&EPA
              United States
              Environmental Protection
              Agency
              Center for Environmental
              Research Information
              Cincinnati OH 45268
EPA-600/8-84-032
January 1985
              Research and Development
User's Guide:
              Fugitive Dust Control
              Demonstration
              Studies

-------

-------
                                        EPA-600/8-84-032
                                        January  1984
                 USER'S GUIDE


  FUGITIVE DUST CONTROL DEMONSTRATION STUDIES
     U.S. ENVIRONMENTAL PROTECTION AGENCY

      Office of Research and Development
 Center for Environmental  Research Information
                    and
  Office of Air Quality Planning & Standards
CENTER FOR ENVIRONMENTAL RESEARCH INFORMATION
     OFFICE OF RESEARCH AND DEVELOPMENT
    U.S.  ENVIRONMENTAL PROTECTION AGENCY
            CINCINNATI, OH 45268

-------
                                  NOTICE
This  document  has been  reviewed  in accordance with  the U.S.  Environmental
Protection Agency's peer and administrative  review policies and approved for
publication.  Mention of trade names or  commercial products does  not consti-
tute endorsement or recommendation for  use.
                                     11

-------
                                 FOREWORD
Under  certain  circumstances,  sources of  air  pollution  may be able  to  meet
their obligations under the Clean  Air Act  by the use of the bubble concept in
which  emission  points  can  be controlled to a  greater or  lesser  degree  than
otherwise  required, as  long as  the  total plant  emissions  are  within estab-
lished limits.  Strict controls  can often be applied to non-process fugitive
(also  called  nontraditional) particulate to  emissions  to  allow a  lesser
degree of control  for  stacks and vents under a  bubble proposal.

State and local  agency officials are  responsible for determining the validity
of bubble  proposals.   Thus,  these officials must  be  in  a  position to deter-
mine  if claimed  reductions  in  fugitive emissions  from such  sources  as
roadways  and  storage   piles  can  be  substantiated for  the control  methods
proposed. There are many studies  available to  the official that  discuss  the
effectiveness  of  various  control  techniques for  nontraditional  particulate
emissions.

This User's Guide provides  state and  local agency officials and  those respon-
sible for conducting fugitive particulate demonstration  studies  insight  into
the  validity  of  recent  studies.    It also stresses  the   considerations  of
importance when  planning this type of  study.   From  the data presently
available, it was  not  possible  to develop quantitative emission  factors  or
control efficiencies for the  various  control  alternatives.

-------
                              ACKNOWLEDGMENTS


Many individuals  contributed to  the preparation and review of this  publica-
tion.   The author  was  Thomas  W. Beggs,  P.E.,  Senior Engineer,  JACA Corp.,
Fort  Washington,  Pennsylvania.   Reviewers  included Brock  Nicholson,   EPA-
OAQPS, Research Triangle Park,  North Carolina,  Kenneth R.  Woodard,  EPA-OAQPS,
Research  Triangle Park,  North  Carolina,  and  Keith  Rosbury,  Denver Branch
Manager, PEDCo Environmental, Inc. (now PEI), Denver,  Colorado.

-------
                              TABLE OF CONTENTS
 Chapter
                                                                     Page
    FOREWORD	     111

    ACKNOWLEDGMENTS.	      1v


    1     INTRODUCTION	       l


    2     GUIDELINES FOR CONDUCTING FUGITIVE DUST CONTROL
            DEMONSTRATION STUDIES	       3

          2.1   Study Design	       3
          2.2   Sampling and Analysis	!  !  !       4
          2.3   Statistical  Analysis of Results	!  !  !  !       6
          2.4   Development  of Emission Factors  and  Control
                  Efficiencies	  .....       7
          2.5   Optional  Investigations.	'.'.'.'.       7

    3     CASE  HISTORIES	       9

          3.1   Recent  EPA Demonstration  Studies  ........       9
                3.1.1    Denver, Colorado  	  „...!       9
                3.1.2    Minneapolis,  Minnesota ....„.!!!     13
                3.1.3    Portland, Oregon	            *     17
          3.2   Other Studies.	    .'.'.".*      20
                3.2.1    Philadelphia, Pennsylvania ....!!!      20
                3.2.2    Bangor, Maine.	      20
                3.2.3    Kansas City,  Kansas	'.'.'.      21
                3.2.4    Lincoln, Nebraska	      21
                3.2.5   Clark County, Nevada	]      22
                3.2.6   Erie County,  New  York	|  [      22
                3.2.7   Allegheny County, Pennsylvania .....*      23
                3.2.8   The Road Carpet Study	      24
                3.2.9   The Iron and Steel Plant Studies . . !  !      24

   4     CONCLUSIONS	        27

REFERENCES	        2g
                                     -v-

-------

-------
                                 Chapter 1

                                INTRODUCTION
 The Clean  Air Act Amendments  of  1977 require states  to  submit to the  U.S
 Environmental  Protection  Agency  (EPA)  a revised  State Implementation  Plan
 (MP)  describing the state s plan for expeditious  attainment of  the National
 Ambient Air  Quality Standard  (NAAQS) for  all  areas  within  the state.  The
 specific EPA  requirements for  the  revised  SIPs   (published  in  the  Federal
 Register,  April  4, 1979 and in  subsequent EPA documents) include  requirements
 for addressing several  classes of particulates, especially in  nonattainment
 areas:   stack emissions,  process  fugitive dust, and nontraditional fugitive
 dust (or  nontraditional  particulate matter).   Nontraditional  fugitive  dust
 includes urban fugitive dust,  dust  reentrained  from road  surfaces, and  dust
 created by construction activities.

 For those  areas  which  have  not  attained the NAAQS  for particulates  and in
 which  fugitive dust  contributes  to  the  violation  of the  particulate NAAQS,
     ^TUNStu adopt the  requirement that  reasonably  available control technol-
 ogy (RACT) be used to  control  traditional  sources  of  fugitive dust (process
 fugitive dust);  in addition the  SIP must  contain a  schedule for the study and
 eventual  adoption  of controls   for  nontraditional  sources  of particulates.
 Many such   RACT-plus"  studies  have  been  conducted across  the  country to
 demonstrate the  effectiveness of  proposed  technologies in controlling these
 nontraditional sources.

 Because  of the  large number of nonattainment  areas  requiring demonstration
 studies  and considering the cost  of these  studies, EPA commissioned a work-
 T£   I,to Provide   guidance  in designing  and executing  a  proper  study  [1].
 The Workbook  was  published in  March  1981.   The guidelines set  forth  in the
 Workbook are  to  ensure  that defensible  results  are obtained  from the  study,
 and  that the results can  be  exchanged among  jurisdictions, thus minimizing
 duplication of  effort.   The  Workbook  stresses two critical  guidelines  that
                         demonstration study to be successful:   (1) The study
                         the  amount  and  type of data generated will be  suf-
                         objectives of the  study;  and  (2)  The data must be
                         rigorous statistical analysis.
must be followed for any
design must  ensure  that
ficient to  realize  the
subjected to the proper,
In  addition  to the  Workbook,  EPA has  also recently sponsored or  partially
funded a number of these demonstration  studies to  evaluate  the  effectiveness
of  a  wide  variety   of  nontraditional   particulate  control  systems.
studies  were  conducted  in  Denver,  Colorado;  Minneapolis,
Portland, Oregon.
                                                                       These
                                                              Minnesota;  and
                                     -1-

-------
Of the  studies  reviewed,  many were found to  be  deficient  in their analyses
when compared to the Workbook  guidelines.  Generally speaking, in the studies
with valid analyses,  control  techniques that  were  evaluated  were  not highly
effective.  Further, many of the studies neglected the critical  time factor;
that is, the study analyzed the  immediate effect of the control but failed to
analyze the decay in effectiveness  over time.
The objectives of this User's  Guide  are to:
       o    Verify that the Workbook is a  firm  foundation  to  build a demon-
            stration study upon

       o    Discuss the study variables that  must be taken into  account'in
            the study design                 .                 -'•....'••

       o    Summarize the major  fugitive  particulate  studies  that have been
            done to date           ,..,.-

       o    Make  state  and  local agency personnel aware that  control  effi-
            ciencies  reported  in some studies may be unduly optimistic.

This Guide is therefore of particular value  to those designing and implement-
ing future studies  as  well as to state and local agency personnel reviewing
permit  applications  in which  the control of  nontraditional particulate
sources is an integral  part.        ;

Chapter 2 of  this Guide reviews the Workbook's  guidelines  on  the design and
implementation of demonstration  studies.  The guidelines include a consider-
ation of:

       o    Study design           :
       o    Sampling and analysis
       o    Statistical  analysis of results
       o    Development of control  efficiencies
       o    Optional investigations
       o    Usefulness to state  and  local agencies.

The review of  case  histories is found  in  Chapter 3,  which  provides -detailed
descriptions of  the three recent EPA-funded demonstration .studies and  brief
descriptions of a number of earlier studies.              •
                                     -2-

-------
                                Chapter  2

    GUIDELINES  FOR  CONDUCTING FUGITIVE  DUST  CONTROL  DEMONSTRATION  STUDIES


 The proper design,  implementation, and  interpretation  of  the  results of  a
 demonstration  study are essential  to the defensibility  and  usefulness of
            S*  JM? S6fi0n  summarizes the  Sidelines set  forth in Workbook
               Studies for Control of Nontraditional Particulate Sources  [1].
       9   In' tif ^^ ?? ?  yardst1ck against  which the project descriptions
 r~    H    I  J   ,Ca!6 hlstones are measured;  The  Workbook was thoroughly
 reviewed  and  found  to provide excellent  guidance  for  a demonstration study
 which,  when rigorously followed,  should yield  defensible results.          Y


 2.1    STUDY DESIGN
 The design  of a demonstration  study is critical  to the study's  ability to
 produce clear, valid,  and defensible  results which can be applied  to  circum-
 stances outside  the  the  study  itself.   The study design  musf begin wUh
 the objectives .of the study,  such  as  "to  determine the  effectiveness of
 roads * W|LnHPPtfS1°n-iChKe^CalS  for Controlling emissions from  unpaved
 roads.    Second, the availability of data  related  to this objective  must be
 ascertained m order to see  whether the demonstration  study  Is  warranted
 L^ht •lsPwarranted1  onl*  wh*n the data needed  are not available  or are
 unusable.    For example, .all  available data on a particular type  of  control
 approaches described below is  most appropriate for determining the effect
                                           ^0-^; ^
                                          uniform emissions  from i


                                                       "el
                                               «.
suited  for  averaging widely  variable emissions  from a large area of  sources
    £u>Yr^ -s^s- ra^^-sssar ^?;  srss. S
directly  to the  areas  in  which  monitors are located.   However,  the number
                                  -3-

-------
and location of  monitors  used are critical elements in the  study  design  if
the test  results  are to be used tp predict area wide  effects.  Most  nontra-
ditional  particulate studies  will  probably be  receptor  oriented since
they involve impacts in a  large  area.

The next  step  is to select  one of three  types  of  study  designs  described
below on  the  basis  of implementability and usefulness  for  subsequent  anal-
ysis.  This step must be closely integrated with the selection of the  statis-
tical  analysis  technique  which  will   be  used later;  these techniques  are
described below.  If the  type of design is ill-chosen  or cannot be followed
meticulously throughout the course[of  the study,  the statistical  analysis  of
field data may be hampered or  even invalidated.

A  Type  1 design  consists  of  a  sihgle  study  area  sampled  during  an  uncon-
trolled   period  and   during a  controlled period.   This design  assumes  that
all parameters  other than the  control parameter  remain  constant from  the
first to the second  period.   Therefore, this design is only effective if the
other variables  (such as  weather £nd site activities)  can  be  shown  through
statistical analysis to not influence  the results.

A Type 2 design consists of simultaneously  sampling two study areas, one with
and one without controls.   In this! manner, the variables that can affect the
uncontrolled and controlled emissions  (other than the control itself)  will  be
nearly identical.

A  Type  3 design consists  of  two study periods:   during the first period one
study area  is  controlled  while  the  other area is not controlled; during the
second period, the roles are  reversed.  Although more expensive to conduct, a
Type  3  design  is usually  preferable to Type 2, and  is  called  for if source
activities  vary  widely or if for other reasons,  such  as  an expectation  of
seasonal effects,  the study areas  cannot  be  considered identical.   The
greatest  disadvantage of this  type  is  the  possibility that the  effect  of
controls  in one area might be carried over into the first part of the second
period when that area is said  to be  uncontrolled.

Another'step  in  the planning  stage is to  detail the  specifications  for
applying  the  controls,  including the area to be  administered,  and the fre-
quency, methods, and  rates of  application.  These specifications must  include
incidental  instructions  that  may japply,  such as  a prohibition  of  parking
during street-cleaning.
2.2    SAMPLING AND ANALYSIS
The  success of  the  study is  dependent  on the coordination  of  all  parties
involved  to ensure that  the  sampling program  is  carried  out in accordance
with  the  study design.   It is particularly important that those responsible
for implementing the source controls do so at  the  agreed-upon  time.  A Type 1
                                     -4-

-------
 or  3  study  may be irreparably damaged if sampling during the control period
 cannot  be  completed because  of  a failure to  completely  implement controls
 during the scheduled sampling period.

 A  primary  requirement  for  the study  is  that enough  samples  are  taken  to
 demonstrate  statistically  that the  control  method does  or does  not  bring
 about  the  desired effect.   If too few  samples  are taken,  results  will  be
 inconclusive.   The Workbook  provides  detailed  instructions for calculating
 the minimum  number of  samples.  The design must also  allow for  the  loss  of
 samples which will be dropped due  to errors.

 The frequency of  sampling generally represents a  compromise between the  time
 period available for the study and the  resources and level  of effort that can
 be  committed  to the project.   For source-oriented studies, any  convenient
 sampling period is  selected,  provided  it is representative.
 oriented studies,  daily sampling  is  preferable,  especially
 effect of variables whose effects  may be transitory,  such  as
  For  receptor-
in tracking the
the application
 of chemical  dust suppressants.  The sampling period's selected are normally  24
 hours in length to correspond with daily cycles of weather,  vehicle  traffic
 and so on.

 Determining the  number of  sampling sites is generally  a  site-specific
 consideration.    Source-oriented  studies can use  a single  monitoring  site.
 However, at least two  monitoring  sites  are recommended for each study area
 for  a  receptor-oriented  study:   normally  one monitoring  site  is   located
 near the source (or sources)  to  measure the  impact  of  the  control method
 in a small  area, and an  additional monitoring  site  is  located up to  several
 hundred meters from the source to  measure the  impact in a larger area
 A background  site  located  upwind  from  the  source is  also advisable  when
 studying ambient impacts  that  are  expected  to be small.   The actual  location
 of the  instruments should  follow guidelines such  as those found in the
 Workbook and similar references.   These  guidelines address concerns  such as
 height  above the ground, distance from  buildings,  and distance  from trees.

 The  standard sampling  instrument  for a  receptor-oriented  study  is  the  high
 volume  sampler,  or hi-vol, which measures total suspended particulates  (TSP
 less  than 30 microns in diameter).   If an important  parameter in the study is
 the  proportion  of participate matter between specific  size  ranges,  such  as
 inhalable particulates (IP, less than  15  microns in  diameter) or fine partic-
 ulates  (FP,  less than  2.5 microns  in  diameter) options  are  available.
 it  two size fractions  are  required,  a dichotomous sampler  can be used-  if
 more  fractions  are desired, a  cascade  impactor with  several  stages  can  be
 used  in conjunction  with  a hi-vol.   It should  be mentioned that  cascade
 impactors suffer  from the problem of "particle bounce"  in which  large  par-
 ticles  migrate   to  lower  stages  of  the instrument,  thereby  significantly
 biasing the  results toward the  small particle size.                 ".cdntiy

 The  specifications for  every demonstration  study  must  explicitly delineate
detailed quality assurance procedures.  A  quality assurance checklist  such  as
that published  in the Workbook must be strictly  observed during the sampling
and analysis phase of the  study.                                      amjjuny
                                    -5-

-------
2.3    STATISTICAL ANALYSIS OF RESULTS
Four methods of statistical analysis have been or could be used to interpret
the results of demonstration studies when a controlled source is compared to
an uncontrolled source.   These  analyses  include  the paired t-test, analysis
of variance, analysis of covariance,  and  multiple  linear regression analysis.
The paired  t-test  compares the averages of  the  controlled source data with
the averages  of  the uncontrolled  source data and  analyzes  the differences
to determine whether or not the control  method has  caused  a real difference.
Analysis of variance compares the  magnitude  of  two  or more sample variances
to determine whether they are different  at a  predetermined level of signifi-
cance, for example, with 90 percent confidence.  Analysis of covariance is an
extension of analysis of  variance; variations attributable to measured
variables  (covariates)  other than  the  control  variable are  subtracted  out
before testing for significance.   This  allows conclusions  to be drawn con-
cerning the  effect of  the  controls  as  well  as  of each  of the covariates.
Multiple linear  regression analysis derives  an equation  relating an effect
(such as a concentration) to a series of independent variables.  The results
of this  analysis   indicate  which,  if any,  of the independent variables  are
correlated to the  effect.  The reader is  referred  to  the  Workbook for more
detailed discussion of these methods  of  statistical  analysis.

The following  is  a  summary of these methods and  their suitability  to  the
three types of demonstration studies:
  Technique
Study Design
Applicability
  Requi rements
   Limitations
Paired t-test
Analysis of
variance
Analysis of
covariance
Multiple
linear
regression
Type 2


Types 1, 2, 3
Types 1, 2, 3
Types 1, 2, 3
Paired data
Sampling must be
done for combi-
nations of vari-
ables
All independent
variables of
interest (covar-
iates) must be
measured

All independent
variables of
interest must
be measured
Only two data sets
can be compared

Period of time re-
quired cannot be
determined because
variables are not
controllable

More difficult to use
for Types 1 and 2
than for Type 3
                                     -6-

-------
 It is  apparent  from the  above table  that  the statistical  analysis  method
 selected  for a demonstration  study ,is  highly dependent on  the  type of
 study design as well as on  the  number  of  variables  the  designer is  able  and
 willing to measure.  Obviously, the study design and corresponding  statis-
 tical  analysis method must  be  selected together in order to meet the  study
 objectives in  the most cost-effective manner.

 Once  selected, the  statistical  analysis  method must be  rigorously  followed
 and the results  compiled  in a  we11-documented  fashion.    Furthermore, each
 method  requires  assumption of a confidence level; for example,  a  finding of
 effectiveness  of a  control  method  at  a  95  percent confidence  level  is a
 stronger endorsement of  the method  than  a similar  effectiveness at an 80
 percent confidence  level.    The  confidence level  must be clearly  delineated
 when  reporting the results in order for the reader to judge  the  impact of the
 findings and  conclusions, as well  as  to  be  able  to compare findings  among
 similar studies, and to decide whether the results  can confidently  be  extrap-
 olated  to  a  full-scale implementation of the control  method.


 2.4    DEVELOPMENT OF EMISSION FACTORS AND CONTROL  EFFICIENCIES


 The results  of a demonstration  study are  usually expressed in the following
 form:   It  can  be  said,  with  a certain confidence  level,  that  the effect of a
 control  system lies  between  A and  B,  where A  and  B  are reductions  in  ambient
 particulate  concentrations  from  the level  when the source was uncontrolled.
 The study may  go on to provide emission  factors or emission  factor algorithms
 relating emissions to variables  such as  traffic density or days since  precip-
 itation.   In order to be  useful,  emission factors  developed  in this manner
 must  be firmly  grounded  in the  results   of  the  study and must not merely
 represent a restatement  of previously developed equations that  are not
 completely verified by the present  study.

 The  result that is  probably most   useful  to  the permitting  process  is the
 development of control efficiencies.  These  should  be based strictly on the
 effects data  described above  and  again, they  should  not  be influenced
 by  previous work or unwarranted  assumptions.  Furthermore, any limitations  on
 the application of the developed  control efficiencies must be clearly  stated.
 For example, a control efficiency for the  use of a chemical dust suppressant
 on  unpaved roads  applies only  at the application doses  and  frequencies
 employed in the demonstration study  and  only for that particular suppressant.


 2.5    OPTIONAL INVESTIGATIONS


Optional investigations are  defined  here  as  activities  other than  general
particle size  analyses (TSP,  IP,  and FP)  that may be  conducted  for  reasons
specific to the demonstration study.   One such study might be  an  elemental
                                    -7-

-------
analysis of samples to  provide  corroborating  data for other similar studies
or  to  identify  and  remove interferences  from the  study.   For  example,  a
significant portion of collected particulate matter in an urban location may
comprise nonsoil  elements,  or  one  or more soil  elements may  be  present in
suspiciously high concentrations.   Either of these phenomena may be traceable
to  an  overwhelming background source,  the contribution  of which,  if quanti-
fiable, could  be subtracted from the  data  in  question  to provide a clearer
estimation of the effectiveness  of control.
                                     -8-

-------
                                 Chapter 3

                              CASE HISTORIES
This  chapter  presents case  histories  of demonstration  studies  designed to
evaluate  the  effectiveness of  control  techniques for  participates.   These
case  histories  are intended  to illustrate the application of the study
guidelines presented in the previous section  to  actual demonstration studies.
The design, implementation, and conclusions drawn from each of the cases are
presented  and  critiqued, with  the hope  that  this  information  will  assist
others in the design of similar studies.   The case histories are divided into
detailed accounts of recent EPA-sponsored  studies  in  Denver, Minneapolis, and
Portland, Oregon, and less detailed accounts  of  earlier  studies.
3.1    RECENT EPA DEMONSTRATION STUDIES
For each of these case histories,  an account  is  provided of the circumstances
prompting the study,  the  reasons  for the choice of particular study design,
the implementation of the study, and the conclusions drawn from the results.
The methods and procedures employed in each study are then compared with the
criteria and guidelines established  in the  Workbook.   In these studies many
results could not  be substantiated due  to such factors  as  the inability to
carry out  a study as designed,  the effect of  interferences,  and  less than
adequate statistical   analysis  of  sampling  results.    State  and  local  agency
personnel  should be aware of these deficiencies both in reviewing compliance
plans based on these  studies  and  in designing  future studies.
3.1.1  Denver, Colorado
The Denver study [2] was  conducted  for the Colorado Division of Air Pollution
Control  during  the  winter  of 1980-81.    The  study  actually  consisted  of two
independent  studies  designed  to  address  the  hypotheses  that  (1)  ambient
concentrations of particulate matter can  be reduced by substituting road salt
for sand;  and (2) ambient concentrations  of particulate matter can be reduced
by later cleaning up the  sand used  on the snowy  roads.
                                    -9-

-------
  No  large-scale demonstration  studies  of this type  had  been conducted  pre-
  viously despite the  widely  held  belief that  the  use  of sand for snow control
  may  be detrimental to  local  air  quality.  The first  hypothesis  was tested  in
  Lakewood,  a suburban  area  near Denver;  the second  was  tested  in the  City
  of Denver.  Both the suburban and urban locations were primarily residential.

  In  both locations, concentrations  of  TSP,  IP, and  FP were  measured  at
  three distances:   near-street  (10 to  11  meters  from  the  street);  off-street
  (50 to 70  meters  from  the  street);  and neighborhood  (greater  than  200 meters
  from the street).  The results of the  studies  indicated  a slight improvement
  in air quality for TSP and FP at the  near-street location when  the sand was
  removed,.   Substituting salt for sand  failed to  show any improvement in air
  quality;  in fact, a slight degradation of TSP air  quality was observed
  at the near-street and off-street  locations.  These results are  summarized
  below.
 Control  Method
                      TSP
                                      Air Quality  Improvement
                      TF
 Salt Substitution (Lakewood,  CO)
   Near street
   Off-street
   Neighborhood
                Negative
                Negative
                NSS
 Sand  Removal  (Denver,  CO)

   Near street          Positive
   Off-street
   Neighborhood

 *Not  statistically significant
NSS*               NSS
Not tested          Not  tested
Not tested          Not  tested
                                  NSS                Positive
                          No conclusions possible
                          No conclusions possible
Study Design.   Both  portions of  the  study  used  the  receptor-oriented  ap-
proacn,  wnich  was appropriate given that a wide  variation  in  emission  rates
was  anticipated.    In order  to  partially overcome  the limitations of  this
approach,  monitors were  located  at  the three  distances  from  the  streets
described  above,  thus eliminating  the  temptation to assume that the  partic-
u 1 ate  concentrations  were  consistpnt  thmunhmit +ha  an+-;^Q  neighborhood
concentrations were  consistent  throughout  the  entire
The  study  design selected for  the  salt substitution case was  Type  3.   Two
sites were monitored concurrently,  one using salt  and  the  other sand; the
process was reversed for the  second  period.   The  sand  removal study was also
to  have  been  a  Type 3 design, but  it was essentially  reduced to  a  Type 2
design  because  street  cleaning activities  were  not  performed  as  planned.
Meaningful  results could have been obtained in the Type  2 design  had the two
areas been shown  to  be  essentially  identical  in terms of meteorology, back-
ground concentrations,  interferences, and  so  on.   However, this was not found
to be true.
                                     -10-

-------
The  actual   areas  selected for  the two  studies  appear to  be appropriate.
Each  was primarily residential, which  tends  to minimize interferences from
background  sources.    The  paired sites  were  similar  in  terms of  land  use
patterns  and number of  paved  streets.   Those  sites  in  the  salt substitute
case  differed  somewhat in  terrain;  site A was  a  valley  and  site  B  a  grassy
hill.  However, because a Type 3 design  was used,  this  inconsistency would be
inconsequential.

The concentration of TSP at each of  the  four monitoring stations was measured
by a  hi-vol  sampler with a standard  glass filter.  A dichotomous sampler with
a  teflon filter  was  used to  measure  concentrations of  IP and FP  at  the
near-street  monitoring stations  only.  ;  For the sand  removal  study, a back-
ground  monitoring station, consisting  of a  hi-vol  sampler  and  a  dichoto-
mous  sampler,  was maintained  in another  part  of  downtown  Denver.  Sampling
was conducted  for 24-hour  periods  from  mid-day to  mid-day,  allowing sample
collection  and restarting  of  monitors  during normal working hours.   The
location  and operation of  monitors  appear to  be  adequate.   The duration of
each  portion of the  study  also appears  to be adequate for generating enough
data  for a rigorous statistical  analysis.

Sampling and Analysis.    As  mentioned  previously,  the  sand  removal  study
suffered from a  lack  of coordination  which resulted  in  a Type  3  design
actually  being conducted as a Type  2  design.   Other  than this,  the imple-
mentation of controls  proceeded  as  planned.   The sampling and analysis plan
provided sufficient samples for  statistical analysis of results, even though
the study  was  terminated after  five months instead of the six months which
had been planned.

Extensive quality assurance  procedures were established  and followed.
These  covered  equipment maintenance  and  calibration,  sampling procedures,
laboratory procedures for filter preparation and analysis, and data handling.
Independent audits  were performed to  check  hi-vol   flowrates and  filter
weights, and to ensure the reproducibility of the hi-vol  results.   The audit
of hi-vol  filters essentially  verified  the procedures used;  some  problems
with  reweighing of dichotomous  filters  were  reported  even  though  a  higher
than  normal  reweighing tolerance was used.  The  report  provides  the  proper
caution  concerning  interpretation of these results in light  of potentially
large weighing errors.

Analysis of Results.    For the  salt  substitution study  the concentrations  of
TSP,   IP,  and  FP  for  all  sampling  days were  calculated  directly  from  the
laboratory results.  For each site and period, the mean concentrations  of  all
data  were then  calculated.  The  differences in ambient concentrations between
each   pair  of sites were averaged for  both periods and  used  in  a  multiple
linear regression  analysis.  The regression  analysis  was used  to determine
the variables,  other  than  salt substitution,  that  may  have  affected  air
quality.
                                    -11-

-------
For the sand removal case, the multiple linear regression  technique  could not
be used because the  expected  Type  3  design  turned  out  to  be a  Type  2 design.
Instead, the paired t-test was used,  requiring the  calculation  of  mean  values
and  standard  deviations for  two  groups of data  (with  and without sand re-
moval) as well as the average difference of paired  data values.  As  mentioned
previously, the  Type 2 design  is  only valid if the two  areas can be shown
statistically  to be  identical.  ;The paired  t-test  was thus  first used
for  this  purpose  and  revealed that  the sites  were  identical only  at the
near-street  locations.   Therefore,   no conclusions concerning  the control
method could be made concerning the other two distances.  Although a somewhat
standard 80 percent confidence level  was used for discussing the significance
of the  salt substitute  results,  the  sand  removal  results are presented in
terms of  a  95 percent confidence level.   It  is not clear why two  different
confidence levels were used.       j

In general, the study  presents  conclusions  properly and does not  extrapolate
the results beyond the constraints of  the study.   The  most useful conclusion
is that neither control method was shown to  provide a significant  air quality
benefit.  Note  on the summary table  (p. 10)  that  14 of the 18 cases  cannot
be discussed in  terms  of the effectiveness of the control.  These  cases are
eliminated for one of three reasons:   samples were  not  taken for that parame-
ter (not tested); the  type  of study1  design  implemented  precluded  conclusions
from  being  drawn  (no conclusion possible);  or  the results were not statis-
tically significant.

Emission Factors and Control Efficiencies.   The  uncertain  nature  of the re-
sults prohibited the development of jemission factors and control efficiencies.
                                   i
Optional Investigations.    Selected [filters  were further  analyzed to  define
the nature  of the particulate matter collected.  Some   hi-vol filter samples
were  ashed  to  determine  percent  volatile  material,   and  some   dichotomous
filter  samples  were subjected to  elemental  analysis  by X-ray fluorescence.

The presence of the ten elements for  which  the filters  were analyzed has been
explained in terms of their expected  sources:   soil, fuel  combustion, automo-
bile exhaust, and application of sand or salt.  However, in general  these ten
elements represented only 10  to 25 percent  of the  total mass of the samples.
Therefore, with  no  further analysis, no conclusions regarding interferences
from  unexpected  elements  are possible.    A  qualitative   discussion of  the
variation of the ten elements across  sites and time  periods  is presented, but
does  not  achieve any  specific objective.   A  brief discussion  of volatile
content is also provided.

Usefulness  to State and Local  Agencies.   As  mentioned earlier, neither
control  method  was  shown to  be effective  in improving ambient  particulate
air quality, given  the limits of  the  test  program.   Therefore,  one obvious
use of  the  results  is  to  encourage  state  and  local   officials to  be  skep-
tical  of favorable  claims made for [these control methods   applied in similar
situations.
                                     -12-

-------
 Future studies of this type would  benefit  from  coordination  and  planning  of
 the implementation  of  controls.   No conclusions for the  off-street and
 neighborhood   locations  were possible  for  the  sand  removal  study  because
 immediate  cleanup of sand was  not practiced.   If  a control  measure  is
 to be  implemented by  other than the study contractor,  a firm  commitment with
 the performing organization must be made before testing begins.   In  addition,
 the contractor's performance must be monitored  throughout the study  period.

 The Denver  sand  removal  study  can  be somewhat useful  in providing background
 data on the  uncontrolled  levels  of  small  particulate in anticipation of a
 size-specific  standard such as the  suggested PM-10 (particles  less than  10
 microns)  standard.    In  both  study  areas at  the near-street  locations, con-
 centrations of IP (less  than  15 microns)  and FP  (less  than 2.5 microns) were
 determined  by dichotomous samplers, which would permit  interpolation to the
 approximate fraction  of PM-10.  However, because  of possible interferences,
 it is  perhaps wise  to  view  this  interpolated  PM-10  value  as  an  order-of-
 magnitude estimate.
 3.1.2  Minneapolis, Minnesota
The  Minneapolis  study  [3]  was conducted for the Minnesota Pollution Control
Agency.  Field measurements took place in the summer of  1982.   The objectives
of this study were to assess the impact of construction-related fugitive dust
on ambient  air  quality  in  an urban area, and to assess the effectiveness of
water  spraying  as a control  method  for this source.   The  study was needed
because  at  the  time,  little data was  available on  fugitive  dust  from con-
struction activities or  on  the  effectiveness of water spraying as a control
method.   Concentrations  of TSP and  IP were measured upwind and  downwind of
the  construction activity;  concentrations  of FP and PM-10 were  interpolated
from U>g-normal plots of  the  data.   Because of problems encountered in
executing the  study, no  firm quantitative  conclusions were  justified  con-
cerning the  effectiveness  of watering to control construction-related dust.

Study Design.   A receptor-oriented approach was planned,  in  which  monitors
were placed  at  distances of 25 and  50 meters  upwind and  downwind  (north to
south  orientation)  of  the   road  construction  right-of-way.    Monitors  were
deactivated  during  periods  when  the wind  direction  differed by more  than
67.5 degrees from the north-to-south  line.

The  study  was  conceived as  a Type  1  design with  one  area to be  monitored
during consecutive  periods  of no  control  followed  by  control.   During  the
early stages of  the road construction, water spraying  of  the dirt  surface,
though planned,  was  not  conducted;  a  modest  amount of watering was done later
when the  road had a gravel  surface.    The  result was  that  the sampling  data
collected before and after  the application were  for two essentially  different
sites, thereby  ruling out a meaningful statistical  analysis of the  effect of
control.    Therefore,  the  effect  of  water  spraying could not be  evaluated
and an alternative analysis  based  partly on  previous work was used.
                                    -13-

-------
The selection of  the  site appears to have been appropriate, as it consisted
of  the full  range  of  activity  frbm site clearing  to paving with  hot  mix
asphalt.   The monitors were  properly oriented, parallel to  the prevailing
wind direction.  Parti oil ate concentrations upwind and  downwind were measured
by  hi-vol  samplers fitted  with  a  15-micron  selective inlet  and  a 5-stage
cascade impactor.  The TSP and IP concentrations were thus measured directly,
while PM-10 and FP were determined by extrapolation or  interpolation from the
particle size curve developed  by the cascade  impactor  results.  Monitors in
both directions were  in areas with  ground cover provided  by  crops  and were
mounted such that the  inlets were two meters above the  ground.

Sampling and Analysis.   As  discussed above,  an  insufficient number of
samples were taken during watering  to provide for  a statistical  analysis of
the effectiveness  of  the' control technique;  furthermore, the  samples  taken
for the  controlled and uncontrolled  conditions were for different  surfaces
(gravel and dirt).  Part  of the prpblem was  the short  duration of that part
of  the construction  which  was  performed on  dirt   (approximately 35  days).
This would have made it difficult to  collect  enough samples even if watering
had been done for as much  as half of this  period.

The sampling and  analysis procedures  followed appear to have  been adequate.
Extensive field measurements  of meteorological  variables  such  as wind  speed
and direction, temperature, and  humidity  were taken, as well  as other  inde-
pendent variables such as  vehicle speed  and number of passes.  Samples of the
actual  working surface were collected for  laboratory analysis.

Extensive and appropriate  quality assurance procedures were followed covering
equipment siting, operation and  maintenace,  instrument  calibration,  and lab-
oratory procedures.  Audits of unexposed and exposed filters were conducted.

Analysis of Results.    The  concentrations of  TSP and  several  particle size
fractions were directly calculated  from the the hi-vol  15 micron inlet,  and
from the  cascade  impactor  catches.   The  data were plotted on a log-normal
distribution curve, from  which the  concentration of PM-10  and FP  were then
interpolated.  During  this procedure, an unexpectedly   large amount  of par-
ticulate matter  bypassed  the  cascade  impactor  stages  and was found on  the
backup  filter  (particle  bounce). '  A  calculation,  presumably  verified  in
previous work,  was used to correct this  problem.

A possible interference was noted in the presence of a  significant amount of
black particulate matter in the smaller size   ranges.  It was speculated that
the source was diesel exhaust from heavy construction equipment.  This
speculation was not verified by laboratory analysis.

Meteorological  and other  independent  variables  were  averaged  and tabulated.
Water application  rates were  highly variable  in duration and  intensity  and
were thus normalized and  reported as liters  per square  meter per  hour.

Originally, it was  intended that an  analysis of variance would be  used to
statistically analyze  the  effectiveness of the control.  However, because of
                                   ;  -14-

-------
 the problems in carrying out the  original  study  design,  this  technique  could
 not  be used.   Instead,  a  multiple  linear regression analysis  was used to
 attempt to  relate  ambient  particulate concentrations to  various  independent
 variables.
 A regression model was proposed which related concentration  to  three  indepen-
 dent variables — silt content (percent less than 200 mesh), traffic  density
 »»* surface  moisture — that  were selected  from  a preliminary correlation
and
 analysis.   However,  it is not clear why  vehicle  weight was not included in
 this model,  since  this variable  exhibited a greater correlation coefficient
 than did surface moisture.

 The regression  model  was  then run with traffic  density data from the field
 and values  for  the two other  variables from  laboratory tests.   The results
 were expressed as  values  for  a coefficient  and for individual exponents for
 each particle size (TSP,  IP, and PM-10).  Correlation coefficients were then
 calculated which revealed  that  a  large percentage of the variation in concen-
 trations was statistically related  to the  three variables.  Surface moisture
 was found to be the least dominant variable, becoming even less important at
 greater  distances  (50  meters versus 25 meters).

 Emission Factors and Control Efficiencies.    Two  approaches  were  used  to
 analyze  the effect  of controls,  focusing on the  data  for the 25-meter
 sampling distance.   (Apparently  because  of alternate  approaches  used,  the
 50-meter data was not  analyzed in this  manner.)   In  the first approach,  the
 regression  analysis was used to calculate particulate concentrations near the
 gravel surface.   This  analysis was  based on  six data points for watering and
 one for  no control.   The  resulting control  efficiencies are  on  the order of
 bU percent  but are of  limited  use, as they  are based  on such a  small  sample
 size.   Furthermore, no account was taken  of the  decay  in effectiveness  over
 11 me.

 The second  approach involved a laboratory experiment in which three represen-
 tative samples from field sites  were  watered in  an attempt  to  simulate  the
 desired  watering program  that  was  not carried out.   Two  application  rates
 were used:   the  field-measured  normalized  rate  and  a  rate twice  this amount.
 bravimetric  analysis  of  samples  before and  after  watering yielded surface
 moisture  values  which  were inserted  along  with the other  variables into  the
 regression  equations  to yield  concentrations  before  and after watering  and
 thus the  control efficiency.   The resulting  control efficiencies ranged from
 41  to 70 percent  (again without regard to  decay with  time); however, because
 these results  are  based on  a  laboratory  simulation,  they  cannot be applied
 directly to actual  field situations.

 The conclusions presented  in this study must be tempered with  an understand-
 ing  of  the actual  conduct of  the study  in  relation to the original  study
design.   One might justifiably  state that the study  has established that road
construction  activity  causes  a   "significant  and  measurable  impact on  the
ambient  air quality" in terms  of  fugitive  dust.  It is perhaps also fair to
conclude  that the regression  model, complete  with constants  for each particle
size distribution,  reasonably  relates  this  impact  to  surface silt content,
surface  moisture, and traffic density.  However, the study's conclusion that
                                     -15-

-------
approximately 50 percent control  can  be  achieved by watering is not totally
supported by the results of the study because one determination was based on
very limited field data for a  gravel  surface and the other determination was
based  on a  laboratory  experiment  that  may or  may not  accurately reflect
conditions for  an  actual  dirt road ;in a construction  area.   The laboratory
experiment is unreliable  in  that it  represents  a  hybrid  approach that sets
out to calculate an  ambient  concentration  from  data derived partly from the
field  and  partly  from  the laboratory.   This  conclusion  also  says nothing
about decay in effectiveness  with time.

The study's conclusion that watering  is  the only viable control strategy for
fugitive dust emissions  from  construction activities  is not warranted because
this hypothesis was  not  tested by this  study.    Also, the study's conclusion
concerning the  black  particulate  matter  found  on filters  is more of a hypo-
thesis to  be investigated at  a  later date, in  that the  association  of the
black  particles with  diesel  exhaust! is a supposition,  rather than a conclu-
sion supported by analysis.         !

Optional Investigations.   Two  optional investigations were conducted as part
of the demonstration study.  The first was  an emissions inventory  of the area
around the monitoring site.  The  only apparent and  direct  result  was that no
other significant sources of  fugitive dust  were found within  one kilometer of
the  construction  site.   The  contribution  of sources  within ten kilometers
of  the construction site  was  estimated  using  previously  developed emission
factor algorithms  for wind erosion,! vehicular  traffic, and  material storage
piles.   These results were published in an appendix,  but were not directly
applied to the analysis.

The  second optional  investigation was a  determination  of  mud or  dirt carry-
out, that is, reentrainment of dust carried onto paved roads by construction
traffic.   Samples were  taken by portable vacuum  cleaners  at  three nearby
sites.  Surface loadings for  each sample  were calculated,  and each sample was
analyzed in  the laboratory for silt  content.   These results were  presented
but,  again,  were  not  directly applied  to the  analysis.  Rather, they were
collected as the  first part  of a jmore comprehensive  study to have been
completed later.                    ;
Usefulness to State and Local Agencies.   A  true  Type  1  analysis  was  not
achieved in the Minneapolis study primarily  because of a  failure to implement
controls  during the  control  period! of  the study.   As  a  result,  indirect
analyses were used involving a laboratory analysis and the limited  field data
collected.     The latter  included  only  one  sample  for  the  control period
and  one other  for a  different  surface  from  that tested under  the uncon-
trolled  scenario.   Consequently,  the conclusions  drawn  from this  analysis
must  be considered  tentative  at best.   It  is  particularly  inadvisable to
apply the  conclusions  to actual  field sites  because  they are  based on
a hybrid  (field/laboratory) experimental  approach.

A  problem  raised  by the  study  is I that  of interference by  either  a point
source  or a traditional  fugitive source in the study area.   It is important
once  a  potential  interference has  been  found  that the source be  identified
                                     -16-

-------
 and  confirmed  so  that its contribution  can  be  removed  in  the analysis of
 results.    This  should have  been  done  once  a significant  amount  of  black
 particulate matter was  noted  on  the  sample  filters.    Speculation  on the
 nature  and  source  of  potential  interferences does  not enhance the usefulness
 of the  results.
 The  Minneapolis  study  could  provide  useful  data on  the
 particulates  from  road construction  under  uncontrolled
 not  measured directly, PM-10  values inferred from the
 relatively  accurate  because they  are determined  by
 cascade  impactor  particle  size  distribution  consisti
 sizes.   The  top  size  is 7 microns,  which  is relatively
 more,  the extrapolation  is done  for  large  particle
 for  the  smaller  particle  sizes where there is greater
 the  particle  bounce problem.
distribution of PM-10
conditions.   Although
 :test data  should  be
extrapolation from  a
ng  of five  particle
close to 10; further-
diameter rather  than
uncertainty  caused  by
 3.1.3  Portland, Oregon
The  Portland study  [4]  was  conducted  in 1981  under the direction  of the
Department of Public Works of the City of Portland,  Oregon.  The objective of
this study was to determine the effectiveness of daily vacuum sweeping of the
curb lane of city streets in controlling particulate concentrations.  Concen-
trations of particulates greater and less than 2.5 microns were determined in
the two  study  areas.  The  study was unable to demonstrate any relationship
between vacuum sweeping and air quality.
Study Design.   Previous  studies  investigating  the effectiveness  of vacuum
sweeping in  improving  the particulate air quality were  not conclusive
because  they  usually  relied on remote measurements of  TSP and because they
failed  to  account  for  independent  variables  relating  to  meteorology  and
traffic.

This  study employed  a receptor-oriented  approach, which  is appropriate
for this type of  study  because  emissions  are distributed over a  wide area
and are expected to vary significantly.   Dispersion modeling was to have been
used to project the results to  a wider area.

A Type  3 study design was  used,  in  which  two areas  were  monitored  for  two
three-month periods,  with  street  cleaning  practiced  in one  area  during  the
first period and in the other area during the second period.  Two areas were
selected after extensive review of candidates, including preliminary modeling
studies.   Each  study  period was  to  last three  months, and  was  designed  to
provide enough data for subsequent statistical analysis.

Two monitoring sites were located in one study area and one in the other.   A
fourth  site,  the  Continuous Air  Monitoring  Station  (CAMS.) operated by  the
Oregon  Department  of Environmental  Quality,  was selected as a regional
trend site.  Meteorological  and traffic data were collected at this site on a
routine  basis.   Each  monitoring  site was equipped with a  hi-vol  sampler,  a
                                    -17-

-------
low-vol  sampler,  and a  dichotomous  sampler with a  2.5  micron  size cutoff.
Sampling periods  were  24  hours in duration and appear to have  run  from
midnight to  midnight.   The  study was conducted during the dry season of May
through October.

Sampling and Analysis.   In  addition to collecting daily ambient air monitor-
ing and meteorological data within each study area, hourly and daily traffic
counts were  recorded in each study area,  as was vacuum sweeper performance
(daily sweeper logs were maintained  by  the  operators).   Road  dust samples
were collected from curbs, medians,  and traffic  lanes to check the efficiency
of street sweeping and to provide a  basis  for comparing the chemical analysis
of filters collected from the monitoring sites.

Quality assurance procedures  that |Were followed are  not  discussed  in detail
in the report but  are  extensively  referenced  to sources including the
Quality  Assurance  Handbook  published by  the EPA Office of Research and
Development.   No  major problems  associated with precision and repeatability
of instruments were reported.  Some  traffic  count data was lost due to damage
to the counting equipment by vacuum  sweepers.

Analysis of Results.   For each monitoring  site, the  concentrations  of  TSP,
IP, and FP were calculated  directly from the  hi-vol, lo-vol, and dichotomous
filter weights.  Filter  samples were also analyzed by X-ray  fluorescence for
17 elements.  The "geologic  component" was  determined for-composite road  dust
samples  as  well as  for lo-vol  and the  coarse portion  of  the  dichotomous
samples in  an  effort to correlate the two  by a chemical  mass balance  tech-
nique.   (The geologic component is that  component  associated  with elements
common to soil.)   Interferences due to industrial  emissions would be  high-
lighted by a large non-geologic  component.

Two additional  data sets were  generated.   The first was  a  set  including
only "dry" days, defined as days greater than one since the  last rainfall of
0.01 inch or more; the second was a set of  all  days in the period of June 25
to September 17.

The experimental  values were  normalized  to the CAMS  (background) value
by dividing the sample values by  the CAMS values for the same period.  Hence,
the results  of each  site were  normalized  to  those  of  the  regional trend
site.   Data  from each  of the three sites  were  then  subjected to  a  linear
regression analysis.               '
Emission Factors and Control  Efficiencies.
statistically  significant  effect ,on
sweeping.   In  other words,  given  the
differences in air quality between the two areas
variations in the generation of particulate matter
of the controls.   It  is  possible that larger sample sizes
in  statistically  significant  differences;  however, because  the number  of
samples  in  all  cases already  exceeds  30 samples,  it  is more  likely  that,
given the conditions of  the experiment,  vacuum sweeping  had  no significant
	   None of the three sites showed a
particulate  air  quality  from  vacuum
number  of  samples  used,  any measured
           can  be attributed to random
             rather than to the effect
                    may have resulted
                                     -18-

-------
 effect  on particulate  air  quality.   The  geologic  mass  data  suggest that, for
 two of the sites,  sweeping  may  actually increase the concentration of par-
 ticulates in  the ambient air.

 Because  the  results  of the study  did  not demonstrate a significant air
 quality improvement from sweeping,  the planned use of chemical mass balance
 techniques  to combine  meteorologic and  traffic variables to develop  a
 dispersion  model  capable of  predicting the area wide effect of sweeping was
 not conducted.

 Generally  speaking,  the  conclusions presented are well  founded in the
 results of the study.  A pertinent, point is made that an  earlier study citing
 beneficial  results  from sweeping was  based  on  sweeping the traveled areas of
 the roadways  in addition to the curb lanes; only the curb lanes were swept in
 this study.   Perhaps the reentrainment of dust  from the travel  lanes of paved
 roads   is  a  major  source  of  nontraditional  particulate  emissions.   Any
 further research  on vacuum  sweeping  should give   attention to this  area  of
 the road  surface.

 Optional  Investigations.    One  of   the  optional   investigations  undertaken
 in  the Portland study  involved elemental analysis of  samples by X-ray
 fluorescence.   Although conducted  primarily  to  support the  chemical  mass
 balance approach, this analysis was  useful  in identifying interferences.   For
 example,  geologic sources of  particulates generally exhibit aluminum-to-sili-
 con ratios of less than one.   However, it was found that  several samples  from
 one of  the sites in an industrial area showed  much higher ratios (as high  as
 30).  Although the  interfering source of aluminum  particulate was not found,
 all samples  with  ratios  exceeding one Were  deleted from the data base.

 In  order  to be able to rule  out  reentrainment caused  by the vacuum sweeper
 itself  as a  significant   source  of particulate  matter,  a  source  sampling
 method  (Oregon Method  8) was used to test sweeper  emissions under isokinetic
 conditions.  The results  showed sweeper emissions  to be on the order of 0.125
 pound per hour, a relatively  small amount.

 Usefulness to State and Local  Agencies.   As  a result of this study, state and
 local  officials  should approach  claims  made  for  vacuum sweeping of paved
 roads with caution.  The results of this study  firmly support the conclusion
 that daily vacuum sweeping  (at least as practiced  in Portland for"this study)
 does not  significantly improve  the  particulate  ambient air  quality.   An
 important finding may  be that no beneficial effect was  observed  from daily
 vacuum sweeping of curb lanes, as opposed to travel lanes.   Further  study of
 curb plus travel  lanes  is warranted.

Another important aspect  of the study was the observation of an interference
and  its  subsequent  removal from the  results.   The removal  would have  been
even more justifiable  if  the  interfering source  had  been identified.  The
measurement of  the  mineral  content  of collected   samples,  as practiced in
this study to reveal the interference, is  highly  desirable  if  not mandatory
for a study of this nature.  This type of analysis should be weighed  heavily
by  state  and  local  officials in  the evaluation  of  demonstration  studies
 referenced by  permit applicants.
                                    -19-

-------
The  investigation of the  vacuum sweeper itself  as  a possible interference
represents  an  appropriate  optional  investigation.  State and local officials
should scrutinize demonstration studies  for this  type  of evaluation of
whether the control method itself  represents a potential  interference.

3.2    OTHER STUDIES

Several  additional   case  studies are  presented  in  lesser detail than  the
foregoing major  studies.   The purpose is to further define the state of the
art for state and local  agency officials.
                                    i
3.2.1  Philadelphia, Pennsylvania   \

                                    \
The  Philadelphia study [5] was ari early  (1977) study of nontraditional
particulate emissions.   A major portion of  the  study consisted  of a liter-
ature  search,  an  emission  inventory,  an  analysis of  previously collected
ambient TSP concentration data, and ;an analysis of field samples for several
metals and for TSP levels as a function of location, such as heavily traveled
urban streets and major construction: sites.

A  receptor-oriented  demonstration  study of  street washing in  the downtown
area was  quite limited  and  was  intended to  be only  a  minor  portion of  the
study.  The application of  controls \consisted of intensive  water flushing of
all streets in a 10-block  downtown >area for three consecutive days  between
the hours of 7:00 AM and 6:30 PM.   TSP levels were measured by hi-vol  samp-
lers, recorded, and plotted  for the ^periods  before, during, and after flush-
ing.  Traffic volume in the  study area was also measured during this  period.

Because the control  period was onl^  three days,  and  because  no  attempt  was
made to statistically remove other  independent  variables, the results  of this
demonstration  study  were  not verifiable.  The tentative  conclusion  reached
was that  street  flushing  under these  conditions  actually  increased  the  TSP
concentration   immediately   after  flushing.     In  analyzing this  apparently
anomalous  result,  the  researchers  statistically  compared  TSP levels  with
traffic volume,  and found  a strong correlation between the  two.   The  re-
searchers  speculated that  the  street  flushing  program was   not  effective
because the violent flushing of traffic lanes toward curbs  and the splashing
of the as-yet  undrained  water by  vehicles essentially  redistribute previously
concentrated particles  and,  after  'drying,  the  particles  may become  reen-
trained.

Because this study was conducted  several years  ago, guidelines  for its design
and conduct were  not available.  Therefore,  this  control technology  can  not
be evaluated unless further  studies are undertaken and  reinforced with
complete statistical  analysis.
3.2.2  Bangor, Maine
This study  [6],  conducted for the  City  of Bangor, examined the effects of
vacuum sweeping  of  urban streets on  ambient  TSP concentrations during the

                                    -20-

-------
period 1980 to  1983.   The study made use of existing daily TSP measurements
determined by two  hi-vol  samples located in the  downtown  business  district
and presumably operated by the  city.   The data spanned four street-sweeping
seasons (March through October of each year), the first two years represent-
ing the uncontrolled  period.  When  the sweeping program was initiated,  data
were collected for the following two-year period.  A total of 223 data points
were  available for the uncontrolled period,  and 569  for the controlled
period.

The study  report  concludes that  the  vacuum sweeping program  resulted  in  a
reduction in ambient  TSP  concentrations  of  20  percent,  and that the. statis-
tical  analysis of  the data  supports this conclusion.   However, it is diffi-
cult  to evaluate this analysis because statistical  parameters such  as
standard deviations are not presented.  Furthermore, it does not appear that
a statistical  analysis of the independent variables was conducted to explain
differences in  air quality during  the two  periods.   Independent variables
such as meteorological conditions or construction  activity may have accounted
for some  of the  difference  in  TSP concentrations for  the two  data  sets.
3.2.3  Kansas City, Kansas
The Kansas  City  study  [7]  was  conducted to determine  the  effectiveness  of
twice-weekly sweeping and  flushing  of paved streets in  an  industrial  area.
The  study  was  receptor-oriented,  measuring ambient  concentrations  of  TSP
only.  Measurements were made before and  after controls were implemented in a
single downtown area.   A Type 1 design was used because an "identical" test
area could  not be found.  Hence, the  study was  subject  to  greater possible
uncertainties than one using a Type 2  or  3 design.  The study concluded that,
given the parameters  of  the  experiment,  there was  no statistically signifi-
cant reduction in  TSP concentrations  in  the  area due to regular cleaning  of
the streets.  A multiple regression analysis revealed that the four indepen-
dent variables thought to have a major influence  on TSP concentration had,  in
fact, a  relatively minor influence and  that  many other  variables were  in-
volved.
3.2.4  Lincoln, Nebraska
The Lincoln  study  [8]  was conducted by  the  Lincoln-Lancaster  County  Health
Department.  This  was  a receptor-oriented study in which  concentrations  of
TSP and  IP  were  measured  before  and after  application  of liquid  calcium
chloride to  unpaved  gravel  and crushed  rock  roads.   Calcium chloride  is  a
soluble  crystalline  salt  that  is  applied as  an  aqueous  solution for  dust
control.  A line  source formula was then used  to  convert  the measured
ambient  concentrations  into emissions on a mass per-vehicle-mile basis.   A
single application  of  calcium  chloride  was not found  to be  effective  for
TSP,  but to be effective in  reducing the  concentration  of  IP by 70 to  80
percent  (this reduction decayed  to  below  50 percent after about two-and-a-half
weeks).   However,  a 65 to  80  percent reduction  in TSP;  concentrations  was
                                     -21-

-------
 found when applications to  the  unpaved roads  were repeated at three to four
 week intervals.  The study  appears to  have been -a well-designed Type 1 study
 in  which  independent  variables  not  of  interest "we re  properly  removed  by
 statistical  analysis.   Furthermore,  the conclusions appear to be backed by an
 adequate statistical  analysis  of the  results.


 3.2.5  Clark County,  Nevada        !
                                   i
                                   i

 The Clark  County  study  [9]  was 'conducted by  the  Clark  County  Department
 of Health.   This  study  had three  major-objectives:

        o    To estimate the  relative^ost-effectiveness of magnesium chloride
             and Coherex for  dust control  (magnesium  chloride is a  relatively
             soluble crystalline  compound  employed as  an 'aqueous  solution;
             Coherex is  a petroleum resin used diluted with water)

        o    To study the  effectiveness of alternative  housing- construction
             practices in reducing ambient particulate levels
                                   i

        o   To study the effectiveness  of emulsified  asphalt on road shoulders
            in reducing ambient particulate levels.

 All  portions  of the study  were rec;eptor-oriented  and  measured  only  TSP.   The
 chemical  dust  suppression and housing construction  experiments were Type  2
 designs,  and the  road  shoulder experiment was a  Type  1 design.   Magnesium
 chloride was  found  to be  significantly less expensive to apply than Coherex;
 however,  no  information  was  provided concerning the  effectiveness of  the
 suppressants.   The  effectiveness  of  the unspecified alternative construction
 practices could not be  determined because the two  study  areas  were  too close
 to  one  another.  The effectiveness' of  the  road shoulder  technique  could  not
 be determined  because of proximity,of the study area  to  construction activi-
 ties.                                   .
3.2.6  Erie County, New York
The  Erie County  study  [10] was  conducted by  the Erie  County  Division of
Environmental  Control  to assess  the effectiveness of  several  dust control
strategies in  an  industrial  area.  [Three techniques were  assessed:  watering
of storage  piles  (actually, notation of the  effect  of rainfall), oiling of
unpaved  roads  and-parking lots,  and regular vacuum" "sweeping of paved roads.
TSP  was  the pollutant monitored  in this  receptor-oriented  study.   Several
hi-vol samplers were located throughout the  study area,  with  at least one
sampler  located  at each specific test site.  The study presented  great
potential  for interference among experiments, as they were all  located
relatively 'Close  to one  another.    The following conclusions  were offered
concerning the control  efficiencies of the  methods  tested:
                                     -22-

-------
       0
       0
       0

       0
          Watering storage  piles  — no conclusion possible
          Oiling unpaved roads .— 40 to 60 percent reduction in TSP
          Oiling parking-lots  --  40 to 70 percent reduction in TSP
          Vacuum sweeping paved  roads -.- 40 to  60  percent  reduction  in  TSP.

Several factors cast doubt on the  reliability of  these  conclusions.   First,
each test  consisted  of  only, a few measurements  (generally three  to  seven).
It does not appear that  a  rigorous  statistical  analysis  of test  results was
conducted.  Also, while  the .possibilities for.interference among  test areas
were numerous, the investigation  did  not  seek  to quantify the extent  of  this
interference.    Finally, the  long-term  reduction  in the  effectiveness of
oiling was not addressed.
3.2.7  Allegheny County, Pennsylvania
An  exhaustive  study  of nontraditional  particulate  emissions  at  U.S.  Steel's
Allegheny County steel  plants [11]  was conducted as part of a consent
decree with EPA.  A number of control methods were tested by surface sampling
of  road dust as well as by  the  exposure profile technique.  Exposure profil-
ing  is  a  technique  employing isokinetic sampling  of  several  points across -a
dust  "plume"  cross  section.  Statistical  analysis was -performed  on  both of
these source-oriented  results to  obtain  control  efficiencies.   Measurements
were made of FP, IP and JSP.                      ..-

A  control  efficiency of 90 to  95  percent was  given for control  of  unpaved
roads where the control of  paving  the  road and following with daily  vacuum-
ing.   Applying Coherex to unpaved roads also  achieved  90 to 95 percent
control  in  the short  term.   The  effectiveness of  a single application of
Coherex decreased  after approximately  2,0 days; on  the other hand, its
effectiveness was  found to  be cumulative with  successive applications.
However, Coherex is considered to  be relatively expensive.   For  paved roads,
the control efficiencies were as follows:  80 to 85 percent for daily  vacuum-
ing,  and  60 to 70  percent  for twice-weekly vacuuming  of  moderately  dusted
roads,;    •"-•--
roads.
       and  40 to  50 percent  for, twice-daily  vacuuming  of heavily-dusted
The  general  elements  of the study  design were  essentially  fixed by the
consent decree; Jiowever, the  specific  aspects  of  .the-design appear to have
been well  formulated  arid  executed., Of particular usefulness are the  develop-
ment of emission factor equations (through multiple linear  regression analy-
sis) for  the  various .control  techniques .tested,  and the comparison of the
emission factors calculate^ from these equations with those  calculated from
the experimental  data.                                            •
                                    -23-

-------
 3.2.8  The Road Carpet Study       ;


 The Road Carpet Study  [12]  was performed for EPA's  Industrial  Environmental
 Research Laboratory (IERL).   The study assessed  the  effectiveness  of a  fabric
 "road carpet"  to  control  emissions   from  unpaved  haul  roads.    Two Type  2
 design tests were  conducted  using controlled  and uncontrolled  contiguous  road
 segments.   The roads consisted  of ;a  crushed marble  road  at  a western  U.S.
 site and a crushed shale road in the  eastern  U.S.  Both tests  employed  hi-vol
 samplers  and  hi-vols  equipped  with  size-selective  inlets  to  measure  the
 ambient concentration  of TSP  and  IP.  The  results  for  individual  tests of
 control  efficiency were as  follows:
                                             Control  Efficiency
                                                 (percent)
Road type
Ma rbl e
Shale
Regiori
West'
East'
TSP
30-70
50-55
IP
30-50
65-75
 A significant  decay in  the  control  effectiveness  from  these  values  over
 several months  was also noted.  The study apparently did not employ extensive
 statistical  analysis  to  further analyze these results.  The conclusions must
 therefore  be tempered with the knowledge that  relatively  few tests were run
 and  many  must  have  been  of short duration.  (During  one  period, four tests
 were  run  in one day.)   Further testing with a thorough statistical analysis
 of results  is warranted before  employing the  results of this study.


 3.2.9  The  Iron and Steel  Plant Studjies

                                    i
 Two  studies of several  iron  and  steel plants have been conducted  for EPA's
 IERL.   The  first steel  plant study; [13] used exposure  profiling (a  source-
 oriented technique)  to measure the effectiveness  of  several  control  methods
 for open sources  of dust.  For each! experiment  in this  study,  a single area
was tested  both in  the presence  and> absence  of  controls.   The  study  concen-
trated more on the decay  of  control]  effectiveness over time than  on  a  sta-
tistical proof of the initial  effectiveness of each measure.  The researchers
apparently  reasoned   that  the  methods  had   been  previously  established  as
effective, at least over the short term.  Hence,  the purpose of  the study was
to demonstrate  the  change in  effectiveness over time.   The parameters meas-
ured were FP, IP, and TSP; the results may therefore be useful for comparison
with a future PM-10 regulation.

The treatment of  unpaved  roads with Coherex  was found to  be  over 90  percent
effective in the short term,  with a slow decay  over time.   The  efficiency of
                                     -24-

-------
 treatment  with water alone decayed  from  almost  100 percent to 60 percent in
 only five  hours.  The  efficiency  of both vacuum sweeping and water flushing
 of paved  roads  was  found to decay  rapidly from  about  50 percent,  while for
 flushing with  broom sweeping the  efficiency decayed  rapidly  from
 mately  70  percent.
approxi-
 The decays  in  effectiveness  were  not  quantified  or  discussed  explicitly
 in  the  report,  but  they were obvious from the emission factors developed from
 the exposure  profiling  results.    Conclusions  for  control  of  coal  storage
 piles were tenuous  at  best due to numerous experimental difficulties, such as
 interference  from precipitation and the contamination of filters by particles
 that  were not  airborne.   The study appears  to have been well  designed  and
 conducted; however,  future work should concentrate on repeated testing of the
 decay  in control  effectiveness  for  a specific  technique,  with  a  thorough
 statistical analysis of the results.

 The second or extended evaluation study [14] was also conducted for the IERL.
 The study was  designed to provide further information on the effectiveness of
 dust  suppressants  on unpaved  roads.  Exposure  profiling was  used  to measure
 emissions  under controlled and  uncontrolled conditions.   The  results  were
 expressed in  terms  of decay  rates  or lifetime of  effectiveness,  the latter
 expressed in  terms  of the number  of vehicle passes.   Variations  in  vehicle
 weight,  number of wheels and  vehicle  speed were statistically  removed by  a
 normalization  procedure  based on  a previously developed  algorithm  relating
 the  emission   factor  to these  three  independent  variables.    Effectiveness
 decay rates were determined by a  regression analysis, and were plotted  for
 each control  method.   The PM-10 fraction  was measured  in addition  to TSP, IP
 and FP.

 "Petro-Tac"  (an emulsified  asphalt)  on  an  unpaved road  was  found to have
 an  effective  lifetime  of  about 100,000  vehicle  passes for  the control  of
 TSP. Effective lifetime in this case was defined as  the time  required  for  the
 effectiveness  of the  control  to  decay essentially  to zero.   The  control
 efficiency of spraying water  was  initially  high  but  decayed at  a rate  of
 about 8 percent per hour.  For Coherex, the results  indicated  a life  of 7,500
 vehicle passes for a single application and  45,000  for  a repeat  application.
 As  with  the earlier study  of  this type, the  results are contingent upon  the
 formulations  and application  rates  used.

 The study appears  to have been well  designed and conducted.   A significant
 number of samples were taken,  and the analysis  of the  data is comprehensive.
 Results were  reported for IP,  FP and PM-10 as well  as  for TSP,  and the
 variations in  decay  as  a  function  of particle  size are discussed in detail.
 The information  on  effective  lifetimes perhaps would  have been  better pre-
sented as the  time  required to decay to 75 percent or 50  percent effective-
ness because,  as a  practical  matter, a  source would  have  to  reapply  the
controls long  before the  effectiveness  had  decayed to zero.
                                     -25-

-------

-------
                                Chapter 4

                               CONCLUSIONS
The  Workbook:   Demonstration Studies for Control of Nontraditional Particu-
late Sources contains explicit instructions  for the design  and execution of a
conclusive  and  cost-effective demonstration study of  control  technologies.
However, the studies  reviewed  here  exhibited  a  variety of problems in study
design  and  also in  execution  which resulted either in  no conclusions con-
cerning the control technologies  tested  or in  conclusions that were difficult
to support and not broadly applicable.  The following studies were conducted
appropriately without major deviations from  good  study procedures.  These are
Portland, Oregon;  Lincoln,  Nebraska;  Kansas City, Kansas;  Allegheny County,
Pennsylvania; and the  Iron  and Steel  Plant  studies.   However,  the remaining
studies suffered from inadequate  designs, in some cases because the Workbook
guidelines were  not  available.  The  following  cautions  and recommendations
are  offered  to both  study designers and those  who must  evaluate studies,
particularly studies offered in support  of bubble proposals or  similar plans
for particulate compliance:

       o    The Workbook has  been  assessed  and found to be  an excellent
            guidance  document  which,  when  carefully  followed,  will  greatly
            enhance  the  likelihood of  success  of a demonstration study.

       o    A proper demonstration  study is  expensive to  design  and conduct;
            however,  a study that is not done properly will  produce  results
            that  cannot be  applied  in the field.   Therefore,  there must be  a
            firm  commitment  to a  complete  and  proper  study and  effective
            coordination  and management of the study throughout.

       o   A demonstration study must  be extremely  well   designed in  order
           to  produce usable  results;   furthermore,  the  execution  of the
           study must rigorously follow the design.  Cost-cutting  ideas for
           fewer  monitors,  less  frequent  application  of controls  and so
           forth must be avoided  so as not to invalidate the  results and the
           subsequent statistical analysis.

       o    In order to be applicable in  the  field, results  from a demonstra-
           tion study must  be  statistically significant.   The study design,
           therefore, must  ensure that  the  statistical  analysis selected is
           correct  for the type of study  and that  a  sufficient number of
           samples are  taken.  Careful  coordination  and control  of the
           control/no control  arrangements  of  the study  are also  required.
                                    -27-

-------

-------
                                REFERENCES
  1.  Workbook:   Demonstration  Studies for Control of  Nontraditional  Partic-
     ipate  Sources.   Contract No.  68-02-2535,  Work  Assignment No. 15.   For
     U.S. Environmental Protection Agency, Office of Air Quality Planning  and
     Standards, Control Programs Operations Branch, March 1981.
 2.  Denver Demonstration  Study'.   PEDCo Environmental,
     297337.   For  Colorado  Division of  Air  Pollution
     October 1981.
 5.
 8.
 9.
10.
11.
                                                       Inc.   Contract No. C
                                                       Control,  Denver  CO,
3.  Kinsey, John Scott, Midwest Research Institute,  and
    Engineering, Inc.   Study  of  Construction Related
    Minnesota  Pollution  Control  Agency,  Roseville,
                                                         Jirik, Alan  J.,  ETA
                                                         Dust  Control.    For
                                                         MN, December 1982.
 4.  Portland  Road  Dust Demonstration Project.   Seton,  Johnson, and Odell,
     Inc.   For City of Portland, OR, Department of Public Works, July  1983.
    Record,  Frank  A.,  and  Bradway,  Robert  M.,  GCA  Technology  Division.
    Philadelphia  Particulate  Study.    Contract No.  68-02-2345.    For  U.S.
    Environmental  Protection Agency,  Region  III, Air  Programs Branch,
    Philadelphia,  PA,  June 1978.

    Hewitt,  T.R.,  Sirrine Environmental Consultants, Inc.  The Effectiveness
    of Street  Sweeping for Reducing Particulate Matter Background Concentra-
    tions, Research Triangle Park, NC, 1981.

    Study  of Street Cleaning  Impact  on  Particulate Levels in  Kansas  City,
    Kansas.   PEDCo Environmental, Inc.  Contract  No.  68-02-2535.   For  U.S.
    Environmental  Protection  Agency,  Region  VII,  Kansas  City,  MO,  April
    X z/O X •

    Walsh, Gary  L.,  Lincoln-Lancaster  County Health  Department.   Control  of
    Fugitive Dust  From Unpaved Roads Using Liquid Calcium Chloride.   For the
    U.S. Environmental  Protection Agency, October  1979.
    1980  Annual  Report  on Status
    Valley.   Clark County Health
    Las Vegas, Nevada, July 1981.
of Air Pollution Control  in the  Las  Vegas
District, Air Pollution Control  Division,
    Demonstration  of Fugitive Dust  Control  Effectiveness on Selected  Open
    Dust  Sources  in South  Buffalo,  NY.   County  of Erie,  Department  of
    Environment and  Planning, 1981.

    Roffman, Amiram,  et  al,  Energy Impact Associates.  A  Study  of  Control-
    ling  Fugitive  Dust Emissions  from  Non-traditional Sources of the  United
    States  Steel  Corporation Facilities in Allegheny County, Pennsylvania.
    Prepared for United States Steel  Corporation, December 1981.
                                   -29-

-------
12.  Tackett,  K,M>,,  Blackwood,  T.R.,  and Hedley,  W.H.,  Monsanto  Research
     Corporation.   Evaluation of Road  Carpet for Control  of Fugitive Emis-
     sions From  Unpaved  Roads.   Contract  No.  68-02-3107.   For U.S. Environ-
     mental Protection Agency,  Industrial  Environmental Research Laboratory,
     Research Triangle Park, NC, October 1980.

13.  Cuscino,  Thomas,  Jr., Muleski;  Gregory  E., and  Cowherd,  Chatten, Jr.,
     Midwest Research  Institute.   Iron and Steel  Plant Open Source Fugitive
     Emission Control Evaluation.  Contract No. 68-02-3177, Assignment No. 4.
     For  U.S.  Environmental   Protection  Agency,  Industrial  Environmental
     Research Laboratory, Research Triangle Park, NC, August  1983.

14.  Muleski,  Gregory  F., Cuscino,iThomas,  Jr., and  Cowherd,  Chatten, Jr.,
     Midwest Research  Institute.   Extended  Evaluation of  Unpaved  Road Dust
     Suppresants in  the  Iron  and Steel  Industry.  Contract  No.  68-02-3177,
     Task 14.   For  U.S.  Environmental  Protection Agency, Industrial  Environ-
     mental Research  Laboratory,  Research Triangle Park,  NC, February 1984.
                                     -30-
                                              U.S. GOVERNMENT PRINTING OFFICE: 1985-559-111/10762

-------

-------
CD
3  ^.

0)  O
CD

c
en
CD
CO
o
o
m
-o


o
0
CO
oo
o
OJ














C 0. =S

S 01 O
2 o c
«o 0
?3 i
o 5 =r
O — ' *••
^ V> o
S o °
Ml
01 O

_ o>
en v>
£ CD
1 3
|I
o> o
n. X
a. m
a. n
3 ^
8 m
5S
og-3
=•2  m i

                   01
                   N>
                   o>
                   00

-------