section 3.4.  Conservatively, long structures (those longer than
5 ^m) represent approximately 4% of total structures ,in most chrysotile size
distributions encountered.  Thus an estimate of the range of background
concentrations for long structures spans 0.02 to 0.2 s/L.  Presumably, PCM
equivalent structures would be some major fraction of these concentrations.

      3.6.  REPRESENTATIVE TOTAL DUST CONCENTRATIONS

      As addressed in Section 6.2.1, the concentration of total particulate in
a dust affects the ability to characterize the asbestos fraction.
Concentrations of total suspended particulate (TSP) vary over several orders
of magnitude depending on location, time of day, and weather.  Urban and
agricultural sites tend to exhibit significantly higher concentrations of TSP
than rural locations.  In addition to variation in overall concentration,  the
composition of TSP also varies significantly as a function of location.   At
Urban sites and specific rural locations,  the TSP tends to be composed
principally of organic matter that can be ashed or  inorganic substances that
are soluble in acidified media.  Agricultural locations and other rural
locations frequently exhibit higher concentrations  of- refractory silicate
particles.  Due to the wide spatial and temporal .variation in TSP
concentrations,  a general rule for estimating levels at a site can not be
provided without data from a comprehensive survey.
                                      39

-------
      3.7.  ANALYTICAL TECHNIQUES USED TO MONITOR ASBESTOS

      Analytical methods traditionally used to monitor asbestos include midget
impinger (MI),  phase contrast microscopy (PCM),  scanning electron microscopy
(SEM) , and transmission electron microscopy (TEM).   Each method differs in its
ability to detect the various size fractions typical of asbestos dusts
including those most likely to relate to biological activity.

      In the occupational studies used to develop dose/response relationships
(addressed in section 3.2), asbestos dusts were analyzed by counting
structures on a slide or filter using optical microscopes.  Until the early
1970's the procedure involved collecting workplace dust with a midget impinger
using a light microscope at a magnification of 100 to count the number of
particles collected.

      The midget impinger is a device in which a stream of contaminated air is
forced through a restricted opening into a liquid (alcohol) where it emerges
as a jet of bubbles that disperse the asbestos.  The particles suspended in
the  liquid are than transferred to a 1 mm deep well and counted using a light
microscope.

       Because the counting rules employed during the examination of midget
impinger samples require that all observed particles be counted, the resulting
concentrations represent coarse total dust counts and have little direct
relationship to the quantity of fibrous structures present.  In Figure 3.1,
the  lack of significant overlap between the midget impinger circle and the
circle representing fibrous particles illustrates the limited relationship
between analytical results from this method and potentially important asbestos
fibers.  Thus, midget impingers provide only  an  indirect  index of asbestos
exposure.

      In the 1970's the midget impinger was replaced by the membrane filter
method, which has become the standard technique  for monitoring asbestos in
industry.  This latter method involves collecting a sample of airborne dust on
a membrane filter, rendering the  filter transparent with  an appropriate
solvent, and counting fibers using a phase contrast microscope  (PCM) at
magnifications between 400 and 900.

      Because of  the increased magnification  and because  phase-contrast
lighting  increases the sensitivity to narrow  objects  in the field of view, it
is possible  to delineate the general morphology  of the particles being counted
so that measurements can be restricted to fibrous structures  (defined as•those
longer than  5 /ira, exhibiting aspect ratios greater than 3:1,  and having
largely parallel  sides).  However, the limitations imposed are  somewhat
arbitrary and not based on health effects  (Chatfield, 1979).  In addition, PCM
techniques were standardized only recently so that counting procedures and
attendant results changed  over the period spanned by  existing epidemiology
studies (Chatfield, 1985d).
                                       40

-------
      Due to  the  limited resolution of PCM, the internal components of complex
structures  (bundles, clusters, or matrices) are not generally distinguishable
so that all such  structures appear as single, solid particles.  Consequently,
all structures visible by PCM that satisfied the appropriate dimensional
criteria were counted as individual structures (fibers).  The term "structure"
is used here  for  consistency, because frequently a specific "fiber" viewed by
PCM is observed to be a complex structure when viewed with the increased
resolution afforded by transmission electron microscopy.  The purpose is to
preserve the  terminology used to describe a specific particle that may be
viewed using  any  of several microscopic techniques.

      Due to  a combination of resolution and contrast,  the minimum width of
structures visible in the phase contrast microscope reportedly ranges between
0.2 and 0.4 /zm, depending on the configuration and condition of the
instrument.   The  most frequently quoted average limit to visibility is
0.25 jura.  Partially by convention and partially due to  the practical
limitations   associated with classifying structures with aspect ratios less
than 10:1, structure counts were further limited to those longer than 5 fj.m
(See for example: Chatfield, 1985d).

      As indicated in Figure 3.1, structures visible by phase contrast
microscopy correspond to a range that encompasses a significant portion of the
structures believed to be biologically active.  However, structures believed
to be the most biologically active (the longest, thinnest structures) are not
counted by this technique.  Additionally, because the method is incapable of
distinguishing asbestos from non-asbestos minerals, non-asbestos structures
(to the extent that they are present in an exposure setting under study) are
included in PCM counts.  Although it is assumed that the vast majority of
fibrous structures encountered in an occupational setting are asbestos, this
is not the case for environmental samples.  Consequently, these two stated
limitations of PCM render it unsuitable for monitoring environmental asbestos.

      More recently, techniques based on the electron microscope have been
introduced.  The  scanning electron microscope (SEM) and the transmission
electron microscope (TEM) count, respectively, asbestos structures on membrane
filters or structures transferred from the filters to electron microscope
grids.

      Magnifications typically achieved with SEM range between 2,000 and
10,000, while TEM magnifications can easily reach 100,000.   Thus,  TEM is
capable of resolving even the finest fibrous structure.   Although the
magnification implies that the resolution of SEM should be significantly
better than PCM,  in practice,  the visibility of structures in the SEM is
limited by a combination of contrast and electronic noise (Small et al 1983).
Consequently,  the minimum structure width visible in the SEM is only slightly
better than PCM.  Under optimum conditions,- the minimum width of structures
visible in the SEM is perhaps a factor of 5 better than PCM (Walton 1982).
Given the increased expense and other instrumental constraints,  however,  there
appears to be little advantage to SEM over PCM except for the ability to
distinguish asbestos from non-asbestos structures.   At the same time,  SEM does

                                      41

-------
not appear to retain a significant cost advantage over TEM for the analysis of
environmental samples.  Therefore, SEM is not likely a useful technology for
environmental analysis of asbestos.

      Most electron microscopes (both TEMs a SEMs) are equipped to allow
raineralogical and elemental analysis tp -confirm the composition of the
structures counted.  Thus, TEM is; clearly capable of identifying populations
currently believed to', represent the most biologically active fraction of
asbestos, the shaded area of Figure 3.1 (see Sections 3.1,; 3.2, and. 3.3).  In
fact, TEM is theoretically capable of distinguishing whatever appropriate size
or mineralogical fraction that is likely to be linked to asbestos biological
activity in the future.

      3.8.  EXISTING TEM ANALYTICAL METHODS

      Several researchers have published analytical methods based on TEM for
the analysis of asbestos samples  collected on mejnbrane filters.  The principal
features of the Yamate method (1984), NIOSH 7402J(NIOSH 1986), Hayward's
method  (Hayward, no date), and the method in the AHERA regulations (USEPA*
1987),  are summarized in Table 3.6.  A PCM analysis method, NIOSH 7400 (NIOSH
1985) is also included in Table 3.6 for comparison.  It should be emphasized,
however, that procedures employed for PCM analysis have changed over time.
Therefore, the NIOSH method for PCM presented in  the table, which is a
relatively recent method, differs in detail from  the various procedures likely
employed to collect the data used in published epidemiology studies.
Nonetheless, it is included to allow comparison between TEM methodstand the
general features of a PCM .method.         •        .•                 :
                                                  ?•
      As indicated in Table 3.6 the methods incorporate procedures for
preparation, counting rules, and  rules for identifying and characterizing
asbestos mineralogy.  Generally,  the methods also define a fixed area of the
specimen grid to be counted.  By  comparing the requirements of a .method
suitable to support risk  assessment to the features of the methods presented
in Table 3.6, it is possible to determine which of the methods may be
applicable to Superfund field investigations.

      Except for differences in procedural details, the TEM methods presented
in Table 3.6 share many common features.  'All rely on direct preparation,'
although the Yamate method contains a protocol for- an optional indirect  ;
preparation should filters prove  too loaded to analyze the filters when
prepared by a direct  technique.   All of the TEM methods incorporate structure
characterization at a magnification of 20,000.  The principle  differences
between the methods involve counting rules and the inclusion  in the Hayward
method  of a low magnification scan designed to provide a statistically  .
balanced count of large asbestos  structures  (longer than 5 jura).  Because long
structures appear to play a major role  in determining the  biological  activity
of asbestos dusts  (see Sections 3.1 and 3.2), the use of a low magnification
scan to count long structures in  a statistically  balanced  manner(is
incorporated in the method presented in Part 1 of this report.
                                       42

-------
                                TABLE 3.6: COMPARISON OF AVAILABLE ASBESTOS ANALYTICAL TECHNIQUES
METHOD:
NIOSH7400
                                                         NIOSH7402
                                                 YAMATE
                                                                                                      HAYWARD
                                                                                             AHERA
Analytical PCM
Technique
Preparation Direct
Methodology
Magnification 450x
Dimensions (/im)
Length (1): l>5
Width (w): w>0.25b
Aspect Ratio (ar): ar > 3



TEM TEM TEM TEM
Direct Direct Direct Direct
(Optional Indirect)
High Magnification:
10.000X 20.000X 20.000X - 50.000X 15.000X -20,OOOX
Low Magnification:
400X-4000X
High Magnification:
l>1 I > 0.06 '" l>0.06 * l>0.5
3.0>W>0.04C w>0.02 W>0.02 ฐ W>0.02
ar>3 ar>3 ar>3 ar>5
Low Magnification:
W > 0.25
ar > 3
Reported
Sensitivity
            s/cm
            s/mm
                                                                                              0.005 s/cm
                                                                                              70    s/mm

-------
                                                                   TABLE 3.6 (continued)
METHOD:
NIOSH 7400
NIOSH 7402
YAMATE
                                                                                                                    HAYWARD
                                                                                                             AHERA
Counting Rules:
Structures Count all structures Count all structures
exhibiting 1 > Sfim, exhibiting 1 > 1/im,
w < 3.0 /im, and aspect w < 3.0/jm, and
ratios > 3. aspect ratio > 3.
Nota PCME fraction
within count.






Count all structures Count all structures
exhibiting an exhibiting an
aspect ratio > 3. aspect ratio > 3.
Use low magnification
scan to count
PCME fraction.






Count all structures
longer than 0.5pm that
exhibit an aspect
ratio > 5. Record
Individual fibers
within all groupings
with fewer than 3
Intersections. Count
structures longer
than Sfan separately
(PCME).
              Bundles         Bundles meeting overall
                              dimensional criteria
                              generally counted as
                              single fibers unless
                              up to 10 Individual
                              fiber ends can be
                              distlnqulshed within
                              the bundle (representing
                              5 individual fibers).
                          Bundles meeting overall
                          dimensional criteria
                          generally counted as
                          single fibers.
                         Bundles meeting overall
                         dimensional criteria
                         generally counted as
                         single entities and noted
                         as bundles on the
                         count sheet.
                        Bundles meeting overall
                        dimensional criteria
                        generally counted as
                        single entitles and noted
                        as bundles on the
                        count sheet.
Bundles of 3 or more
fibers that meet the
overall dimensional
criteria are counted
as single entitles
and noted as bundles
on the count sheet.

-------
                                                                             TABLE 3.6 (continued)
        METHOD:
                                           NIOSH 7400
                                                                        NIOSH 7402
                                                                                                     YAMATE
                                                                                                                               HAYWARD
                                                                                                                                                           AHERA
                     Clusters          Within a cluster, count
                                      up to 10 Individual
                                      fiber ends from
                                     - (up to 5) fibers that
                                      meet the overall
                                      dimensional criteria.
                                      Otherwise, count a cluster
                                      as a single entity.
                               Within a cluster,
                               count up to 3 individual
                               fibers that meet the
                               overall dimensional
                               criteria. Otherwise,
                               clusters that contain more
                               than 3 fibers that meet
                               the overall dimensional
                               criteria are counted as
                               single clusters.
Within a cluster,
count up to 3 individual
fibers that meet the
overall dimensional
criteria. Otherwise,
clusters that contain more
than 3 fibers that meet
the overall dimensional
criteria are counted as
single clusters.
Ul
Within a cluster,
count up to 3 Individual
fibers that meet the
overall dimensional
criteria. Otherwise,
clusters that contain more
than 3 fibers that meet
the overall dimensional
criteria are counted as
single clusters. Clusters
are defined in the low
magnification scan by
their overall dimensions.
 A collection of fibers
 with more than 2
 Intersections where at
 least one. Individual
 projection meets the
 overall dimensional
 criteria Is counted as
 a single cluster.
                     Matrices
                                      Count up to 5 fibers
                                      emanating from a clump
                                      (matrix). Each individual
                                      fiber must meet the
                                      dimensional criteria.
                               Count individually
                               Identifiable fibers
                               within a matrix. Fibers
                               must individually meet the
                               dimensional criteria.
Count as a single matrix,
all matrices with at least
one protruding or
embedded fiber that
meets the dimensional
criteria.
Count as a single matrix,
all matrices with at least
one protruding or
embedded fiber that
meets the dimensional
criteria.
Count as a single matrix,
all matrices with at least
one protruding fiber such
that the protruding
section meets the
dimensional criteria.
                    Maximum
                    Number
                    Counted
100 structures
                              100 structures
                                                           100 structures
                                                                                      100 structures at high
                                                                                      magnification and 100
                                                                                      long structures at low
                                                                                      magnification.
                                                      50 structures

-------
                                                                         TABLE 3,6 (continued)
METHOD:
                                   NIOSH 7400
                                                                NIOSH7402
                                                          YAMATE
                          HAYWARD
                                                                                                                                                AHERA
 Mineralogy
 Determined?
               Area
               Scanned
                                   100 fields
                           100 openings
                                   no
                             yes
10 openings
yes, except matrix
particles
15 openings" at high
magnification
distributed over 4
grids from each sample
and 100 openings at low
magnification
distributed over 4
grids from each sample.


yes, except matrix
particles
Blanks: 10 openings
Samples: 10
openings (given
the defined
sensivity and the
recommended
air volumes).
yes, except matrix
particles
 Statistical
 Balanced
 Counting?
no
                                                                no
                                                                                            no
                                                                                                                     yes
                                                                                                                                               no
    a. The minimum fiber length to be counted has not been defined in this method. Presumably the minimum fiber length counted would correspond to 3 times the resolution
      limit of the width (due to the aspect ratio requirement). Since the presumed resolution limit Is O.OOIyu m, the corresponding minimum length fiber that would be counted
      under this method Is 0.003/Jm.

    b. Width restrictions for PCM are due to limits In resolution. A width of 0.25 fan represents the average resolution reported for PCM.

    c. The width restriction reported for TEM is based on an estimate  of the resolution limit associated with typical magnification settings for the technique. The presumed limit
      of resolution at a typical magnification of 20,000x reported for this technique Is 0.001/*m.

    d. Bundles are defined as a parallel arrangement of fibers separated by distances smaller than one fiber diameter.

    e. Clusters are a collection of fibers in a random arrangement such that all fibers are intermixed and no single fiber is isolated from the group.

    f. Matrices (termed "mats" by Hayward) are one or more fibers embedded within or protruding from another particle.

    g. The 100 grid opening limit for scanning  is not stated directly in this method but can be inferred from the information provided.

    h. This has been converted to a 200 mesh equivalent to be consistent with all of the other criteria presented in this row.

-------
                                            TABEE 3.6  (continued!
      METHOD
                              REFERENCES
1.  NIOSH 7400
2.  NIOSH 7402
3.  YAMATE
4.  HAYWARD
5.  AHERA
Niosh, 1985.
Niosh, 1986.
Yamate, et al., 1984.
Hayward, no date.
US EPA, 1987.

-------
      Changes in counting rules as the methods evolved (each one building on
its predecessors) were designed to minimize subjective decisions by the
analyst using the method.  This serves to minimize observer-dependent
variation, one of the major sources of variation associated with TEM analysis.
Of the published methods, the counting rules presented in AHERA provide the
best defined counting procedures for minimizing observer-dependent variation.
However, the recording of structure dimensions is not required under AHERA,
which potentially limits the ability to use such results in a risk assessment.
A method designed to support a risk assessment must count structures in a
manner that is consistent with the characteristics that relate to biological
activity, which depend on structure size.  Therefore, the counting rules
defined in AHERA were modified for this method to better detect and record the
range of characteristics that potentially relate to.biological activity (see
Sections 3.1 and 3.2).

      One of the limitations common to most of the TEM methods presented^in
Table 3.6 is that the sensitivity of the analytical method cannot be defined.
This is a consequence of defining a fixed area of the specimen to scan without
simultaneously defining a fixed volume of air to be collected during sampling.
One of the critical elements of this method is that the target sensitivity is
defined and the combined effect of sample loading and the area to be scanned
during analysis are addressed  (see Section 6.2).

      Another critical modification incorporated into this method that is only
partially addressed by the TEM methods presented in Table 3.6, is a recording
procedure that preserves sufficient information to allow extensive and
flexible interpretation of the results without the necessity to re-analyze the
specimen.
                                       48

-------
4.    RISK ASSESSMENT OBJECTIVES AND METHOD REQUIREMENTS

      To support a risk assessment, a method must address three objectives:

      (a)   to provide measurements that relate to the biological activity of
            asbestos;                        •

      (b)   to provide sufficient sensitivity to measure asbestos at the low
            concentrations typically found in the environment near potential
            sources ;

      (c)   to provide sufficient precision to elucidate spatial and temporal
            trends in asbestos concentrations.

An additional consideration to be addressed is the need to control sampling
and analysis costs.  Each of these objectives imposes a set of technical
constraints on a method for asbestos sampling and analysis.  Such constraints
served as a basis for defining performance criteria for this method.
      4.1.  CONSIDERATION OF BIOLOGICAL ACTIVITY

      Although published risk factors are expressed in terms of PCM
measurements, other studies indicate that the biological activity of asbestos
is a function of a broader range of the distribution of asbestos structure
types and sizes, which PCM is not capable of distinguishing (see Sections 3.1,
3.2, and 3.7).  Therefore, to properly address current concepts concerning the
biological activity of asbestos, this method is designed to:
(a)



(b)

(c)


(d)


(e)

(f)
            track asbestos structures of all lengths (defined as all entities
            with components exhibiting an aspect ratio exceeding 5:1 that are
            longer than 0 . 5
            separately track asbestos structures longer than 5 /zm;

            track such entities at sufficient magnification to resolve the
            thinnest asbestos structures (approximately 0.02 ^m in diameter);

            distinguish among asbestos fibers, bundles, clusters,  and matrices
            and independently track their concentrations ;

            characterize the mineralogy of each structure counted;

            record the lengths and widths of each structure counted to allow
            for later selection and tracking of size fractions with particular
            emphasis on long structures (longer than 5
                                      49

-------
      (g)   provide a system for recording analytical results that preserves
            sufficient information concerning the mineralogical determination
            and morphological characteristics of each structure to allow for
            later re-interpretation without the need to re-evaluate the
            original specimen.

Regarding aggregates (bundles, clusters, and matrices), it is also important
to estimate and record, if possible, the number of asbestos components present
within each aggregate.

      NOTE

      Although it appears that the size fraction represented by "b" above can
      be derived from the count of total structures by selecting those
      exhibiting the appropriate length and width criteria, it is listed
      separately to emphasize the need to count long structures separately for
      statistical validity.

      Recording the length and width of each structure preserves the ability
      to select size fractions out of the total structure count that may be of
      special interest.  Examples of size fractions likely to be of special
      interest include the PCME fraction (defined as structures longer than
      5 fan and thicker than 0.25 /wn) or "Stanton" fibers (defined as
      structures longer than 8 pm and thinner than 0.25 /im) .

      4.2.  PRECISION REQUIREMENTS

      To support a risk assessment, the precision of this method should be
sufficient to delineate spatial and temporal trends in the field data
collected at a Superfund site.  Primarily, this means distinguishing among
environmental concentrations of asbestos attributable to local sources at a
site from concentrations associated with general background.  In the absence
of site-specific information, it is assumed desirable to distinguish with high
confidence a factor of five difference  (half an order of magnitude) between
concentrations.  The purpose of this assumption is simply to define a target
performance requirement from which the method could be developed.  However,
numerous sources of variation associated with asbestos measurement potentially
contribute to the uncertainty of an analytical result so that a specific,
desired level of precision may be difficult to achieve.

      In practice, the precision of a method for the determination of asbestos
is limited by several factors including the distribution of asbestos on sample
filters, the characteristics of sampling and analysis tools, and the
subjectivity of the analyst.  In the absence of a database sufficient to
establish levels of uncertainty (and, hence, the ability to quantify
precision), a lower limit to the number of structures that must be counted in
a measurement to achieve a desired level of precision may be estimated from
consideration of the distribution of asbestos on an analytical filter,
ignoring variation introduced by other  factors.   The variation potentially
introduced by these other factors would tend to increase the minimum number of

                                      50

-------
structures that must be counted to achieve a desired level of precision.  A
more detailed discussion of the factors affecting precision can be found in
the literature (see, for example, ISO 1981).

      Assuming, as suggested above, that it may be important to distinguish
concentrations that differ by a factor of five, the minimum number of
structures that must be counted to achieve the defined level of precision can
be estimated as follows.  If it is assumed that structures deposited on a
filter exhibit a Poisson distribution and that analytical contamination is
zero so that observed asbestos represents sampled asbestos, the test statistic
for evaluating whether two means ("m" and "n") can be distinguished is:

                              t = (m - n)/(m + n)1/2.

The object is to find the lowest value of the two parameters (m and n) such
that m = 5n and the difference between m and n is significant based on the
standard normal distribution to the Poisson (Miller and Freund 1965).   For n -
1 and m = 5, t = 1.6 which is not quite significant at the 5% level (where the
cutoff is 1.65).  For n - 2 and m =10, 't - 2.3, which is significant at the
5% level.  Thus, since the variability in structure counts is probably larger
than that predicted by the Poisson distribution due to contributions from
other factors, it is reasonable to assume that a minimum of 10 structures need
to be counted at the concentration of interest to distinguish concentrations
that differ by a factor of five.

      Determinations based on multiple samples can be based on fewer counts
per sample as long as the precision of the aggregate array of samples is
comparable to the precision indicated above for single sample determinations.
This requirement applies both to the count of asbestos structures of all
lengths and to the count of asbestos structures longer than 5 fim.  However,
these two size fractions must be counted separately to provide a statistically
balanced representation (Sebastien et al, 1982).

      NOTE

      Although the foundation for the above assumptions may be a subject of
      some debate, the practical impact of the application of these
      assumptions to the development of the method were minimal (see Section
      4.3).

      As indicated above,  method precision is also limited by instrument
      characteristics and analyst subjectivity, among other factors.  To
      minimize the extent of variation contributed by these two factors,
      instrument settings and characteristics must be specified in the method.
      and counting rules must be specified sufficiently to minimize the
      opportunity for subjective decisions by the analyst.
                                      51

-------
      4.3.  SENSITIVITY REQUIREMENTS

      Analytical sensitivity is defined as the estimated airborne
concentration corresponding to the observation of one asbestos structure.   The
required analytical sensitivity for this method may be defined by considering
the range of concentrations over which measurement is likely to be required.

      Based on a qualitative evaluation of the data presented in Section 3.5,
estimates of median background concentrations range between 0.5 and 5 s/L for
total asbestos structures and 0.02 to 0.2 s/L for asbestos structures longer
than 5 ;im.  Such median concentrations are reasonable targets for analysis
because they likely represent the concentrations above which contributions
from local sources may be distinguished.

      To achieve the desired precision, as indicated in section 4.2, the
assumed requirement for this method is the detection of 10 structures at the
target concentrations specified.  However, this requirement must be tempered
against the probability that a particular asbestos concentration measured in
the environment includes contributions from anthropogenic contamination over
local background.  At the low end of the range of median concentrations
reported for background, it is highly unlikely that such a concentration may
be attributed to anthropogenic contamination.  At the high end of the range of
background concentrations, however, the probability that the measurement of
such a concentration represents contamination from anthropogenic sources is
high.  At the high end of the range for background, therefore, it may be
important to distinguish among small changes in measured concentrations such
as might occur between upwind and downwind samples in the vicinity of a
potential source.

      Setting the analytical sensitivity for this method at 0.5 s/L for total
structures and 0.02 s/L for structures longer than 5 pm, achieves the duel
purposes of providing sufficient sensitivity to measure concentrations down to
levels at which anthropogenic contamination is unlikely while providing that
at least 10 structures will be counted when measured concentrations fall into
a range where it is important to distinguish among small differences in
concentrations.  However, it is also important to consider analytical
background.

      Should asbestos be observed during analysis, it is generally important
to distinguish whether such asbestos originated in the sampled medium or if it
was introduced as contamination during analysis.  Asbestos that can be
attributed to the sampled medium is generally considered to have been
"detected".  Thus, a detection limit is defined as the smallest measurement
that is unlikely (probability less than a specified value) to be due entirely
to contamination from sources other than the air being sampled.  Detection
limits are generally quantified by considering the magnitude and frequency of
occurrence of the analytical background associated with a particular method.
                                      52

-------
However, unless  the distribution of analytical background is known, detection
limits are difficult  to quantify.  Consequently, an alternate method to
account for analytical background is incorporated into this method: a
statistical test to distinguish blank measurements from sample measurements.
Factors affecting analytical background are addressed in Section 6.3.

      As addressed in Section 4.2, it should be emphasized that the desired
analytical sensitivity defined above does not have to be achieved for
individual samples if such samples are part of a set of multiple samples that
were collected so that they are representative of the same sampling
environment.. Under such conditions, it is sufficient that the desired
analytical sensitivity be achieved by the aggregate of the set of samples as a
whole.  For example,  assuming that 10 samples were collected in a manner
assuring that they represent the same sampling environment, analyzing each
individual sample so  that the analytical sensitivity of the measurement is
5 s/L, yields an analytical sensitivity for the arithmetic mean of the 10
samples of 0.5 s/L.

      4.4.  REPORTING REQUIREMENTS

      It is necessary to understand the potential sources of variation in a
measurement before the results of an analysis can be properly interpreted.
Sources of variation  that potentially contribute to the result of an asbestos
measurement include particularly the location of the subsection selected for
analysis on the TEM specimen, the characteristics of the instrument used in
the analysis, and the subjectivity of the analyst.   Variation introduced by
each of these factors means that there is a probability that the results of
two or more consecutive measurements obtained from the same sample may differ
within a finite range.  The degree of variation introduced by such factors to
the results of a measurement obtained' using a specific method is usually
represented by specifying a set of confidence limits in association with each
reported measurement result.   Consequently,  rules for constructing appropriate
confidence limits are incorporated as part of this  method (see Appendix E of
The Method, Part 1 of this report).

      In addition to factors  that potentially contribute to variation in the
result of the measurement of a specific specimen,  asbestos observed during the
analysis may have originated either in the air sampled or may have been
introduced from contamination during any of several phases of sample handling,
preparation,  and analysis (see Section 4.3).   Therefore,  it is important to
distinguish contributions to  a measurement that may be attributed to sampled
asbestos from contributions that may be attributed  to analytical background.
Consequently,  rules for conducting a statistical test to distinguish sampled
asbestos from analytical background are incorporated as  part of this method
(see Appendix E of The Method,  Part 1 of this report).
                                      53

-------
      4.5.  METHOD SPECIFICATIONS

      The method presented in Part 1 of this report is a method where samples
are collected on membrane filters and analyzed by TEM.  None of the published
TEM methods satisfy all of the above requirements; accordingly, several
alternatives had to be evaluated to develop a procedure that is capable of
satisfying the entire set of method requirements defined above.  The following
procedures for sample collection, preparation, handling, and analysis were
combined.
                                      54

-------
5.  OVERVIEW OF METHOD

      In the method presented in Part 1 of this report, samples are collected
and prepared for TEM examination by either one of two techniques.  The
majority of air samples (denoted Phase 1 samples) will be analyzed using an'
indirect procedure for preparation of TEM specimens, optimized to provide the
required analytical sensitivity and precision.  A small number of samples
(denoted Phase 2 samples) will be collected in such a way that they can be
analyzed using both indirect and direct procedures for preparation of TEM
specimens, to allow comparisons to be made between the results from the two
specimen preparation procedures.  TEM examination procedures used for the two
sets of samples also differ.

      5.1.  SAMPLE COLLECTION

      A sample of airborne particle is collected by drawing a measured volume
of air through a 25 mm diameter, 0.45 fj,m pore size MCE membrane filter by
means of a pump. Air volumes collected on Phase 1 samples will be maximized.
Air volumes collected on Phase 2 samples will be limited to provide optimum
loadings for filters to be prepared by a direct-transfer procedure.

      5.2.  SAMPLE PREPARATION

      TEM grids will be prepared according to either 5.2.1 or 5.2.2 below.

            5.2.1.  Indirect TEM Specimen Preparation

      This preparation will be applied to all Phase 1 and Phase 2 samples.
Half of the filter is ashed in a low-temperature plasma asher.  The residual
ash is ultrasonically dispersed in freshly-distilled water.   The suspension is
acidified using hydrochloric acid, and immediately filtered through a 25 mm
diameter, 0.1 jum pore size MCE filter.  The filter is dried and the filter
structure is collapsed using a mixture of dimethyl formamide, acetic acid and
water.  A thin film of carbon is evaporated onto the collapsed filter surface
and small areas are cut from the filter.   These areas of filter are supported
on TEM specimen grids and the filter medium is dissolved away by a solvent
extraction procedure.

      NOTE

      An alternate procedure for indirect preparation,  which incorporates
      washing the deposit off of MCE filters and ashing of the wash-suspended
      deposit,  may be substituted into this method subject to the results of a
      pilot study.

            5.2.2.  Direct TEM Specimen Preparation

      One quarter of the remaining filter sections from all  Phase 2 samples
will be prepared by this procedure.   The quarter filter is collapsed using a
mixture of dimethyl formamide,  acetic acid and water.   The collapsed filter is

                                      55

-------
etched for a short time in a low temperature plasma asher to remove the
surface layer of filter polymer which may have encapsulated asbestos
structures during the collapsing procedure.  A thin film of carbon is
evaporated onto the collapsed filter surface and small areas are cut from the
filter.  These areas of filter are supported on TEM specimen grids and the
filter medium is dissolved away by a solvent extraction procedure.

      5.3.  ANALYSIS

      The TEM specimen grids are examined at both low and high magnifications
to check that they are suitable for analysis before carrying out a quantita-
tive examination on randomly-selected grid openings.  In addition to isolated
fibers, ambient air samples often contain more complex aggregates of fibers,
with or without equant particles.  Some particles are composites of asbestos
fibers with other materials.  Individual fibers and more complex structures
are collectively referred to as "asbestos structures".  A coding system is
used to record the type of fibrous structure, and to provide the optimum
description of each of these complex structures.

      The method requires that separate examinations be made for asbestos
structures of all sizes (incorporating asbestos fibers with lengths greater
than 0.5 /m) and for asbestos structures longer than 5 urn.  In both cases,
asbestos structures are defined as structures containing components exhibiting
mean aspect ratios equal to or greater than 5:1.  This TEM examination
procedure allows for specification of a lower analytical sensitivity for the
measurement of the concentration of asbestos structures longer than 5 pm.

       In the TEM analysis, electron diffraction (ED) is used to examine the
crystal structure of a fiber, or fibrous components of complex structures, and
the elemental composition is determined by energy dispersive X-ray analysis
(EDXA).  For a number of reasons, it is not possible to identify  (confirm the
mineralogy of) each structure unequivocally and structures are classified
according to the techniques that have been used to  identify them.  A simple
code is used to record the manner in which each structure is classified.

       The classification procedure is based on successive inspection of the
morphology, the electron diffraction pattern, and the energy dispersive X-ray
spectrum.  Confirmation of the identification of chrysotile is only by
quantitative ED, and confirmation of amphibole is by a combination of
quantitative EDXA and quantitative zone-axis ED.

       Several levels of analysis are specified, the higher levels providing a
more rigorous approach to the identification of fibers.  The procedure permits
a minimum required asbestos identification procedure to be defined on the
basis  of previous knowledge, or lack of it, about the particular  sample.
Attempts are then made to achieve this defined minimum procedure  for each
asbestos structure, and the degree of success is recorded for each.  The two
codes  remove from the microscopist the requirement  to interpret observations
made during the TEM examination, and allow this evaluation to be  made later
without the requirement for re-examination of the TEM specimens.

                                      56

-------
      The lengths and widths of all classified asbestos structures are
recorded.  The number of asbestos structures found on a known area of the TEM
specimen grids, together with the equivalent volume of air filtered through
this area, are used to calculate the airborne concentration in asbestos
structures/liter of air.

      This method specifies minimum analytical sensitivities of 0.5 s/L and
0.02 s/L for the measurements of asbestos structures of all sizes
(incorporating structures longer than 0.5 /*m) and asbestos structures longer
than 5 fj,m, respectively.  In both cases, asbestos structures are defined as
structures containing components exhibiting mean aspect ratios equal to or
greater than 5:1.

      It will not always be possible to achieve the defined analytical
sensitivities, because the volume of air that can be sampled is dictated by
the nature and concentration of the suspended particulate in the atmosphere
being sampled.  To some degree,  this limitation can be overcome by selective
concentration of asbestos structures during the specimen preparation
procedures and by examination of a larger area of the TEM specimens.   However,
the ease and cost of achieving a specific value for the analytical sensitivity
will vary from sample to sample.
                                      57

-------
6.
METHOD COMPONENTS
      The detailed protocols incorporated in this method were selected based
on the state of knowledge of their presumed capabilities and limitations.   In
some cases, data are lacking so that additional laboratory work is needed to
properly evaluate the efficacy of the proposed procedures.

      6.1.  FACTORS AFFECTING CHARACTERIZATION OF ASBESTOS STRUCTURES

      The proper characterization of asbestos structures requires use of an
analytical technique capable of resolving the thinnest asbestos structure and
capable of distinguishing asbestos from non-asbestos minerals.  Counting rules
must be designed to facilitate distinguishing fibers, bundles, clusters, and
matrices.  Various fractions of potential interest must be easily extracted
from analysis results and counts must be recorded in sufficient detail to
allow later reinterpretation.

            6;1.1.      Analytical Technique

      Because one of the goals of environmental sampling and analysis for the
determination of asbestos is to provide a measurement that is comparable with
published risk factors, which are expressed in terms of PCM counts, an obvious
question  that arises is whether to simply use PCM as the analytical technique
for the analysis of environmental samples.  However, environmental samples can
not be properly characterized using PCM due to a combination of the
limitations of PCM and the  characteristics of environmental dusts.  It is
unfortunate that PCM can not be used to evaluate environmental samples because
PCM is significantly less expensive than TEM.

      PCM is inherently less sensitive than TEM at detecting asbestos
structures.  The sensitivity of PCM is limited both by  increasing obscuration
as filters become more loaded and by increasing observer-dependent variation
as fibrous structures become less concentrated on the filter  (Chatfield,
1985a).   In one round-robin study of PCM laboratories (Crawford,  1985),
observer-dependent variation approached a factor of  300.

      Observer-dependent variation may be due to a combination of instrument
limitations, differences in preparation techniques,  and the subjective
judgments of the analyst.   In common with any technique in which  measurements
are made  close to  the  lower limits of sensitivity, PCM  results vary as  a
function of the condition of the  instrument.  Differences may be  due  to such
factors  as misalignment of  the phase ring,  failure to scan the full depth of
focus, and differences in the interpretation of  irregular fibers.  Evidence
for such variation is  provided in several studies including  the  following:

       (a)  fiber  counts made on  identical  samples were shown to  increase by  a
            factor of  two if the  count was  made  at twice the  magnification
             (Lynch et  al 1970);
                                       58

-------
       (b)
       (c)
       (d)
counting with a graticule verses full field counting increases
counts by a factor of three (Beckett et al, 1976);

for chrysotile, novice counters frequently count only 25% of the
fibers observed by experienced counters (Beckett and Attfield
1974);

there are large inter-laboratory differences in counts of the same
sample (Beckett and Attfield 1974).
 In contrast to  the last observation,  inter-laboratory agreement among TEM
 results  has been achieved without the extensive inter-laboratory discussions
 that  have  been  required to normalize  the  optical work.

       The  practical limit of sensitivity  for  PCM is  approximately 10 s/L at
 the maximum total dust  loading that can be tolerated in occupational samples
 (Chatfield,  1985d).   PCM can be applied in occupational settings because
 contaminated work sites tend to exhibit elevated asbestos  concentrations in
 relationship to total dust concentrations.  This allows asbestos to  be
 deposited  on the filter at concentrations  within the range of  the sensitivity
 limits of  PCM at the  same time that dust  deposited on the  filter remains below
 the level  at which  obscuration precludes  a proper PCM count.

       In contrast,  the  ratio  of asbestos  structures  to  total dust particles is
 generally  lower in  environmental samples  than in occupational  samples.   The
 absolute concentration  of asbestos also tends  to be  lower  in environmental
 samples.   The lower ratio of  asbestos  to dust  limits PCM sensitivity so  that
 the typical  asbestos  concentrations found  in  environmental samples (ranging
 between  0.01 and 0.. 1  PCME s/L)  can not be  detected.   The net result,  in  the
 absence  of other factors,  is  the creation  of  large numbers  of  false  negative
 analyses when PCM is  used to  analyze  environmental samples.

      PCM  analysis of environmental samples can  create  false positive  results
 due to the  inability  of PCM to  distinguish  asbestos  structures  from  non-
 asbestos structures.  Non-asbestos fibers  (such  as cellulose and gypsum)  are
 ubiquitous  in many environmental  settings.  In fact,   asbestos generally
 constitutes a minority  of the  total fibrous structures  typically present  in
 environmental samples (Chatfield, 1985d).   Because they can not  be
 distinguished from asbestos, non-asbestos fibers  that exhibit the appropriate
 dimensional criteria will be  included as asbestos in  a PCM count, creating
 falsely  inflated  asbestos counts.

      The inability of  PCM to track asbestos concentrations in environmental
 samples has been documented in numerous studies.  In a comparison of four
 sampling and analysis methods  (Gibbs et al, 1980), three EM methods generated
data that showed trends in asbestos concentrations,  which decreased with
distance from known sources.  The one PCM method included in the study was not
capable of distinguishing between high environmental  asbestos concentrations
                                      59

-------
near sources and low concentrations at background locations.   This is
presumably because the majority of structures observed by PCM were not
asbestos.

      Asbestos measurements by TEM and PCM were also compared in a series of
studies of asbestos abatement projects (Tuckfield et al, 1988; Chesson et al,
1985; and Karaffa et al, 1986).  Both indoor air and outdoor (ambient air)
were monitored in these studies.  In none of these studies was PCM capable of
reliably determining indoor or ambient concentrations.  In fact, it was
reported in one study (Chesson et al, 1985) that PCM counts appeared to relate
to human activity rather than total asbestos (as measured by TEM), suggesting
that non-asbestos fibers were interfering with the analysis.

      Perhaps the most important limitation associated with PCM is the
inability to detect asbestos characteristics that best relate to biological
activity and, therefore, determine risk (see Sections 3.1 and 3.2)  Due to the
limited resolution of PCM, only structures longer than 5 /zm with diameters
exceeding 0.25 /zm can be counted.  Therefore, the long, thin asbestos
structures  (which are believed to constitute the most biologically active
structures) can not be resolved by PCM.  In addition, PCM is incapable of
resolving the internal detail of the structures counted so that it is
frequently not possible to distinguish aggregates .from simple fibers.  Thus,
PCM  is not capable of characterizing many of the aspects of asbestos exposures
generally believed to affect biological activity.

      Over the range of conditions typical of environmental samples, PCM  is
not  capable of providing measurements of asbestos in environmental settings
that relate meaningfully to risk factors in occupational settings.  PCM can
not  be used to properly evaluate environmental samples because:

       (a)   it is not sufficiently sensitive to detect asbestos at
            concentrations typical of environmental samples;

       (b)   it suffers  from observer-dependent variation to a degree that  is
            unacceptable for the level of precision required  for environmental
            samples;

       (c)   it is not capable  of distinguishing asbestos from non-asbestos
            structures, a requirement that is critical  to  the proper analysis
            of environmental samples; and

       (d)   it is not capable  of distinguishing among  the various types and •
            sizes of asbestos  structures that impact  the biological  activity
            of asbestos.

Therefore,  TEM is the only analytical technique capable of characterizing
asbestos  exposures  in a manner that  is consistent with the method requirements
defined in  Section  4.   TEM is  the  analytical technique to be  used for  the
determination of asbestos  in this  method.                          — -
                                       60

-------
      NOTE

      SEM was also evaluated and eliminated from consideration  for use  in  this
      method  (see Section 3.7).
            6.1.2.
Magnification/Resolution
      To properly characterize asbestos as dictated by the method requirements
presented in Section 4, it is necessary to resolve all asbestos structures
present in an airborne dust.  The thinnest asbestos fibrils that retain the
recognizable crystalline character of asbestos  (chrysotile fibrils) generally
range in diameter between 0.02 and 0.04 pm (see Section 3.4).  Therefore, the
resolution of the method must exceed 0.02 /zm in order to detect all of the
asbestos structures on a sample specimen.  A magnification of 20,000 is more
than sufficient to resolve asbestos structures over the total range of lengths
and diameters common to asbestos.  A magnification of 10,000 is sufficient to
resolve TEM structures longer than 5 pm over the entire range of diameters
common to asbestos.  The advantage of conducting the scan for long structures
at the lower magnification is the decreased time required and the
corresponding cost savings.  Thus, magnifications of 20,000 and 10,000 are
employed in this method to scan for asbestos structures of all lengths and
asbestos structures longer than 5 ^m, respectively.

      Method requirements presented in Section 4 also indicate that the width
of detected asbestos structures must be properly delineated.  At a
magnification of 10,000,  structures 0.02 fjaa. in diameter appear as thick as
0.2 mm on the viewing screen of the TEM, which is easily seen, but which can
not be easily distinguished from dimensions that vary by factors of less than
2.  Thus,  although structures of such dimensions may be detected at the
magnification specified,  structure dimensions must be characterized at higher
magnifications to adequately distinguish among diameters.   The objective is to
discriminate among diameters that differ by 0.05 /^m.   Such precision can be
obtained at magnifications of 20,000 and greater.   The primary purpose for the
clear delineation of diameters is to facilitate later classification of
candidate size fractions  of potential interest for assessing risks from among
the structures recorded.

      NOTE

      The possibility of  employing lower magnification (less than 10,000)  to
      detect the long (greater than 5 /^m)  structure fraction was considered
      during method development.   However,  this tends to increase variation in
      the  results due to  a combination of instrument  variation and differences
      in operator experience.

      Asbestos counts derived at  lower magnifications (500 -  2000)  appear  to
      vary as a function  of magnification and other microscope character-
      istics.   For example,  at magnifications typically employed in a PCM
      analysis,  small changes  that affect  the average visibility of asbestos
      structures potentially yield large changes in the number of structures

                                      61

-------
      that can be detected.  The visibility of small asbestos structures
      varies from instrument to instrument and with changes in operating
      conditions.  For PCM, a visibility test slide is available for
      standardizing the visibility of each instrument despite differences in
      operating conditions.  Use of the slide is specified in the method.
      However, no such test procedure is available for TEM.  Consequently,  if
      the TEM is to be used at these low magnifications,  there is no mechanism
      for standardizing the visibility of structures to be counted.

      Due to the increased variation in results that generally accompany the
counting of asbestos structures at reduced magnification, such an approach for
determining the longer structure fractions were not incorporated in this
method.  Rather, longer fractions of asbestos are characterized by counting
structures longer than 5 pm at a magnification of 10,000, characterizing their
diameters at increased magnifications, and distinguishing the various
fractions of interest in the count by selecting the structures listed on the
count sheet that exhibit  appropriate lengths and diameters, with diameters
determined to a precision of +/- 0.05 ^m.
            6.1.3.
Rules for Counting, Characterization,  and Recording
      Counting, characterization, and recording rules for this method are
designed to provide:

      (a)   a count of total asbestos structures (individual entities) whose
            distribution Is best described by Poisson statistics;

      (b)   a count of structure components that is not skewed by an arbitrary
            cutoff for the number of individually recognizable components
            present in each structure; and

      (c)   counts of structures in the range of dimensional fractions that
            are believed to best relate to biological activity.

      To achieve these objectives, several modifications to published
procedures are incorporated in this method.  First, the criterion for meeting
specified counting limits is defined in terms of asbestos structures of all
lengths, which represents a count of individual asbestos entities.  Structure
components are not counted individually.  This is to assure appropriate
statistical validity.  Second, when structure components are counted (to
establish an estimate of the total equivalent number of fibers)  the count is
not arbitrarily truncated at 5 components per structure.  Rather, structures
exhibiting more than 5 individually recognizable components are  specially
noted on the count sheet.

      6.2.  FACTORS AFFECTING SENSITIVITY

      For TEM analysis, analytical sensitivity (the concentration
corresponding to the observation of a single asbestos structure) is limited by
the amount of asbestos that can be deposited on a given area of  filter.

                                      62

-------
The amount  (loading) of asbestos  that can be deposited  on  a  filter  is  limited
by two  factors: sampling constraints and the limit at which  a  filter becomes
too loaded  to allow for proper analysis.  Although the  second  of  these factors
is generally more restrictive than the first,  it  is also influenced by the
type of filter preparation employed.  This latter limitation is due,
primarily,  to the concentration of total dust  deposited on the filter  in
coincidence with the asbestos.  Thus, the volume  of air that can  be usefully
collected for asbestos analysis is restricted  by  the concentration  of  dust  in
the air and the quantity of dust  that can be tolerated  on  the  filter before
analysis is prevented.  This limit is higher for  filters prepared by an
indirect technique than for filters prepared by a direct technique.

      The analytical sensitivity  is a combined function of the volume  of air
sampled, the total area of the analytical filter, the dilution or
concentration factor introduced during specimen preparation, and  the area of
the TEM specimen over which asbestos structures are counted.   The
interrelationship between each of these latter factors  determines the  relative
cost of sampling and analysis.

      The relationship between analytical sensitivity,  the volume of air
sampled, the total area of the analytical filter, the dilution or
concentration factor introduced during specimen preparation, and  the area of
the TEM specimen over which asbestos structures are counted  can be  summarized
by the  following relationship:
                            Cs = Af/(N x Ag x V x F)
                                  (6-1)
      where:
                        As
                        N
                        V
                        F
= Analytical sensitivity in structures/liter
= Total area of the analytical filter in mm2
= Area of a TEM specimen grid opening in mm2
= Number of grid openings examined
= Volume of air sampled in liters
ซ- Concentration factor
      The concentration factor, F, simply reflects the dilution or
concentration of asbestos structures resulting from an indirect preparation.
During indirect preparation, asbestos structures deposited on the original
sample filter are dispersed in water and redeposited on the analytical filter.
The density of structures on the final filter is a function of the relative
area of the two filters and the fraction of the total dispersion that is
refiltered:
                             F - Aa x Vf/(Af x Vr)
                                 (6-2)
      where:
                        Aa = Area of filter  ashed in mm2
                        Af = Area of analytical  filter in mm2
                                      63

-------
                        V  — Volume of water used to  redisperse  ashed
                             particles in liters
                        Vf - Volume of dispersion filtered through analytical
                             filter in liters

For samples prepared by direct preparation, the concentration factor is always
unity.

      Actually, equation 6-1 is a contraction of a more general equation
relating the number of structures counted on a filter to the total airborne
asbestos concentration:
      where:
                          C• - Af x S/(N x Ag x V x F)  (6-3)
                        C  - Airborne concentration of asbestos in
                              structures/liter
                        S  - Number of structures counted
                        Af = Total area of the analytical filter in mm2
                        A  = Area of a TEM specimen grid opening in mm2
                        N  = Number of grid openings examined
                        V  - Volume of air sampled in liters
                        F  — Concentration factor
The original equation  6-1 can be obtained from equation 6-3 simply by setting
the number of structures counted equal  to one and substituting the analytical
sensitivity for  the  airborne concentration of asbestos.

      NOTE

      The required analytical sensitivities  for  total asbestos structures and
      for structures longer than 5  too. were derived  as indicated  in Section 4.3
      using equation 6-3 to relate  target concentrations  to analytical
      sensitivity.   The range of target concentrations were derived  as
      indicated  in Section 3.5.
             6.2.1.
Filter Loading
       TEM analysis can proceed as  long as  the  majority of the  particles  on  a
 filter are deposited directly on the filter and do  not overlap other particles
 present.   Both asbestos and non-asbestos  structures may potentially^obscure
 observation of other asbestos structures,  but  at concentrations typical  of
 environmental samples, non-asbestos structures (total  suspended particulate)
 generally determine the upper limit to filter  loading.

       Based on experience,  analysis of a  TEM grid becomes impossible at  a
 relative  coverage of 25%.   Analysis is difficult when  the fraction of the
 specimen  grid covered by particles exceeds 10%. Based  on private conversations
 with several microscopists, experience suggests that a loading of 10 pg/cm  on

                                       64

-------
a  filter will result  in  10% of the filter area being covered with particles.
The data of Steen et  al  (1983) is in general agreement with this estimate but
indicates a range of  potentially appropriate values.  Steen et al (1983)
report that the maximum  tolerated dusts on filters are in the range of
50 /zg/cm2 for filters prepared.by direct preparation and 250 /ig/cm2  for
filters prepared by indirect preparation.  Higher loadings can be tolerated by
indirect preparation  because much of the interfering particle is removed
during filter preparation.

      The reported variation in tolerable loading is not surprising.  The
fraction of filter covered per unit weight of deposit varies as a function of
the average size and  density of the particles.   For spherical particles, the
fraction of coverage  expected as a function of the size of the particles can
be estimated by taking the quotient of the calculated mass of a particle, (the
product of volume and density) and the calculated circumference of the
particle.  The majority  of the non-asbestos fraction of any dust found in the
environment is expected  to be composed primarily of particles that are
approximately spherical.  Based on such calculations, 10% of the filter will
be covered at a loading  of 10 /ig/cm2 when the average particle diameter on the
filter is greater than approximately 1 fj,m at an average particle density of
2.3 g/cm3 (the average density of many silicates).   However,  in different
environments, this value may vary significantly.  Generally, the smaller the
particles, the less the  mass that can be tolerated before the filter becomes
overloaded.  Much of  the variation in loadings reported as tolerable is likely
due to differences in the average size of the particles present in the dust.
            6.2.2.
                        Sampled Air Volumes
      The volume of air that can be collected during asbestos analysis is
limited both by the technical constraints of the sampling system and by the
tolerable limit of total dust that may be deposited before asbestos analysis
becomes impossible due to overloading (as described in Section 6.2.1).  The
volume of air that can be collected per unit area of filter is related to the
tolerable loading by the airborne concentration of total suspended particulate
(see Section 3.6)..  However, because of the extent of variation in the
concentration of airborne particulate (TSP),  it is difficult to derive usable
general averages suitable for deriving the tolerable upper limit to the volume
of air that can be sampled for an asbestos analysis.  At the same time, some
data are available from limited published studies that provide an indication
of reasonable upper limits.

      (a)   In a series of studies of airborne asbestos both in the vicinity
            of the Quebec mines and at remote urban and rural locations,
            Sebastien et al  (1984 and 1986) report that on the order of 1
            of air may be collected per cm2 of filter  for  filters  that are
            prepared by a direct technique.  The tolerable limit for
            indirectly prepared samples is approximately an order of magnitude
            higher.
m
                                      65

-------
      (b)    In a study of building atmospheres  (Chatfield,  ed.  1985)  an
            optimum loading of 0.3  m3 of air per cm2 of filter was recommended
            when filters  were to  be prepared by a  direct  technique.
      (c)
            In one of a series of abatement studies,  Tuckfield et al (1988)
            report that 2 of 24 filters could not be  analyzed due to
            overloading when 0.6 m3 of  air  was collected per  cm2
            the filters were prepared by a direct technique.
                                                                of filter and
      (d)
      (e)
            In a study of rural,  suburban,  and urban air,  Chatfield (1983b)
            reports that filters  prepared by direct analysis were loaded to
            1.2 m3/cm2.

            At the South Bay Superfund site in California,  only a small number
            of filters out of a total of more than 60 filters could not be
            analyzed due to overloading when 1.1 m3 of air  was  collected per
            cm2 of filter and the filters were prepared using a direct
            technique (unpublished results).

      (f)   Burdett (1985a) reports from a study in England that directly
            prepared samples of amosite and crocidolite from abatement sites
            could be loaded for analysis to a level of 0.5 m3/cm2.

      The above data indicate that slightly more than 1 m3  of air can be
collected per cm2 of filter before environmental samples become overloaded so
that analysis is prevented for filters prepared by a direct technique.   The
corresponding limit for abatement samples appears to be on the order of 0.5 m3
per cm2 of filter.  Better estimates for optimum loading can be derived from
data on local TSP concentrations, when such information is available.
            6.2.3.
                        Area to be Scanned for Analysis
      Equation 6-1 may be graphed to depict the combination of sampling
requirements and analytical requirements necessary to obtain a particular
analytical sensitivity.  The two curves presented in Figure 6.1 depict the
relationship between the volume of air to be sampled and the number of grid
openings to be counted to achieve an analytical sensitivity of 0.5 s/L and
0.02 s/L.  These, in turn, represent the analytical sensitivities required to
monitor, respectively, total asbestos structures and long asbestos structures
in environmental samples  (see Section 4.3).  To generate the figure, it is
assumed that an average grid opening size is 0.0081 mm2,  the area of the
analytical filter is 25 mm2,  and the concentration factor is unity.   The
vertical axis for the curve representing an analytical sensitivity of 0.02 s/L
is presented on the right vertical axis of the figure while the vertical axis
for the 0.5 s/L curve is presented on the left axis.

      Based on cost considerations, it is optimal to minimize the number of
grid openings to be counted during analysis.  Experience also indicates that
analysts tire more rapidly when forced to count an excessive number of grid
openings on a single sample, so that precision suffers.  It is recommended

                                      66

-------
that no more than 20 grid openings be counted at the magnification of 20,000
recommended for characterizing total asbestos structures.  Similarly, 100 grid
openings is a recommended target for the 10,000 magnification employed to
characterize long asbestos structures.

      Based on the information presented in Figure 6.1, the recommended limit
for counting grid openings can be satisfied by collecting a sample volume of
5,000 liters for characterizing total asbestos structures..  For a 25 mm
filter, this corresponds to 1.2 m3/cm2,  which  in most  cases  probably  will  not
cause excessive overloading of filters prepared by direct analysis.  However,
a sample volume of 25,000 liters would be necessary to remain within the
recommended grid opening count for long asbestos structures, corresponding to
a loading of 6.5 m3/cm2.  This clearly  exceeds  the  recommended maximum
loadings defined above for filters prepared by direct techniques.  Therefore,
indirect preparation is recommended to keep the number of grid openings that
must be counted within reason (see Section 6.2.4).

      NOTE

      These calculations ignore the limitations imposed by filter blank
      contamination, which are addressed in Section 6.3.1.
            6.2.4.
Filter Preparation
      To provide the required analytical sensitivity at asbestos levels
typically found in the environment, it is necessary either to selectively
concentrate asbestos, or to count structures over a much greater area of a TEM
specimen grid than has traditionally been required.  To the extent that it can
be employed, selective concentration is the least costly of the alternatives
for increasing sensitivity.  The desired degree of concentration requires
filter loadings in excess of what can generally be analyzed using direct-
transfer methods for TEM specimen preparation.  Consequently, an indirect
technique for TEM specimen preparation is employed in this method.

      In most ambient environments, a proportion of suspended particulate is
organic, consisting of soot, spores, pollens and other debris from vegetation.
Organic materials such as these can be removed from the analysis by
low-temperature ashing.  It is also common to find substantial numbers of
calcium sulfate fibers (gypsum) in airborne particles collected in buildings
and urban environments, and particularly in samples collected where demolition
or construction work is in progress.  The fibers are readily released when
plasters and cement products are disturbed.  Gypsum is also generated on air
filters as a consequence of the chemical reaction between collected calcium
carbonate particles and atmospheric sulfur dioxide.  Gypsum can be removed by
dissolution in water.  Another major component of exterior atmospheres is
limestone, either as calcite or dolomite.   Such carbonates, along with some
oxide species,  can be removed by extraction with hydrochloric acid.
                                      67

-------
      FIGURE 6.1: THE RELATIONSHIP BETWEEN ANALYTICAL SENSITIVITY, SAMPLED

                    AIR VOLUME, AND THE NUMBER OF GRID OPENINGS
CO
(U
co  300-
CO

i
(D

00    s
     ง
     u
     s
     CO

     I
200-
    100-
           \
      0
                            Structures
               Total

               Asbestos

               Structures
                                                       Target Analytical Sensitivities:

                                                        0.5 s/1 for total stuctures

                                                        0.02 s/1 for long structures
                                                                          -  600
                                                                             -  400
                                                                          - 200
                10,000        20,000        30,000

                          Volume of Ak Sampled (liters)
                                                                 40,000
50,000
            CO


            I


            GO


            ง
            H-l

            |


            I
            u

            B
            co
            toO
            .a
            g
                                                                                   &
                                                                              D. Wayne Berman

                                                                              ICF Technology

                                                                              Novembers, 1988

-------
      NOTE

      Since the mineralogy of airborne dusts is related to local geological
      and meteorological conditions, the ability to concentrate asbestos from
      dusts by removal of soluble and ashable components varies as a function
      of location.

      Removal of a large proportion of the suspended particulate by use of
these techniques permits the asbestos present in the sample to be concentrated
on to a smaller area of the TEM specimen.  Consequently, the area of the TEM
specimen that must be examined to achieve a particular analytical sensitivity
is proportionately reduced.  Also, many of the non-asbestos fibrous structures
normally found in a directly-prepared TEM specimen, each of which must be
identified and rejected from the asbestos structure count, do not appear on
the indirectly-prepared TEM specimen.

      A selective concentration of the asbestos structures, incorporating low
temperature ashing, re-suspension in water, and acidification with HC1 is
therefore capable of removing substantial amounts of interfering particle from
the analysis, and, for a particular analytical sensitivity, reduces the area
of the TEM specimens that must be examined.  If these procedures are carried
out correctly, no chemical or crystallographic degradation of asbestos can be
detected by routine methods of TEM analysis.

      Indirect TEM specimen preparation techniques offer several advantages
over direct preparation:

      (a)   air samples can be collected without regard to the amount of
            deposit on the filter surface.  The filter loading can be adjusted
            in the laboratory to provide satisfactory TEM specimens;

      (b)   interfering particulate material can be completely or partially
            removed through a combination of dissolution and oxidation
            (ashing);

      (c)   the uniformity of distribution of asbestos structures on the
            filters to be analyzed is improved.

The middle of the above advantages has been addressed above.   The first of the
above advantages is addressed in greater detail in Section 6.2.1.   The last of
the advantages listed is addressed in Section 6.4.3.

      Despite the advantages,  use of indirect preparation techniques  must be
considered carefully because they also suffer from the following
disadvantages:

      (a)   the size distribution of asbestos structures is modified;

      (b)   there is increased opportunity for fiber  loss or introduction of
            extraneous contamination;
                                      69

-------
      (c)   when sample collection filters are ashed, any fiber contamination
            in the filter medium is concentrated.

The second of the above disadvantages can be minimized by maintaining a clean
laboratory and following proper experimental procedures (see Section 6.3).
The last of the above considerations is addressed in Section 6.3.1.  The
extent that structure size distributions are modified during the indirect
preparation of a TEM specimen has been addressed in several studies.

      (a)   In a set of studies, Sebastien et al (1984 and 1986) found that
            samples prepared by an indirect technique exhibit 20 to 50 times
            as many short fibers as paired specimens prepared by a direct
            technique.  Changes in the number of structures longer than 5 fjaa
            were not significant.  It was noted in the study that the absolute
            concentration of aggregates encountered remained the same whether
            samples were prepared by the indirect or the direct technique.
            However, the average dimensions of the aggregates encountered on
            the indirectly prepared specimens appeared to be smaller than
            those observed on the directly prepared specimens.  Because of the
            increase in the number of short structures, the fraction of total
            structures represented by long structures decreased.  Possible
            sources of the increased number of short structures presented in
            the paper are losses during analysis of the directly prepared
            specimens due to obscuration, liberation of short structures  from
            soluble matrices dissolved during indirect preparation, and
            disassociation of short structures loosely bound to the larger
            aggregates.

      (b)   Chatfield  (1985d) in a comparison of specimens prepared by the
            direct and the indirect techniques found that structures shorter
            than 2.5 pm were 4  to 20 times more plentiful on specimens
            prepared by the indirect technique than on specimens prepared by
            the direct technique.  The conversion factor for structures longer
            than 2.5 ^m is less than a factor than 2.  He also noted agreement
            in results between  direct preparations of polycarbonate and MCE
            filters and indirect preparations of  polycarbonate and MCE
            filters.  For the indirect preparations, the deposits  on
            polycarbonate filters were washed into suspension while deposit-
            on MCE filters were ashed.  Data were not reported for  the process
            of washing MCE filters.

      (c)   In another study, Chatfield  (1985b)  reports that for single fibril
            suspensions of chrysotile, there is  no difference  in results
            obtained  from specimens prepared by  direct and  indirect techniques
            either on polycarbonate or MCE filters.  For suspensions that
            contain aggregates, which are more representative  of the types of
            samples likely to be encountered in  the  environment, PCME counts
            were comparable for direct and indirect preparations on both
            polycarbonate and MCE filters but counts of structures  shorter
            than 2.5  pm increased on indirectly  prepared specimens.  Chatfield

                                      70

-------
      did not comment on the fate of long, thin structures in this
      study.  It is unknown whether the source of the additional short
      structures observed following indirect preparation is from
      disaggregation or disintegration.

(d)   In a .study reported by Cook and Marklund (1980), the ratio of
      amphibole asbestos structure counts on paired specimens prepared,
      respectively, directly and indirectly varied between 0.3 and 5.1
      with a mean of 2.  Results from the direct and indirect
      preparations correlated to 95% significance.  Cook and Marklund
      also report that results with chrysotile are much more
      problematic.  Chrysotile samples that contained a high fraction of
      aggregates exhibited counts varying by four orders of magnitude
      from different laboratories (possibly due to lack of
      standardization).  Unfortunately, there is insufficient
      documentation in this paper to determine the exact conditions
      under which specimens were prepared.  However,  they do report
      significant increases in the total mass of chrysotile observed on
      indirect preparations subjected to "harsh" preparation conditions.

      NOTE

      There may be a range of problems contributing to the observations
      presented in this study.   At the time the study was conducted,
      little was known of the effects of organic materials on aqueous
      suspensions of asbestos so that four orders of magnitude variation
      could well have resulted from loss of asbestos  from suspension
      onto container walls,  etc.

(e)   In a series of studies of abatement sites,  Tuckfield et al (1988),
      Hatfield et al (1988), and Chesson et al (1985),  directly prepared
      specimens consistently generated lower structure counts than
      indirectly prepared samples.   Too few structures were counted on
      any of the preparations to derive size specific information.
      However,  they did report that counts from paired direct and
      indirect preparations did not correlate.

(f)   Chesson et al (1989b)  performed a statistical evaluation of data
      from five independent studies where samples prepared,
      respectively,  by direct-transfer and by indirect-transfer
      techniques could be paired for comparison.   Results indicate  that
      samples prepared by an indirect technique uniformly exhibit higher
      total structure counts and higher total fiber counts.
      Unfortunately,  the data were  not sufficient to  determine the
      effect of indirect preparation on specific  size fractions such  as
      the long structures.

      The Chesson et al study also  indicates that,  although counts  on
      directly and indirectly prepared specimens  are  not  strictly
      proportional across studies,  they both track similar trends in

                                71

-------
            concentrations so that the choice of preparation method is
            unlikely to affect comparisons between samples from the same
            study.   A statistically significant relationship was found in all
            five studies between the two sets of specimens (direct and
            indirect preparations).   Quantitative relationships between
            samples prepared by the two techniques are expected to depend both
            on the characteristics of the dusts sampled and on the specifics
            of the sampling and analysis protocols employed in a particular
            study.

      (g)   In a study of amosite, Burdett (1985b) reported a high.correlation
            between samples prepared by a direct technique and those prepared
            by an indirect technique.  The observed ratio of indirect to
            direct varied between 1.7 to 1.  The slight increase in counts on
            the indirectly prepared specimens was attributed to splitting of
            aggregates.  Correspondingly, the average length and diameter
            measured for structures observed on indirect preparations were
            reported to decrease slightly.  It is noted that surfactant was
            used during indirect preparation in this study.

      Based on these findings, specimens prepared by an indirect technique are
appropriate for comparing relative asbestos concentrations at a site.  In
addition, such specimens are likely to yield representative counts of long
structures that correspond to counts on samples prepared by  direct transfer,
provided that gentle conditions are employed during the preparation.  However,
counts of asbestos structures of all sizes (which include short structures) or
counts of individual components within complex asbestos structures of any size
fraction likely vary when performed on specimens prepared by the indirect
technique compared to specimens prepared by the direct technique.  In either
case, quantitative relationships between directly and indirectly prepared
specimens are likely to be study specific due to a dependence on the sampling
and analysis protocols employed.

      The magnitude of differences between counts of long, thin structures
(potentially the most biologically active) performed on directly prepared and
indirectly prepared specimens has not been adequately addressed.
Relationships between counts of asbestos structures, other than single, long
fibers, obtained from specimens prepared by a direct technique or an indirect
technique, respectively, appear to depend on the fraction of aggregates
present in the asbestos dust.  Therefore, they are likely to be site specific
as well as study specific.

      The question as to whether direct or indirect specimen preparation
yields the "correct" result  (in terms of representing biological activity) is
currently unresolved.  It can be argued that the direct methods yield an
under-estimate of the asbestos structure concentration because many of the
asbestos fibers present are  concealed by other particulate material with which
they are associated.  Conversely, the indirect methods can be considered to
                                      72

-------
yield an over-estimate because some types of complex asbestos structures
disintegrate during the preparation, resulting in an increase in the numbers
of structures counted.

      Because of critical limitations imposed by the sensitivity requirements
for sampling environmental asbestos and because there is little reason to
believe that direct preparations track relative risks better than indirect
preparations,  an indirect preparation technique is incorporated in the method.
presented in Part 1.  However, assessing absolute risks requires that current
measurements be compared with risk factors derived from historical
measurements (PCM) that correspond to direct preparations.  Therefore, the
method also incorporates a procedure for evaluating the relationship between
results from directly and indirectly prepared samples for any study employing
this method.,

      A small subset of sample filters will be split so that one half of each
filter may be prepared by a direct and one half by an indirect preparation
technique to allow detailed evaluation of the effects of preparation.  The
total loading on this subset of filters will be restricted to assure that the
level of dust can be tolerated by analysis of the directly prepared filter
section.  Consequently, a larger number of grid openings will be counted on
these samples to achieve desired sensitivity and precision.

      NOTE

      The effect of an indirect preparation is heavily dependent on the
      specific procedures employed.  A detailed protocol has been incorporated
      into this method.

      As indicated by the Burdett study (1985b),  variation in fiber counts
      between samples prepared by direct techniques versus indirect techniques
      appear to occur primarily for preparations of chrysotile.  Preparations
      of amphiboles likely are relatively unaffected by the choice of
      preparation technique.
            6.2.5.
Existing Analytical Methods
      Equation 6-3 is graphed in Figure 6.2 to illustrate the relationship
between sampling and analysis constraints and the concentrations of airborne
asbestos.  In the log-log plot of Figure 6.2, the vertical axis represents the
volume of air sampled, the horizontal axis represents the number of grid
openings counted, and the diagonal lines represent the airborne concentrations
(analytical sensitivities) of asbestos at which at least 1 structure can be
expected to be encountered during analysis.  Because it is a log-log plot, the
hyperbolic curves depicted in Figure 6.1 are represented by straight lines in
Figure 6.2.  The upper curve corresponds to the required analytical
sensitivity for structures longer than 5 (im (0.02 s/L) and the lower curve
corresponds to the analytical sensitivity required to characterize asbestos
                                      73

-------
structures of all sizes In environmental samples (0.5 s/L).   As Figure 6.1,  it
is assumed that an average grid opening is 0.0081 mm2,  that  the analytical
filter is 25 mm2,  and the concentration factor is unity.

      The horizontal line representing 5 m3 has been emphasized in Figure 6.2
because this volume of air corresponds to 1.2 m3/cm2 on a  25 mm filter, which
is the maximum volume of air that can be tolerated for a direct preparation
(see Section 6.2.1).  Published TEM methods discussed in Section 3.8 are
graphed along the 5 m3 line at points corresponding to the number of grid
openings specified to be counted in each method. Thus, each of these points
represent the effective analytical sensitivities for the methods at the
maximum reasonable volume of air that can be collected on a 25 mm filter.  The
Hayward method is depicted twice to represent the analytical sensitivities
corresponding, respectively, to the high magnification scan and the low
magnification scan defined in this method.

       As is readily apparent from the figure, only NIOSH 7402 yields
sufficient sensitivity to detect asbestos structures of all sizes over the
range of concentrations expected to be found in environmental samples.  The
Yamate method, the method specified in AHERA, and the high magnification scan
of the Hayward method would not provide sufficient sensitivity to detect the
entire range of concentrations expected for environmental asbestos.  None of
the methods depicted in Figure 6.2 (including the low magnification scan of
the Hayward method) are sufficiently sensitive to characterize the expected
range of concentrations of long structures  (longer than 5 pm) in environmental
samples.  In the method presented in Part 1 of this report,  the number of grid
openings to be counted is determined by specifying the analytical sensitivity,
specifying the volume of air to be collected, and calculating the required
number of grid openings to be counted using equation 6-1.
            6.2.6.
Analytical Technique
      TEM is potentially capable of achieving the level of analytical
sensitivity required for this method.
            6.2.7.
Magnification/Resolution
      As noted  in Section 3.4 and 4.2, total asbestos structures and asbestos
structures longer than 5 urn are counted separately in this method because of
the disparity in their abundance. Based on the size distributions discussed in
Section 3.4, the probability of encountering an asbestos structure longer than
5 (Jtm in an asbestos dust is only one tenth to one hundredth as likely as
encountering an asbestos structure of any size.  Given the disparity, the most
cost-effective  manner for analyzing the population of total structures and
long structures representatively is to count them separately.  If they were to
be counted together and the limit for total structures was employed as the
defined level of sensitivity, then the chance of encountering even 1 long
structure while counting 10 total structures is very small.  On the other
hand, using the median for long structures as the defined limit of sensitivity
                                      74

-------
FIGURE 6.2: THE RELATIVE ANALYTICAL SENSITIVITIES OF PUBLISHED TEM METHODS
                                Y

                      Method from-AHEIW
                      3   456789 10       2     3

                        Number of Grid Openings Counted
6 7 8 9 100
          D. Wayne Berman
          ICF Technology
          November 3,1988

-------
means that likely 500 total asbestos structures would have to be counted at
the same time that 10 long structures are identified.  Characterizing such a
large number of structures is extremely time consuming and expensive.

      6.3.  FACTORS AFFECTING THE LIMIT OF DETECTION

      The limit of detection is defined as the smallest measurement that is
unlikely to be due entirely to contamination from sources other than the air
being sampled.  More simply, it is the minimum environmental concentration
that can be distinguished (with specified probability) from analytical
background.  To this point, the discussion of performance requirements (in
terms of the analytical sensitivity) have been developed assuming zero
analytical background.  However,  several published studies indicate that the
analytical background for this method is potentially a critical factor
affecting method performance.

      Analytical background may derive both from filter blank contamination
and from contamination contributed by the method of filter preparation.
During an indirect preparation, for example, background contamination can
arise from within the asher chamber, the containers in which the filters are
ashed, the distilled water supply used for re-dispersing the ash, the pipettes
used for transfer of the dispersion for filtration, and the filtration
funnel2.   However,  while contamination introduced during filter preparation
can be minimized by adherence to a rigid program of laboratory cleanliness,
filter blank contamination is simply a function of the quality of commercially
available filters.
            6.3.1.
Filter Blank Contamination
      Asbestos contamination observed on membrane filters has varied
historically.  The ranges of filter blank contamination also differs for MCE
and polycarbonate filters and may differ for different size filters of each
type.  In addition, the extent to which asbestos contamination on such
membrane filters may contribute  to background contamination during analysis
depends on the procedure employed for preparing the sample. Unfortunately,
studies reporting levels of asbestos filter blank contamination frequently
omit critical information concerning either the filter type, size, date of
purchase, or the preparation technique used in the analysis.  The available
             Some microscopists  also  suggest  that  asbestos  losses may occur
             during  sample  collection and  transport when asbestos structures
             are shaken off of the  filter  and redeposited on  the cowl or
             cassette.   Others contend that this is not a significant problem
             and that  the data used to identify it is  flawed  by not  controlling
             for the effects of  the differences between direct and indirect
             preparation.   The problem has yet to  be confirmed in a  published
             study.

                                       76

-------
data appear contradictory.  Consequently, it is difficult to quantify a
representative range of asbestos contamination levels for filter batches that
are available today.
                  6.3.1.1.
Contamination on MCE Filters
      Prior to 1975, the level of contamination on MCE filters prepared by the
direct method was reported in some batches to be sufficient to preclude use in
standard asbestos analysis (Chatfield, 19,85a).  Prescreening of filter batches
was therefore recommended.  According to current consensus, levels of
contamination detected on MCE filters prepared by a direct preparation now
appear to be minimal and these filters ar.e acceptable for most asbestos work.
However, published data is limited and published data addressing 25 mm filters
specifically is lacking.

      (a)   In a study of indoor asbestos contamination, Hatfield et al (1988)
            report 1 chrysotile asbestos fiber detected on 1 of 19 MCE filter
            •blanks (37 mm, 0.45 /im pore size).  This corresponds to an upper
            limit to background of less than 12 s/mm2.

      (b)   The EPA study on filter blank contamination (USEPA, 1986b)
            concluded with the observation that contamination on MCE filters
            appeared to be composed predominantly of short chrysotile
            structures and that MCE filters are recommended for asbestos
            sampling and analysis.

      (c)   Chatfield (1985d) also reports that most of the contamination on
            filter blanks is chrysotile and that current levels of blank
            contamination are acceptable.  This paper also suggests that short
            fibers predominate, although this observation ,may be due simply to
            lack of a statistically balanced count.

       For MCE filters prepared by an indirect technique,  the situation is
less clear.  General consensus indicates that levels of contamination are
acceptable when MCE filters are ashed.  But published data, especially for
recent batches of 25 mm filters are limited.   Published data that address
background contamination resulting from the washing of MCE filters are
lacking,

      (a)   In a study of abatement techniques, Chesson et al (1985) report
            detection of no asbestos contamination on laboratory blanks and
            field blank MCE filters (47 mm,  0.45 ^m pore size)  prepared by
            ashing the filter and resuspending the deposit in distilled water
            by sonication.   This means that background contamination on the
            filters is less than 12 s/mm2 based on the  analysis  performed.
            Note that the filtrate was redeposited on a polycarbonate filter
            (25 mm,  0,2 /j.m pore size)  so that the background could also
            represent an upper limit to contamination on polycarbonate filters
            subjected to direct preparation.
                                      77

-------
      (b)   In a followup study of abatement techniques,  Chesson et al (1986b)
            report levels ranging between 0 and 58 s/mm2 with  a  mean of  30
            s/mm2 on laboratory blank MCE filters  (47 mm,  0.45 /im pore size)
            prepared by ashing the filter and resuspending the deposit in
            distilled water by sonication.  Field blanks  appear to  exhibit a
            slightly higher average asbestos contamination level than the
            laboratory blanks.  Note that the filtrate  was redeposited on a
            polycarbonate filter (25 mm,  0.2 /im pore size) so  that  the
            background could also represent contamination due  to direct
            preparation of a polycarbonate filter.

      (c)   In another followup to the series of abatement papers,  Tuckfield
            et al (1988) report levels ranging between 0  and 50 s/mm2 with a
            mean of 20 s/mm2 on 30 field  and laboratory blank  MCE filters
            (47 mm, 0.45 p.m pore size)  prepared by ashing the  filter and
            resuspending the deposit in distilled water by sonication.  Note
            that the filtrate was redeposited on a polycarbonate filter (25mm,
            0.2 pm pore size) so that the background observed could also
            represent contamination due to direct preparation of polycarbonate
            filters.

      (d)   Individual microscopists (for example, G.  Dunmeyer,  1988) have
            indicated in private conversations that indirect preparation by
            washing MCE filters results in efficient transfer of asbestos and
            that blank contamination is low (less than 15 s/mm2) .  Filters
            considered were 25 mm, 0.8 /im pore size MCE filters.  However,
            there are no published 'data to confirm such reports.

                  6.3.1.2.    Asbestos Contamination on Polycarbonate Filters

      The level of asbestos contamination on polycarbonate filters  has
historically been reported to be higher than MCE filters.   Available data
addressing current levels of contamination on polycarbonate filters is
inconclusive for filters prepared by direct preparation and published data
addressing background contamination on polycarbonate filters prepared by an
indirect technique  (either ashing or washing) appears to  be lacking.  General
consensus indicates that background contamination on current batches of
polycarbonate filters is generally higher than MCE filters when both are
prepared by the direct technique.  However, published data are limited.

      (a)   In an abatement study, Tuckfield et al (1988) prepared 47 mm
            polycarbonate filters (pore size not reported) by a  direct
            technique and measured the background on a series of•laboratory
            and field blanks.  Levels reported range between 0 and 40 s/mm
            with an arithmetic mean of 7  s/mm2.  Note that an amphibole fiber
            was detected on one of the blanks.

      (b)   In the workshop on blank contamination (USEPA, 1986b),  it was
            reported that polycarbonate filters prepared by a direct  technique
            exhibit variable levels of blank asbestos contamination.  Levels

                                      78

-------
            appear to average about 2-6 structures/10 grid openings (25-75
            structures/mm2)  but particular lots may contain 10 times this
            amount.  It was recommended that filter lots with more than 5
            structures/10 grid openings (60 s/mm2)  be returned to the
            manufacturer.  In the same study, use of MCE filters was recom-
            mended because polycarbonate blank contamination appears to
            include both chrysotile and amphibole and a significant fraction
            of structures longer than 5 /im were encountered.

      (c)   Chatfield (1985b) reports that although contamination of MCE
            filters appears to have been minimized, contamination of
            polycarbonate filters still appears to be a problem.

      (d)   The studies reported in Section 6.2.1.1 indicate that an upper
            limit to filter blank contamination on polycarbonate filters
            prepared by a direct technique is 30 s/mm2 based on contamination
            observed following an indirect preparation of MCE filters in which
            polycarbonate filters were used to filter the final suspension.

      (e)   There appears to be a consensus among several microscopists that
            at least one group of microscopists consistently report lower
            values for polycarbonate filter contamination.  However, there is
            no published data addressing this discrepancy and it has not been
            resolved.
                  6.3.1.3.
Estimating Filter Blank Contamination Levels
      Although qualitative conclusions may be drawn from the available data,
additional research is recommended in this area.  General consensus indicates
that most current batches of MCE filters, when prepared by a direct technique,
should yield filter blank contamination levels below 30 s/mm2.   MCE filters
prepared by ashing should also yield background below 30 s/mm2.   Washing of
polycarbonate filters yields comparably low levels of background
contamination.  Published data, however, is too sparse to support any of these
numbers as typical.  Therefore, at a minimum, duplicate filter blanks from
every batch used in a study should be prepared and analyzed as per the
technique to be used in the study to quantify the level of background.

      If the range of filter blank contamination found during a preliminary
evaluation is not clearly distinguishable from typical levels expected to be
encountered during the study, either the filter batch should be returned for a
new one (which must also be tested) or the study protocol should be
redesigned.  It is assumed for this method that filter blank contamination can
be maintained below 30 s/mm2.   It is further  assumed that the size
distribution of asbestos appearing as filter blank contamination parallels
typical chrysotile distributions as presented in Section 3.4.  Therefore, if
the count of structures of all sizes exceeds background, the count of any size
fraction will also exceed the level of background corresponding to that
                                      79

-------
particular size fraction.  To minimize interference with filter blank
contamination, 60 s/mm2 is set as a target minimum concentration for this
method.

      NOTE

      The potential impact of asbestos contamination contributed by the
      membrane filter can be minimized by maximizing the volume of air
      collected per unit area of filter.  However, collected air volumes must
      be limited to prevent overloading, especially for filters to be prepared
      by a direct-transfer technique (see Sections 6.2.1 and 6.2.4).

      Within the range of the maximum air volume that can be tolerated on a
      filter due to constraints associated with filter loading, there is also
      the problem of the limit of the rate at which air can be pumped through
      a filter given the capabilities of available pumps (see Section 2.2 of
      The Method, Part 1 of this report).   If the capabilities of the pump are
      limiting in a given situation, the total volume of air to be collected
      can theoretically be minimized without compromising analytical
      sensitivity by selecting the smallest filter (25 mm) that is practical
      for sample collection.
            6.3.2.
Conclusions
      Although specification of a detection limit for this method requires
knowledge of the range and variation of analytical background and is therefore
study specific, general conclusions concerning the ability to achieve a
reasonable limit of detection may be drawn from a comparison of the relative
contributions to asbestos counts from environmental and analytical sources.
For filters prepared for TEM analysis in a particular study, the absolute
limit to detection is unlikely to be lower than the concentration at which the
density of structures deposited on the filter from the air is just equal to
the mean density of structures present on the blank filter (30 s/mm2).
Ideally, however, sampled asbestos should be deposited on the filter at
several times the concentration of asbestos present due to analytical
background so that distinguishing sampled asbestos from analytical
contamination is trivial.

      Given the upper range of the target concentrations derived in S'ection
3.5 (5 s/L for total asbestos structures and 0.2 s/L for long asbestos
structures) , an air volume of 1 m3 of air per cm2  of  filter will  deposit  50
asbestos structures of all sizes per mm2 of filter and 2 long structures  per
mm2 of filter respectively.   The fact that the first of these numbers exceeds
projected filter blank contamination levels by less than a factor of 2
indicates that preparing these filters by a direct technique, which is subject
to the 1 m3/cm2 limitation,  likely presents  a marginally acceptable method  for
characterizing asbestos structures of all lengths, at best.   Direct
preparation is clearly unacceptable for characterizing the concentrations of
long structures, because the major contribution to observed asbestos will be
                                      80

-------
from analytical background.  Therefore, an indirect preparation protocol is
incorporated in this method.  The consequences of the selection of an indirect
preparation procedure are addressed in Section 6.2.4.

      NOTE

      At a limit of 30 s/mm2 on a 25  mm filter with 1 m3/cm2 air collected, 1
      structure observed corresponds a concentration (analytical sensitivity)
      of 3 s/L for structures of all sizes and 0.1 s/L for structures longer
      than 5 pm.  Clearly, limiting the quantity of air that can be collected
      to 1 m3/cm2  (as  is  the  case  for a direct preparation)  is not  sufficient
      to achieve the required analytical sensitivities defined for this method
      (see Sections 4.3 and 6.2).

      6.4.  FACTORS AFFECTING PRECISION

      To maximize precision, potential sources of variation within the method
must be controlled.  To minimize systematic variation during sampling and
analysis, the method specifies detailed procedures.  Random variation is
minimized by maximizing the number of structures to be identified and counted,
which includes increasing the probability of encountering asbestos structures
during analysis, as discussed in Section 6.2.  In addition, detailed and
unambiguous counting rules and rules for identification are specified in the
method to minimize variation due to subjective interpretation.  Specific
factors that must be controlled to maximize precision are:

      (a)   generating a uniform deposit on the analytical filter;

      (b)   minimizing subjectivity in the interpretation of counting rules;

      (c)   minimizing variation introduced by small changes in instrument
            performance;  and

      (d)   minimizing statistical variation from counting too small a number
            of structures.
            6.4.1.
Analytical Technique
      TEM is the analytical tool selected for this method so that analysis can
be performed at sufficient magnification to minimize variation due to small
differences in the performance characteristics of the instrument or the
experience of the analyst.  However, several performance criteria must be
specified to minimize the impact on precision.

      In a comparison study, Steel and Small (1985) report that TEM
instruments vary with age and type.  Differences between instruments that
contribute the most to analytical variation include image quality
(resolution), contrast, and brightness.  Electron beam dose and mechanical
stage operation are also important.  Resolution was judged by the ability to
distinguish the hollow core of chrysotile fibrils.  Contrast and brightness

                                      81

-------
are controlled by adjusting the voltage and aperture sizes of the instrument.
In the study, not all instruments could be configured to perform equally well.
Some of the oldest TEM instruments could not be adjusted to provide sufficient
resolution to observe the central canals of chrysotile fibrils clearly.

      The beam dose effects the counting process by causing damage.  If too
high a beam current is required by an instrument to achieve the required
brightness or resolution, than the potential to damage the sample may become
so great that the ability to scan properly may be lost.  This is particularly
true when attempting to generate SAED spectra.

      The operation of the mechanical stage is crucial to achieve good
precision in counts.  Approximately 20-30 traverses are needed to complete a
scan of a single grid opening.  Thus, the extent that a stage wanders during a
traverse may affect the time necessary and the ability to precisely count all
fibers in a grid opening.  Some of the stages tested by Steel and Small
wandered by 30 /im during a single traverse.  However, the TEM used for the
operator tests in the study wandered by less than 1 /^m in a 100 nm linear
traverse.

      Based on the work of Steel and Small, it is recommended that TEM
instruments to be used with this method exhibit a resolution at 20,000
magnification that is sufficient to clearly resolve the central canal of the
finest chrysotile fibril.  In addition, the mechanical stage of the instrument
roust be shown to wander by less than 1 pm per 100 pm linear traverse.
Finally, it is recommended that specific performance criteria addressing beam
dose, brightness, and contrast be developed for instruments to be used with
this method.
            6.4.2.
Magnification/Resolution
      The precision of this method is affected by the magnification selected
to the extent that the ability to detect asbestos structures lies close to the
limit of the associated resolution.  Microscopists vary in their ability to
detect very small asbestos structures at magnifications used for scanning in
this method, 10,000 and 20,000, (Steel and Small, 1985).  It has been reported
that only 50% of asbestos structures shorter than 0.5 pm were detected by all
analysts in a round-robin study.  Below this length, the fraction of analysts
who regularly detect such structures decreases rapidly.  At the same time,
proper characterization of the size fractions present in an asbestos dust
requires that the mode of the distribution be included in the count.  The
published size distributions presented in Section 3.4 indicate that the mode
of asbestos length distributions in most dusts falls within a range from
0.8 pm to 1.2 pm.  Therefore, as a reasonable compromise allowing the
detection of the mode of asbestos distributions while addressing the limited
ability of analysts to detect short structures, a minimum length of 0.5 /^m has
been defined as the shortest fiber to be incorporated in the reported results
for this method.
                                      82

-------
      The minimum detectable width for asbestos structures counted in this
method is determined by the ability of an operator to detect them in a routine
examination at the specified magnification of the method.   The minimum
magnification specified in this method (10,000) is more than sufficient to
assure visibility of the thinnest chrysotile asbestos fibrils.
            6.4.3.
Uniformity of Filter Deposit
      To meaningfully extrapolate structure counts performed over a limited
area to the concentration of structures on an entire filter, the nature of the
distribution of structures on the filter must be known and the distribution
must be reasonably uniform.  Chatfield (1985b) has shown that the distribution
of structures on analytical filters prepared by a direct technique are not
uniform.  Using a chi-squared test, Chatfield found that 100% of the directly
prepared filters that exhibited more than 30 structures/ 10 grid openings had
to be rejected as non-uniform at a 1% level of significance.  The fraction of
filters that passed the chi-squared test for uniformity increased as the
loadings on the filters decreased, but this was attributed to the presence of
an insufficient number of structures to properly determine a distribution.  In
contrast, nearly all of the filters prepared by indirect preparation
techniques exhibited uniform deposits at a 1% level of significance and 100%
bf the indirectly prepared samples could be shown to be uniform at reduced
levels of significance.

      Due to the apparent characteristics of air deposited asbestos, directly
prepared filters do not exhibit uniform deposits.  Consequently, indirect
preparation has been incorporated into this method.

            6.4.4.  Rules for Counting, Characterization, and Recording

      Based on a study by Steel and Small (1985), operator error is one of
several potential sources of variation.  The problem is worse if counting
procedures are not standardized.  It is also reported that this variation can
be reduced when counting rules are standardized and specified in detail.
Performance also varies as a function of observer experience and the degree of
professional time pressure under which an analysis is performed.  To account
for these observations, detailed standardized counting rules have been
incorporated in this method.  It is also recommended that minimum requirements
for experience be developed for the analysts.  This may include, for example,
mandatory participation by analysts in round-robin studies.

      Once the concentration of structures collected on a filter is sufficient
to be observed, the number of structures detected affects the precision of the
measurement.  Precision tends to increase as the square root of the number of
structures counted.  This implies that the incremental improvement in
precision derived from each individual structure counted decreases as the
total number of structures counted increases.
                                      83

-------
      Burdett (1985a) got significant improvement in precision by increasing
counts between 20 and 50 structures.  He reports, however, that the
incremental improvement in precision for structures counted beyond 50 are
minimal.  The characterization of structures to be counted in this method
includes distinguishing among counts of several sub-types (bundles, clusters,
matrices, and fibers).  Based on Burdett's data, improvements in the precision
of the count of each sub-type is expected until 50 structures of each sub type
are counted.  However, this would require counting more than of 400 total
asbestos structures, an excessive requirement.  Therefore, an upper limit to
the count of total asbestos structures of 50 is defined for this method and a
lower level of precision is considered acceptable for the count of each sub-
type.  A similar strategy is adopted for structures longer than 5 urn.

      NOTE

      A count of 100 total asbestos structures and 100 structures longer than
      5 ftm is defined in this method for Phase 2 samples because of a need for
      increased precision in the counts of sub-types in these samples (see
      Section 6.4.2.
                                      84

-------
7.
RECOMMENDED PROCEDURES FOR MANIPULATING ANALYTICAL DATA
      Because the principle objective of the sampling and analysis method
presented in Part 1 of this report is to provide results suitable for
supporting a risk assessment, this section addresses the kinds of data
manipulation typically performed to complete a risk assessment.  The
procedures presented here are recommended for general application.  It is
recognized, however, that alternate valid approaches are also available so
that the recommended procedures should not be viewed as strict requirements.
In addition, specific circumstances may present special computational needs
beyond the scope of what is addressed here.  To preserve the integrity and the
intent of data derived from this method, it is recommended that a statistician
be consulted prior to the adoption of alternate computational procedures for
manipulating data as part of a site evaluation.

      In a risk assessment, field results are typically combined to provide
average concentrations representative of a given location or a given period of
time.  Frequently, long-term (multiple-year) time averages are desired because
lifetime rather than short-term risks are generally addressed.  Depending on
the particular purpose for the computation, average concentrations may be
derived either from homogeneous samples (samples intended to represent the
same sampling environment) or from non-homogeneous samples (samples collected
in different sampling environments).   Procedures for deriving averages from
homogeneous and non-homogeneous sets of samples are provided in Section 7.1
along with instructions for constructing confidence limits around averages of
homogeneous samples (Section 7.2) and non-homogeneous samples (Section 7.3).

      Once average concentrations are computed, they are generally compared to
distinguish anthropogenic contamination from background or to distinguish
among levels of contamination potentially contributed from different sources.
Such comparisons examine the relative differences of various measurements.
However, long-term average exposure concentrations are also typically compared
to existing risk factors to establish the magnitude of absolute risk.  This
last use of analytical data relies on the absolute concentration derived from
a particular set of measurements.  A procedure for conducting a statistical
test to distinguish among the average concentrations derived from various
sample sets is presented in Section 7.4.

      The significance of considering the need for a relative verses an
absolute measurement is to determine whether analytical results must be
adjusted to account for analytical background contamination (see Section 11.4
of Part 1 of this report).  While unnecessary and not recommended for the
manipulation of relative values, the adjustment for analytical background is
required to derive an absolute measurement.  A procedure for adjusting
measurements to account for analytical background is provided in Section 7.5.

      NOTE

      Statistical tests should all be performed on unadjusted analysis
      results.
                                      85

-------
      7.1.  COMPUTING AVERAGE CONCENTRATIONS FROM MULTIPLE SAMPLES

      As indicated above, it is frequently helpful to summarize results from a
set of samples by computing an average.  For several reasons, it is
recommended that arithmetic averages be used exclusively.  First, averages are
often used to judge risk to health and such risk is more likely to be a
function of the arithmetic mean than, say, a geometric mean (e.g., the risk
from breathing 0.2 f/ml 12 hours per day and 0.00002 f/ml 12 hours per day is
probably closer to the risk from constantly breathing 0.1001 f/ml [the
arithmetic mean] than breathing 0.002 f/ml [the geometric mean]).  Second, the
geometric mean is zero whenever one of the individual samples results is zero,
and ad hoc adjustments to address this problem (such as replacing zero
estimates by detection limits) can introduce serious biases.

     If samples are considered to be homogeneous (designed to represent the
same sampling environment), it is recommended that simple un-weighted
arithmetic averages be calculated.  For example, denote the calculated
airborne concentrations from a group of K samples by Clt  ...,  CK.  The
estimated average concentration, (J, would then be

      C - (G! + . . .  + CK)/K.

      If the analytical sensitivities of the samples to be averaged  differ
considerably, it may be appropriate to compute a weighted average with weights
set equal to the inverses of the analytic sensitivities.  This is equivalent
to computing the average by dividing the total number of structures  found in
all of the samples by the total volume of air passed through all of  the sample
filters.

      Frequently, it will also be necessary to compute an average for a
collection of non-homogeneous samples.  For example, one might wish  to
calculate the average exposure of an individual who frequents several areas
with varying estimated airborne concentrations of asbestos.  It is recommended
that such an average be obtained as follows.  First, calculate an average
concentration from each homogeneous area by averaging the set of samples
collected in each such area as indicated above.  Averages derived for the
homogeneous areas should then be combined in a weighted average, with weights
equal to the proportions of time the individual spends in each of the various
areas.  Thus, if TJj_,  (T2ป • • -^k are *-he average concentrations computed from K
groups of homogeneous samples using the previous expression, then the overall
average exposure concentration experienced by the individual under
consideration, based on the respective weights Wlf...WK,  would be

      CH - (Wi^j. 4- . . . + WK'CK)/(W1 +  ... +• WK) .

If desired, one can test whether groups of samples are homogeneous using the
Wilcoxon and Kruskal-Wallis tests (see Section 7.4).               :
                                      86

-------
      7.2.  COMPUTING CONFIDENCE LIMITS FOR AVERAGES OF HOMOGENEOUS SAMPLES

      If there are no more than 30 structures in a combined group of K  samples
then the confidence limits for the average concentration  is computed based on
the Poisson distribution.  Let x be the total number of structures on the
combined filters (the number of structures in a single sample  is the estimated
concentration divided by the analytical sensitivity).  Let Xj, be the 95% upper
confidence limit on the total count x, calculated from the Poisson
distribution as described in detail in Section E.4.2 (Appendix E of The
Method, Part 1 of this report).  The upper 95% limit on the average
concentration, Cy,  is then x^,  divided  by  the  sum of the  reciprocals  of the
analytical sensitivities, i.e.,
                        + 1/SK).
The lower limit is determined from the same formula except that a  suitably
defined lower limit XL replaces x^.

      If more than 30 structures are found in a combined group of  K  samples
then confidence limits are also calculated by the following procedure and the
most extreme values are reported.  Let Ct be the calculated concentration for
the ith samplej  let C" be the arithmetic average, and let sc2  be the
corresponding sample variance,
            k-1
k
2
                         - a")
Then the 95% confidence limits on the average concentration are computed as

      Cj, =  C" + 1.65 sc      .

and

      CL equal to the larger of C" -  1.65 sc or zero.


      7.3.  COMPUTING CONFIDENCE LIMITS FOR AVERAGES OF NON-HOMOGENEOUS
            SAMPLES

      Suppose C'-L, . . . C"K  are  average concentrations computed from K groups of
homogenous samples, scl2,...scK2 are  the  corresponding variance terms,  to
compute confidence limits for the weighted average concentration estimated by
                                      87

-------
      CH - (W^ +  . . . 4- WKCK)/(W1 +  . . . + WK) .

Define

      SH2 -  (W^S^2 +  . . . + WK2ScK2)/(Wl +  ... + WK)2.

The 95% upper and lower limits, respectively,  are then

      CHu -  CH + 1.65  sw

and CHI  equal to the larger  of

      CH - 1.65 sw  or zero.
      7.4.  TESTING FOR DIFFERENCES BETWEEN CONCENTRATIONS DERIVED FROM
            DIFFERENT SETS OF SAMPLES

      It is recommended that the Wilcoxon non-parametric test (Hollander and
Wolfe 1973) be used to test whether airborne concentrations measured from one
set of samples are significantly different from those measured by another set.
Such a test can be used, for example, to determine whether concentrations are
higher in one environmental area than another (e.g., near an asbestos disposal
site verses a more distant control area).  The test is also recommended to
distinguish "sampled" asbestos from contamination attributed to analytical
background.

      The test should be applied directly to the unadjusted means reported for
measurements of airborne concentrations.  A mean of zero, "0", should be used
in place of all measurements reported as "NF" for not found.  A detailed
description of the application of the Wilcoxon Test is provided in Section E.5
(Appendix E of The Method, Part 1 of this report).

      Care should be taken to exclude possible sources of systematic bias when
applying the Wilcoxon test or similar tests.  Such bias can result, for
example, if:

      (1)   different laboratories analyzed the groups of samples being
            compared;

      (2)   the groups of samples were analyzed by the same laboratory but by
            distinct analysts;

      (3)   the samples were analyzed by the same analysts at the same
            laboratory but before and after a modification to laboratory
            procedure was adopted;

      (4)   the samples from the two groups were collected on different types
            or batches of filters; or
                                      88

-------
      (5)   the groups of samples were collected or prepared using radically
            different procedures.

Such sources of bias can be minimized by keeping experimental conditions as
similar as possible when collecting groups of samples intended for comparison,
randomizing filters among laboratories or among analysts whenever multiple
laboratories or analysts within laboratories are used to analyze filters, and
withholding from analysts the identity of the source locations of the filters.

     To test for non-homogeneities among airborne concentrations from several
sites the non-parametric Kruskal-Wallis test can be used (Hollander and Wolfe
1973) .   This test is a generalization of the Wilcoxon test that can be applied
to more than two groups of data.  If analyses involve samples analyzed at
separate laboratories, it is strongly suggested that laboratory be used as a
blocking factor (Lehmann 1975),  so that direct comparisons are made only among
samples analyzed at the same laboratory.

      7.5.  ADJUSTING MEASURED CONCENTRATIONS FOR CONTRIBUTIONS FROM
            ANALYTICAL BACKGROUND CONTAMINATION

      When it becomes necessary to determine the absolute magnitude of a
measured asbestos concentration (as when measured concentrations are compared
to existing risk factors, to establish the magnitude of risk), mean
measurements of airborne concentrations must be adjusted to account for
contributions from analytical background contamination.  As addressed in
Section E.5 (Appendix E to The Method, Part 1 of this report), it is strongly
recommended that such adjustments be applied to averages of groups of samples
rather than individual sample analysis results.

      To adjust an average concentration computed for a group of samples
weighted by the inverse of their individual analytical sensitivities, proceed
as follows.  Let x be the total number of structures detected in the used
filters, AT the total area examined from these filters,  and BC the surface
concentration of structures obtained from an appropriate set of blanks.  The
adjusted airborne concentration is calculated as

        (x/AT - BC)  AT  .
      I/Si + ...  + 1/SK

      As indicated in Section E.5, blank analyses may be obtained either from
filters analyzed from the same batch used for sample collection or from a
cumulative database of historical blank analyses maintained by the laboratory
performing the analyses.  Depending on the given application, one or the other
of these data sets may be considered most appropriate.  When adjusting sample
analysis results for contributions from analytical background, the source of
the blank data must be specified.
                                      89

-------
                                  REFERENCES
Altree-Williams, S., and Preston, J.S. (1985):  Asbestos and other fiber
      levels in buildings.  Annals of Occupational Hygiene, Vol. 29, No. 3,
      pp. 357-363.

Beckett, S.T., and Attfield, M.D. (1974):  Inter-Laboratory comparisons of the
      counting of asbestos fibers sampled on membrane filters.   Annals of
      Occupational Hygiene, Vol. 17, p. 85.

Beckett, S.T., Hey, R.K.,  Hirst, R., Hunt, R.D.,  Jarvis, J.L.,  andRickards,
      A.L. (1976):  A comparison of airborne asbestos fibre counting with and
      without an eyepiece graticule.  Annals of Occupational Hygiene, Vol. 19,
      pp. 69-76.

Beckett, S.T., and Jarvis, J.L.  (1979):  A study of the size distribution of
      airborne amosite fibres in the manufacture of asbestos insulation
      boards.  Annals of Occupational Hygiene, Vol. 22, pp. 273-284.

Bertrand, R., and Pezerat, H. (1980):  Fibrous glass: carcinogenicity and
      dimensional characteristics.  In: Biological Effects of Mineral Fibres,
      Vol. 2, (J.C. Wagner, ed.).  IARC Scientific Publications, Lyon, France.
      Pp. 901-911.

Bonneau, L., Malard, C. and Pezerat, H. (1986):  Studies on surface properties
      of asbestos.  Environmental Research, Vol.  41, pp. 268-275.

Burdett, G. J. (1985a):  The measurement of airborne asbestos releases from
      damaged amosite insulation subjected to physical attrition.  In:
      Asbestos Fibre Measurements in Building Atmospheres: Proceedings  (E.J.
      Chatfield, ed.).  Ontario Research Foundation, Mississauga, Ontario,
      Canada, (1986).  Pp. 209-298.

Burdett, G.J. (1985b):  Use of membrane-filter, direct-transfer technique for
      monitoring environmental asbestos releases.  In:  Asbestos Fibre
      Measurements in Building Atmospheres: Proceedings (E.J. Chatfield, ed.).
      Ontario Research Foundation, Mississauga, Ontario, Canada, (1986).  Pp.
      87-114.

Burdett, G.J., and Jaffrey, S.A.M.T.  (1986):  Airborne asbestos concentrations
      in buildings.  Annals of Occupational Hygiene, Vol. 30, No. 2, pp. 185-
      199.

Chatfield, E.J. (1979):  Measurement of asbestos fibres in the workplace and
      ฃn the general environment.   In:  MAC Short Course in Mineralogical
      Techniques of Asbestos Determination (R.L. Ledoux, ed.).   Mineralogical
      Association of Canada, Toronto, Canada.  Pp. 111-164.
                                      90

-------
Chatfield, E.J. (1983a):   Short mineral fibres in airborne dust.  In:  Short
      and Thin Mineral Fibres (E.J.  Chatfield, ed.).  National Board of
      Occupational Safety and Health, Solna,  Sweden.  Pp. 9-93.

Chatfield, E.J. (1983b):   Measurement of asbestos fiber concentrations in
      ambient atmospheres.  Royal Commission on Matters of Health and Safety
      Arising from the Use of Asbestos in Ontario.  Publications Mail Order
      Service, 880 Bay St., 5th floor, Toronto, Ontario, Canada M7A-1N8.

Chatfield, E.J., ed. (1985):  Asbestos Fibre Measurements in Building
      Atmospheres: Proceedings.  Ontario Research Foundation, Mississauga,
      Ontario, Canada. (Printed in 1986).

Chatfield, E.J. (1985a):   Overview of measurement procedures for determination
      of asbestos fibres in building atmospheres.  In:  Asbestos Fibre
      Measurements in Building Atmospheres: Proceedings  (E.J. Chatfield, ed.).
      Ontario Research Foundation, Mississauga, Ontario, Canada, (1986).  Pp.
      7-24.

Chatfield, E.J. (1985b):   Limitations of precision and accuracy in analytical
      techniques based on fibre counting.  In:  Asbestos Fibre Measurements in
      Building Atmospheres: Proceedings (E.J. Chatfield, ed.).  Ontario
      Research Foundation, Mississauga, Ontario, Canada.  Pp. 115-137.

Chatfield, E.J. (1985c):   Airborne asbestos levels in Canadian public
      buildings.  In:  Asbestos Fibre Measurements in Building Atmospheres:
      Proceedings.  Ontario Research Foundation, Mississauga, Ontario, Canada.
      Pp. 177-207.

Chatfield, E.J. (1985d):   Asbestos measurements in workplaces and ambient
      atmospheres.  In:  Electron Microscopy in Forensic, Occupational, and
      Environmental Health Sciences  (Basu and Millette, eds.).  Ontario,
      Canada.  Pp. 149-186.

Cherrie, J.W., Dodgson, J., Groat, S., and Carson, M. (1987):  Comparison of
      Optical and Electron Microscopy for Evaluating Airborne Asbestos/Report
      No. TM-87-01.  Institute of Occupational Medicine, Edinburgh.  U.S.
      Dept. Commerce, National Technical Information Service Reprint.

Chesson, J., Margeson, D.P., Ogden J., Bauer, K., Constant, P.C., Bergman,
      F.J., and Rose, D.P. (1985):  Evaluation of Asbestos Abatement
      Techniques, Phase 1:  Removal.  U.S. EPA Report No. 560/5-85-109,
      October, 1985.  U.S. EPA, Washington, D.C., U.S.A.

Chesson, J., Margeson, D.P., Ogden J., Bauer, K., Constant, P.C., Bergman,
      F.J., and Rose, D.P. (1986):  Final Report on Task 4:  Evaluation of
      Asbestos Abatement Techniques, Phase 2:  Encapsulation with Latex Paint.
      U.S. EPA Report No. 560/5-86-016, July, 1986.  U.S. EPA, Washington,
      D.C., U.S.A.
                                      91

-------
Ghesson, J.,  Rench, J., Schultz,  B.,  and Milne,  K.  (1989a):   Interpretation of
      Airborne Asbestos Measurements  (Draft).   Chesson Consulting,  Washington,
      D.C., U.S.A.  Pp. 1-45. (In press)

Chesson, J.,  Hatfield, J., and Leczynski, B. (1989b):   Comparison of Airborne
      Asbestos Levels Determined by Transmission Electron Microscopy (TEM)
      Using Direct and Indirect Transfer Techniques.  Prepared for the
      Exposure Evaluation Division, Office of Toxic Substances,  Office of
      Pesticides and Toxic Substances,  U.S. EPA, Washington,  D.C.,  U.S.A.,
      Contract No. 68-02-4294.

Constant, P.O., Bergman, F.J., Atkinson, G.R.,  Rose, D.R., Watts, D.L.,  Logue,
      E.E., Hartwell, T.D.,  Price, B.P., and Ogden, J.S.  (1983):  Airborne
      Asbestos Levels in Schools, U.S.  EPA Report No.  560/5-83-003, June 1983.
      U.S. EPA Office of Pesticides and Toxic Substances, Washington, D.C.,
      U.S.A.

Cook, P.M. and Marklund, D.R. (1980):  Sample preparation for quantitative
      electron microscope analysis of asbestos fiber concentrations in air.
      In:  National Bureau of Standards Special Publication 619: Asbestos
      Standards: Materials and Analytical Methods (issued 1982).  U.S. Dept.
      of Commerce, Washington, D.C.,  U.S.A..  Pp. 53-67.

Crawford, N.P. (1985):  Fibre assessment standards of U.K. laboratories
      engaged in asbestos monitoring.  In:  Asbestos Fibre Measurements in
      Building Atmospheres:  Proceedings (E.J.  Chatfield,  ed.).  Ontario
      Research Foundation, Mississauga, Ontario, Canada.   Pp. 61-67.

Dement, J.M. , and Harris, R.L. (1979):   Estimates of Pulmonary and
      Gastrointestinal Deposition for Occupational Fiber Exposure.   U.S.
      Department of Health,  Education and Welfare,  National Institute for
      Occupational Safety and Health (NIOSH) Report No. 79-135.   U.S. DHEW,
      NIOSH,  Cincinnati, Ohio, U.S.A.

Elmes, P.C. (1983):  Health hazards of short mineral fibres.   In:  Short and
      Thin Mineral Fibres (E.J. Chatfield, ed.).  National Board of
      Occupational Safety and Health, Solna, Sweden.  Pp. 163-180.

Felbermayer,  W. (1983):  Abwitterung von asbestzementprodukten--
      immissionsmebergebnisse aus Osterrich.  In:  Verein Deutscher Berichte
      475.  VDI-Verlag GmbH, Dusseldorf.  Pp.143-146.

Friedrichs, K.H., Hohr, D.,  and Grover, Y.P. (1983):  Ergebnisse von nicht
      quellenbezogenum Immissionsmessungen von Fasern in der Bundesrepublik
      Deutschland.  In:  Verein Deutscher Berichte 475.  VDI-Verlag GmbH,
      Dusseldorf.  Pp. 113-116.
                                      92

-------
Gibbs, G.W. , Rowlands, N., and Brulotte, R. (1980)  A Pilot Study on the
      Measurement of Airborne Asbestos Fibre Concentrations in Ambient Air.
      Institute of Occupational Health and Safety, McGill University, Quebec,
      Ontario, Canada.

Gibbs, G.W., and duToit, R.S.J. (1979):  Environmental considerations in
      surveillance of asbestos miners and millers.  Annals of the New York
      Academy of Sciences, Vol. 330, pp. 163-178.

Gibbs, G.W., and Hwang, C.Y. (1975):  Physical parameters of airborne asbestos
      fibres in various work environments--preliminary findings.  American
      Industrial Hygiene Association Journal,  Vol. 36, No. 6, pp. 459-466.

Gibbs, G.W., and Hwang, C.Y. (1980):  Dimensions of airborne asbestos fibres.
      In:  Biological Effects of Mineral Fibers, Vol. 1 (J.C. Wagner, ed.).
      IARC Scientific Publications, Lyon, France.  Pp. 69-78.

Hatfield, J., et al.  (1988):  Assessing Asbestos Exposure in Public Buildings.
      U.S. EPA Report No. 560/5-88-002, May.1988.  U.S. EPA,  Washington,  D.C.,
      U.S.A.

Hayward, S.B. (No date):  Methodology for the Analysis of Ambient Levels of
      Airborne Asbestos by Transmission Electron Microscopy.   Air and
      Industrial Hygiene Lab, California Dept. of Health Services, Berkeley,
      CA, U.S.A.

Hollander M., Wolfe D.  1973.  Nonparametric Statistical Methods.  John Wiley
      and Sons,  Inc., New York.

Hwang, C.Y., and Gibbs, G.W. (1981):  The dimensions of airborne asbestos
      fibres--crocidolite for Kuruman area, Cape Province, South Africa.
      Annals of Occupational Hygiene, Vol.  24, No. 1, pp.  23-41.

John, W., Berner, A., Smith, G.,  and Wesolowski, J.J. (1976):  Experimental
      Determination of the Number and Size of Asbestos Fibers in Ambient Air,
      Report No. PB-254086.  U.S.  Dept. of Commerce, National Technical
      Information Service, Springfield, Virginia, U.S.A.

Karaffa, M.A.,  Amick, R.S., Crone, A., and Powers, T.J. (1986):   An evaluation
      of asbestos removal effectiveness:  the potential for exposures.   In:
      Asbestos:   Its Health Risks, Analysis,  Regulation and Control.  APCA SP-
      57, Pp. 59-70.

Lehmann E.  JL975.  Nonparametries.  Statistical Methods Based on Ranks.
      Holden-Day, Inc., San Francisco.

Litistorf, G. ,  Guillemin, M., Buffat, P., and Iselin, F. (1985):  Ambient air
      pollution by mineral fibers  in Switzerland.  Staub Reinhaltung der  luft
      45(6), pp. 302-307.
                                      93

-------
Lynch, J.R., Ayer, H.E., and Johnson, D.L. (1970):  The interrelationships of
      selected asbestos exposure indices.  American Industrial Hygiene
      Association Journal, Vol. 31, No. 5, Sept./Oct.,  pp. 598-604.

Marconi, A., Menichini, E.,  and Poaletti, L.  (1984):  A comparison of light
      microscopy and transmission electron microscopy results in the
      evaluation of the occupational exposure to airborne chrysotile fibres.
      Annals of Occupational Hygiene, Vol. 28,  No. 3,  pp. 321-331.

Miller I., Freund J.  1965.   Probability and Statistics for Engineers.
      Prentice-Hall, Inc., Englewood Cliffs,  New Jersey.

National Institute for Occupational Safety and Health (1985):  Method for
      Determination of Asbestos in Air Using Positive Phase Contrast Electron
      Microscopy.  NIOSH Method 7400.  NIOSH, Cincinnati, Ohio, U.S.A.

National Institute for Occupational Safety and Health (1986):  Method for
      Determination of Asbestos in Air Using Transmission Electron Microscopy.
      NIOSH Method 7402.  NIOSH, Cincinnati,  Ohio, U.S.A.

Nicholson, W.J., Rohl, A.N., and Ferrand, E.F.  (1971):, Asbestos air pollution
      in New York City.  In:  Proceedings of the Second International Clean
      Air Congress, December, 1970, Washington, D.C.  Academic Press, New
      York, New York, U.S.A.  Pp. 136-139.

Nicholson, W.J., Rohl, A.N., and Weisman, I.  (1975):  Asbestos Contamination
      of Air in Public Buildings, U.S. EPA Report No.  PB-250-980, October,
      1975.  U.S. EPA, Washington, D.C.,  U.S.A.

Nicholson, W.J., Swoslowski Jr., E.J., Rohl,  A.N., Todard, J.D., and Adams, A.
      (1979):  Asbestos contamination in United States schools from use of
      asbestos surfacing materials.  Annals of the New York Academy of
      Sciences, Vol. 330, pp. 587-596.

Nicholson, W.J. (1988):  Airborne Levels of Mineral Fibers in the Non-
      Occupational Environment.  Mt. Sinai School of Medicine, New York, New
      York, U.S.A.

Rendall, R.E.G., and Skikne, M.I.  (1980):  Submicroscopic fibers in industrial
      atmospheres.  In:  Biological Effects of Mineral Fibers, Vol. 2 (J.C.
      Wagner,ed.).  IARC Scientific Publications, Lyon, France.  Pp. 837-843.

Rickards, A. (1972):  Estimation of trace amounts of chrysotile asbestos by
      x-ray diffraction.  Analytical Chemistry, Vol. 44, No. 11, pp. 1872-
      1873.

Ruch, R.B., and Serper, A. (1977):  Ambient measurements of asbestos in the
      vicinity of asbestos sources.  From the Fourth Joint Conference on
      Sens. Environmental Pollut.
                                      94

-------
Sebastien,  P.,  Billon,  M.A.,  Dufour,  G.,  Gaudichet,  A.,  and Bonnaud,  G.
      (1979):   Levels of asbestos air pollution in some environmental
      situations.  Annals of the New York Academy of Sciences,  Vol.  330,  pp.
      401-415.

Sebastien,  P.,  Bignon,  J., and Martin, M. (1982):  Indoor airborne asbestos
      pollution: from the ceiling and the floor.  Science, Vol. 216,  pp.  1410-
      1412.

Sebastien,  P.,  Plourde, M. ,  Robb, R. , and Ross, M. (1984):  Ambient Air
      Asbestos  Survey in Quebec Mining Towns, Part 1:  Methodological Study.
      EPS Report No. 3/AP/RQ/IE.  Environmental Protection Service,
      Environment Canada.  Pp. 1-41.

Sebastien,  P.  (1985):  Assessing asbestos exposures in buildings.   In:
      Asbestos  Fibre Measurements in Building Atmospheres: Proceedings (E.J.
      Chatfield, ed.).   Ontario Research Foundation, Mississauga,  Ontario,
      Canada.   Pp. 139-158.

Sebastien,  P.,  Plourde, M.,  Robb, M., Ross,  M., Nadon, B., and Wypruk, T.
      (1986):   Ambient Air Asbestos Survey in Quebec Mining Towns, Part 2:
      Main Study.  EPS Report No. 5/AP/RQ/2E.  Environmental Protection
      Service,  Environment Canada.  Pp. 1-50.

Small, J.A., Newbury, D.E.,  and Myklebust, R.L. (1983):   The visibility of
      asbestos  fibers in the scanning electron microscope.  In:  Proceedings
      of the 18th Annual Conference of the Microbeam Analysis Society  (R.
      Golley,  ed.).  San Francisco Press, San Francisco, CA, U.S.A.   P.  148.

Snyder, J.G.,  Virta, R.L., and Segreti, J.M. (1987):  Evaluation of the phase
      contrast microscopy method for the detection of fibrous and other
      elongated mineral particulates by comparison with a STEM technique.
      American Industrial Hygiene Association Journal, Vol. 48, No.  5, pp.
      471-477.

Spurny, K.R.,  and Stober, W.  (1978):   Asbestos measurement in remote ambient
      air.   In:  Proceedings of the International Clean Air Conference (E.T.
      White, P. Hetherington, and B.R. Thiele, eds.).  Ann Arbor Science, Ann
      Arbor, Michigan,  U.S.A.  Pp. 373-389.

Steel, E.B., and Small, J.A.  (1985):   Accuracy of transmission electron
      microscopy for the analysis of asbestos in ambient environments.
      Analytical Chemistry,  Vol. 57,  pp.  209-213.

Steen, D.,  Guillemin, M.P.,  Buffat, P., Litistorf, G. (1983):  Determination
      of asbestos fibers in air transmission electron microscopy as a
      reference method.  Atmospheric Environment, Vol. 17, No.  11, pp. 2285-
      2297.
                                      95

-------
Tuckfield, R.C., Tsay, Y., Margeson, D.P., Ogden, J.,  Chesson,  J.,  Bauer,  K. ,
      Constant, P.C., Bergman, F.J., and Rose, D.P.  (1988):   Draft  Final
      Report for Tasks 1-6, Evaluation of Asbestos Abatement Techniques,  Phase
      3: Removal, U.S. EPA Contract No. 68-02-4294.  U.S.  EPA,  Washington,
      B.C. , U.S.A.

U.S. EPA (1974):  A Preliminary Report on Asbestos in  the  Duluth, Minnesota,
      area.  Office of Enforcement and General Counsel,  Office  of Technical
      Analysis, Duluth, Minnesota, U.S.A.

U.S. EPA (1986a):  Airborne Asbestos Health Assessment Update,  Report No.
      600/8-84/003F.  U.S. EPA, Washington, D.C., U.S.A.   P.  198.

U.S. EPA (1986b)  Filter  Blank Contamination  in Asbestos Abatement  Monitoring
      Procedures:  Proceedings of a Peer Review Workshop.   Contract No.  68-03-
      3264.  Water Engineering Research Laboratory,  Office of Research and
      Development, Cincinnati, Ohio, U.S.A.

U.S. EPA (1987):  Asbestos-containing materials in schools:  final rule and
      notice (Appendix A: AHERA Method).  Federal Register,  40  CFR  763,
      October, Vol. 52, No. 210, pp. 41826-41903.

Walton, W.H. (1982):  The nature, hazards and assessment of occupational
      exposure to airborne asbestos dust: a review.  American Occupational
      Hygiene, Vol. 25, No. 2, pp. 117-247.

Winer, A.A., and Cossette, M. (1979):  The effect of aspect ratio on fiber
      counts: a preliminary study.  Annals of the New  York Academy  of
      Sciences, Vol. 330, pp. 661-672.

Yamate, G., Agarwal, S.C., and Gibbons, R.D.  (1984):   Methodology for the
      Measurement of Airborne Asbestos by Electron Microscopy,  U.S.  EPA Report
      No.  68-02-3266.  U.S. EPA, Washington,  D.C., U.S.A.
                                                 i U.S. GOVERNMENT PRINTING OFFICE:1990-748-159/00476
                                      96

-------